Deliverable D7.3 - Distributed generation on European ...s3.amazonaws.com/zanran_storage/ · - In...

DISPOWER Del_2005_0079 - 2005-12-27

Project funded by the European Commission under the 5th (EC) RTD Framework Programme (1998 - 2002) within the thematic programme

"Energy, Environment and Sustainable Development"

Project: DISPOWER

Contract No. ENK5-CT-2001-00522

Deliverable D7.3 - Distributed generation on European islands and weak grids - Public Report

Document Type: Deliverable

Author: Regine Belhomme

Company: EDF Electricité de France

Address: 1, Avenue du Général de Gaulle, 92141 Clamart Cedex, France

Tel.: +33-1-47 65 3860

Fax: +33-1-47 65 3218

Email: regine.belhomme@edf.fr

Further Authors: S. White, N. Hatziargyriou, A. Tsikalakis, D. Lefebvre, T. Ranchin, F. Fesquet, G. Arnold, S. Tselepis, A. Neris, T. Degner, P. Taylor

Reviewer: Eduardo Navarro

Approver: Regine Belhomme

Document Information

Document Name: del_2005_0079 Rev.Date: 2006-01-17

Classification: R0: public Status: S0: approved

Abstract: This document is the public version of Deliverable 7.3 “Distributed generation on Islands and weak grids” of the DISPOWER European project. It describes the work done and the results obtained in the 6 case studies carried out in Work Package 7b of DISPOWER: - Increased wind energy penetration on weak networks in rural northern England (Task 7.7). - Diesel/hydro generation on the Scottish island of Rum (Task 7.8a). - Islanded operation of a wind farm at Blyth in the UK (Task 7.8b). - Application of MORE CARE control system on the islands of Crete and Kythnos (Task 7.9). - Increased penetration of renewable energies in the Guadeloupe power system (Task 7.10). - Interconnection of solar powered mini-grids to the main grid on the island of Kythnos (Task 7.11).

DISPOWER Del_2005_0079 - 2005-12-27

EXECUTIVE SUMMARY

In the context of the DISPOWER European project, the main objective of Work Package 7 (WP7) was to demonstrate the implementation of distributed generation (DG) technology in national, regional or local grids in Europe and to apply the tools and concepts developed in other work packages of the DISPOWER project.

More specifically, WP7 was divided into two parts :

- WP7a dealing with distributed generation in European interconnected grids, and coordinated by IBERDROLA. Five case studies were carried out in WP7a on the interconnected grids in Germany, France, Spain and Austria. The work done and the results obtained are described in detail in DISPOWER Deliverable 7.2 and summarized in a paper presented at 10th Kasseler Symposium on Energy Systems Technology, 2005.

- WP7b dealing with distributed generation on European islands and weak grids, and coordinated by EDF. Six case studies were carried out in WP7b on weak grids and island power systems in the United Kingdom, Greece and the French West Indies. The work done and the results obtained are summarized in a paper presented at 10th Kasseler Symposium on Energy Systems Technology (2005) and described in the present report which constitutes the public version of DISPOWER Deliverable 7.3.

WP7b is thus dedicated to the implementation of DG technology on islands and weak grids and its specific objectives inside the DISPOWER project are : - to apply tools and concepts developed in the DISPOWER project on islands and weak grids in

different European countries, - to demonstrate the implementation of DG and renewable energy sources (RES) technology on

this type of power systems, - to contribute to the dissemination and exploitation of the DISPOWER results.

Due to the particular structure and characteristics of weak grids and islands, the integration of DG and RES faces important problems and constraints which lead to limitations of their penetration level on this type of power systems. This is particularly true for renewable energy sources (such as wind energy) because of their high variability and rather unpredictable nature.

Therefore case studies carried out in DISPOWER have investigated the problems and constraints met by DG and RES on weak grids and islands and have studied different solutions to overcome some of the existing barriers to a larger development of DG in such grids.

Three main types of case studies were done. The first type dealt with the use of load management and 3 applications were investigated in the UK. More specifically:

- A load management technique was considered to mitigate voltage rise on a rural network in the North East of England in order to facilitate connection of increased capacity of wind generation (Task 7.7). The results showed that load control is indeed an effective strategy. For the studied case, if 20% of the loads are controlled, the network could accommodate 900 kW of wind power instead of 300 kW without load control.

- The use of Distributed Intelligent Load Controllers (DILC) was investigated to control system frequency on the Scottish Island of Rum (Task 7.8a). The simulations showed that the approach was feasible. However, extensive site tests have shown larger voltage changes than anticipated and further developments of the DILCs are needed.

- A solution based on load control combined with a synchronous compensator was designed for the islanded operation of a wind farm (Task 7.8b). Testing was carried out on an islanded network composed of test equipment. The results demonstrated that voltage and frequency on the islanded network could be controlled within acceptable limits but wind turbine start up might be difficult.

The second type of studies focused on the impact of increased DG and RES penetration on islands, the identification of the resulting problems and constraints, and the analysis of possible solutions :

DISPOWER Del_2005_0079 - 2005-12-27

- In the Les Saintes islands case study, existing solutions to improve wind energy penetration have been studied, using the new control and safety concepts developed in DISPOWER (Task 7.10). The following aspects were considered: voltage rise and deviations, flicker, DG coupling/ decoupling, fault ride through capability, short-circuits. In particular, new design and control techniques were applied in order to decrease the out of limit voltage and the high flicker level on the grid and the impact of various wind turbine technologies was evaluated. The possible contribution of GIS tools was examined to devise a complete decision tool, interfaced with a Network Management System.

- For the Kythnos island, grid connection issues for an increased penetration of RES were examined (Task 7.11). Simulation results showed that the power system would present a stable behavior, as long as an adequate spinning reserve exists. Due to the high cost of energy produced by diesel generators a reduction of this reserve is highly desirable without sacrificing the safety of the operation. Two solutions were then examined: the use of power electronic inverters with disturbance ride-through capability and the use of battery storage (to replace the spinning reserve provided by the diesel generators). Both solutions have proved to be feasible.

Finally, the MORE CARE advanced control software developed in previous European projects was applied on two Greek island power systems with high RES penetration, namely the islands of Crete and Kythnos (Task 7.9). The impact and the benefits of the economic scheduling, forecasting and on-line dynamic security assessment functions of MORE CARE were evaluated and compared to the actual operation of the system. The savings obtained in operation and fuel costs were estimated.

To summarize, the results of the case studies have shown that:

- Load management could be an effective strategy to increase DG and RES penetration in islands and weak grids.

- Depending on the problems and constraints met, different solutions can be applied to increase DG and RES on islands and weak grids. In particular, the use of energy storage systems, of power electronic interfaces (of the DG units) or of power electronic devices (for instance Statcom) may prove to be very effective. However, economic aspects also have to be taken into account, and simpler but cheaper solutions are often chosen at the end even if they are not so optimal.

- In islands and weak grids, DG could prove profitable from the economic point of view.

- The increased integration of DG and RES in weak grids and particularly in island grids depends on several important factors such as: • The provision of ancillary services by DG and RES plants, in particular concerning

contribution to the control of network voltage and frequency. • The Fault-Ride-Through (FRT) capabilities of DG and RES units, that’s to say their

capability to withstand network disturbances such as voltage dips and frequency variations. • The use of appropriate monitoring and control systems with relevant communication means

for the DG and RES plants or for the network as a whole. • The availability of appropriate forecasting tools for RES power generation. • And last but not least the motivation and willingness of all the players involved not only at

the local or regional levels but also at the national level.

Very promising results have thus been obtained in the case studies. However further work is still needed for instance concerning:

- load management: load control strategies should still be improved and more thorough investigations on all the potentials and applications of load control should be carried out,

- solutions to achieve a better integration of DG and RES in islands and weak grids: • technical and economic issues related to the use of energy storage systems should be further

investigated,

DISPOWER Del_2005_0079 - 2005-12-27

• ancillary service provision by DG and RES may be critical on islands and weak grids; appropriate and economic solutions are thus needed,

• appropriate monitoring and control systems for DG and RES plants, as well as for the grid operation should be further investigated and tested,

• even if some DG technologies already have FRT capabilities, it is not the case for all of them. In particular RES technologies intended for small islands do not generally have such capabilities and disconnect for voltages dips. Again, effective and economic solutions are needed.

The case studies in WP7b were carried out by the following DISPOWER partners: - Task 7.7, Task 7.8a and Task 7.8b: Econnect, - Task 7.9: ICCS/NTUA, CRES, - Task 7.10: Vergnet, ISET, EDF, ENSMP-CENERG, University Kassel, - Task 7.11: CRES, Econnect, ISET. Full contact information for the above partners is given in the report.

DISPOWER Del_2005_0079 - 2005-12-27

Table of contents

1 INTRODUCTION............................................................................................................ 8

2 INCREASED WIND ENERGY PENETRATION ON WEAK NETWORKS IN RURAL NORTHERN ENGLAND (ECONNECT) .............................................................. 9

2.1 Introduction.................................................................................................................. 9

2.2 The Haydon network.................................................................................................. 10 2.2.1 Simulation Model ............................................................................................................ 10

2.2.1.1 Case study 1: No load control system..................................................................... 11 2.2.1.2 Case study 2: Load control used to control 50% of the total load ......................... 13 2.2.1.3 Case study 3: Load control used to control 20% of the total loads......................... 15

2.3 Conclusions................................................................................................................ 17

3 DIESEL/HYDRO GENERATION ON THE SCOTTISH ISLAND OF RUM (ECONNECT)......................................................................................................................... 18

3.1 Introduction................................................................................................................ 18

3.2 The electrical system on the island of Rum ............................................................... 18

3.3 Development of a model of the Rum network........................................................... 20

3.4 Simulating new control concepts for Rum................................................................. 22

3.5 Installing control concepts on the Rum network........................................................ 23

3.6 Communications ........................................................................................................ 24

3.7 Summary of simulation and site test work................................................................. 26

4 ISLANDED OPERATION OF A WIND FARM AT BLYTH IN THE UK (ECONNECT)......................................................................................................................... 27

4.1 Introduction................................................................................................................ 27

4.2 Design Considerations ............................................................................................... 27 4.2.1 Requirements ................................................................................................................... 27 4.2.2 Frequency control ............................................................................................................ 27 4.2.3 Reactive power compensation and voltage control ......................................................... 28

4.3 System Simulation ..................................................................................................... 28 4.3.1 Wind Turbine................................................................................................................... 28 4.3.2 Synchronous Compensator and Automatic Voltage Regulator (AVR) ........................... 28 4.3.3 Control Loads and Distributed Intelligent Load Controllers ........................................... 29 4.3.4 Simulations carried out .................................................................................................... 29 4.3.5 Simulation results ............................................................................................................ 29

4.4 Islanding Demonstration Using Windmaster Wind Turbine ..................................... 30 4.4.1 Overview of the demonstration system ........................................................................... 30 4.4.2 Windmaster wind turbine ................................................................................................ 30 4.4.3 System design and development...................................................................................... 31 4.4.4 Equipment commissioning and installation..................................................................... 32 4.4.5 Test results....................................................................................................................... 34

4.5 Analysis of starting performance - updated modelling results .................................. 39

4.6 Discussion .................................................................................................................. 41

4.7 Conclusions................................................................................................................ 42

DISPOWER Del_2005_0079 - 2005-12-27

5 APPLICATION OF THE MORE CARE ADVANCED CONTROL SYSTEM ON THE ISLANDS OF CRETE AND KYTHNOS (ICCS, CRES) ......................................... 43

5.1 Introduction................................................................................................................ 43

5.2 Results from the Case studies .................................................................................... 45 5.2.1 Kythnos Case Study......................................................................................................... 45

5.2.1.1 Economic Dispatch (ED) results............................................................................. 46 5.2.1.2 Adequacy-Reliability of supply .............................................................................. 46 5.2.1.3 Impact of Forecasting Errors in operating costs .................................................... 47 5.2.1.4 Impact of energy Storage ........................................................................................ 48 5.2.1.5 Operating Scenario 1- OS 1 .................................................................................... 48 5.2.1.6 Operating Scenario 2 – OS 2 .................................................................................. 49 5.2.1.7 Operating Scenario 3 – OS 3 .................................................................................. 49 5.2.1.8 Results..................................................................................................................... 49 5.2.1.9 Discussion of the results ......................................................................................... 50

5.2.2 Crete Case Study ............................................................................................................. 50 5.2.2.1 Results from Cretan power system operation ......................................................... 50 5.2.2.2 Comparison of the scenarios studied ...................................................................... 52 5.2.2.3 Conclusions for Cretan power system .................................................................... 53

5.3 Conclusions................................................................................................................ 54

6 INCREASED PENETRATION OF RENEWABLE ENERGIES IN THE GUADELOUPE POWER SYSTEM (VERGNET, ISET, EDF, ENSMP-CENERG, UNI KASSEL)................................................................................................................................. 55

6.1 Introduction................................................................................................................ 55

6.2 Grid calculations of the “Les Saintes” network ......................................................... 57 6.2.1 Calculation tools .............................................................................................................. 57 6.2.2 Basic data......................................................................................................................... 57 6.2.3 Assumptions .................................................................................................................... 58 6.2.4 Voltage levels .................................................................................................................. 59 6.2.5 Flicker levels.................................................................................................................... 61 6.2.6 Voltage variations due to switching effects..................................................................... 62 6.2.7 Short circuit levels ........................................................................................................... 62 6.2.8 Grid capacity utilization .................................................................................................. 62 6.2.9 Fault ride through ............................................................................................................ 64 6.2.10 Results evaluation with respect to regulations of different countries .............................. 64

6.3 Solutions to improve wind penetration ...................................................................... 65 6.3.1 Possible measures to limit the grid voltage deviation...................................................... 65

6.3.1.1 Grid reinforcement.................................................................................................. 66 6.3.1.2 Connection of the Wind Farm to another Grid Node of the Island......................... 66 6.3.1.3 Adjustment of the Wind Turbines Power Factor .................................................... 66 6.3.1.4 Variation of the High Voltage Transformer Tap Position ...................................... 67 6.3.1.5 Limitation of wind power ....................................................................................... 67 6.3.1.6 Change of wind turbine technology ........................................................................ 68 6.3.1.7 Distributed load control .......................................................................................... 69

6.3.2 Possible measures to reduce the Long Term Flicker ....................................................... 69 6.3.2.1 Change of wind turbine technology ........................................................................ 69 6.3.2.2 Additional Statcom device ...................................................................................... 69 6.3.2.3 Pitch regulation improvement................................................................................. 70

6.3.3 Fault ride through capabilities ......................................................................................... 71 6.3.3.1 Use of variable speed or statcoms........................................................................... 71 6.3.3.2 Advanced pitch regulation ...................................................................................... 71

DISPOWER Del_2005_0079 - 2005-12-27

6.4 Possible uses of G.I.S. tools ....................................................................................... 73

6.5 System operation after the loss of sub-sea cable........................................................ 74 6.5.1 Description of the system ................................................................................................ 74 6.5.2 Grid control techniques ................................................................................................... 75

7 INTERCONNECTION OF SOLAR POWERED MINI-GRIDS TO THE MAIN GRID ON THE ISLAND OF KYTHNOS (CRES, ECONNECT, ISET) ......................... 78

7.1 Introduction................................................................................................................ 78

7.2 Kythnos power system ............................................................................................... 78

7.3 Examination of steady-state issues ............................................................................ 79 7.3.1 Preliminary study............................................................................................................. 79 7.3.2 Kythnos island study ....................................................................................................... 79

7.3.2.1 Study Cases............................................................................................................. 80 7.3.2.2 Summary of the simulation results.......................................................................... 81

7.4 Examination of dynamic issues.................................................................................. 81 7.4.1 Disconnection of PV distributed systems due to under voltage protection ..................... 81 7.4.2 System response with wind turbines................................................................................ 82

7.5 Conclusions................................................................................................................ 82

8 CONCLUSIONS............................................................................................................. 84

9 CONTACT INFORMATION ....................................................................................... 86 9.1 Tasks 7.7, 7.8a and 7.8b (Chapters 2, 3 and 4) .......................................................... 86

9.2 Task 7.9 (Chapter 5)................................................................................................... 86

9.3 Task 7.10 (Chapter 6)................................................................................................. 86

9.4 Task 7.11 (Chapter 7)................................................................................................. 87

10 REFERENCES............................................................................................................... 88

DISPOWER Del_2005_0079 - 2005-12-27

1 Introduction In the context of the DISPOWER European project, the main objective of Work Package 7 (WP7) was to demonstrate the implementation of distributed generation (DG) technology in national, regional or local grids in Europe and to apply the tools and concepts developed in other work packages of the DISPOWER project.

More specifically, WP7 was divided into two parts:

- WP7a dealing with distributed generation in European interconnected grids and coordinated by IBERDROLA. Five case studies were carried out in WP7a on the interconnected grids in Germany, France, Spain and Austria. The work done and the results obtained are described in detail in DISPOWER Deliverable 7.2 [1] and summarized in [2].

- WP7b dealing with distributed generation on European islands and weak grids and coordinated by EDF. Six case studies were carried out in WP7b on weak grids and island power systems in the United Kingdom, Greece and the French West Indies. The work done and the results obtained are summarized in [2] and described in the present report which constitutes the public version of DISPOWER Deliverable 7.3. A full detailed version of Deliverable 7.3 was also prepared for the DISPOWER consortium [3].

WP7b is thus dedicated to the implementation of DG technology on islands and weak grids and its specific objectives inside the DISPOWER project are : - to apply tools and concepts developed in the DISPOWER project on islands and weak grids in

different European countries, - to demonstrate the implementation of DG and renewable energy sources (RES) technology on

this type of power systems, - to contribute to the dissemination and exploitation of the DISPOWER results.

Due to the particular structure and characteristics of weak grids and islands, the integration of DG and RES faces important problems and constraints which lead to limitations of their penetration level on this type of power systems. This is particularly true for renewable energy sources (such as wind energy) because of their high variability and rather unpredictable nature.

The case studies of WP7b investigate the problems and constraints met on weak grids and islands and study different solutions to overcome some of the existing barriers to a larger development of DG and RES in such grids. More specifically,

- 3 case studies investigate the use of load control for different applications in the UK : 1) a load management technique is applied to mitigate voltage rise on a rural network in

order to enable increased wind energy penetration (Chapter 2), 2) Distributed Intelligent Load Controllers (DILC) are used to control system frequency on

the Scottish island of Rum (Chapter 3), 3) a solution based on load control combined with a synchronous compensator is designed

and tested for the islanded operation of a wind farm at Blyth (Chapter 4).

- The MORE CARE control software developed in previous European projects is applied on the Greek island power systems of Crete and Kythnos, and the benefits of its use are estimated (Chapter 5).

- Grid connection issues for increased RES penetration are studied on the two island power systems of Les Saintes in the French West Indies (Chapter 6) and Kythnos in Greece (Chapter 7). Different problems are identified and possible solutions are studied, e.g. battery storage, use of the power electronic inverters of the RES units, new design and control techniques, improved RES technologies (in particular for wind turbines), …

Finally, the use of geographical information system (GIS) tools to devise a complete decision tool is also examined in the les Saintes case (Chapter 6).

DISPOWER Del_2005_0079 - 2005-12-27

2 Increased wind energy penetration on weak networks in rural northern England (Econnect)

2.1 Introduction This chapter is a summary of the work undertaken under Task 7.7 in WP7b of the DISPOWER European project. Task 7.7 related to a weak network scenario, to investigate the potential for increased wind penetration in combination with distributed intelligent load controllers.

In many rural areas that are ideal for wind energy generators the distribution network is weak. This limits the amount of wind energy that can be connected to the network due to problems such as steady state voltage rise. Load management has the potential to enable increased amounts of wind generation to be connected onto a weak distribution network. This study used computer simulations to investigate this technique for a weak network in the North East of England. This technique has the potential to substantially increase the practicable wind energy resource of the European Union. The load control strategies for grid-connected systems developed in WP1 were applied in this study.

The area of grid used in the study covered a wide and sparsely populated rural region to the west of Newcastle upon Tyne. The primary substation at Hexham was supplied by 66kV circuits. From these, a series of 20kV feeders supplied outlying villages and rural dwellings in the South West of Northumberland. A schematic of the Haydon feeder is shown in Figure 1.

Figure 1 - Schematic of Haydon feeder used in the study

DISPOWER Del_2005_0079 - 2005-12-27

2.2 The Haydon network The Haydon network is owned and operated by CE Electric. It is a 20kV distribution feeder in a rural area with several thousand customers fed from 280 MV/LV transformers with ratings from 7kVA to 600kVA as shown in Figure 1. This network will be used to study how load controllers can be used to facilitate the growth of wind turbines on the distribution network.

2.2.1 Simulation Model A power system analysis tool called Viper was used to model the Haydon network and to study the effectiveness of using load management in promoting the penetration of renewable sources on the network. This network model consists of 16 MV busbars and 45 LV distribution feeders. Each MV busbar is connected to 3 LV feeders. The details of how each MV busbar is connected to each LV feeder are shown in Figure 2.

Figure 2 - LV distribution feeder

The parameters of the Haydon network required for simulation studies were obtained from CE Electric (formerly Northern Electric Distribution). It is worth noting that minimum load demands are used in the simulation studies. This is because the most significant voltage rise effects are relevant to these studies and are likely to happen during minimum load demand conditions.

