BASED ON NUMERICAL WEATHER PREDICTION, SATELLITE DATA, AND POWER MEASUREMENTS Detlev Heinemann, Elke Lorenz Energy Meteorology Group, Institute of Physics, Oldenburg University Workshop on Forecasting, Balancing and Scheduling of Renewable Energy Sources in India Shangri-la Hotel, New Delhi 5 May 2014
CONTENT (I) German power sector today (II) Components of solar power forecasting system (III) Irradiance forecasting based on satellite data (IV) Combination of numerical weather models and satellite data (V) PV power forecasts (VI) Forecasting of Direct Normal Irradiance (DNI) (VII) Advanced techniques 2
GERMAN POWER SECTOR TODAY NET POWER PRODUCTION FROM WIND AND SOLAR, 2013 Installed capacity, end of 2013: PV: 35,7 GW Wind: 34,7 GW (compare with average load: ~63 GW) Power production 2013: PV: 29,7 TWh (5,3 % of net electricity consumption) Wind: 47,2 TWh (8,4 % of net electricity consumption) Max. combined production from wind & solar (18.04.2013): 35,9 GW = 52 % of load (PV 19,2 GW, Wind 16,7 GW) Partial supply of load from wind and solar up to 50% each Problem: peak load Excess production very likely to occur soon Data: Fraunhofer ISE, BSW, DEWI 3
GERMAN POWER SECTOR TODAY NET POWER PRODUCTION FROM WIND AND SOLAR Maximum of combined production from wind & solar 18 April 2013 35,9 GW Data: EEX 4
GERMAN POWER SECTOR TODAY NET POWER PRODUCTION FROM WIND AND SOLAR Maximum of combined production from wind & solar 14 April 2014 37,8 GW Data: EEX 5
GERMAN POWER SECTOR TODAY MONTHLY PRODUCTION SOLAR, WIND 2013 TWh Max: 8,2 TWh 7,0 6,0 5,0 4,0 3,0 Min: 3,9 TWh SOLAR 2,0 1,0 WIND Min: 1,7 TWh Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Sources: B. Burger, Fraunhofer ISE, Leipziger Strombörse EEX 6
GERMAN POWER SECTOR TODAY DAILY PRODUCTION SOLAR, WIND 2013 SOLAR Max: 0,2 TWh Min: 0,002 TWh WIND Max: 0,56 TWh Min: 0,006 TWh SOLAR & WIND Max: 0,58 TWh Min: 0,022 TWh Sources: B. Burger, Fraunhofer ISE, Leipziger Strombörse EEX 7
GERMAN POWER SECTOR TODAY MAXIMUM* POWER SOLAR, WIND 2013 * 15-min averages SOLAR 24,0 GW WIND 26,3 GW Sources: B. Burger, Fraunhofer ISE, Leipziger Strombörse EEX 8
BALANCING OF SOLAR POWER ACCORDING TO RENEWABLE ENERGY SOURCES ACT Mandatory purchase of all available renewable electricity by grid operators Marketing and balancing of PV power by transmission system operators (TSOs) based on their respective share of the all German electricity supply (horizontal burden sharing) Additional option of direct marketing of PV power is getting more important Need for regional forecasts control areas of German transmission system operators 9
BALANCING OF SOLAR POWER BY TRANSMISSION SYSTEM OPERATORS PV power feed-in, Germany day-ahead: selling of PV power at the European Power Exchange (EPEX) - hourly contingents - 12:00 for the next day hourly forecast for next day 10
BALANCING OF SOLAR POWER BY TRANSMISSION SYSTEM OPERATORS PV power feed-in, Germany day-ahead: selling of PV power at the European Power Exchange (EPEX) - hourly contingents - 12:00 for the next day intra-day: trading of electricity at the EPEX - hours or 15minute periods - until 45 minutes before delivery forecast errors influence prices 15min resolution forecast for next hours 11
BALANCING OF SOLAR POWER BY TRANSMISSION SYSTEM OPERATORS PV power feed-in, Germany day-ahead: selling of PV power at the European Power Exchange (EPEX) - hourly contingents - 12:00 for the next day intra-day: trading of electricity at the EPEX - hours or 15minute periods - until 45 minutes before delivery forecast errors influence prices online-estimation from measured values remaining deviations are adjusted with balancing power 12
REGIONAL PV POWER PREDICTION OVERVIEW OF SCHEME Cloud motion from satellite Numerical weather prediction PV power measurements Solar irradiance forecast Simulation of PV power Regional PV power forecast 13
IRRADIANCE FORECAST USING SATELLITE DATA CLOUD TRACKING/CLOUD MOTION VECTORS Cloud index from Meteosat images (Heliosat methode) Meteosat Second Generation (HR VIS) Resolution: MSG (HRV): 1.2 km x 2.2 km (Germany) 15 Minuten 14
IRRADIANCE FORECAST USING SATELLITE DATA CLOUD TRACKING/CLOUD MOTION VECTORS Cloud index from Meteosat images (Heliosat methode) Meteosat Second Generation (HR VIS) Cloud motion vectors (CMV) from identification of pattern in consecutive images Extrapolation of cloud motion to forecast next cloud index image 15
IRRADIANCE FORECAST USING SATELLITE DATA CLOUD TRACKING/CLOUD MOTION VECTORS Cloud index from Meteosat images (Heliosat methode) Satellite-based solar irradiance map Cloud motion vectors (CMV) from identification of pattern in consecutive images Extrapolation of cloud motion to forecast next cloud index image Solar irradiance from forecasts of cloud index images using the Heliosat method 200W/m 2 900W/m 2 16
