Accuracy evaluation of solar irradiance forecasting technique using a meteorological model

Accuracy evaluation of solar irradiance forecasting technique using a meteorological model October 1st, 2013 Yasushi MIWA Electric power Research & Development CENTER

Electric Power Companies in Japan 10 Electric Power Companies for 10 Regions Hokkaido EPCO 60Hz Hokuriku EPCO Kansai EPCO Chugoku EPCO 50Hz Tohoku EPCO Tokyo EPCO Chubu EPCO Kyushu EPCO Shikoku EPCO Okinawa EPCO 2

Today's contents 1. Background of Irradiance Forecasting --- For Large-Scale Penetration of PV Power Generation 2. Estimation for Current Irradiance 3. Forecasting of Irradiance and Accuracy Evaluation 3

Background of Irradiance Forecasting 4

Government Targets for Solar Power Penetration Household: About 70% Home use: About 5.3 million homes 28 GW (7 Gl) PV power generation is installed over widespread area 14 GW (3.5 Gl) Technical system development required About 20 times more than in 2005 About 10 times more than in 2005 Initiation of purchase system Home use: About 0.32 million homes It is 1.4 necessary GW (0.35 Gl) to analyze how it will affect Residences: About 80% Initiation of subsidy scheme Non-residential for residential solar power the buildings: power About 20% system Household: About 70% Non-Household buildings: About 30% Reference: Ministry of Economy, Trade and Industry. 5

In Preparation for Large-Scale Penetration of PV Power Generation The variation of PV power generation brings about the issue of power quality. Therefore, research is conducted in order to maintain the power quality. Demonstration project To be stable frequency Surplus electric power 2009 2010 2011 2012 2013 Accumulation/Analysis of PV generation data Forecast / estimate of PV power Variation Micro-grid demonstration project in islands Power system simulator Development of technology for demand-supply balance Smart power system technology Reference:Report of the Next Generation Transmission and Distribution Network Research(April 2010) 6

Electric power co. Participants of 3 next generation demonstration projects Development of voltage control technology on the distribution network Development of demand and supply planning and control technology of a total power system Development of control technology of Demand-side facility Demonstration Projects for Next Generation Optimum Control of Power Transmission and Distribution Network (28 companies) Development of solar irradiance estimation and forecasting Development of PV generation estimation Demonstration project of forecast technologies for photovoltaic generation (17 companies) analysis of solar irradiance Development of equipment for voltage control on the distribution network Research of present communication standardizations and communication security Demonstration Projects for Power Control Systems by Two-Way Communications Development of PCS with Two-Way communications Demonstration Projects for Next Generation Power Control Systems by Two-Way Communications (33 companies) 42 companies participate University, Laboratories The University of Tokyo, Tokyo institute of Technology, Waseda University, Central Research Institute of Electric Power Industry, Japan Weather Association Manufacturers IBM, Itochu, Itochu techno solutions, NEC, NRI Secure technologies, NTT docomo, Oki, Omron, Kandenko, KDDI, Sanyo, Sharp, Sumitomo Electric, Solar Frontier, Daikin, Takaoka Electric, TMEIC, Toshiba, Nissin Electric, Panasonic SS Japan, Panasonic SS nfrastructure, Hitachi, Fujitsu, Fuji Electric, Mitsubishi Motors, Mitsubishi Electric, Meidensha Electric power companies Hokkaido,Tohoku,Tokyo,Chubu, Kansai,Hokuriku,Chugoku, Shikoku,Kyushu,Okinawa

In Preparation for Large-Scale Penetration of PV Power Generation In order to analyze variations in PV power generation, pyranometers, etc. were placed for measure & accumulate the solar irradiance data. Time-synchronized data is measured every 10 seconds. Pyranometers were placed at 321 points. Among them, PV power generation data is also measured at 116 points. Pyranometer locations (image) Pyranometer Thermometer PV panels 8

Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Smoothing Effect on long term variation Although short term variation in solar irradiance are smaller, there is a significant change in solar irradiance due to the change in weather. 1.2 1.0 0.8 0.6 0.4 0.2 1.2 1.0 0.8 0.6 0.4 0.2 1 0.0 6 9 12 15 Time 18 38 Average solar irradiance of 42 points 1.2 1.0 0.8 Average of 42point Ideal Max. values 0.0 6 9 12 15 Time 18 200 150 100 50km 0.6 0.4 1.2 1.0 57 0.2 0.8 0.6 0.4 0.2 1.2 1.0 28 0.0 6 9 12 15 Time 18 0.0 6 9 12 15 Time 18 :Pyranometer measuring point :Waveform measuring point 0.8 0.6 0.4 0.2 0.0 6 9 12 15 Time 18 Weather condition in Nagoya Sunny then rainy 9

Load Load Impact on demand-supply balance due to large-scale penetration of PV When forecasted value of PV generation was overestimated, Electric Power Companies have to supply from their traditional supply source... Thermal, Hydro, etc. Or, Would Renewable Energy supplier apply energy storage for their RAMP? Or would Energy Storage Service be viable? PV Gen Time Forecasted Load curve & PV Gen. Electric Power Companies would like to know how much they had to prepare their resources from the forecasted PV generation. Current PV generation is also essential for power system operation. PV Gen Time Actual Load curve & PV Gen.

Generation planning & PV generation Forecast & Estimation Six Days before Five Days before Four Days before Three Days before Two days before Previous day The Day Generation Planning Weekly ahead planning Day ahead planning Hour ahead planning & operation Predictive operation Real time operation Photovoltaic Generation Forecast & Estimation Weekly ahead forecast Day ahead forecast Hour ahead forecast short-term forecast Real time estimation

Estimation for Current Irradiance

Approach to estimate for current solar irradiance Estimate current solar irradiance using meteorological satellite Meteorological satellite solar irradiance Caution: These figures are for different day Reference : JMA http://www.jma.go.jp/ Solar irradiance observation data Data source JWA GPS precipitable water Pyranometer 13

Correlation between measured and estimated irradiation Correlati on factor # of mesh Ratio >0.9 213,934 56.7% >0.8 347,238 92.0% Total 377,443 100% Good estimation Except northern area, small islands. There are Fewer households in those area, less PV, smaller impact on amount of PV generation. correlation coefficient

Nagoya ---One of the Largest City Area in Japan Nagoya is picked up for evaluation, estimate & forecast of Irradiation 60Hz 50Hz Osaka Tokyo Nagoya 15

Accuracy of Estimation for Current Irradiation (monthly) Oct. Sep. Aug. Jul. Jun. May Apr. Mar. Feb. Jan. Dec. Nov. Oct. Sep. Aug. Jul. Jun. May Apr. Mar. Feb. Jan. Dec. Nov. City Center City City & suburb City Center City City & suburb In July and August, higher RMSE were found because of higher Irradiation during the summer in Japan. In May and June, lower RMSE were found due to the rainy season in Japan, which give lower irradiation.

Accuracy of Estimation for Current Irradiation (hourly) City Center City City & suburb City Center City City & suburb At noon( 12-15), higher RMSE were found because of higher Irradiation during the daytime. (Similar to the irradiation at the top of atmosphere) City Center of Nagoya, City area of Nagoya is almost same RMSE, also, West of AICHI area, which is wider than City area of Nagoya, shows higher RMSE, due to the higher density of Irradiation measuring in City center and City area than in West of AICHI area.

Forecasting of Irradiance and Accuracy Evaluation 1st day 2nd day 3rd day 4th day 5th day 6th day 7th day 7days forecast of Irradiance

Approach to forecast for solar irradiance Weekly, Day-ahead solar irradiance forecasts using weather forecasting model GSM data from JMA Improvement of weather forecasting model to estimate for solar irradiance Reference GSM : Japan Meteorological Agency(JMA) http://www.jma.go.jp/ Solar irradiance : Japan Weather Association (JWA) Caution : These figures are for different day 19

