Smart Houses in the Smart Grid: Developing an interactive network

2325-5987/14/$31.00©2014IEEE 2325-5987/14/$31.00©2014IEEEIEEE Electr i f icat ion Magazine / MARCH 2014 IEEE Electr i f icat ion Magazine / MARCH 2014PB 81

Ima

ge

co

ur

te

sy

of

sto

ck

.xc

hn

g/s

vIl

en

001.

rivate households constitute a considerable share of europe’s electricity consumption. the current electricity distribution sys-tem treats them as effectively passive individual units. in the future, how-ever, users of the electricity grid will be involved more actively in the grid operation and can become part of intelligent networked collaborations.

they can then contribute the demand and supply flexibility that they dispose of and, as a result, help to better integrate renewable energy in-feed into the distribution grids.

to achieve energy efficiency and sustainability, a novel smart grid information and commu-nication technologies (icts) architecture based on smart houses intelligently interacting with

Digital Object Identifier 10.1109/MELE.2013.2297032 Date of publication: 18 March 2014

By Aris Dimeas, Stefan Drenkard, Nikos Hatziargyriou, Stamatis Karnouskos, Koen Kok, Jan Ringelstein, and Anke Weidlich

Smart Houses in the Smart Grid

Developing an interactive

network.

P

IEEE Electr i f icat ion Magazine / MARCH 201482

smart grids is needed. along these lines, the european com-mission cofunded the research project smarthouse/smart-Grid (sh/sG) (see www.smarthouse-smartgrid.eu), a consortium of leading parties in ict for energy, has adopted an innovative approach. the ict architecture developed by the consortium introduces a holistic concept and technology for smart houses as they are situated and intelligently man-aged within their broader environment (see figure 1). the approach treats smart homes and buildings as proactive cus-tomers (prosumers) that negotiate and collaborate as an intelligent network in close interaction with their external environment. context is king here: the smart home and building environment includes a diversity of other units: neighboring local energy consumers (other smart houses), the local energy grid, associated available power and service trading markets, as well as local producers [local environ-mentally friendly energy resources such as solar and (micro) combined heat and power (chP)].

the architecture is based on a mixture of innovations from recent r&d projects in the forefront of european smart grids research. these innovations include:

xx in-house energy management based on user feed-back, real-time tariffs, intelligent control of appliances, and provision of (technical and commercial) services to grid operators and energy suppliers

xx aggregation software architecture based on agent technology for service delivery by clusters of smart houses to wholesale market parties and grid operatorsxx usage of service-oriented architecture (soa) and strong bidirectional coupling with the enterprise sys-tems for system-level coordination goals and han-dling of real-time tariff metering data, etc.

the project capitalizes on several smart grid concepts developed at different research institutes:

xx the bidirectional energy management interface (beMi) was developed at the fraunhofer institute for Wind energy and energy system technology (iWes), Germany.xx the multiagent intelligent control (MaGic) system was developed at the Power system laboratory of the national technical university of athens, Greece.xx PowerMatcher was developed at the energy research center of the netherlands (ecn). [the PowerMatcher team at ecn was recently taken over by the nether-lands organization for applied scientific research (tno).]

these three technologies were further developed within the project, and synergies between the approach-es were identified. they all share one control paradigm, which can be summarized as:

Transaction Platform

MarketplaceAuctions

Optimization Service

Buy and Sell

Legacy Providers

BusinessIntelligence

AlternativeEnergy

Providers

Smart Meters

Future Service-BasedEnergy Infrastructure

Internet

Internet

Internet Internet

Home AppliancesManagement

Figure 1. The service-based ecosystem based on smart houses and smart grids.

IEEE Electr i f icat ion Magazine / MARCH 2014 83

TABLE 1. SH/SG Business Cases.

Number Name Brief Description

1 Aggregation of houses as intelligently net-worked collaborations

When smart houses are able to communicate, interact, and negotiate with both customers and energy devices in the local grid, the electricity system can be operated more efficiently because consumption can be better adapted to the available energy supply, even when the proportion of variable renewable generation is high. A commer-cial aggregator could exercise the task of jointly coordinating the energy use of the smart houses or commercial consumers that have a contract with him (either via direct control of one or several participating devices or through providing incentives to the participating devices so that they will behave in the desired way with a high probability but not with certainty).

2 Real-time imbalance reduction of a retail portfolio

This business case is rooted in the balancing mechanism as applied in Europe and defined by the European Network of Transmission System Operators for Electric-ity (ENTSO-E) Scheduling System. It focuses on the balancing actions by a balance responsible party (BRP) during the balancing settlement period. The key idea of this business case is the utilization of real-time flexibility of end-user customers to balance the BRP portfolio instead of using traditional power plants. For each control zone, the BRP aggregates all of its contracted flexible distributed generation (DG) and responsive loads in a virtual power plant (VPP). The BRP uses the VPP for its real-time balancing actions.

