4 th INTERNATIONAL CONFERENCE ON ENERGY & METEOROLOGY A methodology for DNI forecasting using NWP models and aerosol load forecasts AEMET National Meteorological Service of Spain Arantxa Revuelta José Luis Casado-Rubio María Postigo Isabel Martínez 27-29 June 2017 Bari, Italy
1) Introduction 2) Methodology 3) Results 4) Summary and future work 5) References
1) Introduction 2) Methodology 3) Results 4) Summary and future work 5) References
1 - Introduction SOLAR ENERGY IN SPAIN Provides ~5% of the electricity consumed in mainland Spain (REE 2015) Concentrating solar power plants (CSP) DNI Photovoltaic power plants GHI Currently 50 operative CSP plants in Spain (http://www.protermosolar.com/proyectos-termosolares/ mapa-de-proyectos-en-espana/) REE, 2015. The Spanish Electricity system 2014. Red Eléctrica de España. Tech. Rep., 150 pp. [Available online at http://www.ree.es/sites/default/files/downloadable/the_spanish_electricity_system_2014_0.pdf] 1
1 - Introduction Factors which influence surface solar irradiance values: CLOUDS GASES AEROSOLS ALBEDO Aerosol loading is a critical parameter in the Mediterranean and northern Africa, causing 30% of total DNI extinction reaching even 100% on dust outbreak events (Wittmann et al, 2008) Wittmann, M. Et Al. Case studies on the use of solar irradiance forecast for optimized operation strategies of solar thermal power plants. IEEE J- STARS 1(1), 18-27.(2008) 2
1 - Introduction TEMPORAL SCALE: Very Short < 6h : nowcasting (sky cameras, satellite images, neural networks, cloud-tracking image analysis...) Short 6h 48h : High resolution NWP models are the best tool (Perez et al, 2010) Medium 3 days 10 days : High resolution NWP models or ensembles Seasonal 1 month 6 months : Probabilistic forecasting (ensembles) Long years : climatology Perez, R., Et Al, 2010: Validation of short and medium term operational solar radiation forecasts in the US. Sol. Energy, 84, 2161 2172 3
1 - Introduction This study: Short range DNI forecasting NWP models But... Meteorological forecast models do not incorporate predictions of atmospheric aerosols yet (only use climatological values) So... Other option: Combination with radiative transfer models 4
1) Introduction 2) Methodology 3) Results 4) Summary and future work 5) References
2 - Methodology First ideas on combining models (selecting clear and cloudy situations separately) Breitkreuz et al, 2009 Our method DNI is obtained through the product of two terms: A clear-sky term calculated through libradtran, including dependency with water vapour and aerosols A clearness index. Dimensionless term in the range [0,1], including dependency with clouds. Obtained through the NWP model (Casado-Rubio Et Al, 2017) Validation with observational data from the Red Radiométrica Nacional Breitkreuz, Et Al (2009). Short-range direct and diffuse irradiance forecasts for solar energy applications based on aerosol chemical transport and numerical weather modeling. J. Appl. Meteorol. Clim., 48(9), 1766-1779 Casado-Rubio, J.L.; Revuelta, M.A.; Postigo, M.; Martinez-Marco, I.; Yagüe, C. (2017). A post-processing methodology for direct normal irradiance forecasting using cloud information and aerosol load forecasts. J. Appl. Meteorol. Clim., 56(6), 1595-160 5
2 - Methodology Process diagram Aerosols: MACC AOD550 (dust, seasalt, OM, sulphate and BC) or BSC-DREAM8b (dust) + monthly background Retrieval and interpolation 3h 1h Clear-sky DNI with libradtran Water vapour (ECMWF) Retrieval and interpolation 3h 1h DNI forecast DHI DHI clear-sky (ECMWF) Retrieval and interpolation 3h 1h Clearness index 6
2 - Methodology Black curve: Combination of the irradiance simulated by libradtran and extracted from the NWP model 7
2 - Methodology Area of study : Iberian Peninsula and Canary Islands 8 sites : Albacete, Arenosillo, Badajoz, Caceres, Madrid, Murcia, Maspalomas and Tenerife Time period : 2013 and 2014 Hourly forecasts 24h and 48h forecasts 8
1) Introduction 2) Methodology 3) Results 4) Summary and future work 5) References
3 Results: clearness index grid size RMSE Vs. Grid area 9
3 Results: clearness index grid size RMSE Vs. Grid area 9
3 Results : Annual scores 24h rrmse = RMSE DNI obs rrmse = rrmse EC rrmse cmacct rrmse EC Site Year rrmse (EC) rrmse (cmacct) (**) rrmse (cmaccd) rrmse (cdream8b) Δ(rRMSE) % (**) rrmse (pers) Arenosillo 2013 42,4 36,9 40,79 41,46 12,8 64,53 Badajoz 2013 40,7 35,5 39,23 40,01 12,7 63,26 Madrid 2013 43,2 38,7 41,29 41,66 10,4 69,64 Murcia 2013 42,5 39,2 41,87 42,7 7,8 68,81 Maspalomas 2013 56,4 50,2 52,96 55,06 11,0 60,67 Tenerife 2013 65,8 62,2 63,84 64,75 5,4 71,59 Arenosillo 2014 42,7 37,1 40,47 41,02 13,1 63,44 Badajoz 2014 47,7 42,3 46,19 46,46 11,4 70,95 Madrid 2014 43,1 39,5 41,24 41,68 8,3 70,43 Murcia 2014 41,7 37,4 39,78 41,1 10,5 65,83 Maspalomas 2014 60,8 55,8 58,37 59,43 8,3 64,46 Tenerife 2014 58,6 55,1 56,69 57,36 5,9 72,25 Average 47,9 43,4 46,1 46,9 9,6 67,7 10
3 Results : Annual scores 24h % improvement rrmse = rrmse EC rrmse cmacc rrmse EC 2013 2014 Circles: cmacct Diamonds: cmaccd 11
3 Results: Monthly scores 12
