TutorCristina CornaroDavid Moser

MACRO AREA DI INGEGNERIADOTTORATO DI RICERCA IN INGEGNERIA PER LA PROGETTAZIONE E

PRODUZIONE INDUSTRIALE

Anno Accademico 2019/2020XXXII CICLO

Photovoltaic power forecasting methods and applications to support high solar power

penetration

Marco PierroCoordinatore Ettore Pennestrì

Summary

Objectives

Motivations

Methodology

Results

Publications / workshop / conferences

1

Objectives

To set up new “state of the art” PV power,residual load forecasting models/methods

To go beyond the application of solar forecastsonly showing the technical/economic benefits ofimplementing a new category of photovoltaicsystems, which we have called "flexible" solarsystems

To show the technical/economic benefits ofaccurate forecasting in electrical gridmanagement

2

Motivationso In Italy, the electric mix in 20

years moved from 20% to 35%of RES generation and Oil products for thermoelectric generation almost disappear.

o In further 10 years the electric mix has to reach 55% (National Energy and Climate Plan –PNEC).

20% 35% 55%

o In only 10 years, solar became thesecond renewable energy source afterHydropower and it should become thefirst at 2030

3

MotivationsIntermittent Since PV power generation is locally produced and

consumed, it can completely modify the daily shape of the electric load

𝑷𝑷𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵=𝑷𝑷𝑳𝑳𝑵𝑵𝑵𝑵𝑵𝑵-𝑷𝑷𝑾𝑾𝑾𝑾𝑾𝑾𝑵𝑵-𝑷𝑷𝑷𝑷𝑷𝑷

5

Motivations

Accurate solar forecast is essential

to limit the imbalance and its related costs on

the Balancing Energy Market

Variable Residual electric demand (Net load) becomes dependent on the solar/wind stochastic variability, thus it is more

difficult to predict

Source T

erna S

pa

Il merca

to p

er il servizi d

i disp

accia

men

to

(semin

ario

RSE)

𝑷𝑷𝑰𝑰𝑰𝑰𝑰𝑰𝑵𝑵𝑵𝑵𝑵𝑵𝑾𝑾𝑰𝑰𝑵𝑵 = 𝑷𝑷𝑮𝑮𝑵𝑵𝑾𝑾𝑵𝑵𝑮𝑮𝑵𝑵𝑵𝑵𝑾𝑾𝑵𝑵𝑾𝑾𝑺𝑺𝑰𝑰𝑺𝑺𝑵𝑵𝑵𝑵𝑺𝑺𝑵𝑵𝑵𝑵𝑵𝑵 −𝑷𝑷𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑰𝑰𝒐𝒐 = 𝑷𝑷𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝒇𝒇𝑵𝑵𝑮𝑮 -𝑷𝑷𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑰𝑰𝒐𝒐

6

Solar Plants imbalance benchmark and related costs

Site PV power forecast

- best forecast- probabilistic

forecast

Upscaling methods for regional PV power

best/probabilistic forecast

PV power forecast blending methods

Dispatchable PV generation 24/365 and related costs

Imbalance mitigation strategies and related cost

100% renewable transition scenario

Reserves assessment

Net-load scheduling

Single plant scale

Regional scale

National scale

Flexible PV power forecast

Solar Plants imbalance benchmark and related costs

MethodologyResults relevant at national level

100% renewable transition scenario

Dispatchable PV generation 24/365 and related costs

Imbalance mitigation strategies and related cost

7

Results - Balancing market analysis 8

Italy is divided into 6 zones defined by physical energy transit limits of the NTG.Each Market Zone has its Energy Markets:• Day-Ahead/Intra-day markets (bulk energy trading)• Dispatching (MSD)/Real-time (MB) markets

(dispatchable energy trading)

• Relevant PV farms (capacity > 10 MW) have to deliver to the Italian TSO the day-ahead scheduling to evaluate their imbalances

• They are not allowed to provide balancing ancillary services and bid on the Balancing Markets (MSD/BM)

Results - Balancing market analysis 9

To define energy prices on the Balancing Market Italy is divided into 2 macro-zones

These prices are computed by the “single pricing” rule that can produce economic revenues or costs for PV

producers/traders depending on the macro-zone imbalance

Zone over-generation / PV under-generationZone under-generation / PV over-generation

IMBALANCE SIGNS

(+) (-)