3 Offering (secondary) reserve capacity to the transmission system operator (TSO)

This business case is rooted in the ancillary services as initiated by TSOs throughout the world. In this business case, the BRP should be able to offer its flexible demand and supply on the reserve market. To enable BRPs to offer flexible demand and supply on the reserve market, their bids have to fit into the above market structure. The key idea of this business case is the utilization of real-time flexibility of end users (prosumers) in balancing a control zone. For each control zone, market parties aggregate these flexible DG and responsive loads in a VPP. The TSO contracts in real time part of these flexible loads for its real-time balancing actions.

4 Distribution system con-gestion management

This business case aims at deferral of grid reinforcements and enhancement of network utilization. The need clearly arises in areas with a large amount of DG near one location. Noncoordinated control of (new) electric devices (e.g., heat pumps and electric cars) may lead to a sharp rise in needed capacity on lines and transform-ers. By coordination of these devices, there can be allocated timeslots for operation that are spread out over time. Furthermore, coordination can increase the simul-taneousness of local supply and demand in case local generation is integrated. Congestion management as a service can be used to better match own generation and consumption for prosumers; also, distribution system operators (DSOs) may be interested in improving the quality of supply in areas with restricted capacity in lines and transformers.

5 Variable tariff-based load and generation shifting

In well-functioning and liquid markets, the expectations of all market participants about the generation and consumption situation of the next day are well reflected in day-ahead power exchange prices. If these wholesale prices are passed over to the end users, these have an incentive to shift loads from high-price times to times of lower prices. The key idea of the business case is, thus, to provide the customer with a variable price profile on the day before power delivery. At the customer’s premise, an energy management system should receive the price signal and determine the optimal timing for the energy consumption (or generation, for prosumers) of those appliances that can be shifted in time or that have a storage characteristic. The main value driver from the customers’ perspectives is to receive a tariff and a technology that reduces their energy bills. The value driver from the retailers’ perspectives is the opportunity to reduce his procurement.

6 Energy usage monitoring and optimization ser-vices for end consumers

Awareness of one’s energy use can stimulate behavioral changes toward energy savings. Personalized and well-targeted advice on how to save energy can help further exploit the savings potential. This business case, therefore, suggests providing customers with detailed and comprehensible information about their own energy con-sumption. The additional value to the customer provided by the described information services can either be remunerated through additional fees or through enhanced customer loyalty. A combination of both is also conceivable.

(Continued)


TABLE 1. SH/SG Business Cases (Continued).

Number Name Brief Description

7 Distribution grid cell islanding in case of higher system instability

The key idea of this business case is to allow the operation of a grid cell in island mode in case of higher system instability in a market environment. This business case considers that the islanding procedure is performed automatically. Technically, it involves monitoring and forecasting the available DG and loads and creating a load shedding schedule based on to the criticality of the consumption loads and on the customer’s willingness to pay for running the appliance during island mode. During an event, decisions are taken how many and which loads must be shed to maintain island mode steadily. Grid cell islanding is of value to the DSO. Islanding helps him to quickly restore system stability within his grid area.

8 Black-start support from smart houses

The key idea of this business case is to support the black-start operation of the main grid. It considers that after a blackout, the local grid is also out of operation and the main goal is to start up quickly in island mode and then to reconnect with the upstream network to provide energy to the system. Black-start support is of value both to the DSO and the consumer. Flexible demand helps the DSO restore system stability.

9 Integration of forecast-ing techniques and tools for convenient participa-tion in a common energy market platform

The volatility of the production level of distributed energy resources (DERs) makes forecasting a necessary tool for market participation. The actor with the lowest fore-casting error will have the most efficient market participation. This business case pro-vides benefit for both the consumer and the aggregator. The aggregator has the ability to participate accurately in the wholesale market and gain by reducing the uncertain-ties. The consumer benefits from lower prices. However, it requires the participation of the consumers since an accurate forecast requires online monitoring of the DERs and not simply reading from the smart meter. The business case comprises the data collection, which is the most critical part that may lead to a correct forecast. The second part is the data evaluation and processing, e.g., for extracting a wind power prediction valid for a certain region.

TABLE 2. An Overview of the SH/SG Technologies.