Three case studies were performed based on the following assumptions:

1. Installation of wind turbines is assumed to take place at a busbar (busbar 7) in the vicinity of which, according to historical wind speed measurements from 1995 to 1997, wind speed is relatively high.

2. In each simulation study, the wind turbine exports its maximum power to the network.

3. None of the loads controlled by load management make any contribution to the minimum power demands.

4. Power factor at each node is assumed to lie between 0.94 and 0.97, which happen during the summer period.

5. Tap positions of all MV/LV transformers are set such that the voltage levels at LV feeders are close to their upper statutory limit.

6. The upper statutory limit is assumed to be 1.038pu on MV and LV networks.

The details of each case study are described in the following subsections.

DISPOWER Del_2005_0079 - 2005-12-27

2.2.1.1 Case study 1: No load control system In the first case study, a number of load flow calculations were carried out without using load management. For each load flow calculation, the size of the wind turbine at busbar 7 was specified and then voltage magnitudes across the network were obtained and plotted. Figure 3 shows the voltage profile during no load condition.

Figure 3 - Voltage profile during no load condition

Figure 3 shows that, during no load condition, voltage magnitudes at all LV busbars are 1.038 pu, which is the upper statutory limit. This is because the tap positions of all MV/LV transformers were set such that the voltage magnitudes at LV networks are near to 1.038 pu. Figure 4 shows the voltage profile across the network during the minimum load condition.

Figure 4 - Voltage profile during minimum load condition

Figure 4 shows that the voltage magnitudes across MV and LV networks are below the statutory upper limit (1.038 pu). This voltage profile is likely to be seen during the summer period. Figure 5 shows the

DISPOWER Del_2005_0079 - 2005-12-27

voltage profile during a maximum generation and minimum load demand if a 300 kW wind turbine is connected to busbar 7.

Figure 5 - Voltage profile during maximum generation and minimum load conditions, for a

300kW wind turbine

Figure 5 shows that, as a result of the maximum power output of the 300kW wind turbine at busbar 7, the voltage profile across MV and LV networks becomes higher than that shown in Figure 4. However, the voltage profile remains within statutory limits. Figure 6 shows the voltage profile during maximum generation and minimum load demand conditions if a 900 kW wind turbine is connected at busbar 7.

900kW wind turbine

DISPOWER Del_2005_0079 - 2005-12-27

Figure 6 shows that some of the voltage magnitudes are slightly higher than the statutory limit (1.038 pu). CE Electric may not permit this size of wind turbine (900 kW) to be installed on busbar 7 because the maximum power injection from this wind turbine will cause some of the voltage magnitudes to rise above the limit during minimum load condition. Figure 7 shows the voltage profile when a 1.5 MW wind turbine is connected to busbar 7.

1.5MW wind turbine

This figure shows that the voltage profile due to the installation of the 1.5 MW wind turbine is much higher than that shown in Figure 6. Figure 5, Figure 6 and Figure 7 show that, without any load control, the network can accommodate a 300kW wind turbine at busbar 7 without experiencing any voltage rise problems.

2.2.1.2 Case study 2: Load control used to control 50% of the total load In this second case study, a number of load flow calculations were performed. In each load flow calculation, the voltage profile of the Haydon network was determined during the following conditions.

- The wind turbine specified on busbar 7 exports maximum power to the network.

- The load management system adds 50% of the total load to the minimum power demand, which was specified in the model before load management was used. Therefore, in periods of minimum demand and maximum power generation, the load management system increases the load on the system to a level such that an additional 50% of total load is added during these conditions.

Figure 8 shows the voltage profile of the Haydon network when a 900 kW wind turbine is installed on busbar 7. Figure 8 shows that, by using load management to add 50% of the total load to the LV networks, the voltage profile is maintained below the statutory upper limit (1.038pu) as compared to that without using load management.

Figure 9 shows the voltage profile if a 1.2 MW wind turbine is installed on busbar 7. Figure 9 shows that the voltage profile of the network is close to the statutory upper voltage limit if a 1.2 MW wind turbine is connected to busbar 7.

DISPOWER Del_2005_0079 - 2005-12-27

Figure 8 - Voltage profile during maximum generation and 50% controlled load, for a 900kW

wind turbine

Figure 9 - Voltage profile during maximum generation and 50% controlled load, for a 1.2MW

wind turbine

Figure 10 shows the voltage profile if a 1.5 MW wind turbine is connected to busbar 7. This figure shows that some of the voltage magnitudes are slightly above the statutory upper limit. This means that, if load management is used to control 50% of the total loads on the network, the maximum size of the wind turbine that can be installed to busbar 7 is about 1.5 MW, which is 5 times greater than that without using load management.

DISPOWER Del_2005_0079 - 2005-12-27

Figure 10 - Voltage profile during maximum generation and 50% controlled load, for a 1.5MW

wind turbine

2.2.1.3 Case study 3: Load control used to control 20% of the total loads In this third case study, a number of load flow calculations were performed. In each calculation, the voltage profile of the Haydon network was determined based on the following conditions. - The wind turbine specified on busbar 7 exports its maximum power to the network. - The load management system adds 20% of the total loads to the minimum demand level, which

was specified in the model before load management was used.

Controlling 20% of the loads is more realistic than controlling 50% of the total loads because it may be difficult to find a sufficient number of loads for the load control system to control. Figure 11 shows the voltage profile when a 300 kW wind turbine is installed on busbar 7.

wind turbine

DISPOWER Del_2005_0079 - 2005-12-27

Figure 11 shows that the voltage profile is maintained below the statutory upper limit if a load management system is used to add 20% of the total loads to the network when a 300 kW wind turbine exports its maximum power to the network.

Figure 12 shows the voltage profile across the network if a 900 kW wind turbine is installed on busbar 7. Figure 12 shows that the voltage profile is maintained below the statutory upper limit for the 900 kW wind turbine on busbar 7 if the load management system is used to add 20% of the total loads.

Figure 13 shows the voltage profile if a 1.2 MW wind turbine is installed on busbar 7.

wind turbine

Figure 13 - Voltage profile during maximum generation and 20% controlled load, for a 1.2 MW

wind turbine

DISPOWER Del_2005_0079 - 2005-12-27

This figure shows that this voltage profile rises above the statutory upper limit if the 1.2 MW wind turbine is on busbar 7. This means that, if the load management system is used to control 20% of the total load, the maximum size of the wind turbine that can be installed on busbar 7 is about 900 kW, which is 3 times greater than that without using load management.

2.3 Conclusions A number of case studies have been performed to demonstrate that load management is able to mitigate voltage rise problems and so facilitate the growth of renewable sources on the Haydon network. The results of all the case studies show that if the amount of loads controlled by the load management system is increased, then the maximum size of wind turbines that can be installed on the network will be increased.

Figure 14 shows the relationship between the maximum size of wind turbine that can be installed on the network and the amount of controlled loads. It is shown that the maximum size of wind turbine is almost directly proportional to the amount of controlled loads. However, there are several problems with regard to the use of load management. One of potential problems is the difficulty of finding a large amount of controlled loads (>20%). This means that controlling 20% of the total loads on the network is more realistic than controlling 50%. Therefore, if a load management system is used on the network, the realistic size of wind turbine that can be installed on the network without voltage rise problems is about 900kW. Without any load control, the size of wind turbine that can be installed on the network without voltage rise problems is 300kW.

The results of the study have shown that load control is an effective strategy for mitigating voltage rise on weak rural networks, in order to facilitate connection of an increased capacity of renewable generation.

Figure 14 - Maximum size of wind turbines versus the amount of controlled loads, without

voltage rise

DISPOWER Del_2005_0079 - 2005-12-27

3 Diesel/hydro generation on the Scottish Island of Rum (Econnect)

3.1 Introduction

Figure 15 - Kinloch Castle, Rum.

The Rum hydro electric scheme is a small run of river scheme serving the community of 30 permanent residents living and working on Rum National Nature Reserve, and providing power for Kinloch Castle. The network is owned and operated by Scottish Natural Heritage. Scottish Natural Heritage are interested in learning more about the mini-grid on the Island of Rum, particularly ways in which they might gain greater efficiencies. The population and load on the island is set to increase in the coming years, therefore it is important to gather as much information about the system as possible so that increasing demand for power can be managed in controlled, energy efficient and sustainable ways. New control concepts developed in DISPOWER Work Package 1 have been simulated on the Rum Island system. Benefits of using storage, inverter technology and load controls have been quantified.

Econnect have installed and tested simple load control devices on the island of Rum scheme over a number of years. The objective of this task was to continue this work by applying the distributed intelligent load control devices with communications to the Rum network. The project would enable Econnect to demonstrate the use and benefits of more sophisticated load control devices on the island power scheme.

The main activities which were carried for this task were as follows. - Data logging on the system to measure existing power quality and power system efficiency and

performance - A case study looking at the power system as it stands and looking ahead approximately 10 years,

to see how the community on Rum may develop, and how its energy needs can be served whilst reducing the impact on the environment

- The installation and testing of innovative load control devices

The main benefits of this work were expected to be as follows. - Power quality, reliability and efficiency gains - Fewer interruptions to supply - Development of the power system such that the electricity needs of the community can be served

in an environmental way and sustainable way - Raising the profile of the community and its environmentally friendly power system.

3.2 The electrical system on the island of Rum A diagram showing the principle components of the power system on the Island of Rum is shown below.

DISPOWER Del_2005_0079 - 2005-12-27

Figure 16 - Schematic of Rum’s power system.

The power system on Rum used two hydro turbines. These were run of river systems, fed from a common penstock. The water source was sufficient to run both turbines during the winter months. These two turbines and generators were manufactured by Gilkes, and were located in a power house originally built to house a turbine and DC dynamo. Subsequently the power station had been converted to AC, and new switchgear had been installed. At the same time as this had taken place the original overhead line network had been removed and replaced by buried cables.

When there was insufficient water to operate the two hydro turbines, a single diesel set was used to provide power. Under these circumstances frequency regulation was provided by an electronic governor built into the diesel set. Occasionally there was a requirement to run the diesel set in parallel with the hydro turbines to provide up to 60kW of power, however this arrangement was unstable when the loading was reduced.

Whilst the hydro turbines were running, frequency control was provided by a Headley controller. This controller used a binary load bank to add loads in proportion to the frequency error. These loads were used to preheat the water in the central heating boiler circuit. In this way any generation from the hydro turbines in excess of the consumption of the other electrical loads was used to reduce the consumption of heating oil.

The electrical load on the island power system was principally domestic loading, such as kettles, heating and lighting. Frugal use of electricity, and a level of cooperation contributed to the successful operation of this power system.

During a previous project Econnect Limited had installed load shedding devices. These devices measured the frequency of the power system. If the frequency fell below a threshold then interruptible loads were disconnected from the network. These interruptible loads were often kettles and refrigerators. These load control devices had reduced the number of blackouts, especially during periods of peak loading. A typical load profile is shown below (Figure 17).

Figure 17 - Typical load profile for Rum.

DISPOWER Del_2005_0079 - 2005-12-27

This chart shows that there is a base load on the power system of approximately 10kW, which rises to 25kW. The Headley controller regulates the frequency by increasing the deferrable heat load on the power system when the consumer loading is reduced. Therefore the output from the two hydro sets remains steady, even though the consumer loading is changing.

To protect the alternators against a potentially damaging overspeed condition, the two turbines were equipped with governors. These governors played no part in the regulation of the frequency of the power system.

A proposal was made to replace the ageing Headley controller with a distributed intelligent load control scheme. To assess the performance of this scheme models of the network of the power system on the Island of Rum were created.

3.3 Development of a model of the Rum network To produce a model of the power system on Rum, the penstock and hydro turbine were modelled. A diagram showing the penstock and hydro turbine is given below.

Figure 18 - Penstock with needle valve used on the Island of Rum.

From this diagram a torque based computer model was constructed. Torque based models were chosen as they provide adequate insight into the behaviour of the control systems of a small power system, without requiring excessive computer resources. In a torque based model the power system is regarded as a rotating mass acted on by torques. The torques which accelerate the mass represent generators, those which decelerate the mass represent loads.

Figure 19 - Penstock and hydro turbine simulation diagram.

In the above diagram G denotes the ratio between the area of the needle An and the penstock Ap. When the needle valve is closed the pressure at the needle valve is increased. This causes a transient increase in mechanical power from the hydro turbine. This is shown in Figure 20 taken from the computer model. This shows that the instantaneous head is significantly increased during a rapid change in valve position, which causes a brief increase in the power exported. This result shows that the new control concepts developed as part of this project must be able to accommodate large changes in generation.

DISPOWER Del_2005_0079 - 2005-12-27

Figure 20 - Step change on a hydro turbine with penstock.

Subsequently the Headley controller was modelled. This device switched electrical loads to match demand to generation and control the electrical frequency on the power system. For this representation the Headley controller was modelled as a proportional only system as shown below.

Figure 21 - Headley controller simulation diagram.

Therefore the loading was quantized into 16 steps from no load to full power. The loads were represented as a retarding torque. The simulation was run and provided good frequency control. At 15 seconds there was a step change in valve position. This caused a transient increase in torque. Subsequently the torque from the hydro turbine reduced. As a result of the change in driving torque the speed was reduced. The Headley controller then applied a reduced retarding torque. The angular speed was slightly reduced, but remained stable. This behaviour was expected, and matched experience gained during site investigations.

Figure 22 - Results of model of the Headley controller.

DISPOWER Del_2005_0079 - 2005-12-27

The models developed during this work were subsequently used for the simulation of the new control concepts developed as part of the DISPOWER project.

3.4 Simulating new control concepts for Rum New control concepts for autonomous power systems were developed for this project. As an initial step these had been simulated using blocks in Simulink, such as the diagram shown below.

Figure 23 - Self tuning governing load controller implemented in Simulink using a windows DLL

This particular model used a self tuning system, which had been shown to provide improved frequency regulation. To provide a realistic simulation the self tuning software used in the load controller was loaded into the computer model. In this work angular speed and rate of change of frequency (ROCOF) were given by measurements of the rotating mass.

When the simulation was run under the same conditions as the Headley load controller the following results were obtained.

Figure 24 - Results of model of self tuning load controller.

These results show that the load controllers developed as part of this work were able to control the frequency. However the system did result in significant changes in load, and the frequency control was not as good as with the Headley controller. However, the new controllers allowed loads to be distributed around the network without installing additional cabling.

After this simulation work had been completed experimental equipment was constructed and a site test was planned for the Island of Rum.

DISPOWER Del_2005_0079 - 2005-12-27

3.5 Installing control concepts on the Rum network Prior to the DISPOWER project, 50 load controllers had been installed on the Island of Rum. These load controllers acted to disconnect loads during a frequency excursion. One of these devices is shown below.

Figure 25 - PowaPlug load controller.

For this new work it was intended to replace the Headley controller with the load controllers simulated during the earlier phase of this work. An experimental unit was built and installed into the basement of Kinloch Castle.

Figure 26 - Load control cabinet installed at Kinloch Castle.

This unit included 21 load controllers and contactors to provide changeover facilities between the existing controller and the new cabinet. This cabinet was commissioned and tested on site.

Initially the cabinet was configured in the same way as the load controllers in the torque based models. However, tests showed that both the frequency and voltage were unstable. It appeared that changes in load caused larger changes in voltage than had been anticipated from the model. Although the frequency control appeared to be adequate, the initial site tests had to be abandoned as the voltage on the power system was fluctuating by an unacceptable amount, and there was a risk of damage to connected loads.

Different configurations for the load controllers were tested. Although the system could be tuned to suit any particular operating condition, when the consumer load changed, the control began to become unstable.

DISPOWER Del_2005_0079 - 2005-12-27

Whilst the smallest loads were performing the switching the frequency was well controlled. However, as soon as a large load switched the resulting frequency excursion caused an oscillation to begin. This oscillation resulted in the generator protection operating.

This behaviour was attributed to the binary arrangement of the loads. On this site the smallest load was 250W and the largest load was 7500W. As the loads were very different sizes from each other different load controller switching actions had very different effects on the power system frequency and voltage. During the site tests no solution was found for this problem.

3.6 Communications An experimental rig at CRES, near Athens, had been used for investigating the use of communications to improve the performance of load control on autonomous power systems, in particular to prevent load controllers cycling on and off during periods of high consumer load. These tests had shown that even low speed communications were capable of making a significant difference to the performance of autonomous power systems, in particular the load factor.

Work was carried out to investigate the performance of load controllers with communications. Detailed models of the performance of the communications system were created in Simulink. These models were used to develop and test the communications protocol which was used for the load controllers. A simulation was run in order to test the communications software, and to establish the delay between a command being sent on the communications network, and the command being acted on. The results of this simulation are given below.

Figure 27 - Results of modelling of load control with communications.

The performance of this communications system was similar to the performance of the networks used for the tests at CRES. Therefore an experimental load controller was constructed to allow investigations to be carried out into the performance of a system with communications.

DISPOWER Del_2005_0079 - 2005-12-27

Figure 28 - Prototype load controller with communications.

This prototype load control system used power line carrier to provide a low speed communications link. These load controllers could be configured by messages sent over the communications link. Power transducers were connected to the network and to a central controller. The central controller measured the consumer load on the system and compared this to the amount of generation available. When additional generation was available the central controller sent signals to turn on the load controllers. The load controllers continued to measure the frequency and disconnected loads during a frequency excursion, for example if there was a sudden increase in the consumer load. However the central controller prevented oscillations from taking place.

Figure 29 - Performance of load controllers with communications.

DISPOWER Del_2005_0079 - 2005-12-27

The central controller was configured to keep the difference between the available generation and the loading within 17 and 20kW. The system was run over a period of 9 hours. During this time six load controllers, connected to various loads, performed as expected to keep the loading within range. In this way blackouts were avoided, but the load controllers connected loads when power was available.

For the purpose of these tests the load controllers were all in the same room. However subsequent tests were carried out where the load controllers were installed across the island’s power network. Communications were reliable across the network to all places apart from one office where there were several computers.

3.7 Summary of simulation and site test work The torque based simulation work showed that torque based models could be used in the development of load control systems. The site work, however, showed that the torque based simulations did not accurately represent the behaviour of the power system and load controllers used for governing the frequency of the power system.

The tests carried out as part of this work were particularly arduous as the loads were of very different sizes. This test had never been carried out before. The site tests showed that load controllers were not capable of controlling the frequency on a power system where the loads were of very different sizes. On these systems the characteristics of the AVR seemed to play a significant role in the performance of the load controllers, but only when large loads were switched. When only small loads were switched the frequency and voltage were adequately controlled. Therefore frequency regulation cannot be provided with load controllers for all sizes of loads. This scheme is best suited to systems where there are many loads of approximately the same size.

Simulations and laboratory tests were used in the development of a communications system for load control. The simulations and laboratory tests suggested that a load control system with communications and a central controller could provide better performance than a system without communications.

Subsequently the experimental equipment was installed on the Island of Rum. This equipment showed that the load controllers with communications performed well over the power line network. Although the total load controlled in these tests was small relative to the size of the power system, the results suggest that load control with communications could increase the amount of power supplied to consumers on a power system, by improving the load factor of existing plant.

DISPOWER Del_2005_0079 - 2005-12-27

4 Islanded operation of a wind farm at Blyth in the UK (Econnect)

4.1 Introduction This chapter is a summary of the work undertaken under Task 7.8b in WP7b of the DISPOWER project. This task investigated the use of load management and synchronous compensator to demonstrate wind turbine operation in islanded mode.

Safe and stable operation of embedded generation on islanded mode under network fault conditions could offer significant benefits in improving security of supply to customers, especially in remote rural areas. The aim of Task 7.8b was to apply the techniques developed in WP1 to assess the viability and options for control of a wind farm during islanded operation due to network faults.

Haverigg wind farm was originally selected as a suitable site to demonstrate islanded operation. However by the time equipment became available for testing, Haverigg was undergoing a repowering exercise and the turbines were not available. A wind farm with 300kW pitch regulated turbines on the Harbour wall at Blyth in Northumberland was selected as an alternative, and permission to work on turbine no. 1 on this site was obtained. A computer model of this system was developed and a synchronous compensator and load control solution was designed to permit islanded operation of the turbine. Equipment was purchased, assembled, installed and commissioned on site. Testing of islanded operation and data collection was carried out during June 2004. The equipment was decommissioned in July 2004.

4.2 Design Considerations

4.2.1 Requirements The main technical operating requirements for an islanded electrical power system incorporating non-dispatchable generation such as wind turbines are to: - maintain a continuous supply with voltage and frequency within required limits [6]. - operate generating equipment safely and within its limitations, to maximise its operating life, - ensure adequate protection against failures within the system, - maximise use of the available wind power.

A solution was developed to achieve these requirements and to demonstrate the viability of operating such an islanded system. Additional equipment to provide frequency and voltage control of the islanded grid was specified. This equipment was chosen to be low cost, robust and reliable, with minimal, and straightforward, maintenance requirements. The aim of the work was to achieve islanded operation with the minimum of modification to the configuration of the wind turbine and its control system.

Details of the transition between grid-connected and islanded operation did not form part of the scope of the work.

4.2.2 Frequency control Frequency control is identified in [7] as a key issue in an islanded system, to the extent that frequency excursions are a reliable method of identifying islanded operation. This is due to the mismatch that will usually occur between the system load and the active power generated within the island, unless enough of the available generation is dispatchable and is configured to balance generation with load using governors. An alternative approach is to manage frequency control in the islanded wind turbine system by balancing the total system load with the system's input power from the wind, using load controllers, and this was the approach taken here.