RMSE & ME of Forecast of Irradiation (monthly) Day ahead Forecast Nov. Dec. Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. City Center City City & suburb City Center City City & suburb Hour ahead Forecast Oct. Sep. Aug. Jul. Jun. May Apr. Mar. Feb. Jan. Dec. Nov. Oct. Sep. Aug. Jul. Jun. May Apr. Mar. Feb. Jan. Dec. Nov. City Center City City & suburb City Center City City & suburb RMSE depends on season, higher Irradiation in Summer (Jun. Jul. Aug)

RMSE & ME on Forecast of Irradiation (by weather conditions) Day Ahead Forecast Clear Fine Slightly Cloudy Rainy Clear Fine Slightly Cloudy Cloudy Cloudy Rainy City Center City City & suburb City Center City City & suburb Hour ahead Forecast Clear Fine Slightly Cloudy Rainy Clear Fine Slightly Cloudy Cloudy City Center City City & suburb City Center City City & suburb Higher RMSE at Cloudy, Rainy days, compare to Clear & Fine days Cloudy Rainy

Error Variance of Forecast of Irradiation (monthly) Error (W/m 2 ) Error (W/m 2 ) Error Ratio Error Ratio Day Ahead Forecast Nov. Jan. Apr. Jul. Oct. Nov. Jan. Apr. Jul. Oct. Hour ahead Forecast Nov. Jan. Apr. Jul. Oct. Nov. Jan. Apr. Jul. Oct. Larger variance for Day ahead Forecast, than Hour ahead Forecast Larger variance in Summer (Jun. Jul. Aug.) due to Higher Irradiation

Error Variance of forecast of Irradiation (by weather conditions) Error (W/m 2 ) Error (W/m 2 ) Error Ratio Error Ratio Day Ahead Forecast Clear Fine Slightly Cloudy Cloudy Rainy Clear Fine Slightly Cloudy Cloudy Rainy Hour ahead Forecast Clear Fine Slightly Cloudy Cloudy Rainy Clear Fine Slightly Cloudy Cloudy Rainy Lower Actual Irradiation on Cloudy & Rainy days, even Higher Forecast Irradiation

;Irradiation Reliability of Forecast of Irradiation ;Legends Estimated Actual 99% Confidence Interval 60% Confidence Interval Ex. of Reliability of Forecast of Irradiation (at Nagoya area) Reliability Range of Forecast of Irradiation is deeply depending on the day, season, weather conditions. Further Study on Reliability Range of Forecast of Irradiation is required.

Thank you Polycrystalline HIT CIS Amorphous

Blank

Blank

Additional Slides

Error Variance of forecast of Irradiation (by Time) Day Ahead Forecast Today's Forecast Higher ME at 9-15, due to high Irradiation at noon

Error Variance of forecast of Irradiation(by month) 3 different methodology applied for Irradiation Forecast in the Project. In Jun. Higher ME were found due to Rainy Season

Error Variance of forecast of Irradiation(by Time) Relatively small difference between in the morning, noon, afternoon at ME Each methodology(jwa, denken, CTC) has its own character

Error Variance of forecast of Irradiation(by weather conditions) ME depends on weather conditions 3 methodologies also depends on weather conditions. Weather Conditions Must consider at Forecasting of Irradiation

Point for changing Time Scale

About "Accumulation/Analysis of PV generation data(pv300)" Ministry of Economy, Trade and Industry project From 2009 To 2011(3Years) R&D Items 1 Accumulation of PV generation data 2 Analysis of PV power variation Project organization Hokkaido Electric Power Co., Inc. Tohoku Electric Power Co., Inc. Tokyo Electric Power Co., Inc. Chubu Electric Power Co., Inc. Hokuriku Electric Power Co., Inc. Kansai Electric Power Co., Inc. Chugoku Electric Power Co., Inc. Shikoku Electric Power Co., Inc. Kyushu Electric Power Co., Inc. Okinawa Electric Power Co., Inc. 34

In Preparation for Large-Scale Penetration of PV Power Generation Measuring equipment Basic system PV power measuring system Horizontal pyranometer Thermometer Horizontal pyranometer Thermometer Inclined pyranometer PV panel 35