PowerMatcher BEMI MAGIC

Basic concepts

■ Decentralized decision making about consumption and production

■ Decision making based on central-ized market equilibrium of all bids

■ Real-time mapping of demand and supply

■ Automated control of production and consumption units

■ Scalable architecture


■ Decision making based on central-ized tariff decision

■ Mapping demand to available supply■ Automated control of consumption

units■ User information for manual control

of consumption behavior


■ Decision making based on central-ized negotiation of requests

■ Mapping of demand and supply■ Automated control of production and

consumption units

Methodology

■ Market-based concept for demand and supply management

■ General equilibrium theory■ Market is distributed in a tree

structure■ Participants: devices, concentrators,

objective agents, and auctioneer■ Device agents submit bids/demand

and supply functions■ Auctioneer determines prices■ Round-based marketplace

■ BEMI enables decentralized decisions based on tariff information

■ Decision consists of local information about devices and central informa-tion about variable prices

■ Pool-BEMI sends price profiles■ “Avalanching” can be avoided by giv-

ing different price profiles to different customer groups

■ Day-ahead announcement of price profiles

■ Multi-agent system (MAS) based using Java Agent Development Frame-work (JADE) (negotiation based)

■ Grid announces selling price/buying price

■ Microgrid tries to agree on “better” prices

■ Maximum of internal benefit■ Auction algorithm such as the sym-

metric assignment problem■ Agents also may use reinforcement

learning (Adapted MAS Q-Learning)■ Number of involved agents differs

with the action to take


the concept of the project is to combine central-ized and decentralized control approaches with the following philosophy: let the end- customer decide as much as possible within his or her private grid. therefore, offer the end customer the online tools with appropriate boundary conditions and incen-tives to optimize his or her energy interface to the outside world according to actual (dynamic) prices and energy efficiency considerations that reflect the real-time needs of the public grid. to this end, pro-vide centralized information but allow for decen-tralized decisions.

Business Casesthe technological developments in the sh/sG project have been based on nine business cases that describe how smart grid approaches could be applied by single stakeholders in the electricity supply business. as shown in figure 1, not all business cases are applicable to all stakeholders, but each stakeholder can apply more than one business case. table 1 summarizes the nine business cases.

The Overall Architecturethe sh/sG architecture has to account for the heterogene-ity of concepts adapted and tested within the project. one major overarching paradigm that has to be reflected is the distributed control paradigm. following this, there needs to be some distributed decision making at the house level, which is facilitated through an appropriate in-house archi-tecture in combination with global coordination. the latter, in turn, facilitates a business case of some involved enter-prise. table 2 summarizes the main characteristics of the three technologies employed in the project: PowerMatcher, beMi, and MaGic.

there are some important commonalities between these technologies. as already depicted in table 2, it can be recog-nized that the common idea of the sh/sG implementation follows a unified approach: PowerMatcher, beMi, and MaGic manage demand and supply on the basis of a centralized optimization tool that works with decentralized decision making. this is highly important for the acceptability of these technologies since each participant keeps full control over his devices but has incentives to align the device opera-tion with the global status of the overall system.

each of the three technologies is based on the concept of mapping the demand to the producible or produced energy. it is possible to adjust the amount of energy to be consumed by deploying features like automatically switching on and off consuming devices or indirectly influencing the consumer’s behavior via price incentives. these features are part of all three trials (which are based on three different technology approaches), and the auto-mated switching of the controllable devices in the house-holds plays a significant part. the control of the shiftable production of energy is in a similar way possible by means

of automated on and off switching features for chP pro-ducers, as an example.

each of the concepts includes a centralized negotiation or calculation mechanism that tries to map the producible energy to the consumable energy for all actors (smart houses and production sites) within the smart grid. exter-nal production sites producing and providing a certain amount of energy can be included in the negotiation pro-cess as a fixed and uncontrollable amount of energy. therefore, the architecture of all three setups contains a central coordination mechanism.

the way the three coordination mechanisms are designed is similar from a high-level perspective, but differs in the details. each tool either collects information or fore-casts the desired amount of energy to be consumed or produced from all participating smart houses and produc-tion sites. each tool is able to understand not only the desired energy amounts but also some indicators about the conditions energy will be consumed or produced, namely, price incentives to shift demand. based on all offers and requests, the tool analyses how the equilibrium can be reached under the given conditions.

one major difference between the negotiation proce-dures is the time cycle of the negotiations and, therefore, the consideration of unforeseeable changes. PowerMatcher and the MaGic system work in (near) real time. the advan-tage is that for unforeseeable demand or production requests, a short reaction time can be expected to map the complementary production or demand requests. the beMi technology, in contrast, works on a time scale of a day, where day-ahead production and consumption patterns are considered to define the price levels that are used as decision-guiding signals.

the field trials described in next few sections aim to investigate the appropriate time scale of equilibrium cal-culations. the near real-time negotiation demands a high degree of scalability and performance requirements. the PowerMatcher tool performs real-time negotiation using a multilevel approach realized by the use of agents, cluster-ing several smart houses or concentrator levels stepwise. for a small number of smart houses, the concept of real time could scale easily, but for a higher number of smart houses, the concept has yet to be proven.

decentralized decisions about consumption and pro-duction decisions are decentralized, i.e., the control of switching on or off of a certain producing or consuming device is always done within the smart house itself. even when for the smart houses a central control is established, the decision remains within the house. of course, the decision is guided by a centrally determined and provided signal (e.g., virtual price signal or a real-time tariff/price structure or direct control signals).

because of the difference between the technologies employed, sh/sG does not have a common architecture in the classical notion, but an amalgamation of heteroge-neous approaches that are glued together by an soa, as


shown in figure 3. this is compatible with the future smart grid vision as it is not expected that a single archi-tecture will prevail but several heterogeneous approaches will be applied. all of them will exchange information at a higher level via common standardized approaches, such as those enabled by Web services.