The load management strategy employed a number of small, individually-controlled deferrable loads, distributed around the islanded grid. This approach is scalable to any size of system. These loads

DISPOWER Del_2005_0079 - 2005-12-27

would typically be water or space heaters, switched on and off by individual governing load controllers, developed by Econnect. Each controller sampled the supply voltage waveform and calculated the frequency and rate of change of frequency of the supply. This information was then processed using a “fuzzy logic” algorithm, to produce an appropriate switching decision. Load switching was achieved using a solid state TRIAC device that switched at zero crossing points, ensuring negligible supply distortion.

The global combination of the independent switching decisions from all the load controllers formed a decentralised, dynamic load management system. The use of multiple small loads produced a controllable load with good resolution. The resulting system was flexible and had high redundancy; it would also permit easy expansion to cope with increased supply capacity. In a typical islanded application, deferrable loads around the network would make effective use of the excess wind power.

4.2.3 Reactive power compensation and voltage control Islanded operation of a section of network runs the risk of voltage excursions outside acceptable operating limits, in a similar manner to frequency. This is usually avoided by managing reactive power flows in the network. Wind turbines incorporating induction generators generally do not have this capability, and in fact act as consumers of variable amounts of reactive power, which in an unregulated island system will cause the voltage to collapse unless reactive power is provided by an alternative source. In the system tested, voltage control was provided by a synchronous compensator, which regulated the flow of reactive power within the islanded network. The power factor correction capacitors already installed on the wind turbine provided a fixed amount of reactive power, and the synchronous compensator was used to meet the additional fluctuating reactive power requirements. The synchronous compensator exhibited great advantages, such as flexibility of operation under all load conditions, and an essentially inductive source impedance that could not cause harmonic resonances with the network.

4.3 System Simulation The first step in the design process was to develop a computer model of the islanded power system. The models were developed using MATLAB / Simulink, and incorporated all the components in the proposed test system by employing Simulink's Power Systems Blockset and Fuzzy Logic Toolbox.

An islanded grid containing a wind turbine, a Newage synchronous compensator with a backup power source, variable consumer loads, and control loads under the control of Distributed Intelligent Load Controllers (DILCs), was simulated. Variable wind speed inputs and variable consumer loads were applied to investigate the power quality and stability of the system.

4.3.1 Wind Turbine The wind turbine was modelled in two sections, the aerodynamic performance and the generator. The windspeed was input from a data file in the form of windspeed versus time data. Various sets of windspeed data were used, to investigate the effects of varying mean windspeed and turbulence levels. A simplified approach to calculating the relationship between windspeed and aerodynamic torque was adopted, and the calculated aerodynamic torque was applied to a predefined asynchronous machine block, available from the Power Systems Blockset library. The parameters used in the model of the asynchronous machine were obtained from the generator data sheet for the wind turbine [8].

4.3.2 Synchronous Compensator and Automatic Voltage Regulator (AVR) The synchronous compensator subsystem comprised two major components, the synchronous machine and the AVR. The model initially employed a simplified synchronous machine block from the Power Systems Blockset library [9]. The simplified synchronous machine block modelled both the electrical and mechanical characteristics of a synchronous machine [10], using parameters from the datasheet for the machine selected.

DISPOWER Del_2005_0079 - 2005-12-27

A simple model of the AVR was developed. The AVR model sensed all three voltage phases and used a mean of all three root-mean-square (rms) values as its input voltage, outputting a field voltage which was applied to the synchronous machine model block.

4.3.3 Control Loads and Distributed Intelligent Load Controllers Single-phase control loads were simulated in line with the system design. Each phase had a number of switchable resistive loads, each with its own Distributed Intelligent Load Controller (DILC). The controllers were implemented using the MATLAB Fuzzy Logic Toolbox library [11]. This permitted easy integration of the controllers into the SIMULINK simulations.

4.3.4 Simulations carried out Simulations were carried out over a range of operating conditions. Some results are described below. Overall the simulation results were encouraging, and gave confidence that the system design would perform satisfactorily when implemented. The results also highlighted areas for investigation during wind turbine testing.

4.3.5 Simulation results Results for the base Windmaster computer model (post-startup) are shown in Figure 30 and Figure 31. A varying wind speed of the order of 10m/s was applied to the model. Figure 30 shows the predicted system frequency resulting from management of the active power flows using distributed loads, and the corresponding power flows. The control load profile clearly makes up the difference between the varying wind turbine input power and the varying base load power requirement. Predicted system frequency remains between approximately 48Hz and 52Hz. This is slightly outside the BS EN 50160 recommendation [6]. The model showed a slow response to windspeed fluctuations, due to the system inertia, and resulted in relatively infrequent controller switching operations.

Figure 30 - Predicted frequency and active power flows from computer simulation

DISPOWER Del_2005_0079 - 2005-12-27

Figure 31 - Predicted voltages and reactive power flows from computer simulation

Figure 31 shows the predicted voltages and reactive power flows. Predicted voltage control was good, within the required ±10% of 230V stated in BS EN 50160. The AVR model achieved this with variations in active power input between 180kW and 300kW, and total reactive power requirements (from the wind turbine and the base load) fluctuating between 100kVAr and 350kVAr. The largest predicted imbalance between two phase voltages was 20V, or 8.7% of nominal. Some imbalance would be expected, since the control loads were independent single-phase loads. The predicted (approximately constant) reactive power contribution from the power factor correction capacitors can be seen to be around 100kVAr, varying slightly as the voltage and frequency vary. The reactive power from the synchronous compensator makes up the shortfall between the contribution from the power factor correction capacitors and the reactive power required by the wind turbine and the base loads.

4.4 Islanding Demonstration Using Windmaster Wind Turbine 4.4.1 Overview of the demonstration system A demonstration system was designed to incorporate the Windmaster wind turbine. The islanded network was entirely composed of test equipment, i.e. no actual consumers were included in the test network. The wind turbine was disconnected from the main grid for the entire period of the testing. The main aim of the testing was to demonstrate that voltage and frequency on the islanded network could be controlled within acceptable limits when the network was powered only by the wind, and to identify any technical issues associated with islanded operation.

4.4.2 Windmaster wind turbine The Windmaster was a three-bladed, pitch-regulated upwind wind turbine rated at 300kW. The generator was normally connected to the local 11kV / 50Hz mains via a three-phase 11 / 0.380 transformer. The Windmaster's generator was a 4-pole asynchronous ABB machine (wQU 315 L4 DA). Approximately 70% of the no-load reactive power was provided on the low voltage side by power factor correction capacitors.

DISPOWER Del_2005_0079 - 2005-12-27

The Windmaster operated when average windspeed was between 4 m/s and 25 m/s, achieving rated power of 300kW at 13m/s. Local turbulence could result in highly variable instantaneous windspeed and hence highly variable instantaneous power. Above rated windspeed, the wind turbine's power output could easily exceed the rated power for short periods without causing the wind turbine to cut out. This gust power was estimated at 370kW at the Blyth site, based on observations by site engineers, and this defined the total capacity required for the load banks.

Starting resistors Unlike modern wind turbines, the Windmaster was not fitted with a power electronic soft-starter. Inrush currents were reduced by the use of starting resistors when the generator was first connected to the grid. At synchronous speed, the resistors were bypassed to avoid overheating and fire, and to minimise power dissipation.

Blade pitch control The Windmaster, like most wind turbines at this rating and above, had variable-pitch rotor blades, which were adjusted to allow the rotor to self-start, and to regulate the power output above the rated wind speed. The pitch-control mechanism introduced the potential for interaction with the frequency management strategy of the load controllers. The pitch angles of the three rotor blades were controlled hydraulically.

Existing Windmaster control and protection Sequencing control and protection for the Windmaster was carried out by an HMZ Data communications system. During its start up the Windmaster's generator was not connected to the grid. Its variable pitch blades were angled to provide aerodynamic torque to accelerate the rotor up to its operational speed. When the generator reached approximately 1455rpm, the power factor correction capacitors were connected to the grid. When the generator speed reached approximately 1498rpm, the generator was connected to the grid. The initial connection was made through starting resistors in series with the generator, to reduce the inrush current, and then the resistors were bypassed and the generator was connected directly to the grid. This sequence meant that the synchronous machine was not required to motor the generator up to its operational speed, and was assisted in exciting the generator because the power factor correction capacitors were connected before the wind turbine generator was connected. The Windmaster's controller was set up to carry out a shut down on any of the following events (among others): - rotor overspeed / rotor underspeed - excessive vibration - winding overtemperature - over-voltage / under-voltage - over-frequency / under-frequency

Windmaster auxiliary equipment The Windmaster was equipped with a variety of auxiliary devices with various voltage supply requirements. This included a 250bar hydraulic pump supplying the motor for the blade pitch control and the yaw motor.

4.4.3 System design and development Synchronous generator selection A two-bearing 250kVA Newage HCI434C with an MX321 AVR was selected. Typical efficiency was quoted by the manufacturer as 93% at rated output. This gave a predicted full load loss of 17.5kW, providing an order of magnitude estimate of system losses at around 5% of wind turbine rating.

Earthing and protection Under normal circumstances the low voltage side of the wind turbine system is referred to earth by earthing the neutral point of the grid-connected transformer. As the islanded grid was disconnected

DISPOWER Del_2005_0079 - 2005-12-27

from the transformer, the islanded network was provided with an earth reference by earthing the neutral point of the synchronous machine. This provided a path for fault currents, to allow protection devices to be triggered in the event of an earth fault. A central busbar cabinet was included in the system. This distributed the power between the synchronous machine, the wind turbine and the loads. Circuit breakers were included on each line to provide protection against system faults and allow isolation of any section of the system. A central emergency stop system allowed the machines to be de-energised and shut down immediately if required.

Governing load controllers and control loads A total control load of 150kW was specified, which employed seventeen load controllers per phase, switching a range of 1kW, 2kW and 4kW single-phase loads. A 250kW variable base load was employed, fulfilling the role of an uncontrolled consumer load. In addition, two three-phase inductor banks, of 100kVAr and 25kVAr respectively, were employed, to provide reactive load.

Pony motor A mains-powered pony motor fulfilled the role of a back-up generator for the testing. The motoring requirements of the Windmaster's generator during the grid connection process were unknown, but not expected to be significantly larger than the rating of the wind turbine's generator. Therefore the pony motor was sized at 200kW, smaller than the generator but with the ability to deliver higher power for short timescales. The pony motor employed was an ABB induction motor, connected to the shaft of the synchronous machine using a flexible coupling. A 400kW variable speed drive powered from the mains supply was used to control the speed and torque of the pony motor. This high rating was used for availability reasons rather than technical ones.

4.4.4 Equipment commissioning and installation Wind turbine Figure 32 shows the Windmaster wind turbine at Blyth Harbour Wind Farm.

Figure 32 - Windmaster wind turbine

Test equipment The synchronous compensator and load control test equipment was housed in two containers. Container 1, a 20ft (6.09m long) container, housed the control load banks, their associated load controllers, the variable speed drive for the pony motor, the cabinet containing the main AC bus bars which distributed the system power, the control and protection equipment, and all the instrumentation

DISPOWER Del_2005_0079 - 2005-12-27

for the system. Container 2, a 16ft (4.88m long) container, housed the synchronous compensator / pony motor assembly, the base load resistor banks and the inductors. It should be remembered that the equipment was co-located purely for the testing carried out here. In an islanded system, load control equipment would be connected to deferrable loads distributed around the system. Photographs of the equipment are shown in Figure 33 to Figure 36.

pony motor

Newage HCI434C generator

100kW load bank

30kW load banks

Figure 33 - Container 2 : the synchronous compensator and pony motor, plus load banks

distribution cabinet

protection panelconsumer distribution unit

for auxiliary power

100kW load bank control switches

Areva M871 datalogger

current transducers

datalogging PC M871 interface PC

Figure 34 - System equipment in container 1

DISPOWER Del_2005_0079 - 2005-12-27

20kW resistor load bank

Figure 35 - Resistive control load banks

load controller cabinet

Figure 36 - Cabinet housing 30 load controllers

4.4.5 Test results Low wind input mode Satisfactory operation with low wind power input was achieved as shown in Figure 37 and Figure 38. Interaction between the pony motor's frequency setpoint (on the variable-speed drive) and the load controllers was avoided by ensuring that the pony motor setpoint was lower than that of the load controllers. The motor drive would therefore only power the pony motor if the wind was insufficient to maintain this lower frequency. Figure 37 shows the active power flows during a period of low wind input and relatively high load. No load controllers were active here. Stable system frequency was maintained despite variable amounts of wind power being input into the system. Figure 38 shows the corresponding reactive power flows and phase voltages. Note that the "wind turbine reactive power" represents the surplus required by the wind turbine generator that has not already been provided by the power factor correction capacitors.

DISPOWER Del_2005_0079 - 2005-12-27

300 320 340 360 380 400 420 440 460 480 500

time, seconds

system frequency

load power

wind turbine power

pony motor power

Figure 37 - Low wind input - active power flows and system frequency

300 320 340 360 380 400 420 440 460 480 500

time, seconds

load reactive power

wind turbine reactive power

synchronous compensator reactive power

voltages

Figure 38 - Low wind input - reactive power flows and voltages

Wind-only operation Satisfactory operation with higher wind power input was achieved as shown in Figure 39 to Figure 41. Again, interaction between the pony motor's frequency setpoint (on the variable-speed drive) and the load controllers was avoided by ensuring that the pony motor setpoint was lower than that of the load controllers. In Figure 39 and Figure 40, the only power input into the system was from the wind turbine, and the only frequency (speed) control was provided by the load controllers. Figure 39 shows the active power flows during a period of higher wind input relative to the load. Stable system frequency was maintained by the load controllers, despite variable amounts of wind power being input into the system. Figure 40 shows the corresponding reactive power flows and phase voltages. Note that the "wind turbine reactive power" represents the surplus required by the wind turbine generator that has not already been provided by the power factor correction capacitors. No adjustment to the load controller firmware was necessary to achieve stable frequency control in wind-only mode, and no additional adjustment was necessary to the AVR to achieve stable voltage control.

DISPOWER Del_2005_0079 - 2005-12-27

550 560 570 580 590 600 610 620time, seconds

system frequency

load power

wind turbine power

synchronous compensator power

Figure 39 - Wind-only operation - active power flows and system frequency

560 570 580 590 600 610 620

time, seconds

load reactive power

voltages

Figure 40 - Wind-only operation - reactive power flows and voltages

Figure 41 shows the control loads connected on each phase and demonstrates that the unbalanced single phase loads did not generate unacceptable voltage imbalance. Wind-only mode could be maintained stably for as long as there was enough wind to overcome the losses in the system, mainly the synchronous compensator's rotating losses. Only a short period of wind-only operation was achieved with the Windmaster due to the restrictive rotor speed limits within the controller. This period of operation has been analysed to evaluate the system's performance. During this period the mean wind turbine power generated was 41.3 kW. Table 1 shows mean, maximum, minimum and standard deviation values for frequency and voltage, as these are the main parameters specified in BS EN 50160. The requirements stated in BS EN 50160 are also shown. Frequency control was good. Assuming a normal distribution, a standard deviation of 0.23Hz would suggest a variation of +/-0.46Hz for 95% of the time, within the BS EN 50160 requirement of +/- 1Hz. The mean value of 50.32Hz means that the overall frequency level was within statutory limits. A slow oscillation in frequency was observed. This is a feature inherent in the use of the distributed load control technique. However the frequency of this oscillation, around 0.4Hz, was well outside the region where flicker would be a problem (8-10Hz).

DISPOWER Del_2005_0079 - 2005-12-27

550 560 570 580 590 600 610 620time, seconds

control loads

voltages

blue phase

red phase

yellow phase

Figure 41 - Wind-only operation - load power flows and system voltages

Table 1 - Frequency and voltage results - wind-only Windmaster operation

Frequency, Hz Phase-to-neutral voltage, V, rms measured BS EN 50160 red yellow blue BS EN 50160 Mean 50.32 50 219.0 217.9 217.1 230 Maximum 50.82 51* 219.9 218.8 218.2 253 Minimum 49.58 49* 218.0 216.7 215.5 207 Standard deviation 0.23 - 0.397 0.295 0.517 -

* for 95% of week

Voltage control was good. The AVR had been adjusted to give a phase voltage of 220Vrms with no load on the synchronous machine, to meet the requirements of the wind turbine. The BS EN 50160 voltage requirements are fairly wide and the voltages on all three phases remained well within them. As the load control employed single-phase loads, with no restriction on the relative loading on each phase, this was especially encouraging, particularly in view of the predictions regarding voltage imbalance obtained from the computer model. Negative sequence voltages were well within the nominal 2% limit in BS EN 50160 (mean 1.1% of positive sequence voltage, standard deviation 0.32%).

Wind turbine starting

Starting the wind turbine on the islanded system proved difficult. Initially the G59 frequency and voltage protection prevented any progress being made past the initial connection of the starting resistors. Temporarily bypassing these limits by employing an alternative G59 relay allowed the bypass contactor K1 to be closed and some successful starts to be achieved. Figure 42 and Figure 43 show the momentary frequency and voltage dips that occurred when connecting the generator. Many starts were terminated just after the main contactor was closed, due to excessive rotor speed excursions, the limits for which effectively mirrored the frequency settings which had been successfully bypassed. Although the original aim was to maintain voltage and frequency within limits at all times, it became apparent that this would not be possible without significant system modification which was not possible with the turbine being tested. The aim then became to demonstrate successful post-start maintenance of frequency and voltage within the islanded system.

Even with the G59 frequency and voltage limits temporarily bypassed, a successful start required all the components of the system (i.e. variable speed drive, AVR and load controllers) to operate appropriately immediately to prevent the wind turbine exceeding any of its protection limits.

DISPOWER Del_2005_0079 - 2005-12-27

A "successful" start is shown in Figure 42 and Figure 43. In order to ensure that no loads were switched on during wind turbine starting, the pony motor speed setpoint was set below the load controller setpoint. The power factor correction capacitors were switched in at 249.5 seconds; the starting resistors were switched in at 269.5 seconds, and the bypass contactor K1 was closed at 262 seconds.

-350-300-250-200-150-100-50

100150200250300350400450500550600650700

249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269time, seconds

10121416182022242628303234363840424446485052

frequency

load active power

wind turbine systemactive power

synchronous compensator active power

Figure 42 - Wind turbine start - active power flows and system frequency

-350-300-250-200-150-100-50

100150200250300350400450500

249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269

time, seconds

load reactive power

voltages

Figure 43 - Wind turbine start - reactive power flows and voltages

The events that occurred during the start had almost contradictory requirements. For example, the moment at which the wind turbine was first connected required a large flow of real and reactive power into the wind turbine system to avoid under-frequency and under-voltage. As soon as that first inrush was overcome and the turbine's rotor became supersynchronous, the turbine started to generate. The power input from the variable speed drive then had to be removed, and large amounts of load had to be applied instantly to prevent an overspeed.

DISPOWER Del_2005_0079 - 2005-12-27

4.5 Analysis of starting performance - updated modelling results The Simulink model was updated so that its behaviour reflected the observed results more closely, in order to investigate the observed starting performance of the wind turbine in more detail. - It was found that the original version of the model predicted a system voltage collapse at the

moment of connection of the wind turbine. This was not reflected in the observed performance of the system, and so the simplified synchronous machine model was replaced with a full synchronous machine model, and the AVR model was replaced with the Power Systems Blockset Excitation block.

- A delay was added into the diesel model to allow for its response time in reacting to a frequency dip.

- Additional measurements were incorporated to allow the power consumption of the starting resistors to be evaluated.

The updated model's behaviour during starting was qualitatively correct, however it did not replicate quantitatively the exact frequency and voltage response of the system, most likely due to the simplifications in the AVR and pony motor controller models. The frequency dip and the overall voltage drop during the start were underpredicted; in contrast, the model suggested that application of out-of-balance single-phase loads would result in larger voltage imbalances between the phases than were seen in practice. The updated model was therefore used to provide an overall indication of the processes involved during wind turbine starting, and to estimate the effects of modifying the starting strategy.

The model indicated that the majority of the power drain on the system during the start was due to the starting resistors, which absorbed a peak of around 200kW, whereas the maximum motoring power required by the generator was only around 50kW. The peak current flow seen in the model was 480A. These values would vary depending on the assistance provided by aerodynamic torque at the moment of connection. Running the model with the starting resistors bypassed gave a predicted maximum current flow of 512A, a peak power drawn by the wind turbine generator of 80kW, and a 50% reduction in the predicted frequency excursion.

The immediate conclusion from these observations is that the starting resistors appeared to be generating as large a problem as they solved for the islanded system - they reduced the instantaneous current demand of the system, but greatly increased the real power demand placed on the backup power source.

The model suggested that complete removal of the starting resistors would be likely to result in a smaller underfrequency but a larger voltage drop (though this would depend on how much the observed voltage drop was a consequence of the under-frequency roll-off protection in the AVR). The risk associated with such a starting strategy might be that in high winds there would be insufficient voltage at the moment of wind turbine generator connection to "pull in" the wind turbine rotor and prevent an overspeed.

Figure 44 shows the predicted real power flows from the model results for the starting process. Figure 45 shows the corresponding predicted reactive power flows, and Figure 46 shows the predicted frequency and wind turbine rotor speed during this time.