PV Power PV Power PV Power Characteristics of Weather and PV power variation Sunny without cloud Sunny with cloud Rainy 6 9 12 15 Time 18 6 9 12 15 Time 18 6 9 12 15 Time 18 Bell curve with little variation PV power vary when clouds pass by Little power with little variation 36

Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Solar irradiance(kw/ m2 ) Smoothing Effect on short term variation Variation in solar irradiance from movements of small clouds differ in every location. So if the variation is large in every point, the total variation of multiple locations becomes small. It is called smoothing effect and is observed by these data. 1.4 1.2 1.0 0.8 0.6 38 Average solar irradiance of 42 points 1.4 1.2 0.4 0.2 1 1.4 1.0 0.8 0.6 0.4 0.2 0.0 1.4 1.2 6 9 12 15 Time 18 0.0 6 9 12 15 Time 18 200 150 100 50km 57 1.2 1.0 0.8 0.6 0.4 1.0 0.8 0.2 0.6 0.4 0.2 0.0 6 9 12 15 Time 18 :Pyranometer measuring point :Waveform measuring point 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 6 9 12 15 Time 18 28 0.0 6 9 12 15 Time 18 Weather condition in Nagoya Sunny (cloud cover: 2-6) 37

PV Power Characteristics of Weather and PV power Variation Sunny then Rainy (or cloudy) 9:00 Sunny 15:00 Rainy 6 9 12 15 Time 18 PV power changes according to changes in weather 38

PV generation forecasting technology from the perspective of Power System Operation 39

Requirements for PV generation forecasting Demand Increased demand Increase generators' power Generator power must be controlled depending on the changes of PV generation. PV power Demand Demand 6 9 12 15 18 Demand-PV power Decreased PV power PV generation Time Ex. Decreased PV generation depend on cloud The "The Demonstration Project of Forecast Technologies for Photovoltaic Generation (METI project)" has been in place since 2011. 40

PV generation Requirements for PV generation forecasting 2When PV power will change. 3How much PV power will change. 1Current PV power 6 9 12 15 18 Time 41

Approach to estimate for PV power Estimate for PV power variation Input Solar irradiance Temperature Convert Tool (Computer) Output PV Power Additional information Installed place Rated value of PV, PCS Front direction of PV Longitude and latitude of PV Type of PV panel(ex., crystalline CIS) Efficiency of PCS Considering these information need or not 42

Solar irradiance(w/ m2 ) Solar irradiance(w/ m2 ) Ex. Results of forecasted solar irradiance Day-ahead forecasts May / 2011 1000 Measurement Forecast 800 600 400 200 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Day 1200 1000 Measurement Forecast 800 600 400 200 0 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Day Measurement : Average of 42 points within a radius of 200km Forecast : By Japan Weather Association (JWA) 43

RMSE(W/ m2 ) Ex. Results of forecasted solar irradiance Month (11/2010-10/2011) 100 RMSE Extra-terrestrial solar irradiance (W/ m2 ) 1000 50 500 0 0 11 12 1 2 3 4 5 6 7 8 9 10 Month The results vary in months? The results depend on extra-terrestrial solar irradiance? Data source : Japan Weather Association (JWA) 44

RMSE(W/ m2 ) Ex. Results of forecasted solar irradiance Weather (11/2010-10/2011) 100 50 0 Sunny without cloud Sunny with scattered cloud Sunny with dense cloud Cloudy Rainy The results also depend on weather Data source : Japan Weather Association (JWA) 45

Error(W/ m2 ) Ex. Results of forecasted solar irradiance Weather (11/2010-10/2011) 300 Percentile 200 100 0 Now we improve 95% 75% Average -100-200 forecasting quality 25% 5% -300 Sunny without cloud Sunny with scattered cloud Sunny with dense cloud Cloudy Rainy Data source : Japan Weather Association (JWA) We are checking using percentile Plus error or Minus error Range of forecasted error There are terrible error or not 46