Field Trials

Win-Win Situations: Trial in Manheim

Trial Setup and Objectivesthe main objective of the trial in Manheim is to test the automated response of household devices and the cus-tomers’ behavior on variable electricity prices within a real utility environment. Participants are located in Mannheim’s ecologically oriented suburbs of Wallstadt and feudenheim with a large number of customers act-ing as consumers and photovoltaic (Pv)-based electricity generators, i.e., prosumers. Part of the solution proposed by the field trial in Manheim is to integrate a demand-side energy management system as an active part of the grid by offering incentives to use electric devices at speci-fied times. this energy management system should be flexible so that it can be extended by control functions for electricity generators, e.g., micro-chP in combination with thermal storage devices to allow electricity genera-tion when needed, while also supplying heat or warm water when requested. therefore, decentralised energy management in households for balancing local consump-tion and local supply of electricity is investigated in 100 households.

the core of the energy management system used is a newly developed device called the “energy butler.” together with a broadband power line modem and peripheral additional modules (e.g., smart meter, data storage/data aggregator, and switchboxes), the energy butler serves to optimize selected household appliances

based on a tariff profile received one day ahead via a broadband power line used for communication with the dso. automatically switched devices include dishwashers, wash-ing machines, refrigerators, freezers, and clothes dryers (with a maximum of two devices switched per custom-er). in addition, other devices could be started or stopped by the custom-er any time according to price incen-tives by a simple click on the Web interface. consumer interaction with the energy butler and the communi-cation among all elements of the smart house is presented in figure 4.

the energy management system tested in this field trial uses two main

units together with supporting additional equipment:1) a smart meter to measure the electricity consumption

digitally2) the energy butler together with switchboxes to start/stop

connected household devices and primary equipment.in addition to these units, the energy management sys-

tem comprises supporting technical infrastructure involv-ing many rather complex processes and equipment within a single household. to keep high customer satisfaction and considering interaction between various stakeholders—between the customer and the local utility at Mannheim, Germany, Mannheimer versorgungs- und verkehrsgesell-schaft energie aG (Mvv) energie between Mvv and the equipment developer, hardware manufacturer and other third parties involved—a failure management system was introduced. the field trial installations were conducted in several phases, starting with a small group of households in the first phase and with further replication to a larger num-ber of the customers in the subsequent phase. the two main phases were:

xx elaboration and installation of the smart meters according to suitability and price and their testing afterwardsxx testing and checking the functioning of the newly developed hardware and software for the energy butler energy management system, including briefing, train-ing, customer care, and installation teams, as well as implementation of a new billing procedure.

four preferred tariffs were introduced for weekends and working days with two different profiles for each of them. each tariff profile used for the field trial had two price levels: low and high tariff with a minimum duration of one hour. the tariff profiles were static, which means that the same sequence (high and low tariff) applied for every working day and every weekend day and holiday, respectively.

the field trial measurements were implemented in three phases adding step-by-step tariff incentives and

WholesaleMarket

EnergyRetailer

DGOperator

Consumer/Prosumer

DSO

TSOLarge Power

Producer

Energy Trade

BalancingEnergy

CommoditySubsystem

TechnicalSubsystem

PhysicalEnergy Flow

BC 2, 3, 4BC 1, 5, 9

BC 6, 7, 8

Figure 2. Mapping the business cases to the market participants.


energy butler devices from october 2010 to august 2011, with further ongoing data acquisition and documentation:

xx the first phase with new energy management/vari-able tariffs started in october 2010. the customers received a first variable tariff on a monthly basis and access to the Mvv metering portal. the custom-ers could shift their loads according to this variable tariff manually.xx the second phase started in december 2010. since the energy butler was still not available to all customers, a new tariff model was introduced to get a better statis-tical basis for the evaluation of the load shift potential. in parallel, the energy butler was installed and tested, first at up to ten selected friendly customers followed by the installation of the other 90 energy butler devic-es with all peripheral devices (e.g., switchboxes, gate-way/modems, and temperature sensors) starting in spring 2011. this included constant software improve-ments and installation of the most recent software updates on the energy butler via remote access.

xx the third phase started in May 2011 and lasted until august 2011 when activating the energy butler soft-ware allowing the customer to use it not only for manual but also for automated load shift.