(-) (+)

REVENUESZone / PV under-generationZone/PV over-generation

COSTS

IMBALANCE SIGNS

(-) (-)

(+) (+)

Results - Balancing market analysisForecasting method

10

1325 locations

ANN ensemblemodel

Physical basedmodel

NWP inputs

SINGLE PLANTPV POWER FORECAST

1325 locations

Methodology

EXSTIMATE PV generation

Satellite derived irradiance

NWPirradiance

Temperature

FORECAST PV generation

computeIMBALANCE

between estimation/forecast

ComputeIMBALANCE ECONOMIC VALUE

Day-ahead MSD markets

prices

Net-imbalance value per unit of PV installed capacity by regions

Region code

k€/M

Wp

Results - Balancing market analysis

1) Net-Imbalance values should be always NEGATIVE (cost) but there are some locations in which it is POSITIVE (revenue)

The net-imbalance value is the value of an imperfect forecast

net-imbalance value= real incomes – ideal incomes (no imbalance)

11

Imperfect forecast could be more profitable than “perfect”

Results - Balancing market analysis

2) The forecast values with respect to persistence should be always POSITIVE(economic gain) but there are some locations in which it is NEGATIVE

(economic loss)

Values of forecast with respect to persistence on DA and MSD markets

Region code

k€/M

Wp

12

Persistence could be more profitable than more accurate forecast

Results - Balancing market analysis

Imbalance Value on MSD can be decomposed in 2 factors:

Imbalance Value= avg imbalance unitary value* imbalance volume

• the average imbalance unitary value 𝑷𝑷𝒁𝒁𝑩𝑩 i (€/MWh per year) that embeds both the MSD energy prices (for upward or downward regulation) and the match between the UP and the Macro Zonal imbalance signs.

• the imbalance volume (MWh/MWp per year) is the accuracy of the forecast

RevenueCost RevenueCost

Good imbalance

signs match

Bad imbalance

signs match

The imbalance values on the

MSD market are much more correlated

𝑷𝑷𝒁𝒁𝑩𝑩 rather than with

𝑾𝑾𝑰𝑰𝑰𝑰𝑵𝑵𝑵𝑵𝑵𝑵𝑾𝑾𝑰𝑰𝑵𝑵𝒗𝒗𝑵𝑵𝑵𝑵𝑺𝑺𝑰𝑰𝑵𝑵

13

Results - Balancing market analysis

current market regulation framework is completely in contrast with the physical

need of reducing imbalances and hence it requires a

significant revision to allow a higher PV penetration.

Largest PV farm of 84,2 MWp in ItalyMontalto di Castro – Italy

(1th in Europe and 25th in the world)

14

3) predict the right sign of the macro-zone imbalance (and than provide a suitable under/over forecast to match the sign) is more profitable than minimize the imbalance volume (with the most

accurate forecast)

Results - RES transition strategies

The proposed RES transition strategies aim:

Remove intermittencyby firm solar generation

Remove variability by perfect forecast

Reduce uncertainty by solar regulation

15

increasingpenetration

of FLEXIBLE

PV plants

Results - RES transition strategies

Battery Energy Storage System (BESS)

Smart inverters& remote control system

PV field

“flexible”PV

16

Ancillary Services

Firm 24/365 generation

Flex PV plants adapt generation to predicted

profile or to load profile

BESSadditional power

Smart inverterpro-active

curtailment

Dispatchable generation

We cost-optimally size the Flexible PV plants to generate at energy costs lower than current

Results - RES transition strategiesUpscaling forecast method

1325 locations

Methodology to test the strategiesItalian

Load-PV-WindNWP

irradiancetemperature

FORECAST Load-PV-Wind power

computeIMBALANCE VOLUME-COSTS / LCOE

compare the obtainedVOLUME-COSTS / LCOE

with TSO current and future value

Day-ahead MSD markets prices

ANN ensemblemodel

Physical basedmodel

NWP inputs

ITALIANPV POWER FORECAST

Pre-processing2650 inputs --> 14 inputs

Post-processing

computeNet-load

Net-load forecast

17

Imba

lanc

e vo

lum

e [T

Wh/

yr]

Imba

lanc

e co

st [M

€/yr

]year year

12131620

2398

24

37

17.3

Current/expected TSO forecast Solar regulation by flex PV Perfect forecast by flex PV