Power Demand-Supply Balance for changes of PV Power Generation In order to provide a stable power supply, the power generation (supply) of generator (thermal, etc.) is finely adjusted according to changes in demand to maintain a certain frequency. If the demand-supply balance is lost, the frequency changes, resulting in problems such as abnormal operation of electrical equipment. New-type power demand-supply It is also necessary to control system maintain balance by controlling the thermal Power control Forecast of Demand power generations, Battery Hydro Thermal 水力火力 etc., in response to Frequency changes in the power of PV generation. Supply Forecast / Estimate of PV power Demand 47? PV Power

Error(W/ m2 ) Results of forecasted solar irradiance Weather (11/2010-10/2011) 300 Percentile 200 100 95% 75% 0 Average -100 25% -200 5% -300 Sunny without cloud Sunny with scattered cloud Sunny with dense cloud Cloudy Rainy Data source : Japan Weather Association (JWA) We are checking using percentile Plus error or Minus error Range of forecasted error There are terrible error or not 48

In Preparation for Large-Scale Penetration of PV Power Generation The variation of PV power generation brings about the issue of power quality. Therefore, research is conducted in order to maintain the power quality. Demonstration project To be stable frequency Surplus electric power 2009 2010 2011 2012 2013 Accumulation/Analysis of PV generation data Forecast / estimate of PV power Variation Micro-grid demonstration project in islands Power system simulator Development of technology for demand-supply balance Smart power system technology Reference:Report of the Next Generation Transmission and Distribution Network Research(April 2010) 49

Forecasting of Irradiation -- Time Scale Forecasting of Irradiation Generation planning Weekly Every 30 min in 168hours+ (same as on the left) Day ahead Every 30 min in 24hours+ (same as on the left) Today Every 30 min for today (same as on the left) hourly Every 30 min for next few hours (Commitment/ELD) minutely Next Few minutes (Governor Free/ELD)

Participants of 3 next generation demonstration projects Development of voltage control technology on the distribution network Development of demand and supply planning and control technology of a total power system Development of control technology of demand-side facility Development of solar irradiance estimation and forecasting Development of PV generation estimation Analysis of solar irradiance Development of equipment for voltage control on the distribution network Research of present communication standardizations and communication security Demonstration Projects for Next Generation Optimum Control of Power Transmission and Distribution Network (28 companies) Demonstration Projects of forecast technologies for photovoltaic generation (17 companies) Demonstration projects for power control systems by two way communications Development of PCS with Two-way communications Demonstration Projects for Next Generation Power Control Systems by Two-Way Communications (33 companies) 42 companies participate University, Laboratories The university of Tokyo, Tokyo institute of technology, Waseda University, Central Research Institute of Electric Power Industry, Japan Weather Association Manufacturers IBM, Itochu, Itochu techno solutions, NEC, NRI secure technologies, NTT docomo, Omron, Kandenko, KDDI, Sanyo, Sharp, Sumitomo Electric, Solar Frontier, Daikin, Takaoka Electric, TMEIC, Toshiba, Nissin Electric, Panasonic SS Japan, Panasonic SS infrastructure, Hitachi, Fujitsu, Fuji Electric, Mitsubishi Motors, Mitsubishi Electric, Meidensha Electric Power companies Hokkaido, Tohoku, Tokyo, Chubu, Kansai, Hokuriku, Chugoku, Shikoku, Kyusyu, Okinawa

About "The Demonstration Project of Forecast Technologies for Photovoltaic Generation" R&D Items 1. PV generation estimation methodology 2. PV generation forecast methodology Project organization The University of Tokyo Prof. Kazuhiko OGIMOTO (Project Leader) ITOCHU Techno-Solutions Corporation Solar Frontier K.K. Japan Weather Association Hitachi, Ltd. Mitsubishi Electric Corporation Central Research Institute of Electric Power Industry Ministry of Economy, Trade and Industry project From 2011 To 2013(3Years) Hokkaido Electric Power Co., Inc. Tohoku Electric Power Co., Inc. Tokyo Electric Power Co., Inc. Chubu Electric Power Co., Inc. Hokuriku Electric Power Co., Inc. Kansai Electric Power Co., Inc. Chugoku Electric Power Co., Inc. Shikoku Electric Power Co., Inc. Kyushu Electric Power Co., Inc. Okinawa Electric Power Co., Inc. 52