Key Findings and Lessons LearnedQuantitative assessment of the load shift potential requires comparison between the load curve when variable tariffs (electricity prices) are applied and the normal reference case with fixed tariffs. since it is impossible to measure concurrently the load curve with the two options for any customer, it was decided to use as the reference case modi-fied standard load profiles considering seasonal effects.

a representative load shift together with the applied, as obtained for february 2011, is shown in figure 5. there are several effects that need to be considered with respect to the load shift. besides the seasonal and price effects, changes in consumer awareness, saturation, or habitual effects can be important when tariffs are constant for sev-eral months.

Centralized

SendingSignals/Prices,ReceivingPreferences/Bids

Decentralized

ReceivingSignals/Prices,DeliveringPreferences/Bids

External Services Smart Retailer/Service Provider

Meter DataAnalysis

Smart Grid ProductDevelopment

HEM SignalOptimization

Energy Feedbackfor Customers

Rating andBilling

Aggregation Level

Smart HouseHEM UserInterface

ConsumptionMonitor Interface

ConcentratorHome Gateway

Devices and DeviceControl

Smart Meter

HEMConcentrator

Metering DataConcentrator

DSLPLC, BPLCGSMIP

ZigBeePLC,Z-WaveIP

LoadForecasting

RES GenerationForecasting

PriceForecasting

RES: Renewable Energy Sources

HEM: Home Energy Management

GSM: Global System for Mobile Communications

PLC: Power Line Communication

BPLC: Broadband Power Line Communication

Figure 3. The architecture of loose coupling via services.


customer feedback showed that during the test, the majority of the participants used devices at times conve-nient for them. the most commonly used flexible appli-ances were dishwashers, washing machines, and clothes dryers. two thirds of the customers indicated that they changed their electricity consumption behavior during the field test. the participants also indicated that they had adapted their electricity consumption to the tariffs and it was estimated that they could save about €5 per month compared to flat tariffs. Participants also reported that they were motivated to reduce consumption by acknowl-edgment of their contribution to climate protection. in summary, two conclusions can be drawn, as assessed by the participating consumers: 1) power consumption was reduced in absolute terms and 2) the power consumption of white goods (e.g., household devices) has been shifted to off-peak hours.

finally, a unique feature of the field trial in Manheim is that it involves many profit-oriented stakeholders. an important lesson learned is that the emergence of these stakeholders, who were once single vertically integrated utilities, greatly increased mediation issues when imple-menting smart grids. Within the project context, this is resolved by project management and common consor-tium decisions. however, if considering implementation of smart grids in today’s unbundled market, such a

consortium would have to be replaced by a trusted cross-partner organization for mediation between con-flicting partner interests, which otherwise would hinder technical implementation. for example, giving the cus-tomer a tariff bonus to motivate a load shift for better utilization of the dso’s grid resources may be an advan-tage for the dso, but it is unattractive for the energy pro-vider as of today. this is due to the introduction of variable tariffs that raise costs for establishing new bill-ing procedures but do not yield direct income. however, without technical services provided by the dso and energy provider, the smart grid cannot prevail, which is a disadvantage for both.

Support in Critical Grid Operation: Trial in Meltemi

Trial Setup and Objectivesthe aim of the field trial in Meltemi is to demonstrate the ability of a decentralized system to handle critical sit-uations, such as the transition to island mode or black start. furthermore, it aims to demonstrate the capability of decentralized resources to provide ancillary services, i.e., load shedding to alleviate network congestions. Melt-emi offers seaside camping near the athens coast, con-sisting of 170 cottages used mostly for summer holidays. because of the small size of each cottage, its electrical

Display for Monitoringand Control

Energy Butler

Energy ManagementNetwork

Energy Management GatewayRouter

Multitariff Metering

Micro-CHP

Meter Gateway

Intelligent Devicesand Systems

www

Figure 4. The smart house as envisioned in the Mannheim trial.


consumption is lower than an ordinary house in Greece. however, the ecological awareness of its habitants and the electrical structure [all houses are connected to the same medium voltage/low voltage (Mv/lv) transformer] of the settlement make it ideal for use as a test bed for functions related to emergency and critical grid situa-tions. the installation of distributed generation (dGs), including a 40-kva diesel generator, 4.5-kW Pv panels, and small residential wind turbines, can partially sup-port the Meltemi campground in a microgrid operation.

the MaGic system installed allows the dGs and the household to negotiate to decide the next sequence of actions. this system is a Java-based software that imple-ments intelligent agents. a critical component of the MaGic system is the intelligent load controller (figure 6) based on an embedded processor that runs linux and can be used to monitor the status of a power line providing voltage, current, and frequency measurements. the con-troller is expandable with several serial ports and a uni-versal serial bus port and has the ability not only to control but also to monitor several appliances. it is designed for indoor installation and is equipped with a display to present messages directly to the consumer. in addition, the consumers are informed online about the status of the system as well as their consumption and energy costs. this information is also available through a Web portal. this information is critical since consumers accept them only when they visualize the potential for energy savings cost benefits.