It can be clearly seen from Figure 44 that the majority of the power drain on the system during the start is due to the starting resistors. They absorb a peak of 200kW whereas the maximum motoring power required by the generator is around 50kW. The peak current flow seen in the model was 480A. The model prediction of a frequency dip to 47Hz does not replicate exactly the results seen in practice. This could be due to a difference in the aerodynamic torque provided by the wind for the start under consideration.

The immediate conclusion from these observations is that the resistors appear to be generating as large a problem as they solve for the stand-alone system - they reduce the instantaneous current demand of the system, but massively increase the real power demand placed on the backup power source.

DISPOWER Del_2005_0079 - 2005-12-27

Running the model with the starting resistors bypassed gives a predicted maximum current flow of 512A; the peak power drawn by the wind turbine generator is 80kW, and the predicted frequency dip is 1.5Hz.

Figure 44 - Model results for wind turbine starting - real power flows

Figure 45 - Model results for wind turbine starting - reactive power flows

DISPOWER Del_2005_0079 - 2005-12-27

Figure 46 - Model results for wind turbine starting - wind turbine speed and system

frequency

4.6 Discussion Potential for improving system performance during wind turbine starting

The following options would be recommended for improving starting performance. - Replacement of the starting resistors with a soft-starter (as is employed in all modern direct-

connected induction-generator wind turbines). This would limit the voltage applied to the generator terminals without affecting the islanded grid system voltage, and hence reduce the inrush current requirement, without increasing the real power demand on the system. Sequencing of the power factor correction capacitors would probably need to be modified in order to avoid interaction with the soft starter.

- Increasing the inertia, and hence the amount of stored rotational energy, within the system. The simplest way to do this would be by connecting a flywheel onto the synchronous compensator's shaft.

Windmaster protection settings

The Windmaster controller's speed, frequency and voltage limits proved restrictive in operating the islanded system. Adapting a grid-connected wind turbine to operate within an islanded system is likely to require some modification of protection limits. This is difficult in the case of the Windmaster, as the controller is old and the manufacturer does not exist in their original form.

Load control

The original design of the Windmaster system allowed for a combination of control load plus base load. It became apparent during testing that the transient events during starting, in conjunction with large and rapid variations in wind speed, required a larger proportion of controllable load to be available, in order to regulate system frequency during connection of the wind turbine.

Practical lessons learnt

The experience of demonstrating an islanded 300kW wind turbine proved onerous compared with a similar exercise previously carried out with a 20kW machine. The exercise highlighted many of the

DISPOWER Del_2005_0079 - 2005-12-27

issues that would be associated with a more permanent application of the technology. The wind turbine required a continuous low voltage supply to be maintained to its auxiliary systems at all times for safety reasons, which would have to be provided by backup power or a UPS in an islanded system. Other aspects of this particular wind turbine system, mainly related to the temporary nature of the test installation and hence the inability to modify its configuration or control parameters, made it unsuitable for longer term demonstration of islanding capability.

4.7 Conclusions Design assumptions and requirements

The satisfactory performance of the islanded system confirms that it is possible to operate an islanded network powered only by wind generation, and vindicates the design assumptions made. Additional equipment is required to provide frequency and voltage control of the system, and the islanded system would need to be carefully defined to meet the following criteria. - Generation capacity within the island must be able to meet transient and steady-state load

demand. Provision must be made within the island for periods when insufficient renewable energy is available to meet the loads. Any dispatchable back-up power source must be sized to meet essential loads, but its rating could be reduced if load management is employed to constrain the total load on the islanded system. Technically this would be very straightforward if load management has already been incorporated into the system to accommodate wind generation or similar.

- Available deferrable load needs to match non-dispatchable renewable generation capacity. This load could include "dump" loads to allow transient events to be accommodated, but would ideally comprise useful loads such as heating to allow best use to be made of available renewable energy. Single phase distributed deferrable loads, operating independently and without any centralised control, have been demonstrated to be capable of achieving frequency control in a wind turbine system with no control over the input power required, and with satisfactory voltage control over all three phases.

- The island must include dynamic reactive power capability, a synchronous generator being an ideal source. Its capacity needs to be sized to meet all reactive power needs of the islanded system, including wind turbine (and any motor) starting, and inductive loads.

Switching between islanded and non-islanded operation

The work described here covers a purely islanded wind turbine. If it were required to switch from a non-islanded configuration to an islanded configuration, significant issues would need to be addressed. These include: - Management of the transition between the two modes of operation, such as switching, and

synchronisation of the synchronous compensator with the grid if required - Disabling and enabling of the load controllers when on- and off-grid respectively. Some of the

communications technologies developed in Work Package 1.4 could be employed here. - Earthing reconfiguration and protection of the islanded system, including possible differences in

fault levels between the two modes.

DISPOWER Del_2005_0079 - 2005-12-27

5 Application of the MORE CARE advanced control system on the islands of Crete and Kythnos (ICCS, CRES)

5.1 Introduction Advanced control systems can help operators of islands with increased RES penetration to operate their system in a more economic and secure way than without having such an operating tool. MORE CARE is an advanced Control System for island systems with increased RES penetration. Within MORE CARE, besides the economic scheduling and the load and RES forecasting functions, algorithms that assess the dynamic security of the system have been developed. Three pre-selected disturbances are assessed in order to provide the operator with information on the secure or not secure operation of the power system for each one of the selected disturbances.

Within Task 7.9 of the DISPOWER project an evaluation of potential benefits that MORE CARE operation in island systems with increased RES penetration can have in their secure and economic operation is performed and the results are described within this document for two specific case studies, Crete and Kythnos. These two, of the many dispersed Greek islands in both Ionian and Aegean Sea, operate autonomously under high RES penetration and were considered the most suitable Greek islands for the study.

More specifically Kythnos is a small island with 2000 inhabitants, 3 hours by boat from the port of Pireaus and has many visitors during the summer especially during August.

The total demand of the island in 2002 was 5630MWh. The peak demand during summer was 1605kW (19/8) and the minimum demand was 120 kW during October. 10.2% of the demand is met by the Wind Turbines production and 1% is met by the PV station production. The total annual RES production penetration exceeds 11 %, but for some hours, the RES penetration reaches 100%. In Figure 47 the RES penetration duration curve for the whole year is depicted. The RES penetration during January is very low due to maintenance of the 500 kW Wind Turbine. It can be seen that the RES penetration exceeds 40% for more than 1000 hours per year.

Figure 47 - RES penetration on the island of Kythnos during 2002

The installed components on the island are described in the table below.

DISPOWER Del_2005_0079 - 2005-12-27

Table 2 - Installed Capacity on the island of Kythnos

Number Type of Unit Capacity 5 Diesel units 5*400=2000kW /2500kVA 1 Wind turbine 500kW 5 Wind turbines 5*33=165kW 1 PV Power station 100kW 1 Battery Bank 400kWh/500kWh 1 Phase Shifter 600KVA 1 Compensation bank 8*100=800kVAr

Crete on the other hand is the largest isolated network in Greece and has high wind penetration up to 10% of the annually consumed energy on the island. A map of the Cretan power system network is shown in Figure 48.

Figure 48 - Cretan Power system Network Map

There are two power plants, one in Linoperamata (LIN) and Hania (HAN) with a variety of units, Steam turbines, diesel units and Gas turbines and one combined cycle unit. The installed capacity of wind parks at the end of 2003 was 81.19MW, of which 11.32MW had been installed during 2003 as shown in Table 3. The annual wind power production for 2003 was 204.6 GWh. Since August 2004 a new thermal power station with 2 Diesel units of 50 MW each has been in operation in Atherinolakos Region.

Table 3 - Installed Wind Parks on Crete during the year of study

Wind Park Name

Installed Capacity(MW)

WP No Wind Park Name

Installed Capacity(MW)

1 PPC-I (TOPLOU) 6.60 6 MARONIA 27.50 2 OAS 0.50 7 PPC-II (XIROLIMNI) 10.20 3 ROKAS 13.20 8 CRETE PLASTIKA 5.94 4 IWECO 4.95 9 WRE 2.40 5 AIOLOS 9.90

The peak demand was 498.4MW on the 1st September 2003, lower than the 505.8MW peak of the previous year when extremely hot days (43 deg. C) were noticed. A summary of the demand data, energy and minimum and maximum power as well as wind power penetration data for the years 2000, 2001 and 2003 is provided in Table 4.

DISPOWER Del_2005_0079 - 2005-12-27

Table 4 - Summary of energy, load and wind production in 2000, 2001 and 2003

2000 2001 2003 Annual Energy 2078.6 GWh 2191.8 GWh 2349.5 GWh

Peak load 435.4 MW (24/8 21.00) 471.4 MW (4/9 21.00) 498.4 MW (1/9 20.30)Minimum Load 108MW (14/3 4:00) 113.6MW (27/2 5:00) 113.6MW

Annual Produced Energy by RES

203.5GWh 220.2GWh 204.67GWh

Maximum Instantaneous Wind Production

62.7 MW (28/12 11:00) 61 MW 62 MW (03/12 16:00)

Maximum Wind Penetration

39.2% (10/12 6:00) 37 % 40 % (27/04 02:56)

Figure 49 provides a graphical representation of the monthly demand and the monthly wind power production as well as the wind power penetration for 2003. Maximum wind power production is during August while the maximum penetration reaches 12% during July.

Demand , wind production and penetration

120000

160000

200000

240000

280000

1 2 3 4 5 6 7 8 9 10 11 12

Load Wind Penetration

Figure 49 - Demand and Wind Penetration Chart in 2003 in Crete

In the following sections the results of our analysis are presented.

5.2 Results from the Case studies

5.2.1 Kythnos Case Study Actual data for 2002 about load, Wind, PV and Diesel plant production were collected at 10 min intervals. Fuel prices data for each month of 2002 were collected as well. Then the following issues have been addressed: 1. The annual operating cost has been calculated and a comparison with the results from the

Economic Dispatch algorithms of MORE CARE has been performed. 2. The possibility of shortage of supply due to each one of the selected disturbances –Diesel unit trip,

Loss of Wind Power or PV has been evaluated with and without energy storage. 3. Proposed scenarios of operation taking into account the available storage device are presented and

have been evaluated as far as economy and adequacy of the power system are concerned. 4. The impact of forecasting accuracy for load and RES production has been evaluated in the

economic operation of the island.

DISPOWER Del_2005_0079 - 2005-12-27

5.2.1.1 Economic Dispatch (ED) results In order to evaluate the ED correctly, the units that have been committed by the operators of the system have been selected as the operating units and their production is re-dispatched among them using the MORE CARE algorithms.

The difference in operating cost from month to month depends on the production and availability of the RES and the operation schedule, as determined by the operators. The total cost, if the optimal operation suggested by the MORE CARE functions is followed, is lower by 0.8% than the actual operation cost as analyzed in Figure 50.

Profit made using MORE CARE ED algorithms

1 2 3 4 5 6 7 8 9 10 11 12Month

Figure 50 - Results from ED evaluation

5.2.1.2 Adequacy-Reliability of supply It has been examined whether the running units are adequate to meet the demand when one of the following disturbances takes place either with available energy storage devices or not. • Machine Trip (loss of one of the committed units) • Loss of Wind power production • Loss of PV production.

It has been assumed that the battery bank can provide 250 kWh for 5 minutes (300 kW peak) that are sufficient to start up one of the units. The number of hours that the system cannot satisfy its load are summarized in Table 5.

Table 5 - Adequacy results from energy storage impact investigation

Without storage With storage Machine

Trip Wind Loss PV Loss Machine Trip

Wind Loss PV Loss

Total (annual) 7308 893 19 720 66 0 Possibility of insecure operating points (%)

83.4 10.14 0.17 8.18 0.72 0

Table 6 - Impact of Battery bank in the reduction of insecure operating points for the selected disturbances.

Operation Reduction of insecure intervals Actual -90.2%

Wind Loss -92.95% PV loss -100%

DISPOWER Del_2005_0079 - 2005-12-27

5.2.1.3 Impact of Forecasting Errors in operating costs The reduced errors in forecasting load, wind or PV may reduce significantly the operating cost of a power system since the uncertainty in estimating the load that the conventional units should meet is reduced. Thus, the necessity for maintaining spinning reserve to cope with these uncertainties is reduced and this sometimes results in avoidance of running additional units to meet the spinning reserve requirements, reducing the operating cost of the system.

The effect of forecasting accuracy on the economic operation of the Kythnos power system has been examined and Table 7 summarizes the examined scenarios for Load, Wind and PV production forecasting. Scenarios I and V are compared to evaluate the impact of an improved Load Forecast model assuming that wind and PV forecasting models are considered sufficiently accurate. Scenarios I, II and III are compared in order to evaluate the impact of wind forecasting accuracy when the PV and load forecasting models are considered sufficiently accurate. Scenarios I and IV are compared in order to evaluate the impact of PV forecasting accuracy when the wind and load forecasting models are considered sufficiently accurate.

Table 7 - Summarizing the scenarios of Forecasting errors.

Load Forecast Error

Wind Forecast Error

PV Forecast Error

Operating cost (€)

Scenario I 6% 20% 50% 506602.02 Scenario II 6% 50% 50% 517774.54 Scenario III 6% 100% 50% 519778 Scenario IV 6% 20% 100% 507391.12 Scenario V 15% 20% 50% 520318.1

The impact of load forecasting error is greater when the PV and Wind forecasting tools are considered as accurate as possible The annual savings due to the improvement of Load forecasting error is 2.63%. Similar results can be obtained for PV forecasting as Table 8 indicates

Table 8 -Comparison of PV and Load forecasting

Scenarios Euro Percentage Load forecast 6%-15% I and V 13716.08 2.63% PV forecast 50%-100% I and IV 789.1 0.16%

Impact of improved wind forecasting

In this case Load and PV production are considered accurate enough –6 % and 50% error respectively so that the greatest impact of wind forecasting error can be evaluated. Figure 51 compares the cost difference among different Wind Forecasting errors.

Cost difference for different WF errors

012345678

1 2 3 4 5 6 7 8 9 10 11 12

Difference 20-50% Difference 20-100% Difference 50-100%

Figure 51 - Savings due to improvement of wind forecasting error

DISPOWER Del_2005_0079 - 2005-12-27

Table 9 - Comparison of annual savings due to increased wind forecasting accuracy.

20-50 20-100 50-100 5172.52€ 1.01% 13176.01€ 2.6% 8003.49€ 1.54%

The highest percentage savings are during June since the highest RES penetration percentage is obtained for this month.

5.2.1.4 Impact of energy storage A study to evaluate the impact of energy storage in the Kythnos power system has been performed.

In order to investigate the impact of energy storage on the operation of Kythnos, three different scenarios regarding the operational scheduling of the five diesel units have been studied. In these scenarios, load and RES production measurements at 20 minutes intervals, as obtained from the SCADA system of the island have been used.

At each time interval, the Economic Dispatch (ED) is solved, considering the following constraints: a) Production covers demand and losses b) Technical minima and maxima of the diesel units c) Ramping rates (kW/min) of the diesel units d) Minimum Spinning Reserve for the Power System.

The demand to be dispatched to the committed diesel units –Edload -is the same in all three scenarios, as defined by:

That is the demand of the island minus the production of the installed RES units.

5.2.1.5 Operating Scenario 1 - OS 1 Accordingly, the committed units are able to meet the demand (5.1), but also compensate for the uncertainty of load and RES production by maintaining sufficient fast reserve. The amount of spinning reserve compensates uncertainty in load demand (Dem_res), Wind power production (Wind_res) and PV production (PV_res). Therefore, the units to be committed should have sufficient capacity to meet the demand defined by:

Ucload=Edload+PV_res+Wind_res+Dem_res (5.2)

It is assumed that the operator accounts for 50% of PV uncertainty, i.e. PV_res is calculated by

It is assumed that the minimum percentage of reserves for wind power is 20% of the nominal capacity of the wind park. This percentage increases linearly to 100% as the wind power production is reduced.

Thus the reserves to compensate for wind power production uncertainty are:

prodWindprodWindwindIns

resWind ___

8.01_ ⋅

⋅−= (5.4)

where Ins_Wind is the installed capacity of the WTs.

A mean square error of 6% has been used for load forecasting error as equation (5.5) shows. .

Dem_res=Demand*0.06 (5.5)

Since all the units to be committed are identical, only the number of units to be committed has to be identified. This number (UC_No) is found by dividing the UCload by the maximum output of the units, reduced by a small percentage kept for spinning reserve.

prodWindprodPVDemandEdload __ −−=

5.0_PV_res ⋅= prodPV

DISPOWER Del_2005_0079 - 2005-12-27

=perc_Spincap_Un

UCLoadNo_UC (5.6)

Un_Cap is the nominal capacity of each unit,

Spin_perc is the nominal capacity of the diesel unit considering its fast spinning reserve, 0.98 in our application, meaning that the upper 2% of the capacity of the unit is used only for emergency.

5.2.1.6 Operating Scenario 2 – OS 2 In this case it is assumed that the operator wishes to maintain sufficient spinning reserve not only to meet uncertainties due to forecasting errors, but also to meet the load, in case of machine trip. Therefore, the committed units are optimally dispatched to meet the Edload (5.1). In case of a machine trip, the rest of the committed units should be adequate to meet the demand. In case of very low demand, in order to avoid extremely low operating points of the operating units, it is assumed that the operating units are not allowed to operate below 30% of their nominal capacity. Therefore, if it is for the reason of avoiding machine trip only, two units operate only if the demand is higher than 240 kW. Due to the above fact some inadequate operating points are expected and their number for the annual operation is given in Table 11.

5.2.1.7 Operating Scenario 3 – OS 3 This scenario sets the same demands as the previous scenario, but the operator has a storage device, a battery bank, at his disposal. The operator does not have the total nominal capacity of the battery available, but he knows the lower limit of its available power, i.e. the limit below which the battery bank should not be discharged, in our case 300 kW for 5 minutes, enough time to start up a unit. It is also assumed that the operator is mainly interested in using the storage device as back-up power during the time necessary to start up a diesel unit, rather than using it as an energy source. It should be noted that for battery banks, the amount of energy derived is decreased when the discharge rate increases as typical battery data sheets describe.

5.2.1.8 Results In order to evaluate the economic operation of the three scenarios, the monthly fuel prices and the fuel consumption of each installed unit are taken into account. The operating cost for each scenario is calculated and the results are shown in Table 10. These results are also compared with the actual operation of the system as evaluated using the actual production time series.

In order to evaluate steady-state security, the number of 20 minutes intervals, that the system would not be capable to meet demand in case of machine trip is calculated as described in Table 11. These results are compared with the number of intervals that the system would not be adequate to meet demand in case of machine trip when the battery storage is available or not in its actual operation.

Table 10 - Comparison of Economic Operation for operating scenarios with energy storage

OS 1 OS 2 OS 3 Operating Cost 507571.5 € 572305.7 € 511585.1 € Difference with actual operation -2.93% +9.46% -2.16% Difference with cheapest scenario 0 +11.87% +0.79%

These results show that the operating cost can be significantly reduced by optimal Economic Dispatch and lower uncertainty in load and RES power forecast. Scenario 1 is the cheapest scenario, since it sets the least reserve requirements and therefore requires the minimum number of units to be committed. In the other two operating scenarios, the number of units to be committed is at least equal to the number of units of operating scenario 1. Committing more units at lower operating points decrease fuel efficiency and increase operating costs. The battery bank increases the limits above which additional units need to be committed. This affects not only the number of committed units, but also the time of operation in order to compensate for the machine trip.

DISPOWER Del_2005_0079 - 2005-12-27

Table 11 - Comparison of adequacy for Kythnos power system

Operation Hours Difference

Actual (w/o battery) 7304 0 Actual + battery 717 -90.2% OS1 6761 -7.43% OS1 + battery 449 -93.86% OS2 558 -92.37% OS2 + battery 0 -100% OS 3 0 -100%

The battery bank helps in decreasing the intervals that the committed units are not adequate to meet demand in two ways: first by reducing the number of intervals of low operating points without sufficient fast reserve, and secondly by decreasing the time of inadequacy, when Edload exceeds 1600 kW. Since the demand never exceeds 1605 kW for 2002 and the battery capacity is higher than the limit for low operating point, the system can adequately meet the demand during the time necessary for starting up a new unit, even if the only committed unit trips. Therefore with the proposed operating scenario 3 the number of insecure operating points is reduced to zero.

5.2.1.9 Discussion of the results The impact of load and wind forecasting errors is significant in the economic operation of Kythnos power system. The savings reach 2.63% for the load forecast and may reach 2.6% for the wind forecast error. As far as energy storage is considered it can be concluded that it has significant impact on the operating cost and adequacy of Kythnos power system. Application of MORE CARE Economic Dispatch algorithms can also reduce the operating cost.

5.2.2 Crete Case Study In [17] the results from the evaluation of MORE CARE operation for the period July 2001 –February 2002 have been presented. Within the framework of Task 7.9 in DISPOWER WP7b, the impact of MORE CARE in the operation of the Cretan power system for the first 6 months of 2003 has been evaluated. For this purpose necessary data about load, wind and diesel plant production have been collected at 20 min intervals. Moreover, data about the fuel prices for each month studied have been collected.

For the Cretan power system the impact of improved wind forecasting error in the economic and secure operation has been evaluated for the following cases:

• actual operation, • 15% load forecast error and 20% wind forecasting error spinning reserve (case 1), • 15% load forecast error and 50% wind forecasting error spinning reserve (case 2),

for their operating cost and as far as dynamic security is concerned.

The spinning reserve for the power system was at least 15% of the load and 20% or 50% of the wind power production and the Unit commitment was performed using these numbers for the economic operation.

