Sun to Market Solutions
S2m has become a leading global advisor for the Solar Power industry 2 Validated solar resource analysis Solcaster pro Modeling Delivery and O&M of weather stations for solar projects Solar Resource Engineering Short and very short term GHI & DNI prediction S2m Nowcasting System Marquesado Solar SL Technoeconomic Assistant
Nowcasting is a short-term and very-short weather forecasting system 3 The S2m Nowcasting system provides continuous GHI/DNI & meteo data streaming for three forecast horizons: Tool Temporal Scale Resolution Sky Camera 0 30 min 5 minutes Satelite Images 0 2h 15 minutes Numerical Model 0 48h 1 hour
The final objective of a Nowcasting system is to improve the predictability of the solar power plant production 4 Provides uncertainty factor control in real-time, supporting grid operator decisions as well as global grid operation strategies. Optimizes battery storage and improves the ramp rate control of CPV and PV power plants. Evaluates operation alternatives in view of equipment breakdowns or unexpected events (Preventive and corrective maintenance) Increases over all efficiency and dispatchability of solar power plant EXAMPLE: Managing combined seawater desalination with PV power generation Sun to Market Solutions 18 February 2014
How does s2m s Nowcasting system work? 5 S2m developed a data acquisition scheme from four different sources with posterior processing on own servers. Sources: SkyCam, Satellite, Meteo-station and Global numerical weather model Meteo Station Stereocopy Fisheye cameras Satellite Meteosat Server Global Forecast /Mesoscale Numerical Models Internet Client observation S2m observation 18 February 2014
Fisheye camera 6 S2m has deployed proprietary sky-cameras and weather stations in order to motorize the sky and measure the needed weather variables Sky images are taken and processed every 30 seconds Acciona CSP plant Palma del Rio, Spain
3D Cloud position 7 By calculating the position at sunrise and with the aim of the skycamera radiation prediction and cloud cover detection at Power plant start up are extremely improved. Stereocopy Fisheye cameras Cloud detection and segmentation applying Red/Blue filters Skycameras installed in Acciona s solar power plant, Palma del Rio I and Palma del Rio II. Stereoscopy estimates cloud height and relative position to the plant. Providing a 3D framework to improve cloud detection and short range forecastings (ongoing).
Data Acquisition Raw Data Labeling and Storing Data Processing Raw data processing and data extraction 8 Raw data have to be properly labeled and stored before the real processing and data extraction. SkyCam Images Cloud Detection Cloud Classification according optical features and height Motion Estimation Satellite Images Cloud Detection Cloud Classification according optical features and wind field. Motion Estimation Meteo-station measurements Numerical Model Automatic synoptic analysis Dynamically downscaling using mesoscale numerical models
Data Processing Irradiation Prediction Data processing allows the extraction of information about physical movement height and characteristics of the clouds to obtain a good radiation prediction 9 Segmentation Motion Estimation 15 min Extract outputs numeriacl model 45 min 240 min
Cloud segmentation allows plant operator to evaluate the cloud situation with a higher precision 10 Clear Medium cover (cyrrus type) Opaque (stratus and cumulus type)
EXAMPLE: GFS - Global longitude-latitude grid 1.0 degree resolution 2014/02/10T18z forecast for the 2014/02/11 9:00:00 GMT 11 % Nubes bajas % Nubes Altas % Nubes medias Sun to Market Solutions 18 February 2014
NWP Cloud Cover Estimates: Nesting downscaling with Numerical Weather Prediction mesoscale models and 3D representation 12 27x27km Cloud cover is obtained employing mesoscale NWP. Cloud transmisivity is obtained employing empirical methods based on climate analysis. 9x9km The cloud transmisivity is applied to state of the art Radiative transfer models to obtain the final DNI and GHI. 3D cloud scheme Meteosat SEVIRI HR channel 20/11/2013 12:00:00 WRF Fine domain 20/11/2013 12:00:00 WRF coarse Domain 20/11/2013 12:00:00
Example: WRF Downscaling 13 2013/12/04 10:00 UTC 2014/02/11 12:00 UTC Sun to Market Solutions 18 February 2014