Measurements and Resultscongestion management is based on the monitoring of the transformer that feeds the camping site. the idea is that when the transformer is close to its critical level of overloading, the dso requests that the aggregator proceed with load shedding.

the MaGic system deals with this request in two steps. during the first step, the agents monitor the system and provide to the other agents and the aggregator infor-mation about its status, i.e., production, consumption, and voltage measurements. the system provides a list of loads that can be shed, as well as the units that are capable of providing extra energy.

the second step concerns the time period just before the transformer overload. the system predicts the over-load using its forecasting functions and proceeds with intelligent load shedding. advanced load-shedding algo-rithms are used ensuring that loads of the consumers are curtailed in a fair way.

to evaluate this function, a simple algorithm that simply reacts to the various disturbances was consid-ered. the difference between the two approaches is shown in figure 7. during the disturbance, the simple algorithm reacts when the system detects the overload. this stimulates load shedding at the Mv substation. on the other hand, the MaGic system has the ability to pre-dict the incoming event and act faster. it has been dis-covered that 15 min early warning is enough time to cope with the disturbance.

Electric Load Shift Within the Field Trial

Ele

ctric

ity P

rice

(cts

/kW

h)

1,000

900

800

700

600

500

Ele

ctric

ity C

onsu

mpt

ion

(W)

400

300

200

26

24

22

20

18

16

14

12

10

00:0

0

01:0

0

02:0

0

03:0

0

04:0

0

05:0

0

06:0

0

07:0

0

08:0

0

09:0

0

10:0

0

11:0

0

12:0

0

13:0

0

14:0

0

15:0

0

16:0

0

17:0

0

18:0

0

19:0

0

20:0

0

21:0

0

22:0

0

23:0

0

Day Time (hh:mm)

Real Load MeasuredActual Electricity Tariff

Predicted Load w/o Bonus Tariff

Figure 5. The load shift in the field test achieved by variable tariffs (February 2011).


the islanding/black start support scenario also has two main steps: the first step takes place before the event that may occur, and the second step is the steady island-ed operation. during the first step, the system monitors both the available dG units and the loads, and forecasts the consumption and the available power and energy in the next hours. a load-shedding schedule is derived based on the criticality of the loads expressed by the cus-tomers’ willingness to pay for their service during the island mode period.

in the first few minutes after the event, the dso (simu-lated agent) allows operation of critical loads depending also on the local power availability. When balance and sta-bility have been ensured, MaGic assumes on its own the energy management, i.e., generation and consumption, within the islanded network. the transition to the island mode is done automatically without interference from the end users or the aggregator.

it should be noted that if the disturbance happens dur-ing the middle of the day, there is sufficient energy and

critical loads can be supplied by the Pvs. the MaGic sys-tem interferes only when the consumption increases sig-nificantly. for example, on 2 august, a cloud reduced the Pv production, as shown in figure 8. if this event hap-pened during the islanded mode, the MaGic system would limit the consumption of the controllable devices. the general conclusion is that effective monitoring (knowledge) and device control improves quality of ser-vice to the consumers.

Key Findings and Lessons Learned from the Trial in Meltemithe primary lesson learned from the Meltemi trial is the importance of enhanced soa capabilities. large-scale implementation requires cooperation with enterprise information systems and, consequently, an adaptation of the negotiation algorithms. another important issue is the importance of the legal framework and the level of market deregulation. all of the scenarios implemented in the field trial assume the implementation of flexible tariffs and the ability of the aggregator/energy service company (esco) to make arrangements with the consumers/ dG owners.

finally, the facilities provided by the intelligent load controller are crucial. the availability of an indoor display is significant since it allows the inhabitants to actively participate in shaping their energy profiles. communica-tion with the users has revealed that information about level of consumption and costs increases their awareness regarding energy savings and, as a consequence, solutions like the one proposed by sh/sG project are easier accept-able. the existence of a Web portal is also significant, although it was not widely used in the Meltemi trial because of the limited access to internet by the residents of the holiday campground.

Mass Scalability: Trial in Hoogkerkthe field trial in hoogkerk aims to demonstrate the mass-scale perspective of automated aggregated control of end-user systems for energy efficiency, combined with testing the information exchange with enterprise systems using data traffic at mass-application strengths. the scale of mass-application is set at 1 million households.

Detection Thatan IncidentHappened

Detection Thatan Incident Is

Ready to Happen

Action Action

Overload Overload

LoadIncrease

(a) (b)

Figure 6. The MAGIC load controller. (Photo courtesy of NTUA.)