For dynamic security the following disturbances have been considered: 1. Short-circuit at one of the substations of the system (Short). 2. Machine trip of one of the largest in capacity units, a steam turbine (Trip) 3. The loss of wind production due to disconnection of the linking transmission line (Wind Loss).

5.2.2.1 Results from Cretan power system operation Table 12 summarizes data concerning the actual economic operation of the power system of Crete for the months studied.

DISPOWER Del_2005_0079 - 2005-12-27

Table 12 - Actual operating cost during each studied month of 2003

Month Cost (Euro) January 11,441,588 February 13,951,810 March 14,861,568 April 13,638,904 May 12,443,750 June 11,455,365 Total 77,792,985

Table 13 summarizes the results from the evaluation of the actual operation for the 6 studied months.

Table 13 - Evaluation of security indices for actual operation during the studied months of 2003

Short Trip Wind Loss Month Hours Percentage (%) Hours Percentage (%) Hours Percentage (%)

January 197 26.48 112 15.05 130 17.47 February 228 33.93 118 17.56 106 15.77 March 206 27.60 136 18.32 222 29.75 April 225 31.25 166 23.01 103 14.21 May 93 12.54 105 14.11 85 11.38 June 235 32.59 160 22.18 148 20.56 Total 1185 27.24 797 18.34 793 18.24

Table 14 and Table 15 summarize the security and economic results for the Cretan power system for the case study 1 while Table 16 and Table 17 summarize the results for the case study 2.

Case 1 - 15% load forecast reserve and 20% wind forecast reserve.

Table 14 - Evaluation of security indices for case 1 during each studied month of 2003

January 189 25.40 313 42.07 145 19.5 February 257 38.24 356 52.98 167 24.85 March 337 45.29 359 48.25 190 25.54 April 260 36.11 322 44.72 192 26.67 May 100 13.44 418 56.18 73 9.81 June 224 31.11 275 38.19 155 21.53

Total 1367 31.46 2043 47.03 922 21.22

Table 15 - Actual operating cost according to case study 1 scenario.

Month Cost (Euro)January 9,960,840 February 13,130,991 March 14,141,634 April 11,731,332 May 11,718,888 June 10,496,532

Total 71,180,217

DISPOWER Del_2005_0079 - 2005-12-27

Case 2 - 15% load forecast reserve and 50% wind forecast reserve.

Table 16 - Evaluation of security indices for case 2 during each studied month of 2003

January 180 25 209 29.03 102 14.17 February 257 38.24 274 40.77 126 18.75 March 259 34.81 296 39.78 140 18.82 April 260 36.11 247 34.31 162 22.5 May 98 13.17 378 50.81 61 8.20 June 222 30.83 179 24.86 104 14.44 Total 1276 29.37 1583 36.44 695 16

Table 17- Actual operating cost according to case study 2 scenario.

Month Cost (€) January 10,100,222 February 13,271,648 March 14,267,670 April 11,846,862 May 11,807,404 June 10,661,891

Total 71,955,697

5.2.2.2 Comparison of the scenarios studied The following tables compare the results of the proposed operation to the ones of the actual operation. Moreover, the case study scenarios of the proposed operation are compared. Negative values mean decrease of the latter compared to the former scenario studied.

Actual operation vs 15% load forecast and-20 % wind forecast

Table 18 - Comparison of security indices and cost between actual operation and case 1 during each studied month of 2003

Month Short Trip Wind Loss CostJanuary -4.06% 179.5% 11.54% -12.94%February 12.72% 201.69% 57.55% -5.88%March 64.12% 163.32% -14.16% -4.84%April 5.56% 94.36% 87.62% -13.99%May 7.14% 298.09% -13.78% -5.82%June -10.94% 72.23% 4.73% -8.37%Total 15.52% 156.44% 16.37% -8.50%

DISPOWER Del_2005_0079 - 2005-12-27

Actual operation vs 15% load forecast and-50 % wind forecast

Table 19- Comparison of security indices and cost between actual operation and case 2 during each studied month of 2003

Month Short Trip Wind Loss CostJanuary -8.63% 86.6% -21.54% -11.72February 12.71% 132.2% 18.87% -4.88March 26.13% 117.12% -36.75% -3.99April 15.555 49.09% 58.30% -13.14May 5% 260% -27.95% -5.11June -0.71% 12.10% -29.73% -6.93Total 7.83% 98.71% -12.28% -7.5%

Comparison of 20% and 50% wind forecast error

Table 20 - Comparison of security indices and cost between cases 1 and 2 during each studied month of 2003

Month Short Trip Wind Loss CostJanuary -4.76% -33.23% -29.66% 1.40%February 0% -23.03% -24.55% 1.07%March -23.15% -17.55% -26.32% 0.89%April 0% -23.29% -15.63% 0.98%May -2% -9.56% -16.44% 0.75%June -0.89% -34.90% -32.90% 1.58%Total -6.66% -22.52% -24.62% 1.09%

5.2.2.3 Conclusions for Cretan power system The proposed operation helps in decreasing the operating cost compared to the actual operation, by 8.5% and 7.5% respectively for each scenario studied, but especially for machine trip disturbance the insecure operating points are increased. This is due to the fact that the operators of the Cretan Power System try to operate the system maintaining, if possible, spinning reserve equal to the largest operating unit of the system. This leads to committing additional units to meet this constraint and thus in increased operating cost especially for low demand periods like March, April and May.

An improvement in wind forecast error improves the operating cost by 1.09%. This improvement however does not bring any improvement in the security indices due to the reduction of spinning reserve. Especially for machine trip and wind loss the indices are deteriorated more than for short-circuit disturbance.

Increased spinning reserve for the wind power forecasting error has a significant impact in reducing the insecure operating points due to wind power disconnection, since more spinning reserve is maintained when the wind power production is higher. That is the reason for reduced insecure operation points in 50% wind forecasting error scenario.

Table 21 summarizes the results form the analysis performed.

DISPOWER Del_2005_0079 - 2005-12-27

Table 21 - Summarizing the security indices and cost comparison of among cases 1 and 2 and actual operation for 2003

Comparison Short Trip Wind Loss Cost15-20% vs 15-50% -6.66% -22.52% -24.62% 1.09%Actual vs 15-20% 15.52% 156.44% 16.37% -8.50%Actual vs 15-50% 7.83% 98.71% -12.28% -7.5%

5.3 Conclusions The above results from the evaluation of operation of MORE CARE Advanced Control System regarding economy and security of islands with increased RES penetration show that economic scheduling functions can decrease the operating cost of the power system. Moreover, the improved performance of the forecasting modules can help in the decrease of the operating cost of an island system, compared to the case of insufficient wind forecasts. As far as secure operation is concerned, MORE CARE functions can timely inform operators in case of a disturbance. According to the spinning reserve percentage maintained the MORE CARE functions can reduce the operating cost for the specific operating scenario reducing at the same time the number of insecure operating points as much as possible. This reduction is higher if an energy storage device exists in the island power system and the economic management is performed using MORE CARE functions taking into account the characteristic of the energy storage device.

Therefore the operation of islands with increased RES penetration using Advanced Control Systems like MORE CARE can help operators in operating in a more economic and secure way their power system with increased RES penetration.

Results from the evaluation of the MORE CARE functions have been published in the following papers:

N. Hatziargyriou, A. Tsikalakis, A. Dimeas, D. Georgiadis, J. Stefanakis, A. Gigantidou, E. Thalassinakis , “Security and Economic Impacts of High Wind Power Penetration in Island Systems,” Presented in the 40th Cigre Session 2004, 27th August 2004.

Impact of Energy Storage in the secure and economic operation of a small Greek island.” by N. Hatziargyriou, A. Tsikalakis and I. Tassiou referring to the power system of Kythnos. Presented at the Med Power Conference, 15th-17th November 2004 in Cyprus.

DISPOWER Del_2005_0079 - 2005-12-27

6 Increased penetration of renewable energies in the Guadeloupe power system (Vergnet, ISET, EDF, ENSMP-CENERG, Uni Kassel)

6.1 Introduction This chapter presents the work done and the results obtained in Task 7.10 of DISPOWER WP7b. This task deals with the technical aspects to be considered while connecting a wind farm to a small island grid, based on a real case on “Les Saintes” islands. This case is typical to Distributed Generation connected to small island grid or “end of grid”.

Case study objectives

Les Saintes is part of Guadeloupe archipelago, in the West Indies. Its location is a typical “end of grid” situation. Problems encountered in this case are representative of the problems that may have a developer of independent power producer when he wants to connect a DG on the existing MV grid in a remote place.

The objective of this case study was to study the interest of using new planning and control tools to increase:

- the penetration of wind power on one particular island of Guadeloupe: a 2 MW wind farm project is under construction. The main limitations to wind penetration are voltage rise and flicker level. New design and control techniques are applied in order to decrease the out of limit voltage and flicker level on the island MV and LV grid. The study should help to define more relevant and precise rules and design methods for the monitoring of wind power output to maintain the grid voltage and flicker within required limits. These methods will ease the connection of wind power on the grid, and therefore increase the wind penetration. Potential uses and applicability of Geographical Information Systems (GIS) for DG planning are also explored.

- the quality on the grid when distributed generators are operated in parallel with the grid: new control techniques are proposed in order to increase the quality of voltage and frequency on the island grid.

- the possibility of fast reaction after a cyclone or the loss of the submarine cable: new control techniques are proposed in order to maximize the wind penetration in case of loss of sub-sea cable while operating on the diesel plant only.

The grid of Les Saintes Islands was modeled to study with existing or new tools the benefit of using the new control and safety concepts developed in other Work Packages of DISPOWER: grid control techniques (WP1), reduction of Wind Turbine Generator (WTG) impact on grid quality (WP2), use of a GIS (WP4), interest of using a battery (WP6).

Les Saintes islands

Les Saintes consists mainly of two islands interconnected by a 2 km long sea-cable. They are connected to a 63 kV/20 kV substation on the Guadeloupe main island by a 20 kV AC 16 km long sea-cable. In case of loss of the submarine cable, a backup diesel plant can be started to power the grid.

In the French overseas departments and territories, « conventional » electricity generation is mainly done in small-sized thermal plants, which costs are higher than those of high-power plants in metropolitan France. Furthermore, significant quantity of renewable energies is available (hydraulic, solar, wind energy because of trade wind) and is scattered throughout the year thanks to the Equator’s proximity (solar).

Les Saintes, like other islands of Guadeloupe archipelago, have a very good wind potential. Wind measurements on the site "Pointe Sud" located on the south coast of Terre de Bas island give an average wind speed of 8.24 m/s at 50 m (k = 4.3) [23].

DISPOWER Del_2005_0079 - 2005-12-27

A wind farm project has been developed for a few years. The loss of the submarine cable to Guadeloupe main island in 2004 have caused further delay. The submarine cable has been replaced in Summer 2005, and the wind farm has been constructed during the second semester of 2005.

The 2 MW wind farm is located on Terre de Bas Island, at Pointe Sud, where was the old airport (now helicopter landing platform). The site has an East-West orientation, and is on top of a cliff on the sea shore, facing the sea on the South. The site has been leveled for aeroplane use, and has a size of 70m x 800m. The heliport will be moved to a more suitable place, proposed by the council and accepted by the local authorities. Average wind speed on site is above 8 m/s.

The wind farm was first made of 8 x 275kW Vergnet wind turbines designed for island grids and cyclonic areas. The number of wind turbines has been reduced to 7, to take into consideration the space available and also grid connection constraints. The power output would be about 4 520 MWh per year, which is above the power consumption of the whole island. Terre de Bas would then be net exporter of electricity.

Limitations to grid connection

The implementation of wind power on Les Saintes is facing various technical limitations. EDF ARD, the grid operator, made preliminary studies in 2003 to evaluate the impact of the wind farm on the grid. These studies were based on standard assumptions, and gave the following results.

Voltage limitation: the connection of 2200 kW at tan phi=0, with a minimum consumption on the grid, would lead to overvoltage levels at various MV/LV transformer secondary that range from 10,7% to 12,4%. A maximum number of only 3 x 220kW wind turbines can be connected in that situation to maintain the voltage below 400V +6%.

Flicker limitation: the connection of 660 kW (4 wind turbines with a power output limited to 165 kW) at tan phi=0, at minimum Pcc level, with a stall regulation and default flicker coefficients, will generate a flicker type 1 equal to 0,46 at the connection point, where the acceptable level is 0,25.

Harmonic limitation: the GEV MP wind turbine is equipped with an induction generator directly connected to the electricity system. According to the IEC 61400-21 it is expected to cause no significant harmonic distortion

Network stability: network stability has to be studied in some specific situations

Reactive power: study of reactive power exchanges, various solutions to compensate for wind turbines reactive power consumption, use of reactive power to correct grid voltage should be studied

Wind farm

16km sub-sea cable

DISPOWER Del_2005_0079 - 2005-12-27

6.2 Grid calculations of the “Les Saintes” network

6.2.1 Calculation tools Grid calculations have been performed by the University of Kassel. The grid of Les Saintes has been modeled in different software tools : DigSilent for loadflow calculations, WindPro for flicker calculations, ATP-EMTP for voltage rise.

6.2.2 Basic data The information and data were provided by VERGNET and EDF [24], [25]. The grid is divided in two feeders, one for each island.

The main criteria concerning the assessment of the operation condition is the voltage level before and after the connection of the wind farm. So, our investigations concerning the grid of Les Saintes are mainly focused on load-flow calculations. Short-circuit calculations also have to be performed in order to check other connection points and to calculate long term flicker values.

The following grid nodes have been selected:

Table 22 - Selected grid nodes

Name Reason for observation Vieux-Fort Feeding point of the 20 kV grid “Les Saintes”

Crawen First connection point to the island “Terre de Haut” / Main coupling point of the island grid

PCC Connection point of the wind farm to the grid

Labas Terminal node of the radial line Crawen – Labas to which the wind farm is connected

Zozio Terminal node of the radial line Crawen – Zozio

Figure 52 - Les Saintes island grid and the selected grid nodes

DISPOWER Del_2005_0079 - 2005-12-27

6.2.3 Assumptions Since some data of the grid components and the wind farm connection point are not available, several assumptions have to be made in order to simulate the grid.

High voltage (HV) grid at the transformer feeder station “Vieux-Fort”: the grid angle of the high voltage network has been assumed to 84.29° (R/X-ratio = 0.1).

HV transformers in the substation “Vieux-Fort”: the following data has been supposed: • I0 (no-load losses) 0.5% • PCU (copper losses) 160 kW • PFE (magnetisation losses) 32 kW • Switching group YNd-5 These data were taken out of or calculated in relation to the German norms VDE 0532 resp. DIN 42523. Due to the transformer capacity utilization of less than 10% in case of maximum load within the grid, we assumed that only one transformer is in operation.

Cables and overhead lines in the “Les Saintes” network: exact data were available concerning the lengths and impedances of the cables and lines, but no data about their capacitive reactances. Based on the German norms DIN 48201 resp. IEC 208-1966 for overhead lines and DIN VDE 0276-620, the following data were used.

Table 23 - Assumptions concerning the cables and overhead lines

Type Cross-section (mm2) B´ (µS/km) Number of parallel lines Cable AL-150 (grid) 150 80.00 1 Subsea Cable (grid) 35 57.00 1 Overhead Line (grid) 54 2.00 1 Cable AL-50 (WTG) 50 56.55 1

Loads: the loads were assumed to be independent from the voltage.

Wind turbines: the following data have been assumed for the transformers and generators:

• Transformers: I0 (no-load losses) 0.222 % PCU (copper losses) 4.5 kW PFE (magnetisation losses) 0 kW Vector group Dyn-11

• Generators: IA/IN (starting current ratio) 5.0 R/X – ratio 0.15

Cabling within the wind farm: the wind park has been implemented as depicted in Figure 53. Hereby, the following data are used for the cables (aluminium cables with a cross section of 50 mm2).

Table 24 - Parameters of the wind farm cabling

R´ (Ω/km) X´ (Ω/km) B´ (µS/km) Number of parallel lines

0.641 0.13823 56.55 1

DISPOWER Del_2005_0079 - 2005-12-27

Figure 53 - Cable length within the wind farm

6.2.4 Voltage levels

Voltage limits: at all MV grid nodes, the voltage should be within the range UN ± 6 % (18.8 kV, 21.2 kV).

Operation conditions: the aim of grid calculations is the determination of voltages at the grid nodes for worst case operation conditions. Normally, the following conditions can be defined as worst-case:

• grid without wind turbines at maximum load, • grid with maximal power output from the wind turbines at light load.

Furthermore, the following grid operations are interesting: • grid without wind turbines at light load, • grid with maximal power output from the wind turbines at peak load.

Calculation results of other grid conditions supplements the worst case examinations. For instance, the knowledge of the short-circuit power at the several grid nodes represents important data for the design of the grid connection of a new wind turbine. The capacity utilization of the grid components after the connection of the wind turbines is also useful for electric and economic analysis.

In this report, mainly the calculation results for the P1min feed-in power of the WTG are presented. To refer to French regulations, they present the node voltages referred to the nominal voltage:

−⋅=∆ 100[%] .

No-load operation: according to the data provided by EDF the no-load voltage on the 20 kV side of the HV/MV transformer is 20.5 kV with a neutral tap position (tap 9). Based on this value, the voltage of the “far generator” has been set to:

DISPOWER Del_2005_0079 - 2005-12-27

Table 25 - Voltages of the “Far Generator” in relation to the initial short-circuit power

Initial short-circuit power at HV Level Corresponding voltage of the Far Generator 207 MVA 61.5851 kV 248 MVA 61.5752 kV 300 MVA 61.5666 kV

Calculation results

The following results are calculated using the neutral tap position of the HV/MV transformer:

Table 26 - Node voltages without wind turbines at peak load (neutral tap position)

207 MVA 248 MVA 300 MVA Node Voltage

[kV] Voltage

[%] Voltage

[kV] Voltage

[%] Voltage

[kV] Voltage

[%] Vieux-Fort 20.3443 1.72 20.3612 1.81 20.3758 1.88 Crawen 19.4334 -2.83 19.4513 -2.74 19.4667 -2.67 PCC 19.3930 -3.04 19.4109 -2.95 19.4263 -2.87 Labas 19.3786 -3.11 19.3965 -3.02 19.4120 -2.94 Zozio 19.3471 -3.26 19.3650 -3.17 19.3805 -3.10

Table 27- Node voltages without wind turbines at light load (neutral tap position)

207 MVA 248 MVA 300 MVA Node

Voltage [kV]

Voltage [%]

Voltage [kV]

Voltage [%]

Voltage [kV]

Voltage [%]

Vieux-Fort 20.5067 2.53 20.5073 2.54 20.5077 2.54 Crawen 20.1217 0.61 20.1222 0.61 20.1227 0.61 PCC 20.1045 0.52 20.1051 0.53 20.1055 0.53 Labas 20.0980 0.49 20.0985 0.49 20.0989 0.49 Zozio 20.0822 0.41 20.0828 0.41 20.0832 0.42

Grid with 8 wind turbine (P1min) at peak load: the wind farm at full power covers 160% of Les Saintes peak load. Table 28 shows that at peak load, the maximum 6% voltage rise limit is kept at any point of the grid.

Table 28 - Node voltages with wind turbines at peak load (neutral tap position)

Voltage [kV]

Voltage [%]

Voltage [kV]

Voltage [%]

Voltage [kV]

Voltage [%]

Grid with 8 wind turbines (P1min) at light load: the wind farm at full power covers 300% of Les Saintes islands light load. Table 29 shows that at light load, the maximum 6% voltage rise limit is not kept at almost any point of the grid. This is one of the main constraints to wind penetration on this grid, as it is usually in most of practical cases.

DISPOWER Del_2005_0079 - 2005-12-27

Table 29- Node voltages with wind turbines at light load (neutral tap position)

Voltage [kV]

Voltage [%]

Voltage [kV]

Voltage [%]

Voltage [kV]

Voltage [%]

6.2.5 Flicker levels

Flicker limits: on MV level, the maximum allowed levels for flicker produced by the wind farm are 0.35 for Pst (probability short term) and 0.25 for Plt (probability long term).

Manufacturer’s data: the following data concerning flicker (type 1) was given by the wind turbine manufacturer:

Table 30 - WT-flicker coefficient for different grid impedance angles Ψk

Network phase angle 30° 50° 70° 85° 6 m/s 12.6 11.4 9.0 6.3 7 m/s 12.4 11.3 8.9 6.2 8 m/s 12.4 11.3 8.9 6.1

Since the calculation of the long term flicker at any grid node requires the knowledge of the flicker coefficient for each possible grid impedance angle, the flicker coefficient has to be generated using these data. This has been done describing the worst-case data (at 6 m/s) by a quadratic polynomial

97.110685.00016.0 2 +Ψ⋅+Ψ⋅= kkc

Basics: based on the calculation algorithm for different wind turbines in the grid, the flicker contribution of each wind turbine is given by

´´1.1k

WTWTlt S

ScP ⋅⋅= .

where Sk” is the initial short-circuit power of one grid node affected by a single wind turbine. This can be neither the grid node to which the turbine is connected nor the node where the long term flicker is being calculated. The derivation of this node is not presented in the report. However, this explains that, the grid nodes PCC, Labas and Petite anse on the one hand, and all the grid nodes from Crawen to Zozio on the other hand, have the same flicker values. The interaction of each single flicker value to the total flicker is described by the formula

mWTiltreslt PP ∑

where factor n represents the number of wind turbines and factor m represents an interaction factor normally set to 2. In case of additional flicker sources in the grid, the factor m can be set up to 3.