EVALUATION OF IRRADIANCE FORECASTS DATA BASIS Period 01.01.-30.09.2012 Measurements irradiance measurements of 270 stations* in Germany Forecasts DWD ECMWF CMV Combination of forecasts * operated by DWD and meteomedia GmbH 17
EVALUATION COMBINATION OF ECMWF, DWD & CMV* FORECASTS * CMV: Cloud Motion Vectors I in W/m 2 date 18
EVALUATION COMBINATION OF ECMWF, DWD & CMV* FORECASTS * CMV: Cloud Motion Vectors I in W/m 2 date Combination of forecasts increases accuracy! 19
EVALUATION: 3-HOUR FORECASTS COMBINATION OF ECMWF, DWD & CMV FORECASTS CMV, CMV- mean Germany rmse ECMWF 41W/m 2 rmse CMV 34W/m 2 3h ahead: Satellite-based forecasts show improved results than NWP-based forecasts (MVF == CMV) 20
EVALUATION: 3-HOUR FORECASTS COMBINATION OF ECMWF, DWD & CMV FORECASTS CMV, CMV- mean Germany rmse ECMWF 41W/m 2 rmse cmv 34W/m 2 rmse combi 25W/m 2 improvement * vs ECMWF improvement * vs CMV 39% 26% Combination shows significant improvement: cc error(ecwmf,dwd) =0.51, cc error(ecwmf,cmv) =0.29 * 21
UNCERTAINTY OF COMBINED FORECASTS RMSE AS FUNCTION OF FORECAST HORIZON CMV forecasts superior to NWP based forecast up to 4 hours ahead CMV forecast superior to persistence of cloud situation from 2 hours onwards significant improvement with combined model Calculation was done only for solar zenith angle < 80 and only for hours for which all models were available (depending on the forecast horizon) 22
UNCERTAINTY OF COMBINED FORECASTS RMSE AS FUNCTION OF FORECAST HORIZON significant improvement by combination of forecasts with different time horizons more pronounced for regionally smoothed forecasts 23
REGIONAL PV POWER PREDICTION OVERVIEW OF SCHEME Cloud motion from satellite Numerical weather prediction PV power measurements Solar irradiance forecast Simulation of PV power Regional PV power forecast 24
REGIONAL PV POWER PREDICTION OVERVIEW OF SCHEME Cloud motion from satellite Numerical weather prediction PV power measurements Solar irradiance forecast Simulation of PV power Regional PV power forecast 24
PV POWER FORECAST DATA BASIS Control area 50Hertz December 2011 November 2012 Hourly data PV measurements: Monitoring data meteocontrol: > 1000 PV systems, approx. 20% of installed power Single site evaluation: 80 selected systems No measurement of overall PV power feed-in in control area! 25
ESTIMATION OF PV POWER FEED-IN IN CONTROL AREA Upscaling using Meteocontrol monitoring data all systems in the control area considering spatial distribution with 2-digits post code resolution* considering distribution of PV system size * overall installed power Pnom (given by 50Hertz) Forecast data basis 06:00, 1-3 days 11/11-05/12: ECWMF only 06/12-11/12: ECMWF & DWD 11:00, intra-day 09/12-11/12: combination of NWP and CMV * based on EEG data, 16.8.2012 26
EVALUATION control area single sites bias rmse bias rmse Intra-day 0.9% 4.9% 1.1% 12.0% Day-ahead 0.7% 5.7% 1.2% 12.8% 3 days 0.5% 6.2% 0.9% 13.4% daylight values only, normalization to installed power 27
EVALUATION ANNUAL COURSE OF FORECAST ERRORS Largest rmse values: February April 2012 28
EVALUATION ANNUAL COURSE OF FORECAST ERRORS Largest rmse values: February 2012 --> Snow 29
EVALUATION COMBINATION OF NWP AND SATELLITE-BASED CMV FORECASTS o NWP, 06:00 x combi, 11:00 o NWP, 06:00 x combi, 11:00 Large improvement by combination of models! 30
THANK YOU FOR YOUR ATTENTION! 31
FORECASTING OF DNI CSP needs a high DNI forecast accuracy especially in cloud-free cases with high DNI (different from, e.g., PV) Good forecasts in case of low DNI are required (-> good water cloud mask forecast) as well as for medium DNI situations ( ->good forecast of cirrus cloud optical properties) CSP technologies generally operate only in areas with high DNI and low cloud cover depending on the geographical region of interest and its vicinity to global aerosol sources, the priority is set either on good aerosol or cirrus forecasts The MACC (Monitoring Atmospheric Composition and Climate) project of ECMWF in the scope of GMES provides analysis and forecast of aerosols integrated into the Integrated Forecast System (IFS) 32
FORECASTING OF DNI PRINCIPLE Measurement data Clear sky model ECWMF irradiance forecast Hourly site-specific GHI forecasts Postprocessing Direct/ diffuse model DNI forecasts 33
FORECASTING OF DNI MODEL FOR DIRECT IRRADIANCE Parameterization for beam fraction b = Ibeam /G in dependence of k* cloudy situation clearsky broken cloud effect k*-range separated in three intervals: Smoothed with weighting function 34
FORECASTING OF DNI EVALUATION (2005 & 2007) Stations in Southern or Central Spain, DNI: rel. rmse: 40% 50% Stations at the Northern Spanish coast, DNI: rel rmse: >70 % 35