Nowcasting system: Web based user interface 14 Our web based software allows to compare results for three different horizons with the in situ measurements from the meteo-station. User friendly interfaces
S2m Nowcasting Timeline 15 Over the last four years we have constantly improved our multi-nowcasting system 15 NWP Satellite Skycamera WRF 3.1 compilation First successful Satellite Reception First S2m Meteostation WRF Parameter's First Output analysis Heliosat III Implementation First Skycamera images WRF 3.3 Update Laine Cloud Mask Cloud segmentation New cloud parameterization Optical-flow analysis Solar Position algorithms WRF wind field integration Optical-flow algorithms New Radiation scheme First Skycamera for Acciona 3D Cloud Analysis New RTM* scheme implementation Cloud segmentation Skycamera Second camera Release 1.2 Installation in P.dR. I For stereoscopy Present day 2010 2011 2012 2013 2014 * Radiative Transfer Model
measured DNI measured DNI measured DNI measured DNI measured DNI measured DNI measured DNI measured DNI measured DNI GHI & DNI Nowcasting Results with horizon until 2 h 16 Low dispersion for both clear and cloudy days plazo: 0 min y = 1.05224*x-36.1335 RMSE: 26.3482 800 600 400 200 0 200 400 600 800 modeled DNI plazo: 45 min y = 0.904338*x+33.1746 RMSE: 37.3 1000 800 600 400 200 plazo: 15 min y = 0.983336*x-7.1439 RMSE: 31.4476 0 200 400 600 800 modeled DNI plazo: 60 min y = 0.897784*x+41.1798 RMSE: 38.5291 1000 1000 500 plazo: 30 min y = 0.930643*x+18.0264 RMSE: 34.7296 0 0 500 1000 modeled DNI plazo: 75 min y = 0.882428*x+58.2533 RMSE: 42.0865 1000 500 500 500 0 0 500 1000 modeled DNI plazo: 90 min y = 0.85634*x+72.7706 RMSE: 44.3152 1000 0 0 500 1000 modeled DNI plazo: 105 min y = 0.814956*x+100.136 RMSE: 48.3299 1000 0 0 500 1000 modeled DNI plazo: 120 min y = 0.815944*x+99.7879 RMSE: 47.7389 1000 500 500 500 0 0 500 1000 modeled DNI 0 0 500 1000 modeled DNI 0 0 500 1000 modeled DNI GHI forecasting: 90% of the cases are under 150 W/m² bias error in a 2h Satellite Forecast DNI forecasting: 75% of the cases are under 200 W/m² bias error in a 2h Satellite Forecast
DNI Nowcasting Results with horizon until 2 h 17 Real project validation (2013, South Spain)
measured GHI measured DNI GHI & DNI Nowcasting Results with horizon until 24 h 18 1000 800 plazo: 24h y=0.838244*x + 6.76005 RMSE: 38.8728%, 221.477 (W/m 2 ) Real performance of the systems during winter (cloudy days, transient situations, etc) 600 DNI GHI 400 200 0 0 200 400 600 800 1000 modeled DNI 500 plazo: 24h y=0.880233*x + 3.40041 RMSE: 26.7686%, 74.6153 (W/m 2 ) Error/Plazo 24h Error/Plazo 24h MBE (W/m²) -75.20 MBE (W/m²) -28.11 MAB (W/m²) 141.09 MAB (W/m²) 47.97 MBE (%) -15.21 MBE (%) -5.68 MAB (%) 56.29 MAB (%) 19.14 RMSE (%) 38.87 RMSE (%) 26.76 RMSE (W/m²) 221.48 RMSE (W/m²) 74.61 400 300 200 100 0 0 100 200 300 400 500 modeled GHI Error/Plazo 24h Error/Plazo 24h P75 (W/m²) 166.05 P75 (W/m²) 43.46 P90 (W/m²) 405.23 P90 (W/m²) 127.85
Nowcasting Example: PV power pant production forecast in Germany International Electricity Trader 19 S2m provides 1 and 2 day-ahead prediction for all 440.000 PV plants in Germany for each 15 min Day ahead and two day ahead forecast for PV park in Germany Cloudy winter day 2014 GHI DNI DNI Service can be provided on global level 18 February 2014
For further information please contact: 20 Madrid Head Office Science & Technology Park Avda. Gregorio Peces-Barba, 1 28918 Leganés-Madrid phone: +34.914.966.189 fax: +34.911.413.368 Bangalore New Bridge Corporate Center 777D, 100 Feet Road, Indiranagar Bangalore - 560038 phone: +91.96.6352.3948 fax: +91 80.3072.3678 Beijing Beijing Pacific Century Place 14th Floor, IBM Tower, Pacific Century Place 2A Workers Stadium Road North Chaoyang District Beijing 100027 phone: +86.10.65876464 fax: +86.10.65391060 Sunnyvale 375 N Fernwood Circle Sunnyvale, CA 95085 phone: +1 650 461 9048 fax: +1 484 313 9504 Pisa Via Giuseppe Montanelli, 100 56121 - Pisa (Italy) phone: +39 050 28529 fax: +39 050 28529 Santiago de Chile Alcántara 200, Piso 6, Las Condes, Santiago C.P. 7550159 phone: + 56228977651 Johannesburg West Tower, 2nd Floor, Nelson Mandela Square, Maude St., Sandton, Johannesburg, 2146 phone: + 27 11 083 9587