Calculation results

As mentioned above, due to the derivation method of the grid node to which a wind turbine is virtually connected, the grid nodes (see Figure 52) PCC, Labas and Petite anse on the one hand, and all grid nodes from Crawen to Zozio on the other hand, have the same flicker values.

DISPOWER Del_2005_0079 - 2005-12-27

The results mainly depend on the grid angles at the grid nodes which even are lower than 30°. The calculated values are depicted in Table 31.

Table 31- Short-circuit power Sk´´ and grid angle Ψk at the selected grid nodes

207 MVA 248 MVA 300 MVA Node Sk

´´ Ψk Sk´´ Ψk Sk

´´ Ψk Vieux-Fort 100.8744 86.2191 110.7010 86.4071 121.0162 86.6046 Crawen 29.6866 29.3728 30.1323 28.1287 30.5212 27.0094 PCC 25.0372 27.9817 25.3428 26.9178 25.6090 25.9656 Labas 23.1110 28.2113 23.3731 27.2328 23.6015 26.3583 Zozio 24.8252 29.6525 25.1392 28.6181 25.4136 27.6915

The flicker values for the selected nodes are given for interaction factors m of 2 and 3 in Table 32.

Table 32 - Long term flicker values at the selected grid nodes

207 MVA 248 MVA 300 MVA PCC, Labas

Vieux-Fort

Crawen, Zozio

PCC, Labas

Vieux-Fort

Crawen,Zozio

PCC, Labas

Vieux-Fort

Crawen,Zozio

m = 2 0.4317 0.0514 0.3635 0.4268 0.0466 0.3586 0.4226 0.0424 0.3544 m = 3 0.3052 0.0364 0.2570 0.3018 0.0329 0.2536 0.2988 0.0299 0.2506

The calculation results show that the limit level of 0.25 is not kept, even if the interaction factor m is set to 3. Using factor m = 2, only two wind turbines could be installed in order to keep the flicker level within the limit at the PCC node.

6.2.6 Voltage variations due to switching effects The maximum voltage variation due to switching effects of each wind turbine can be calculated by:

´´max

max 1.1k

u⋅⋅

⋅=∆λ

where PnWT is the nominal active power of one wind turbine. Factor λ represents the real power factor. In our investigations, we equate this factor with the specified cos°ϕ value, 0.1cos == ϕλ . Factor ki max has been specified to 1.1. The resulting voltage modifications are given in the following table.

Table 33 - Voltage modifications due to switching effects at the selected grid nodes

207 MVA 248 MVA 300 MVA PCC,

Labas Vieux-Fort

Crawen, Zozio

PCC, Labas

Vieux-Fort

Crawen, Zozio

PCC, Labas

Vieux-Fort

Crawen, Zozio

∆umax [%] 1.3290 0.3299 1.1209 1.3130 0.3006 1.1043 1.2993 0.2750 1.0902

6.2.7 Short-circuit levels Short-circuit levels have been calculated with and without wind turbines at different nodes of the grid. The results are shown in Figure 54 and Figure 55. Considering the topology of the grid, the contribution of the wind farm to the short-circuit levels is low, and in any case the total short-circuit level remains far below the capacity of grid components.

6.2.8 Grid capacity utilization Grid components capacity utilization has been checked. It remains far below 100% (see Figure 56). The capacity utilization of the sub-sea cable remains almost constant, but the power flow changes direction after connecting the wind farm.

DISPOWER Del_2005_0079 - 2005-12-27

Figure 54 - Short-circuit levels without wind turbines

Figure 55 - Short-circuit levels with wind turbines

Figure 56 - Grid capacity utilization (in % of In)

Zozio Labas

Vieux-Fort Petite anse

Crawen

Sk“=300 MVA Sk“=207 MVA ψk=0°

Sk“= 30 MVA Sk“= 29 MVA ψk= 27,01°

Sk“= 121 MVASk“= 101 MVA ψk=86,61°

Sk“= 25,6 MVA Sk“= 25,0 MVA ψk=25,97°

Sk“= 23,6 MVASk“= 23,1 MVA ψk=26,36°

Sk“= 24,0 MVA Sk“= 23,5 MVA ψk=26,28°

Sk“= 25 MVA Sk“= 24 MVA ψk= 27,69°

Zozio Labas

Vieux-Fort Petite anse

Crawen

Sk“=300 MVA Sk“=207 MVA

Sk“=30+9,6 MVA Sk“= 29+9,6 MVA

Sk“=121+8,3 MVA Sk“= 101+8,3 MVA

Sk“= 25,6+10 MVA Sk“= 25,0+10 MVA

Sk“= 29,4 MVASk“= 29,2 MVA

Sk“= 30 MVA Sk“= 29,8 MVA

Sk“= 10 MVA Sk“= 10 MVA

Sk“= 30,12 MVA Sk“= 29,8 MVA

Zozio Labas

Vieux-Fort

Petite anse

Crawen 36,4 / 2,7

18,4 / 1,4

17,8 / 2,0 44,1 / 4,9

28,4 / 21,43

10,9 / 8,5

Centrale

Les Santies

With WTGs / Without WTGs

Felicite

T-Felicite

DISPOWER Del_2005_0079 - 2005-12-27

6.2.9 Fault ride through Small grids, and even more small island grids, are already subject to high DG penetration levels for various reasons: - they are usually powered by diesel plants, and electricity produced from oil is very expensive; - they are on islands or remote areas, where RES are abundant: water falls, wind, solar, geothermal.

On grids with high penetration levels, the loss of all DG in case of grid perturbation will create a very hazardous situation for the network operator. That is why the grid operators require that the DG remains connected to the grid in case of voltage drop. Requirements differ in different European countries (cf §6.2.10).

In France, the regulation (Ministerial Order of March 2003) for island grids requires that a power generator remains connected to the grid and continues to operate when the grid voltage is reduced to 0,3 Un during 0,6 sec, and to 0,7 Un during 2,5 sec. This requirement is up to now not applicable to asynchronous generators.

Wind turbine behaviour in case of voltage dips

Vergnet is using simple squirrel cage, direct coupling induction generators. This type of generators usually has no fault ride through capabilities. In case of a voltage dip, the active power delivered to the grid is instantaneously reduced, the resistive torque on the wind turbine generator shaft is reduced accordingly, hence the wind turbine rotor speed increases. The squirrel cage induction GEVMP wind turbine may lose its stability, and the rotor speed may increase until the overspeed protection trips.

At the same time, components of the wind turbine coupling panel cannot continue to operate properly if their supply voltage comes directly from the grid, without any backup. This is the case for contactor coils, control devices. Usually the minimum acceptable voltage for a safe operation is 0,7 Un.

Deep voltage dips occur in case a short-circuit occurs on the grid. Usually the short-circuit is cleared in 100 to 500ms (if it is not on the feeder where the power producer is connected). If the short-circuit lasts longer, it is impossible to maintain the wind turbine on the grid at a reasonable cost, and the wind turbine protections will trip the main contactor and stop the rotor.

In a few 100ms, the wind turbine rotor speed increase is more than 5 to 6% if no other action is taken. The pitch control could be used to reduce the rotor speed increase and contribute to a stable operation of the WTG. An aerodynamic brake using a fast pitch activation already exists on the GEVMP wind turbine. A similar control function could be developed to pitch the blades in case of a voltage dip. Coupled to a source of energy that compensates for the grid loss, the system should be able to withstand a grid voltage dip for a limited period of time.

Wind turbine behaviour in case of frequency variations

In case of grid frequency variations, the rotor speed will change and follow the grid frequency. If the frequency is reduced, the rotor speed will be reduced in the same proportion, and the wind turbine will continue to operate but at a lower speed. The power curve of the wind turbine will be modified and most probably to lower performance. The pitch could be used to adjust to the new speed, but this would require a maximum possible power track function. If the frequency is increased, the rotor speed will be increased in the same proportion, and the wind turbine may reach its maximum speed. There is not much to do if the wind turbine design does not allow the wind turbine operation at the new grid frequency.

6.2.10 Results evaluation with respect to regulations of different countries Connection criteria to island distribution networks in Europe are presented in [26].

Voltage levels

Due to the distance from the main power station, injection of wind energy on the grid at PCC node produces significant voltages rises. In France, voltage fluctuations shall be limited so that the grid

DISPOWER Del_2005_0079 - 2005-12-27

voltage does not exceed the maximum voltage limit. This limit is +/-7 % (reduced to +/-6% to allow for measuring uncertainty) for isolated grids not connected to the mainland, which is the case for Guadeloupe. At light load, the calculation results have shown that this voltage rise is above 6%. French regulations are not complied with.

In Germany and Greece, voltage changes (from no wind power to full wind power) shall be limited to a given value: 5% in Germany and 3% in Greece. Load flow results can be used to show the voltage changes. They are quite similar at low load and peak load. Figure 57 shows the results at peak load.

Voltage fluctuations in points of the grid with maximal loads

25% 50% 75% 100% 109% 127%

PnG (%)

dU (%)

PCC Vieux-Fort Labas Pentite anse Zozio

Figure 57 - Voltage change with peak loads compared to the German 5% limit

The voltage change is above 5% as soon as the wind power exceeds 60% of rated power (1.3MW), and above 3% as soon as the wind power exceeds 35% of rated power (0.75MW).

Flicker levels

Wind speed variations and wind turbines operation produce power fluctuations that generate flicker on the grid. With the default flicker coefficients used for the calculations, the maximum flicker level at PCC node is between 0.42 and 0.43. This would be acceptable in Greece and in Germany.

In Spain, to limit the flicker, the wind farm power shall not exceed 5% of the grid short-circuit power at the connection point. This short-circuit power is around 25 MVA.

Table 34 - Short-circuit power Sk´´ at PCC grid node

Node 207 MVA 248 MVA 300 MVA PCC 25.0372 25.3428 25.6090

That means the wind farm rated power shall not be above 1.25MW.

6.3 Solutions to improve wind penetration

6.3.1 Possible measures to limit the grid voltage deviation Basically, the solutions to keep the grid voltage within the contractual limits at any time are: - Change the grid topology or capacity, - Modify the grid voltage control,

DISPOWER Del_2005_0079 - 2005-12-27

- Change the wind turbine technology, - Add some extra equipment for grid voltage deviation compensation, - Reduce the wind farm power, permanently, or only when required. The voltage rise is only a problem when the wind speed is high and the grid load is low.

6.3.1.1 Grid reinforcement One option is to reinforce the grid in order to reduce the voltage deviations. This is a very expensive option, which cannot be motivated only to allow more renewable power to be fed to the grid.

6.3.1.2 Connection of the Wind Farm to another Grid Node of the Island An important characteristic for the connection of a wind farm to a grid node is the initial short-circuit power at this node. If the initial short-circuit power is much higher (in Germany the factor is 50) than the apparent power of the total wind farm, the grid connection will generally cause no problems. So, the initial short-circuit power at the grid nodes can be used to find a better connection point.

The calculations results show that the connection of the wind farm at Crawen would lead to slightly lower node voltages. But the improvement would be insignificant.

6.3.1.3 Adjustment of the Wind Turbines Power Factor Initial calculations have been performed with a power factor of 1, as the wind farm shall not consume any reactive power. Additional capacitor banks are provided in each wind turbine to comply with this requirement.

If a significant amount of reactive power is consumed by the wind farm, losses will be increased on Les Saintes grid, and the grid voltage will be changed. Calculations were performed with power factors of 0.95, 0.9 and 0.85. The results show that changing the wind farm power factor set point from 1 to 0.85 would reduce the voltage at PCC by 3%. This is enough to maintain the grid voltage within acceptable limits. The calculation results for a initial short-circuit power of 300 MVA are presented at light load in Figure 58.

Such a solution is easy to implement by switching off the compensation capacitors of 1 to 8 wind turbines in operation. It could be done only when necessary, i.e. when the grid voltage is above a given value.

Figure 58 - With wind turbines (109%) at light load (neutral tap position)

DISPOWER Del_2005_0079 - 2005-12-27

6.3.1.4 Variation of the High Voltage Transformer Tap Position Another possibility to reduce the voltages in the island grid is to modify the tap position of the HV/MV transformer. Figure 59 and Figure 60 show the results for the grid with the wind turbines at peak load and light load.

Figure 59 - With wind turbines (109%) at peak load; tap position 11

Figure 60 - With wind turbines (109%) at light load; tap position 11

To avoid risks of low voltage when the wind farm is delivering a low power and the grid load is high, the HV/MV tap changer set point should be changed only in some situations, i.e. when the grid voltage and the wind farm power are above a given value.

In principle, this solution could almost solve the problem of voltage rise. However it should be checked that the voltage limits are not exceeded on the low voltage (LV) side of the MV/LV transformers and on the LV feeders connected to the MV network.

6.3.1.5 Limitation of wind power To maintain the grid voltage within required limits, a simple solution is to reduce the power of the wind farm. Two options are possible: - reduce permanently the maximum output power of the wind farm, - reduce temporarily, when necessary, the output power of the wind farm.

In the first case, to maintain the grid voltage at PCC below UN + 6% (21.2 kV), the wind farm power should be reduced below about 65% of its rated power, i.e. 1.4 MW. In the second case, the power limitation would be done only if the grid voltage is, say above UN + 5.5%, generally when the load is minimum on the grid and the wind speed is high.

DISPOWER Del_2005_0079 - 2005-12-27

Various possibilities exist to reduce the wind farm power: - Wind turbines disconnection: the simplest one is to disconnect some of the wind turbines. This

solution would increase the number of switching of each wind turbine, and therefore the flicker level. But if it is used only when necessary, the flicker level increase will be negligible.

- Wind turbine power limitation: another option is to use the wind turbine pitch mechanism to change the maximum power set point of each wind turbine. This is easy to do if the wind turbines are already fitted with a pitch regulation device. The maximum wind turbine power could be fixed between 70 to 100% of Pn (wind turbine rated power). If it is necessary to limit the power of each wind turbine at a very low value, the disconnection is a better option.

Reducing the wind farm power obviously reduces its yearly energy yield. So, it is important to be able to evaluate the energy losses resulting from such a power limitation. In practice, if the wind farm power is limited only when necessary, these losses are usually low and affordable for the producer.

Measures have been made on a wind farm in Guadeloupe on a site where the average energy yield is 1900h to 2000h at full power, to assess the probability of having the wind farm power exceeding a given value for various wind speeds [28]. These measures show that in practice all wind turbines do not see the same wind speed at the same time and that even for high wind speeds (above 13m/s), the probability to have a wind farm power above 90% of Pn is 10%, above 69% of Pn it is 50% and above 57% of Pn it is 80%. If we consider in the Weibull distribution that on this wind farm, the average wind speed is about 8m/s with k=3.8, the probability to have a wind speed above 13m/s is less than 2%.

Therefore, if the maximum number of WTGs that could be connected to the grid to maintain the grid voltage within acceptable values is defined by considering their maximum power when the grid load is minimum, then we size the whole wind farm for an event having a probability that is negligible.

In practice, for places where the shape factor of Weibull wind distribution is high like in Les Saintes, to reduce the wind turbine maximum power does not lead to high energy losses, and is acceptable by the wind energy producer.

6.3.1.6 Change of wind turbine technology Slow voltage variations are fixed by the wind farm (active and reactive) power output. The impact of the wind turbine technology on the grid voltage deviation mainly depends on its capability to control the reactive power output. Due to their power electronics interfaces, doubly-fed WTG and synchronous WTG generally have the capability to control the reactive power or the grid voltage provided that the power electronics interface is properly sized.

Additional equipment could also improve the overall impact of the wind farm on the grid. Simulations have been made by American Superconductor with a D-VAR equipment. The network was modelled for peak and off-peak loads, as well as full, mid, and zero generation of the wind farm. The objective was to correct the over-voltages by modelling a 3 MVA D-VAR at the connection point of the wind farm to the grid. The reactive power the D-VAR needs to supply for the different load levels and generation levels are summarized in Table 35.

The simulation results show that with the addition of the 3 MVA D-VAR, all the voltages can be kept within the specified range of ± 6%. For instance, Table 36 shows the voltages at various buses obtained with the D-VAR for the off-peak load. This equipment could also be used for other purposes, such as flicker compensation and fault ride through capabilities.

Table 35 - Reactive power to be supplied by the D-VAR to keep voltages within range of ± 6%

Load level Peak Off peak

Generation level (kW/machine) Generation level (kW/machine) 275 137.5 0 275 137.5 0

MVar at PCC -0.783 0 0 -2.625 0 0

DISPOWER Del_2005_0079 - 2005-12-27

Table 36- Voltages at various buses for different generation levels and off-peak load obtained with the D-VAR.

Generation Level (kW/Machine) 275 137.5 0

Voltage

(pu) % Voltage from 1 pu

Voltage (pu)

% Voltage from 1 pu

Voltage (pu)

% Voltage from 1 pu

Vieux 1.017 1.7% 1.028 2.8% 1.028 2.8%Crawen 1.044 4.4% 1.048 4.8% 1.022 2.2%PCC 1.05 5.0% 1.052 5.2% 1.024 2.4%Labas 1.05 5.0% 1.052 5.2% 1.024 2.4%Zozio 1.042 4.2% 1.046 4.6% 1.02 2.0%Wind Farm Bus 1.06 6.0% 1.059 5.9% 1.024 2.4%

6.3.1.7 Distributed load control One way to influence the grid voltage deviation is to have an action on the grid load profile. As the voltage rise occurs when the grid load is low, an increase of this load when the voltage is too high could solve part of the problem.

In practice, transferable loads are heating, pumping, air conditioning. Many hotels on Les Saintes islands use this kind of loads. But switching on devices before they are really needed is not easy to organize except for very short periods of time.

6.3.2 Possible measures to reduce the long term flicker

6.3.2.1 Change of wind turbine technology Studies made in DISPOWER [29] show that wind turbine technologies have an impact on fast voltage fluctuations due to wind speed variations and rotation of blades.

Squirrel cage induction generator: with a fixed pitch, wind speed variations are rapidly followed by corresponding variations in the active power provided to the grid, depending on the inertia of the machine. These power variations create equivalent voltage variations on the grid.

Doubly-fed induction generators: the converter to the rotor can regulate the rotor speed to extract the maximum power from the wind, and can also regulate the reactive power provided to the grid. So it is possible to smooth the variations of the active power due to wind speed variations by adjusting the rotor speed, and therefore to smooth the variations of the grid voltage.

Synchronous generators: synchronous generators act as doubly-fed induction generators with regards to active power fluctuations. They can also smooth the variations of the grid voltage.

When pulse width modulation (PWM) based inverters are used as interface between the generator and the grid, they have the ability to provide many of the power quality functions already achievable with synchronous generators, as control of under- and over-voltages. PWM inverter also presents an opportunity to provide voltage phase unbalance correction, mitigation of voltage variations and fluctuations when configured as a STATCOM or D-VAR [33].

6.3.2.2 Additional Statcom device STATCOM or D-VAR equipment can also be added in parallel with the wind turbines. In practice, this solution is used for large wind farms (at least few 10MW) at high voltage levels (above 100kV). Here are several examples of the DVAR application in the US.

DISPOWER Del_2005_0079 - 2005-12-27

Table 37 - Examples of D-VAR applications

Site name D-VAR size Generation size Additional capacitors Total reactive capability Mendota Hills 4 MVA 50 MW 2x7.2 MVar -4 to +18 Mvar Summerview 8 MVA 63 MW 6x6.6 MVar -8 to +47 Mvar Caprock wind 2x8 MVA 80 MW 9x3.6 MVar -16 to +48 Mvar Rush Lake wind 2x8 MVA 150 MW 8x13.2 MVar -16 to +121 MVar

Unfortunately, this kind of equipment has several disadvantages: - For small applications, the cost of such equipment is too high. - Electronic devices are difficult to maintain in tropical sites like Guadeloupe, unless they are

enclosed in a room fitted with air conditioning, which increases even more its cost. - This type of equipment generates harmonics.

6.3.2.3 Pitch regulation improvement The initial simulations have been made using default flicker coefficients. Since then, pitch regulation algorithms have been improved to reduce the flicker generated, and real coefficients have been measured using IEC 61400-21.

Thanks to the in-house mechanical simulation and control algorithm knowledge, Vergnet developed efficient and sophisticated control functions. A powerful regulation algorithm developed by Vergnet, has significantly improved the performances of the power control, reducing the amplitude of power peaks and activity of pitch actuators, leading to a much more stable power output. The measurement results and improvements are shown on the figure below (purple curve showing power improvements before and after).

Before improvements After improvements

Figure 61 - Power Control Improvements

The measured flicker coefficients are significantly lower than the default ones. With these coefficients, the flicker level in the case of Les Saintes will be reduced to an acceptable level (Table 38).

DISPOWER Del_2005_0079 - 2005-12-27

Table 38 - Measured flicker coefficients

Grid impedance phase angle (ψk) 30° 50° 70° 85° Annual wind speed Va (m/s) Flicker ratio c(ψk, Va)

6.0m/s 6,4 5,13 6,02 6,44 7,5m/s 6,58 5,05 5,9 6,31 8,5m/s 6,64 5,05 5,9 6,31 10m/s 6,67 5,05 5,89 6,31

6.3.3 Fault ride through capabilities

6.3.3.1 Use of variable speed or statcoms A PWM inverter can be configured to provide a STATCOM function for voltage dips but the utilisation of a shunt device requires a significant amount of reactive power. This may require that the rating of the inverter be increased beyond that required for generation.