12.2(-29%)

Solar regulation

778(-36%)

Solar regulation

1694(5%) 2110

(-12%)

Perfect forecast

Highly flexible Relevant PV plants can remove the effect of the PV penetration allowing a

massive share of solar energy

Imbalance increase dramatically with

penetration

11.9(-50%) 9.6(-74%)

Perfect forecast

7% 22% 37%PV penetration unconstrained PV

7% 22% 37%PV penetration unconstrained PV

3.7 GWp

0 GWh

19%

16.2 GWp

32 GWh

25%

33.7 GWp

67 GWh

25%

600$/kWh

DA price€/MWh

400$/kWh

DA price€/MWh

100$/kWh

DA price€/MWh

18

2016 2030 2060 2016 2030 2060

According to a NREL studywe consider:

BESS cost= 100 €/kWh

Utility scale PV cost= 400 €/kWpDistributed PV cost = 920 €/kWp

Wind cost= 1400 €/kWp

LCOE = 40.8 €/MWh

BESS =120 GWh (0.92 GWh/GWp)

PV = 130 GWp

23.2% of generation

WIND = 50 GW

NATURAL GAS = 8% of loadWe perform a cost-optimal sizing of flexible plants

Among 10,000 Energy Plan simulations with different

PV, wind and storagecapacity, we select the

solution with the minimum achievable LCOE

Removeintermittency by

PV-WIND FIRM GENERATION

19

Results - RES transition strategies92% Least-cost RES transition is obtained at 2060 by a progressive

replacement of unconstrained PV-Wind with flexible PV-Wind

20

Results - RES transition strategiesGW

GW13% wind-solar penetration @2016

66.5% wind-solar penetration @2060

PV ~

2X -

Win

d ~2

.5X

the

obje

ctiv

e fo

r 20

3021

ConclusionsThe Italian Balancing Energy Market regulation framework leads toparadoxical and counterproductive effects regarding the need to reducethe imbalance volumes and should be absolutely revised to allow a massivesolar penetration

We demonstrated that flexible PV plants can effectively provide imbalanceregulation services as long as they are directly controlled by the Italian TSO.These flexible plants can reduce or eliminate the forecast errors of thewhole Italian PV generation and they can be suitably dimensioned tominimize regulation costs at or below current imbalance costs.

We demonstrated that least-cost RES transition is fully feasible for Italy under two conditions only: (1) the grid constrains between the market zones have to be removed

allowing a complete share of renewable generation along the whole country and consequently enabling the enlargement of the forecast controlled area;

(2) all the variable energy systems have to be turned in to “flexible” and controlled directly by the Italian TSO.

22

Conclusions

From this study three consequences emerge in contrast to the widespread thinking in the renewable energy sector :

It should be promote RES CURTAILMENT since it is key factor to reduce thecapex of the flexible power plants

Massive RES penetration does not need DISTRIBUTED GENERATIONSbut TSO CONTROLLED large “flexible” wind-solar farms generation

No changes of grid architecture are needed

As solar energy is AVAILABLE TO EVERYONE FREE of charge, so it is muchmore efficient to use it for the BENEFIT OF THE WHOLE COUNTRY (by“flexible” PV controlled by the TSO) than for INDIVIDUAL CITIZENS (by“unconstrained“ grid connected PV)

23

Publications / workshop and conference23 Publications

o n. 8 published on peer review journals

o n. 1 submitted on peer review journals

o n. 10 published on conference proceedings

o n. 2 international reports

o n. 2 disseminating papers

4 Workshop / conferences / seminars

o 2016/2017 n. 2 conferences (1 oral presentation) and n. 4 seminars

o 2017/2018 n. 1 workshop (visual presentation)

o 2018/2019 n. 1 workshop (oral presentation) and n. 1 conference (oral presentation)

3 months of training period abroad11 July-21 October 2018 State university of New York, Atmospheric Sciences Research Center, Albany, USA, Supervisor: Prof R. Perez

24

Thank youCristina Cornaro

University of Rome«Tor Vergata»

David MoserEurac Research

Matteo Giacomo PrinaEurac Research

Matteo De FeliceENEA

Alessandro Perotto - Enrico Maggioni - Francesco SpadaIdeal Srl

Richard PerezAtmospheric Sciences

Research Center, SUNY, USA

Marc PerezClean Power Research,

USA