3 Results: 24h Vs. 48h forecasts The annual improvement decreases ~10% from 24h to 48h forecasts 13
3 Results: high aerosol load High aerosol load situation : AOD 500 AERONET > 85th percentile Average of: Sites: Badajoz, Madrid and Murcia Years: 2013 and 2014 ObsMean no. high AOD hours DNIfall 669,1 221 27,2 % Forecast rrmse (EC) rrmse (cmacct) (*) rrmse (cdream8b) Δ(rRMSE) % (*) Annual Δ(rRMSE) % (*) 24h 21,0 15,2 18,6 27,5 10,2 48h 22,7 17,1 20,4 25,4 9,1 14
3 Results: African outbreaks Murcia. AF outbreak from 14 to 17/6/2013 (source: SDS-WAS) 15
3 Results: African outbreaks Murcia. AF outbreak from 14 to 17/6/2013 (source: SDS-WAS) DNI drop 18 % at noon 15
1) Introduction 2) Methodology 3) Results 4) Summary and future work 5) References
4 Summary and future work Summary : Improvements in DNI forecasts have been made combining libradtran 1D + MACC aerosol forecasts with the ECMWF operative meteorological model (considering an aerosol climatology), compared to the ECMWF meteorological model alone The method proposed is relatively simple, effective and flexible. Useful in operations. Avoids the use of different methods for different meteorological situations. Can be adapted to different models. Annual improvements in rrmse are over 10% in some locations in the Iberian Peninsula and the Canary Islands The monthly evolution of the forecast improvements suggests the prelevance of the aerosol or the clouds in the processes involved in different geographic locations. Under high aerosol load situations the DNI fall is 27% in average. The improvement in rrmse is over 25% 16
4 Summary and future work Future work... DNI forecasting improvement energy production improvement Red Eléctrica de España (REE) is performing tests using production models Other NWP models Harmonie. We need obtaining DNI_clear_sky Ensemble 17
1) Introduction 2) Methodology 3) Results 4) Summary and future work 5) References
5 References REFERENCES Breitkreuz, H., Schroedter-Homscheidt, M., Holzer-Popp, T., & Dech, S. (2009). Short-range direct and diffuse irradiance forecasts for solar energy applications based on aerosol chemical transport and numerical weather modeling. Journal of Applied Meteorology and Climatology, 48(9), 1766-1779. Casado-Rubio, J.L.; Revuelta, M.A.; Postigo, M.; Martinez-Marco, I.; Yagüe, C. (2017). A postprocessing methodology for direct normal irradiance forecasting using cloud information and aerosol load forecasts. Journal of Applied Meteorology and Climatology, 56(6), 1595-1608. Perez, R., S. Kivalov, J. Schlemmer, K. Hemker Jr., D. Renn, and T. E. Hoff, (2010). Validation of short and medium term operational solar radiation forecasts in the US. Sol. Energy, 84, 2161 2172 REE, 2015. The Spanish Electricity system 2014. Red Eléctrica de España. Tech. Rep., 150 pp. [Available online at http://www.ree.es/sites/default/files/downloadable/the_spanish_electricity_system_2014_0.pdf Wittmann, M., Breitkreuz, H., Schroedter-Homscheidt, M., Eck, M. Case studies on the use of solar irradiance forecast for optimized operation strategies of solar thermal power plants. IEEE J-STARS 1(1), 18-27.(2008) 18
ACKNOWLEDGEMENTS The MACC project provided aerosol forecasts. MACC-III is a Coordination & Support Action (2014-2015) funded by the EU-Horizon 2020 Programme. It is coordinated by ECMWF and operated by a 36-member consortium, including AEMET Dust forecasts from the BSC-DREAM8b (Dust REgional Atmospheric Model) model, operated by the Barcelona Supercomputing Center (http://www.bsc.es/projects/earthscience/bsc-dream/) have been used AERONET provided observational data. We thank all the PIs and their staff for establishing and maintaining the sites used in this investigation The Red Radiométrica Nacional, managed by AEMET, provided observational data WMO Sand and Dust Storm Warning Advisory and Assessment System (SDS-WAS) (https://www.wmo.int/pages/prog/arep/wwrp/new/sand_and_dust_storm.html) provided the identification of dust events Horizon 2020 - PreFlexMS project (Predictable and Flexible Molten Salts Solar Power Plant; Grant N. 654984) funded the assistance to this conference 19
ACKNOWLEDGEMENTS The MACC project provided aerosol forecasts. MACC-III is a Coordination & Support Action (2014-2015) funded by the EU-Horizon 2020 Programme. It is coordinated by ECMWF and operated by a 36-member consortium, including AEMET Dust forecasts from the BSC-DREAM8b (Dust REgional Atmospheric Model) model, operated by the Barcelona Supercomputing Center (http://www.bsc.es/projects/earthscience/bsc-dream/) have been used AERONET provided observational data. We thank all the PIs and their staff for establishing and maintaining the sites used in this investigation The Red Radiométrica Nacional, managed by AEMET, provided observational data WMO Sand and Dust Storm Warning Advisory and Assessment System (SDS-WAS) (https://www.wmo.int/pages/prog/arep/wwrp/new/sand_and_dust_storm.html) provided the identification of dust events Horizon 2020 - PreFlexMS project (Predictable and Flexible Molten Salts Solar Power Plant; Grant N. 654984) funded the assistance to this conference 19