In order to fulfill the DNO requirements for the GEVMP 250/275kW wind turbine, Vergnet has been planning to use a power electronic converter. Technical studies carried out with major variable frequency drive manufacturers showed that the GEVMP combined with a power electronic converter could fulfill all the power quality requirements. The major drawback of this solution is its unacceptable cost increase (more than 50%), compared to the classic solution used at the moment (two-speed induction generator and thyristors soft starter).

6.3.3.2 Advanced pitch regulation Vergnet has performed studies to evaluate how its GEVMP 250/275kW wind turbine could be improved to stand voltage variations defined in the regulation (“Ministerial order of March 2003”).

To maintain power and control components in safe operation during the voltage dip, more standard solutions can be used, like backup battery and inverter, capacitors on contactor coils, etc.

To avoid a wind turbine trip due to overspeed after a voltage drop, the possibility of rapid action on the pitch system has been studied.

Voltage reduced to 70% during 2,5 sec

Figure 63 shows that the rotor speeds up very fast and then reaches a stable speed that remains acceptable. No trip occurs but, if the voltage reduction is sharp, a torque variation is observed (Figure 64).

Figure 62 - Voltage profile

DISPOWER Del_2005_0079 - 2005-12-27

Figure 63 - Rotor speed (rad/s) during voltage drop

Figure 64 - Torque variations after a sharp voltage reduction

Voltage reduced to 30% during 0,6 sec

This time, the rotor speed increase is much higher, and the wind turbine is stopped.

Figure 65 - Rotor speed after a voltage drop to 30%

DISPOWER Del_2005_0079 - 2005-12-27

Various types of action on the pitch to reduce the driving and the rotor speed acceleration torque have been studied:

Figure 66 - Rotor speed after a 30% torque

reduction, 0,5 sec after the fault

Figure 67- Rotor speed after a 70% torque

reduction, 0,4 sec after the fault

In the case of Figure 66, the rotor speed is not controlled, but in the case of Figure 67, the rotor speed is controlled and the current increase is acceptable. Such torque control is possible with the existing pitch control mechanism and control technology.

6.4 Possible uses of G.I.S. tools Potential uses of GIS tools for Les Saintes are presented in [35]. GIS could help to have a better understanding of the behavior of a grid with DG, as Les Saintes. It could be used for:

Planning - Compilation of statistics on the production of electricity and energy, and on the demand. - Analysis of the statistics produced. - Study of the economic balance of existing equipments.

Studies on networks based on constraints - Simulation of distributed PV generators - Simulation of distributed loads - Study of different possible scenarios - Definition of best choices - Collect data on the different renewable resources (biomass, hydrologic, solar, wind) - Model the potentials according to different technical solutions for production - Collect data on the different constraints in technical, economic, environmental, network domains - Propose different scenarios of development based on some realistic hypotheses.

Simulations for development and operation of the grid - Study different load development scenarios - Simulate the voltage on a period of time - Determine the actual operation conditions of the network - Simulate operational use of present network and production means based on different demands - Propose production development scenarios based on technical and economic aspects combined

with demand - Propose network development scenarios - Evaluate the impact of the proposed solutions on environment. - Evaluate the impact of various possible solutions to improve the DG penetration

Maintenance preparation - Location of DG generators (if many distributed generators) - Optimisation of maintenance operation

DISPOWER Del_2005_0079 - 2005-12-27

Complete decision support system - Build a complete decision tool, interfaced with a Network Management System

These tools can be very helpful for a reasonable and pragmatic development of DG in Guadeloupe. However a strong collaboration is required between the politicians, the energy providers and the energy distributors in order to obtain the most efficient use of these tools. As the most important time-consuming task in such a project is the collection of data, it is very important not to under-estimate it. A user-driven approach should be privileged in order to fit their needs and expectations. A continuous dialog will favor the use and the development of adequate tools based on GIS for DG.

The main problem in this study was to collect the data to perform simulation using GIS tools. These data have been found confidential and not accessible to private operators.

6.5 System operation after the loss of sub-sea cable

6.5.1 Description of the system If the subsea cable between Vieux Fort and Crawen is out of order, the diesel plant on Terre de Bas is started and supplies the two Les Saintes islands. The existing distributed generators are therefore coupled on a diesel grid.

Figure 68 - Les Saintes grid after a failure on the subsea cable

The wind turbine output varies according to wind speed fluctuations. On a wide interconnected grid these power fluctuations are evened out by powerful power plants and lines, and the grid is able to maintain a fixed frequency and voltage. On a small isolated diesel grid where the wind penetration is significant (say above 10%) specific solutions should be used to maintain the power balance between production and consumption. These solutions are necessary to keep the voltage and frequency within acceptable limits. The balance between production and consumption is normally maintained in spite of wind farm power fluctuations thanks to the diesel engines regulation. But introduction of wind power on the grid reduces the load on the diesel engines. It is only possible to reduce this load down to 30 to 50% of their rated power (so called minimum permanent running load). At low load, diesel efficiency drops significantly, and after more than 10 to 20 hours engine wear may occur in the cylinders. This wear may create severe mechanical faults when the engine is brought back to full load. Therefore, when there is an excess of wind energy, an action is required to maintain the diesel load above the minimum value. We have three options: store this excess energy in batteries, disconnect some of the wind turbines from the grid, or increase the load on the grid. Maintaining system stability at high penetration levels has been widely studied and has been confirmed under test bed conditions but there has been limited demonstration in the field. At high wind power levels the fluctuations in wind power can cause the system frequency and voltage to vary. This is a relatively simple control problem which has been solved in high penetration wind-diesel

Zozio Labas Vieux-Fort

Petite anse

Crawen

Wind farm Diesel plant

DISPOWER Del_2005_0079 - 2005-12-27

systems, using dump loads or wind power regulation for frequency regulation and synchronous condensers or short-term storage with power converters for voltage regulation. Demonstration projects are in the planning stages but there is not yet significant numbers of demonstration or commercial projects. Wind-diesel technology is ready for deployment into the field in many configurations. In low penetration applications, where virtually no impact is made on diesel plant operation, there are no obstacles, other than economic, to wide spread deployment of the technology. In medium to high penetration level applications, most of the technical issues are under study or have been solved. Many of the problems have been found not to be an impediment to the deployment of the technology. But in our case we have to cope with existing diesel engines. When the existing diesel equipment is old, the grid stability may be difficult to maintain without heavy modifications. The state of the existing diesel plant is often a practical limitation to high DG power penetration. The more important the wind level of penetration in the diesel network is, the higher the complexity of the system is. A monitoring system becomes essential with the top from 5 to 10% of wind penetration and with beyond 20% it becomes necessary to set up a system of control.

6.5.2 Grid control techniques Some companies had focused their Research and Development credits to supply renewable electricity in remote areas, where production costs are high. One target is Wind/Diesel systems, for which the companies had developed specific approaches to increase the Wind Power Capacity and improve the economics of these hybrid plants, generally adapted to insular grids. On Les Saintes, the wind farm is connected to the MV grid at the island windiest place. To allow wind diesel joint operation, a remote control links between the wind farm and the diesel plant should be established. A computer should analyze in real time the operation parameters and adjust the number and power of connected wind turbines to fully optimize the wind potential, while keeping the diesel generators running above their “minimal load condition” (in general 30 to 40% of nominal power). This system allows an instantaneous supply of wind power to the grid up to 70% of the total load. The yearly average wind energy supply to the global demand may then be as high as 20 to 50%, depending on the wind potential of the site and the size of the diesel. Several companies use this system of connection/disconnection. For example Vergnet SA proposes this system of connection/disconnection for wind-diesel network and has several wind-diesel projects in activity like Désirade island (Guadeloupe), Lifou island (New Caledonia), Rodrigues island (Indian Ocean), Pine Island (New Caledonia)… On Les Saintes, with a consumption varying from 0.7 to 1.5MW, and 2MW of diesel engines, there is little room for wind energy. For the moment, the grid operator requires to shutdown the wind farm when the diesel plant is started to provide extra guaranteed power to the grid. This possibility is used when one of the main power stations is out of service.

Heat storage

In order to smooth the fluctuations in the excess power and provide a more consistent heating power output for the user, electric thermal storage (ETS) heating units can be used. They are resistance heaters with electric elements encased in ceramic blocks. They are commercially available space heating units designed for individual rooms and equipped with thermostats and fans to dispense the stored heat. Reaching temperatures around 675° C (1250° F), they can provide several hours of thermal storage, depending on the size and rated output of the heater. All heaters listed have a nominal input voltage of 240 V. This type of system is in place in Cuttyhunk Island, New England.

The fundamental objectives of the heating system are to: increase the amount of economically useful wind energy, save heating fuel, maintain quality of primary load.

DISPOWER Del_2005_0079 - 2005-12-27

The concept of controllable distributed secondary loads as a method to dump excess power usefully in wind/diesel systems has existed for several years. The basic idea is to activate or shed the distributed loads individually as the excess power varies. Control is typically the main issue for implementing dispatchable distributed loads, because the challenge is to vary the load to match the available power. Switching distributed loads requires some command to be transmitted from a location in the system where the available excess power is known. In its simplest form, the command should switch a relay to enable or disable the load. This necessitates, for each controllable load a controller with some form of intelligence located near the load to receive and process the command, and a central controller with knowledge of the excess power to create and transmit the commands. There are a number of factors that influence a detailed load control strategy for a particular application, including : Type of load (i.e. space heating, hot-water heating, water pumping), Number and location of the loads on the system, Existing communication carrier infrastructure (i.e. phone line, cable), Expected profile of excess power.

Load adjustment

This system was tested and developed in AWTS (Atlantic Wind Test Site) by Hydro-Quebec in 1988 and is actually installed in various sites. The diesel engine supplies power back-up for the wind turbines. In periods with sufficient wind power the diesel engine is automatically stopped in order to save fuel and the system operates with 100% wind penetration. The stopped diesel engine is preheated in order to facilitate a fast startup and when needed it is automatically restarted. The system uses a standard diesel engine equipped with a Low Load Unit which reduces the fuel consumption under low load conditions and enables the engine to run at idle running for long periods with a minimum of soot formation. The specially developed secondary load circuit absorbs the excess energy from the diesel engine and from the wind turbines and makes it available for production purposes. This energy can be used for pumping water, heat

DISPOWER Del_2005_0079 - 2005-12-27

generation, cold generation, desalination water station… The complete wind/diesel power system achieves a high efficiency as a result of maximum wind exploitation combined with optimal utilization of waste energy. The entire system remains extremely stable because the cooling system absorbs the large fluctuations caused by the extreme variations in the available wind power, the power consumption and the grid. To further compensate for these constantly changing conditions the rotation of the engine is stabilized. This smoothes the operation of the control system and ensures that the engine can achieve a normal, long life.

The control of the grid voltage is maintained by the automatic voltage regulator of the generator. When the engine is disengaged and stopped the generator continues to run in order to supply reactive power and to control the voltage. The control of the grid frequency is maintained by the fast control of the power balance between the wind power, the power consumption and the grid. In periods where the diesel engine is in operation the frequency is controlled by the diesel engine governor. When wind penetration is 100%, the frequency is controlled by absorbing the surplus wind power in a dynamic dump-load.

Wind turbine power adjustment

The wind turbine speed-control solution allows for convenient integration in weak grids and wind-diesel systems. Diesel gensets can be shut down when there is sufficient wind. Standard diesel gensets can be used. Batteries are optional. The rotational kinetic energy of the wind turbine can be used to reduce the number of diesel genset starts (important for systems without batteries).

For example in December 2002, PitchWind Systems AB commissioned a wind-diesel system rated 30 kVA on Osmussaare. The wind-diesel system is the only power source on this isolated island in Estonia. The end user is the Estonian Border Guard and the project was carried out in cooperation with the local partner Empower EEE, who also supplied the tower of the wind turbine. PitchWind's wind turbines are well suited for wind-diesel applications, because of the variable speed operation and special control system. No expensive dumpload, rotary converter or custom-built diesel gensets are necessary with this concept, which was originally developed by Chalmers University of Technology.

On Osmussaare it was decided to include a battery bank in the system to increase fuel savings, because of the very high cost of fuel on the island.

DISPOWER Del_2005_0079 - 2005-12-27

7 Interconnection of solar powered mini-grids to the main grid on the island of Kythnos (CRES, Econnect, ISET)

7.1 Introduction The introduction of a substantial penetration from Renewable Energy Sources in the power mix of the Greek island grids is of great importance. This fact is justified because island grids are autonomous and are mainly electrified by diesel generators, something that dramatically increases the energy cost and the pollution in environmentally sensitive areas. Although the wind and solar potential in these areas is excellent, a number of technical issues pose a barrier. The implementation of a hybrid system with battery storage and an intelligent management system in Kythnos island has been proved to be an effective solution to this problem. However, today a significant part of the renewable energy sources and other power equipment is outdated and additionally it should be worthwhile to further increase the penetration of renewable resources in the island. In this context, this chapter summarizes the work done and the results obtained in Task 7.11 of DISPOWER WP7b, and more precisely it examines interconnection issues for a number of case studies involving the increased penetration of Renewable Energy Sources (RES) in the grid of this island.

These interconnection issues concern the interaction of RES with the grid. They cover the following issues:

• Voltage profile • Active and reactive power flows • Thermal (current) loading on circuit elements • Transient stability (maintenance of synchronism) • Voltage stability and reactive power control • Frequency control

7.2 Kythnos power system The design objectives for the Kythnos power supply system were [40] as follows.

• A large portion of renewable energy generation • Operation without diesel sets in times with low consumer load (diesel-off mode) • Fully automatic operation of the entire system including the diesel sets • Very stable voltage and frequency in all operating modes • Total remote control of the entire system

A schematic diagram of the Kythnos power system is shown in Figure 69. Five diesel generators, producing 400kW each, constitute the main power source of the island. These generators are equipped with electronic speed controllers and automatic load sharing devices for parallel operation with fixed frequency (isochronous mode). The renewable energy is provided by a small wind-farm (5x33kW), a 500 kW Vestas wind turbine and a 100 kWp photovoltaic system. It should be mentioned that the 5x33kW wind farm as well as the photovoltaic unit are outdated. The battery plant combined with a 12-pulse converter can be used to cover the power deficit if a diesel generator, or the large wind turbine, is switched off unexpectedly. In the diesel-off mode the following system is used for frequency control. The dump load unit is interfaced with the power system with another 12-pulse converter. Its main task is to damp power peaks from the wind turbines in order to stabilize voltage and frequency and avoid high loads on the battery storage. The phase shifter machine is used to control grid voltage in the diesel-off mode.

All of these units are connected via step-up transformers to a 15 kV medium voltage overhead grid. The annual average of load served from this grid is about 600 kW. However, strong seasonal fluctuations exist.

DISPOWER Del_2005_0079 - 2005-12-27

Figure 69 - Kythnos power system block diagram.

7.3 Examination of steady-state issues

7.3.1 Preliminary study As a first approach, ISET has carried out a preliminary simulation of a generic medium voltage grid of one of the Greek islands. The simulation was carried out with the software program ATP/EMTP. The objective was to study the grid behavior when several renewable resources, such as solar PV and wind, were integrated at various points. As a result, the voltage at a node increases after the PV injection under lightly loaded network conditions. However, with an injection of 10 to 25 kWp, the increase is well within the EN-50160 limit. Also, in the lightly loaded condition with fully compensated reactive power, the voltage increase at the node of wind generator is about 11% when the turbine is operating at the rated wind power (600 kW rated). This is slightly more than the design limit. The thermal capacity of the main conductors is well within limit. In fact, it is only loaded by about 20% of the limit in the heavy load network condition [36].

7.3.2 Kythnos island study Continuing the work at CRES, a similar approach has been used for the examination of the steady-state situation of the actual Kythnos island grid when a number of PV systems and wind turbines are connected to the island grid at various points. For the system analysis, a software tool developed in the Information Division of CRES has been used [37]. This tool is basically a GIS application that uses an interface with PSSE® software in order to carry out load-flow analysis. The scenarios examined, concerned a significantly larger penetration from renewable resources, compared to the previous approach. However, the results were very similar, regarding the voltage profile of the grid and the thermal loading of the circuit elements.

DISPOWER Del_2005_0079 - 2005-12-27

7.3.2.1 Study Cases As was mentioned earlier, the 5x33 kW wind farm and the 100kWp central photovoltaic system in the Kythnos power system are outdated. This fact in addition to the EU and Greek government incentives for private investments in the field of renewable energy sources were the motivation to examine two scenarios for the increase of the renewable energy penetration in the Kythnos island grid. Both scenarios do not take into account the above-mentioned renewable energy units. The first scenario involves the installation of 10 photovoltaic systems, distributed across the island grid. The maximum power that can be injected from each photovoltaic system is 130 kWp. The second scenario involves the installation of an 800 kW wind generator at the south end of the medium voltage grid. The location of the new wind generator is interesting because it is some distance from the central power station and as a consequence it constitutes a worst-case scenario regarding the impact of the generation on the voltage profile of the network. Additionally, this location is favorable from a wind potential point of view, as was identified in an inspection of the wind potential thematic map in the GIS application.

A diagram of the Kythnos medium voltage grid, with the location of the generation units and those considered by the two scenarios is shown in Figure 70.

Figure 70- Kythnos medium voltage grid and the location of power sources.

DISPOWER Del_2005_0079 - 2005-12-27

The issues examined for these scenarios are as follows. - The influence of the power injected from the renewable energy units on the grid voltage profile. - The influence of the power injected from the renewable energy units on the thermal loading of

the overhead conductors.

Distributed generation units, in general, affect the power flow and thus also the voltage profile in the distribution system where they are connected. In a radial network, which is the case of Kythnos, an increased distributed generation penetration and real power production first relieves the upstream parts of the network since the local distributed generation units supply the downstream loads. The reduced loading of parts of the network tends to flatten the voltage profile. If the real power production is higher than the demand of the downstream loads, the power will be exported up through the overlying network. This power flow will load the upstream network and will create a voltage ramp in the opposite direction than that in a feeder without distributed generation.

7.3.2.2 Summary of the simulation results For the first scenario, a load-flow analysis has been conducted for different levels of power injected from the photovoltaic stations. For a lightly loaded grid, the observed voltage rise is well within the limit posed by the EN-50160 standard, which is ±10%. Considering the heavy load condition for the grid, it was observed, that the voltage is below the nominal value, because the high load currents result in a voltage drop. Incremental increases in power injected from the photovoltaic stations result in an improvement of the voltage profile and the size of the voltage drop is reduced.

For the second scenario, the voltage increase at the wind generator bus is about 6% above nominal for a lightly loaded grid, at the rated wind generator power and fully compensated reactive power. Additionally, the calculated thermal loading of the lines is well within limits. In fact, for the heavy load condition, the maximum thermal loading is about 25% of the conductors thermal limit.

7.4 Examination of dynamic issues For the examination of the grid dynamic behavior, when a significant penetration from renewable resources exists, a model of the Kythnos power system was developed in the Simulink® environment, using the Power System Blockset ® library. This model utilizes a simplified representation of the grid that includes only the main nodes, a model of the local power station with the diesel generators and models for the renewable resources (wind turbine and PV systems) as well as for a battery storage inverter. In order to evaluate the system dynamic behavior, two types of disturbances were considered.

• A three-phase short-circuit, with different clearing times, that was applied at a specific node, away from the local power station.

• Disconnection of the wind turbine when it is producing maximum power. Such an incident can sometimes take place, due to the activation of the wind turbine protection system because of a high-wind condition.

More specifically, to investigate the effects of a considerable penetration of renewable DER units on the Kythnos grid transient behaviour, the following cases were examined.

a. Disconnection of distributed PV units, due to under-voltage protection trip during a three-phase fault and comparison of the system frequency response with the case where the PV inverters are quickly reconnected when the voltage recovers.

b. System behavior for a three-phase fault with a wind turbine and no PV.

c. Disconnection of a wind turbine due to high wind and comparison of the frequency response with or without battery storage.

7.4.1 Disconnection of PV distributed systems due to under-voltage protection Different penetration levels up to 33% have been studied. For the considered PV penetration scenarios and applied disturbance, simulation results show that the system exhibits stable behavior provided that sufficient spinning reserve exists. However, due to the high cost of the energy produced by the diesel

DISPOWER Del_2005_0079 - 2005-12-27

generators it is highly desirable to achieve a reduction of this reserve without sacrificing the safety of the power system operation.

A solution is related to the use of PV power electronic inverters with disturbance ride-through capability. Today these grid interfaces are disconnected quickly when a voltage disturbance takes place in order to avoid islanding situations and protect themselves. Reconnection takes place several minutes later. Disturbance ride-through capability means, among other things, that the grid interface should stay connected during the disturbance or it should be reconnected quickly after the disturbance [38]. The second option of rapid reconnection for disturbance ride-through has been simulated and the results indicated that, in this way, it should be possible to operate the power system with a lower spinning reserve without any safety risk [39].

7.4.2 System response with wind turbines For the examined wind penetration cases, the simulation results show that for the three-phase fault the wind turbine will not become unstable provided that the fault clearing time is shorter than the calculated critical clearing time.

For the wind turbine disconnection scenario, the simulation results show that for the considered conditions, the power system presents a stable behavior as long as enough spinning reserve is available to cover the power deficit.

