Phase 1 Users Group 09/02/2012 Pieter-Jan Marsboom v12.02.09 1
Overview 1. Context & Drivers 2. Forecast & Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 2
Context & Drivers ELIA sells @ 1,1* P_Belpex_DAM ELIA buys @ 0,9*P_Belpex_DAM Imbalance tarrif M W p -in s ta lle d c a p a c ity e v o lu tio n B e lg iu m 1800 1600 1400 1200 1000 800 600 400 200 0 SO LAR W IN D 2005 Ref : 30%-rule in KB Offshore 30/03/ 09 2006 2007 2008 2009 2010 2011 Ref : VREG,CWAPE,BRUGEL 3
Context & Drivers 4
Overview 1. Context & Drivers 2. Forecast & Upscaling Model Forecast Model Inventory Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 5
Forecasting model (physical) DSOLocation + Wind [MW] + type Connection Eliasubstation PV [MWp] Cogen [MWe] Meteo Forecast [D+1 D+7] 4 km x 4 km BLACKBOX Wind Simulation Powercurve i.f.o. turbine-type Substation [D+1 D+7] Wind Temperature [ C] Windspeed [m/s] 10 m & 100 m PV PV Powercurve & η of PV-cells Wind direction [ ] Irradiation [W/m²] Exhaustive inventory With turbine-type & # HUB height coordinates Current content: Total = 930 MW Onshore = 735 MW Offshore = 195 MW Cogen Cogen Installation types industrial households 6
Inventory of windfarms in Belgium 7
Upscaling methodology v v v Realtime : 15/07/11-08/09/11 Realtime + ex-post: v = 70,5% => To compare with Germany : 23% = 78,1% - Not static numbers Try to acquire more measurements 8
Overview 1. Context & Drivers 2. Forecast & Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 9
Forecast Service Working in a probabilistic world MW Probabilistic Forecast Storm Indicator [0/1] 120 100 P90 [MW] 80 MW Measured [MW] 60 40 Forecast [MW] 20 P10 [MW] 0-20 Time (Hours) 10
Overview 1. Context & Drivers 2. Forecast & Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 11
Wind Forecast Tool : 3 GUI s GUI1: forecasts versus measurements For internal & external use (online publication, go-live expected 14th of february) Specifications : -Aggregated forecasts [D+1,D+7] in [MW] updated each day @ 11 A.M. -Upscaled measurements in [MW] updated each 15 min -Filtering possible : onshore <-> offshore, ELIA-connected <-> DSO-connected -Possibility to select period of interest (history) -Extracts in MS Excel possible 12
Wind Forecast Tool : 3 GUI s GUI2: realtime evolution of forecast error For internal use (national dispatching) Specifications : -Running average of forecast error [MW], updated each 2 min -Absolute forecast [MW] -Storm indicator [0/1] = which indicates a possible cut-off risk in the next 4 hours GUI3 : detailed dashboard with wind farm resolution & quarter hour time scale For internal use (national dispatching) exports possible 13
Overview 1. Context & Drivers 2. Forecast & Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 14
Forecast Quality Day-ahead wind forecast-error over 1 month RMSE% i.f.o. DA-prediction time aggregate of [437MW] ~ Belgium [1000MW] 16 Offshore Wind BE Onshore Wind BE 11 7 Aggregate Wind BE 5,5Aggregate Wind 50Hz 4,5 Aggregate Solar 50 Hz Predictability & Observability of Renewables : challenges for the TSO 15 15
Overview 1. Context & Drivers 2. Forecast & Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 16
Challenges wind power forecasting - Gain experience : ex. correlations with other control zones BE <-> GE 17
Challenges wind power forecasting Plausible correlation forecast error with system imbalance 18
Challenges wind power forecasting Storm management issues EWP V1 = cut-in windspeed V2 = Pmax windspeed V3 = cut-in after EWP(10 avg) V4 = cut-off in EWP (10 avg) EWP - Hysteresis [V4 V3] Typically [25 m/s 20 m/s] for one WT [22.5m/s 18m/s] for one WF RR = 4 à 5MW/min Ref: Twenties project Assessment of storm forecast Deliverable nº: 6.1 19
Challenges solar power forecasting Variability wind > solar Forecast-error wind > solar Importance of aggregation effect Germany > Belgium Ref: CORESO 20
Overview 1. Context & Drivers 2. Forecast & Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 21
Next Steps wind power forecasting Red = high wind zone Blue = low wind zone 22
Overview 1. Context & Drivers 2. Forecast & Upscaling Model 3. Forecast Service 4. Wind Forecast Tool 5. Wind Forecast Quality 6. Challenges 7. Next Steps 8. Conclusion 23
Conclusion Expected increase in installed renewables capacity (2015: wind > 2GW, solar > 2GWp) Considering a minimum load of 6 GW during summer : at times >50% of load will be covered by wind & solar alone Quid incompressibilities considering nuclear production & other not flexible units in base load First steps have been made regarding predictability & observability of wind power Go-live of external publication on ELIA website : expected 14th of February Next steps will have to deal with: an analogous project for solar power forecasting (considering the significant volume) improved reserve dimensioning based on gained experience further integration into decision support tools (congestion & balancing management) renewables dispatching // traditional dispatching MWp-installed capacity evolution Belgium 3000 2500 2000 Wind [MW] 1500 Solar [MW] 1000 500 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 24
Phase 1 Users Group 09/02/2012 Pieter-Jan Marsboom v12.02.09 25