Figure 7. The operation of the (a) simple versus (b) advanced algorithm.


one subgoal of the trial was to demonstrate that the performance of automated control of 1 million house-holds is adequate with respect to the business case on real-time Portfolio imbalance reduction. this business gives a brP the ability to control the flexible demand and supply of household appliances to improve its overall demand and supply balance within a settlement period. since this settlement period varies in europe from 15 to 30 min, the control actions at the brP enterprise level should preferably be in the order of 5 min or fewer. in this way, a brP will have at least three moments during a set-tlement period to exert its control.

a further subgoal was to demonstrate the ability of the system to handle variable tariff-based metering data from the smart meter through the smart grid infrastructure to the enterprise system for billing and rating. this business cases assumes real-time tariffs, which implies that prices can change at any moment in time. billing is based on integrated volumes and prices over fixed periods, i.e., 15 min. therefore, the metering interface should be able to handle both the real-time varying tariffs on the usage side and the integration into periodic volumes and prices on the meter reading side.

the field trial in hoogkerk is built around the Power-Matcher technology. the PowerMatcher is designed to be scalable and applicable on a large scale. the objectives have proven this in two ways:

xx Performance of control: to control a cluster of house-holds for a business case, the control signals should reach the households and the devices fast enough to support the business case.xx Performance of metering and billing: variable pricing leads to large amounts of detailed measuring data, which has to be processed by the enterprise system.

a main challenge for the trial was to reach 1 million households. because of the large numbers involved, it is impossible to include 1 million real households in the trial. therefore, a large part of the trial was based on simu-lated entities. for the same reason, it is not feasible to sim-ulate every single household as a separate entity. the setup of the trial is shown in figure 9. it shows the hierar-chical structure of the PowerMatcher technology with two levels of concentrators between the enterprise system and the smart house gateways. the metering data are commu-nicated through the same layered structure.

in figure 9, several components can be distinguished. on the left side are the real-world components in the test (encircled in red). concentrator 1.1 is dedicated to the control of the real smart houses that are part of Pow-erMatching city, an existing test field in hoogkerk in the north of the netherlands. concentrator 1.2 is connected to 100 actual smart meters; this part is dedicated to delivering metering data without implementing Power-Matcher control. concentrator 2.1 connects concentra-tors 1.1–1.100 to the enterprise system located in Karlsruhe, Germany.

a simulation was set up running 1 million virtual households on the sh/sG side of the setup. this system was connected to the PowerMatching city field test to measure latency of the communication of the control sig-nal coming from the enterprise and reaching the lower level device agents in the smart houses, both real and simulated. to connect to the PowerMatching city field test, some adjustments have been made, which added artificial delays to the latency. to accommodate for software discrepancies, two objective agents and several Web ser-vices were used to enforce the sh/sG price on the Power-Matching city cluster (figure 10). bids and prices signals were recorded with the corresponding time at each level of the simulation to measure the latency.

the results of the latency test can be seen in table 3. the influence from an objective agent on the sh/sG side of the simulation reaches a real device in fewer than 5 min (on average 4 min), thus meeting the desired target. how-ever, there was a large amount of artificial latency, which influenced this result and could require 1 min to reach the agents. this artificial latency was caused by the polling and measurement tactics enforced on the PowerMatching city cluster. further, an additional auctioneer, Web services, and objective agents are required to complete this test.

on the metering side, it was discovered that a lot of time is spent on internal processing of the meter reading assessment process (e.g., validation, sanity checks, and storage) inside the application server itself. for example, the total request/response time for one connection, for a single meter reading submission to the metering server, was approximately four times longer than the time required to insert the metering data into the database. this difference was the first sign that the application server load should be balanced over multiple nodes. as such, further performance enhancements should focus on reducing the request processing time, e.g., a meter data concentrator to collect meter readings could be used and submit them in bulk to the main server. this way, the request processing time per meter reading can

1,800

1,6001,400

1,2001,000

800

600

400200

030 July 2010

0:001 Aug 2010

0:003 Aug 2010

0:005 Aug 2010

0:00

Active Power-Parking PV (W)

Active Power

Figure 8. The measurements at the Meltemi PVs.


be reduced, yielding higher efficiency. however, even more aggressive performance-related strategies might provide better results, such as usage of in-memory data-bases or strategic (on-demand or periodic) committing to the database.

Key Findings and Lessons Learnedartificial latency in the PowerMatcher communication will impact the latency between an objective agent and device agent is well within the settlement period of the brP. therefore, it was proven that there is potential to

Figure 10. The latency test.

SmartHouse

SmartHouse

SmartHouse

Concentrator 1.1

Agent forSmartHouse

vIntegralInterface

Concentrator 1.2

Concentrator 2.1

vDataConcentrator

SmartMeter

SmartMeter

SmartMeter

P5516

Concentrator1.3

Concentrator2.2

Concentrator2.100

Concentrator1.100

...