In order to reduce the costs and to reduce the spinning reserve without sacrificing the safety of the power system operation, a solution based on the use of battery storage was studied to replace the spinning reserve provided by the diesel generators.

As it was mentioned earlier, battery storage already exists in the power system of Kythnos island (see Figure 69) but it is interfaced with the grid using a thyristor-based inverter. Today, taking into account the evolution of power electronics, the replacement of the existing inverter with a shelf-commutated one based on IGBTs (Insulated Gate Bipolar Transistors) would add a number of benefits related to more reliable behavior and the capability to assist the voltage control of the system.

A model of such a battery inverter was introduced in the Kythnos power system model. Simulation results show that, because of the fast response capability of the battery inverter, the frequency dynamic response of the power system is considerably improved, and in addition there is better fuel utilization for the diesel generators.

7.5 Conclusions The grid connection issues related to the increase of renewable sources penetration in the autonomous power system of Kythnos island have been examined. These issues covered steady-state as well as dynamic conditions caused by abnormal events such as short-circuits and loss of power sources.

For the examination of the island grid steady-state operation, a model of the entire distribution system has been developed with the help of a GIS application, created for such studies [37]. This model was used for a number of scenarios, varying the point of interconnection of renewable sources (photovoltaic systems and wind turbines) and the level of penetration. The GIS application is interfaced with PSSE to enable load flow studies. Utilizing this functionality, the load flow studies for the examined scenarios show that the voltage increase on the Kythnos grid, due to injection of active power at light load conditions, is well within the limits posed by the European standard EN50160. Under heavy load conditions the injection of active power from renewable sources improves the voltage profile. Additionally, the calculated thermal loading of the lines is well within limits. In fact, for the heavy load condition, the maximum thermal loading is about 25% of the conductors’ thermal limit.

For the examination of dynamic conditions, a model of the Kythnos power system in Simulink has been developed. This model includes a simplified representation of the distribution grid, taking into account only the main buses, and models for the diesel generators, renewable sources, battery storage and loads. Using this model, the frequency and voltage response of the Kythnos power system, for

DISPOWER Del_2005_0079 - 2005-12-27

different PV penetration levels and the connection of a wind turbine, has been studied for the following disturbances.

• A three-phase short-circuit away from the power station. • Disconnection of an operating wind turbine.

For the considered PV penetration scenarios and applied disturbance, simulation results show that the system exhibits stable behavior provided that sufficient spinning reserve exists. In addition, if the PV inverters are equipped with disturbance ride through capability allowing a rapid reconnection after voltage recovery, simulation results show that this feature could facilitate secure and stable operation of the Kythnos power system with a lower spinning reserve.

For the examined wind penetration cases, simulation results show that for the three-phase fault the wind turbine will not become unstable provided that the fault clearing time is shorter than the calculated critical clearing time. For the wind turbine disconnection scenario, the power system exhibits stable behavior provided that enough spinning reserve is available in order to cover the power deficit. For this disturbance, simulation results show that the frequency response of the system can be improved if a battery inverter is used in order to replace spinning reserve provided by the diesel generators. This is in addition to the fuel cost savings, the operation and maintenance cost savings, as well as the environmental benefits. This is very important, especially for the presented case of the wind turbine disconnection due to high wind, which can be a very common event in Greece due to gusting winds.

DISPOWER Del_2005_0079 - 2005-12-27

8 Conclusions In this report , the work done and the results obtained have been presented for 6 case studies carried out in Work Package 7b of the DISPOWER European project: - Increased wind energy penetration on weak networks in rural northern England (Task 7.7). - Diesel/hydro generation on the Scottish island of Rum (Task 7.8a). - Islanded operation of a wind farm at Blyth in the UK (Task 7.8b). - Application of MORE CARE control system on the islands of Crete and Kythnos (Task 7.9). - Increased penetration of renewable energies in the Guadeloupe power system (Task 7.10). - Interconnection of solar powered mini-grids to the main grid on the island of Kythnos (Task 7.11).

The case studies have investigated the problems and constraints met by DG and RES on weak grids and islands and have studied different solutions to overcome some of the existing barriers to a larger development of DG in such grids.

Mainly three types of case studies were carried out. The first type dealt with the use of load management and 3 applications were investigated in the UK. More specifically: - A load management technique was considered to mitigate voltage rise on a rural network in the

North East of England in order to facilitate connection of increased capacity of wind generation (Task 7.7). The results showed that load control is indeed an effective strategy. For the studied case, if 20% of the loads are controlled, the network could accommodate 900 kW of wind power instead of 300 kW without load control.

- The use of Distributed Intelligent Load Controllers (DILC) was investigated to control system frequency on the Scottish Island of Rum (Task 7.8a). The simulations showed that the approach was feasible. However, extensive site tests have shown larger voltage changes than anticipated and further developments of the DILCs are needed.

- A solution based on load control combined with a synchronous compensator was designed for the islanded operation of a wind farm (Task 7.8b). Testing was carried out on an islanded network composed of test equipment. The results demonstrated that voltage and frequency on the islanded network could be controlled within acceptable limits but wind turbine start up might be difficult.

The second type of studies focused on the impact of increased DG and RES penetration on islands, the identification of the resulting problems and constraints, and the analysis of possible solutions: - In the Les Saintes islands case study (Task 7.10), existing solutions to improve wind energy

penetration have been studied, using the new control and safety concepts developed in DISPOWER. The following aspects were considered: voltage rise and deviations, flicker, DG coupling/ decoupling, fault ride through capability, short-circuits. In particular, new design and control techniques were applied in order to decrease the out of limit voltage and the high flicker level on the grid and the impact of various wind turbine technologies was evaluated. The possible contribution of GIS tools was examined to devise a complete decision tool, interfaced with a Network Management System.

- For the Kythnos island (Task 7.11), grid connection issues for an increased penetration of RES were examined. Simulation results showed that the power system would present a stable behavior, as long as an adequate spinning reserve would exist. Due to the high cost of energy produced by diesel generators a reduction of this reserve is highly desirable without sacrificing the safety of the operation. Two solutions were then examined : the use of power electronic inverters with disturbance ride-through capability and the use of battery storage (to replace the spinning reserve provided by the diesel generators). Both solutions have proved to be feasible.

Finally, in the third type of studies, the MORE CARE advanced control software developed in previous European projects was applied on two Greek island power systems with high RES penetration, namely the islands of Crete and Kythnos (Task 7.9). The impact and the benefits of the economic scheduling, forecasting and on-line dynamic security assessment functions of MORE CARE were evaluated and compared to the actual operation of the system. The savings obtained in operation and fuel costs were estimated.

DISPOWER Del_2005_0079 - 2005-12-27

To summarize, the results of the case studies have shown that:

- Load management could be an effective strategy to increase DG and RES penetration in islands and weak grids.

- Depending on the problems and constraints met, different solutions can be applied to increase DG and RES on islands and weak grids. In particular, the use of energy storage systems, of power electronic interfaces (of the DG units) or of power electronic devices (for instance Statcom) may prove to be very effective. However, economic aspects also have to be taken into account and simpler but cheaper solutions are often chosen at the end even if they are not so optimal.

- In islands and weak grids, DG could prove profitable from the economic point of view.

- The increased integration of DG and RES in weak grids and particularly in island grids depends on several important factors such as: • The provision of ancillary services by DG and RES plants, in particular concerning

contribution to the control of network voltage and frequency. • The Fault-Ride-Through (FRT) capabilities of DG and RES units, that’s to say their

capability to withstand network disturbances such as voltage dips and frequency variations. • The use of appropriate monitoring and control systems with relevant communication means

for the DG and RES plants or for the network as a whole. • The availability of appropriate forecasting tools for RES power generation. • And last but not least the motivation and willingness of all the players involved not only at

the local or regional levels but also at the national level.

Very promising results have thus been obtained in the case studies. However further work is still needed for instance concerning:

- load management: load control strategies should still be improved and more thorough investigations on all the potentials and applications of load control should be carried out

- solutions to achieve a better integration of DG and RES in islands and weak grids: • technical and economic issues related to the use of energy storage systems should be further

investigated, • ancillary service provision by DG and RES may be critical on islands and weak grids;

appropriate and economic solutions are thus needed, • appropriate monitoring and control systems for DG and RES plants, as well as for the grid

operation should be further investigated and tested, • even if some DG technologies already have FRT capabilities, it is not the case for all of

them. In particular RES technologies intended for small islands do not generally have such capabilities and disconnect for voltages dips. Again, effective and economic solutions are needed.

DISPOWER Del_2005_0079 - 2005-12-27

9 Contact information

9.1 Tasks 7.7, 7.8a and 7.8b (Chapters 2, 3 and 4)

Econnect

S. White Econnect Limited Energy House 19 Haugh Lane Industrial. Estate, HEXHAM, Northumberland, NE46 3PU, UNITED KINGDOM Tel.: +44 (0) 1434 613600, Fax: +44 (0) 1434 609080 e-mail: Sara.White@econnect.co.uk

9.2 Task 7.9 (Chapter 5)

ICCS/NTUA

N. Hatziargyriou, A. Tsikalakis Institute of Communication and Computer Systems National Technical University of Athens, 9, Heroon Polytechniou str., 157 73 Zografou, ATHENS, GREECE Tel.: +30210-7723661, Fax: +30210-7723659 e-mail: nh@power.ece.ntua.gr , atsikal@corfu.power.ece.ntua.gr

Vergnet

D. Lefebvre AEROWATT 6 Rue Henri Dunant, 45140 INGRE, FRANCE Tel. : +33 (0)2 38 88 14 31 e-mail: d.lefebvre@aerowatt.fr

T. Degner Institut für Solare Energieversorgungstechnik e. V. Königstor 59, D-34119 KASSEL, GERMANY Tel.: +49 561 7294 232, Fax: +49 561 7294 200 e-mail: tdegner@iset.uni-kassel.de

F. Fesquet, R. Belhomme Electricité de France, S.A., EDF R&D , 1 Avenue du Général de Gaulle, 92141 CLAMART CEDEX, FRANCE Tel.: +33 147 65 54 94, +33 147 65 38 60, Fax : +33 147 65 32 18 e-mail : floriane.fesquet@edf.fr, regine.belhomme@edf.fr

DISPOWER Del_2005_0079 - 2005-12-27

ENSMP-CENERG

T. Ranchin Ecole Nationale Supérieure des Mines de Paris Group Teledetection & Modelisation BP 207, 06904 SOPHIA ANTIPOLIS CEDEX, FRANCE Tel: +33 493-957453, Fax: +33 493-957535 e-mail : thierry.ranchin@ensmp.fr

University Kassel

G. Arnold University Kassel Wilhelmshoeher Allee 71-73, D - 34121 KASSEL / GERMANY Tel.: +49 561 804 6512, Fax: +49 561 804 6521 e-mail: garnold@iset.uni-kassel.de

S. Tselepis and Aristomenis Neris Center for Renewable Energy Sources 19th km Marathonos Ave., Pikermi, 19009, ATHENS, GREECE Tel. +30 210 6603369, Fax +30 210 6603318 e-mail: stselep@cres.gr , mneris@cres.gr

Econnect

S. White Econnect Limited Energy House 19 Haugh Lane Industrial. Estate, HEXHAM, Northumberland, NE46 3PU, UNITED KINGDOM Tel.: +44 (0) 1434 613600, Fax: +44 (0) 1434 609080 e-mail: Sara.White@econnect.co.uk

T. Degner Institut für Solare Energieversorgungstechnik e. V. Königstor 59, D-34119 KASSEL, GERMANY Tel.: +49 561 7294 232, Fax: +49 561 7294 200 e-mail: tdegner@iset.uni-kassel.de

DISPOWER Del_2005_0079 - 2005-12-27

10 References [1] E. Navarro, A. Badelin, F. Schlögl, K. Rohrig, R. Mackensen, G. Arnold, M.H. Chowdhury, A.

Tamzarti, R. Belhomme, P. Bousseau, E. Gautier, A. Berenguer, I. Chocarro, C. Materazzi-Wagner, M. Haslinger, B. Trajanoska, W. Pospischil, “Deliverable 7.2. DG in European interconnected grids”, DISPOWER Deliverable D7.2, del_2005_0071, December 2005.

[2] R. Belhomme, P. Bousseau, E. Navarro, A. Badelin, T. Degner, G. Arnold, A. Berenguer, I. Chocarro, C. Materazzi-Wagner, S. White/Econnect, N. Hatziargyriou, S. Tselepis, A. Neris, D. Lefebvre, “Case Studies on the Integration of Renewable Energy Sources into Power Systems”, 10th Kasseler Symposium Energy Systems Technology 2005, Kassel, Germany, November 10-11, 2005.

[3] R. Belhomme, S. White, N. Hatziargyriou, A. Tsikalakis, D. Lefebvre, T. Ranchin, F. Fesquet, G. Arnold, S. Tselepis, A. Neris, T. Degner, P. Taylor, “Deliverable D7.3 - Distributed generation on European islands and weak grids”, DISPOWER Deliverable D7.3, del_2005_0078, December 2005.

[4] N. C. Scott, “Limitation Of Distributed System Voltage By Decentralised Load Control”, PhD Thesis, 2000.

[5] Northern Electric, “Statement of the basics of charges for use of Northern Electric’s Distribution System”, Northern Electric PLC, 1999.

[6] European Standard, EN50160, “Voltage Characteristics of electricity supplied by public distribution systems.” CENELEC, November 1994.

[7] Econnect Ltd, “Islanding Protection for Rotating Generation Plant Embedded in the Distribution System”, DISPOWER Work Package 2.2, 22 November 2004.

[8] ABB wQU 315 L4 DA generator data sheet. [9] Matlab/Simulink Power Systems Blockset / SimPowerSystems library, The Mathworks, Inc. [10] Newage International “Technical Reference Manual” Edition 4 / 2000, data sheet for HCI434C. [11] Matlab/Simulink Fuzzy Logic Toolbox library, The Mathworks, Inc. [12] N. Hatziargyriou, et al., “MORE CARE Overview”, Proceedings of the IEEE/IEE

MedPower2002 Conference, Athens, Greece, 4-6 November 2002. [13] NTUA, “MORE CARE” Final Report, April 2003. [14] Regulatory Authority of Greece, www.rae.gr . [15] N. Hatziargyriou, M. Papadopoulos, ”Developments and perspectives of wind energy in the

Aegean sea islands”, Dissemination of the advanced control technologies and SCADA systems for the isolated power networks with increased use of Renewable Energies Book of Proceedings Ajaccio-Corsica, 3rd of March 2000.

[16] PPC data kindly provided by Islands Department. [17] N. Hatziargyriou, A.Tsikalakis, A. Dimeas ,D. Georgiadis, J.Stefanakis, A.Gigantidou, E.

Thalassinakis , “Security and Economic Impacts of High Wind Power Penetration in Island Systems,” Presented in the 40th Cigre Session 2004 27th August 2004.

[18] N. Hatziargyriou, A. Tsikalakis and I. Tassiou “Impact of Energy Storage in the secure and economic operation of a small Greek island”, Presented at the Med Power Conference 15th-17th November 2004 in Cyprus.

[19] G.A. Vokas, A. Nikolopoulos ,”Financing And Implementing The Hybrid Wind - Hydroelectric Project Of Ikaria”,Proceedings of the 4th Med Power Conference MED02/122, Athens, November 2004,CH39/04.

[20] Renewable Energy on Small Islands. Second edition august 2000, Forum for Energy & Development.

DISPOWER Del_2005_0079 - 2005-12-27

[21] G.C. Bakos and M. Soursos, “Technical Feasibility and economic viability of a grid-connected PV installation for low cost electricity production.” Energy and Buildings J., vol. 34, pp. 753–758, July 2002.

[22] P. N. Korovesis, G. A. Vokas, I. F. Gonos, F. V. Topalis, ”Distortion of line voltage in weak low-voltage networks due to large scale installation of energy saving lamps”, Proceedings of the 3rd Med Power Conference MED02/122, Athens, November2002.

[23] C. Lainé, WINERGY, “Etude du potentiel éolien de la ferme éolienne de Les Saintes – Terre de Bas », 2003.

[24] F. Fesquet, EDF, “Data collection for Les Saintes case study”, DISPOWER tech_2003_0013 [25] D. Lefebvre, Vergnet SA, “Required wind and wind turbine data for the grid simulations in Les

Saintes case study”, DISPOWER. [26] E. Monnot, EDF, “Connection criteria to island distribution networks in Europe for distributed

generation”, DISPOWER tech_2005_0047. [27] RISOE, “Conceptual survey of Generators and Power Electronics for Wind Turbines”,

December 2001. [28] Université des Antilles et de la Guyane, “Wind velocity analysis for times scales smaller than 1

hour”. [29] P. Bousseau, E. Gautier, I. Garzulino, Ph. Juston. R. Belhomme, “Grid impact of different

technologies of wind turbine generator systems (WTGS)”, European Wind Energy Conference (EWEC), Technical Session “Grid Integration”, Madrid, Spain, June 16-19, 2003.

[30] T. Ranchin, F.-P. Neirac, M. Vandenbergh, ARMINES, “Survey of the different GIS, standardised data structures and databases”, DISPOWER, June 2003.

[31] T. Ranchin, F.-P. Neirac, M. Vandenbergh ,ARMINES, “Present GIS capabilities for Distributed Generation”, DISPOWER, June 2003.

[32] T. Ranchin, ARMINES, “Recommendations for the use of GIS as a planning tool for Distributed Generation”, DISPOWER, August 2003.

[33] University of Strathclyde, Technical University of Lodz, “Report describing and comparing opportunities and effectiveness of generator control, load management, and additional storage and power electronic controller options for managing LV networks with distributed generation”, DISPOWER, January 2004.

[34] D. Lefebvre, Vergnet SA, “Advanced pitch control to improve the quality of supply of a wind turbine coupled on weak grids”, 2005.

[35] T. Ranchin, ARMINES, F.-P. Neirac, M. Vandenbergh “Potential uses and applicability of Geographical Information System for Distributed Generation planning in the case study of Les Saintes”, DISPOWER tech_2005_0044.

[36] T. Degner, P. Taylor, D. Rollinson, A. Neris, S. Tselepis, “Interconnection of solar powered mini-grids - a case study Kythnos island, Greece”. Visual presentation 7BV 3.40 at the 19th European Photovoltaic Solar Energy Conference and Exhibition, Paris, 7.–11, June 2004.

[37] M. Psalidas, D. Agoris, E. Pyrgioti, V. Kilias, P. Stratis, K. Tigas :” Geographic Information System for the Digitization and Management of Electrical Networks”, Proceedings of the 38th UPEC 2003, Thessaloniki, Greece.

[38] J.G. Slootweg, “Wind power modelling and impact on power system dynamics”, PhD Thesis, 2003, Technical University of Delft, http://eps.et.tudelft.nl.

[39] S. Tselepis, A. Neris, “Dynamic behaviour of the autonomous grid of the island of Kythnos, Greece, due to large penetration of PV and wind systems”. Visual presentation 6DV.4.13, Proccedings of the 20th European Photovoltaic Solar Energy Conference and Exhibition, Barcelona, Spain, 6-10 June 2005.

[40] Ph. Strauss, W. Kleinkauf, J. Reekers, G. Gramer, G. Betzios, “Kythnos island- 19 years experience of renewable energy integration”, International Conference “Renewable energies for islands, towards 100% RES supply”, Chania-Crete, Greece, 14-16 July, 2001.

Deliverable D7.3 - Distributed generation on European ...s3.amazonaws.com/zanran_storage/ · - In...

Documents

LAUTEL DE SAINTES

WP7 D7.3 Exploitation plan - initial version

e Festival de Saintes - Abbaye aux Dames

Douze Nuits Saintes Et Les Hierarchies Spirituelles

Saints Et Saintes de Dieu (Tome 1)

D7.3 CombCkt Verilog-2

Leçon 2 : Chapitre 1 : Les Saintes Écritures

Deliverable D7.3 - Policy report on open peer review ...€¦ · Deliverable D7.3 - Policy report on open peer review, impact measurement and novel dissemination practices OPENingUP

Saintes - France - europan10 RECONQUÉRIR UN BELVÉDÈRE

Les Saintes Ecritures

La vie des saintes presentation du projet

Mercredi 25 septembre 2019...TRANSPORTS REGIONAUX COMITÉ DE LIGNES Royan–Saintes- Angoulême Royan –Saintes-Niort COMPTE-RENDU Mercredi 25 septembre 2019 Saintes PRÉSIDÉ PAR

D7.3 Communication Plan - INBOTSinbots.eu/wp-content/uploads/2018/08/deliverables... · D7.3 Communication Plan This project has received funding from the European Union’s Horizon

D7.3 HEIMDALL End-Users’ manual

LES FEMMES SAINTES · TABLE DES MATIERES LES FEMMES SAINTES James W. Knox Guide Pour Femmes Chrétiennes

D7.3 Waternomics Water Billwaternomics.eu/wp-content/uploads/D7.3-Waternomics-Water... · 2017-05-02 · Waternomics is to explore how Information Technology can help households,

COMMUNAUTE D'AGGLOMÉRATION Saintes visioconférence, …

Les Saintes : Extrait de « Nos paroisses »

Deliverable D7.3 Mid-term Cross4Health Communication Report · Deliverable D7.3 Mid-term Cross4Health Communication Report Dissemination Level: PU Deliverable Type: R Date: 04 June

D7.3 Application Final Results - CORDIS · D7.3 / Application Final Results 3 Executive Summary The project TrendMiner1 aims at developing a social media monitoring platform for analysing