Enterprise System

vPerformancePC

vDummyDataProvider

vDummyDataConcentrator

Data Provider,One or More PCs

Data Provider,One or More PCs

...

AgentMimicking

ConcentratorCluster 100

AgentMimicking

ConcentratorCluster 2

AgentMimickingHouseholdCluster 100

AgentMimickingHouseholdCluster 3

Figure 9. An overview of the components in the Hoogkerk field trial.

SAPSetpoint

Web Service

Web Service

ObjectiveAgent

ObjectiveAgent

ObjectiveAgent

Auctioneer

AuctioneerAgent

SH/SG

Integral

Latency Measurement

Concentrator

Concentrator

Concentrator

Price Price Price

Price

Artificial Latency

Price Price

Bid

Bid


balance control via demand-side man-agement at a household level. further-more, there is also the potential for balancing ancillary services for the tso using this same method.

the sh/sG project made the assump-tion that there will be no common archi-tecture penetrating all layers of the envisioned smart grid. it is proven that this is a viable approach and the interme-diate communication and data exchange layers are appropriately wrapped. to this end, we must once more underline the importance of stan-dardization of smart meters, smart services, and energy management systems toward energy-related services and toward data models depicting the energy data acquired.

on the metering side, it turned out that there was no tool available for asset management of the smart meters used in the trial. as the measurement data are accumulat-ed values, it was not a big problem when values were miss-ing or duplicated. once in a while, the meshed radio- frequency network delivered duplicate data, i.e., with the same origin, value, and timestamp. the application should be robust enough to handle this situation. duplicate meter reading is a nonmalicious error; however, it high-lights the need to check data for compliance to specific for-mats and also to perform logical checks. although the focus of the project was not on security, in a real-world deploy-ment, this would be of paramount importance. at this instance, it would be quite possible to inject erroneous metering data in an attempt to destabilize the system, lower its performance, or even reduce the monthly bill.

Conclusionthe infrastructure that will exist in the future smart house is expected to be highly heterogeneous. however, it seems that at some level, all devices—either by them-selves or via gateways—will be able to communicate over the internet protocol and participate in bidirectional col-laboration with other devices and enterprise services. similarly, multiple concepts for monitoring and control-ling the smart houses and the smart grid will emerge, with different optimization and control algorithms. it is therefore imperative not to focus on a single one-size-fits-all approach but rather to prove that an amalgamation of existing approaches should be developed. the sh/sG proj-ect can be seen as a first step to developing mechanisms for gluing different monitoring and control approaches as well as empowering the next-generation enterprise services and applications. this is done by using Web ser-vices and open standards, as applied by the PowerMatch-er, beMi, and MaGic systems.

innovative technologies and concepts will emerge as we move toward a more dynamic, service-based, market-driven infrastructure, where energy efficiency and savings can be facilitated by interactive distribution networks. a

new generation of fully interactive ict infrastructure has to be developed to support the optimal exploitation of the changing, complex business processes and to enable effi-cient functioning of the deregulated energy market for the benefit of consumers and businesses.

AcknowledgmentsWe would like to thank the european commission for their support and the partners of the smart Grid projects sh/sG, integral (www.integral-eu.com), Modellstadt Mannheim (www.modellstadt-mannheim.de), and nobel (www.ict-nobel.eu) for the fruitful discussions.

For Further Readingsmarthouse/smartGrid. [online]. available: http://www.smarthouse-smartgrid.eu/

smartGrids european technology Platform. (2010, apr.). smartgrids: strategic deployment document for europe’s electricity networks of the future,. [online]. available: http://www.smartgrids.eu/documents/smartGrids_sdd_final_aPril2010.pdf

BiographiesAris Dimeas ([email protected]) is a researcher at the national technical university of athens, Greece.

Stefan Drenkard ([email protected]) is a senior engineer at Gopa-intec, Germany.

Nikos Hatziargyriou ([email protected]) is a pro-fessor at the national technical university of athens. for six years, he was deputy ceo of the Public Power corpora-tion of Greece.

Stamatis Karnouskos ([email protected]) is a research expert at saP in Karlsruhe, Germany.

Koen Kok ([email protected]) is a researcher at the nether-lands organization for applied scientific research (tno), the netherlands.

Jan Ringelstein ([email protected]) is a researcher at the fraunhofer, institute for Wind energy and energy system technology, Germany.

Anke Weidlich ([email protected]) is a professor at the hochschule offenburg, Germany.

TABLE 3. The Latency Test Results for Controllability.

NodeMeasured

PerformanceNet

PerformanceArtificial Latency

Virtual network

Enterprise 0 s — —

Concentrator level 2 70 s 10 s (30 + 30) s

Real network

Concentrator level 3 142 s 22 s (30 + 30) s

Home gateway 205 s 25 s (60) s

Device 240 s 30 s (30) s

Documents

Smart Houses in the Smart Grid: Developing an interactive network