PROSOL: Integrated model for solar energy forecasting (2016-2019)

 

Aumentar Tamaño del texto Disminuir Tamaño del texto

Project (ENE2014-56126-C2) funded by Ministry of Economy and Competitivity (Spain)

A key issue to increase the competiveness of the solar energy and to increase their share in the electric systems is the improvement in the reliability of the solar energy forecasts. Along the last years solar resources forecasting methodologies have showed a notable development. Particularly, a wide range of forecasting methodologies have been developed, with very different characteristics as the spatial and temporal resolution or their forecasting horizon. Nevertheless, reliability of these forecasts is still limited. In addition, there have been scarce efforts to combine the different forecasting methodologies and, therefore, take advantage of the eventual synergies. Furthermore, few works have conducted neither comprehensive evaluation of solar power forecast nor probabilistic analyses of these forecasts.

This projects aims to contribute to explore the former issues that, we think, will improve the reliability of the forecast and provided an added value for the society. Notably, this project aims to provide an integrated solar power forecasting tool, able to forecast the power of both PV and thermoelectric solar plants in time horizons ranging from minutes to one month. The tool will be developed based on evolutionary computation and machine learning techniques. The hypotheses are that 1) the combination of different forecast may improve the reliability and that 2) the former techniques may provide an optimum combination of these forecasts. The achievement of the objectives of the project will need a close collaboration of two groups, one with a background on solar radiation forecasting and the other on artificial intelligence and machine learning techniques.

In a first part of the project, an improvement of the current solar radiation forecasting methodologies (bases on sky camera, satellite images and numerical weather models) is aimed. The improvement is focused in the increment of the forecasts horizons of the different techniques. This will allow the forecast horizons of different independent forecasts to overlap and, therefore, will allow these independent forecasts to be combined in the integration tool.

In a second step of the project a comprehensive data base of both forecasts and validation data will be generated. This database will be used for the development of the integration tool.

Finally, the output of the tool will be used as input of solar plant models in order to generate solar power forecasts. Probabilistic forecasts will be generated.

As a result of the project, an overall increment in the reliability of the solar power forecasts is expected. We expect also that this increment may lead to improvement in the competitiveness of the solar energy through: 1) the improvement in the plant managing, facilitating the decision making in the electricity market and the 3) facilitating the grid integration

 

 

Groups

MATRAS Group  Modelización de la ATmósfera y RAdiación Solar. Universidad de Jaén.

MATRAS Group

EVANNAI Group Evolutionary Algorithms and Neural Networks and Artificial Intelligence. University Carlos III de Madrid (Spain).

EVANNAI Group

Problems

Machine Learning Cloud Classification (UJAEN+UC3M) 

Machine Learning Cloud Classification (UJAEN+UC3M)

Integration of Predictions to Improve Solar Forecasting (UJAEN+UC3M) 

Integration of Predictions to Improve Solar Forecasting (UJAEN+UC3M)

Probabilistic Forecasting (UC3M) 

Probabilistic Forecasting (UC3M)

Publications

  • Journal articles:

     

    • J. Huertas-Tato, F.J. Rodríguez-Benítez, C. Arbizu-Barrena, R. Aler-Mur, I. Galvan-Leon and D. Pozo-Vázquez. Automatic cloud type classification based on the combined use of a sky camera and a ceilometer. Geophysical Research – Atmospheres. In press. (IF=3.454)
    • Aler, R., Valls, J. M., Cervantes, A., and Galván, I.M. Multi-objective evolutionary optimization of prediction intervals for solar energy forecasting with neural networks. Information Sciences, 418, 363-382. 2017. (IF=4.832)
    • Martín, R., Aler, R., & Galván, I. M. A filter attribute selection method based on local reliable information. Applied Intelligence, 1-11, 2017. (IF=1.215)
    • Aler, R., Galván, I. M., Ruiz-Arias, J. A., & Gueymard, C. A. Improving the separation of direct and diffuse solar radiation components using machine learning by gradient boosting. Solar Energy, 150, 558-569, 2017. (IF=4.018)
    • Martin, R., Aler, R., Valls, J. M., and Galvan, I. M. Machine learning techniques for daily solar energy prediction and interpolation using numerical weather models. Concurrency and Computation: Practice and Experience. 2016. 28:1261–1274. (IF=0.997)
    • C Arbizu-Barrena, JA Ruiz-Arias, FJ Rodríguez-Benítez, D Pozo-Vázquez, J. Tovar-Pescador. Short-term solar radiation forecasting by advecting and diffusing MSG cloud index. Solar Energy 155, 1092-1103, 2017
    • FJ Santos-Alamillos, DJ Brayshaw, J Methven, NS Thomaidis, JA Ruiz-Arias, D Pozo-Vázquez,. Exploring the meteorological potential for planning a high performance European Electricity Super-grid: optimal power capacity distribution among countries. Environmental Research Letters. 2017
    • JA Ruiz-Arias, CA Gueymard, S Quesada-Ruiz, FJ Santos-Alamillos, D Pozo-Vázquez. Bias induced by the AOD representation time scale in long-term solar radiation calculations. Part 1: Sensitivity of the AOD distribution to the representation time scale. Solar Energy 137, 608-620. 2016
    • JA Ruiz-Arias, CA Gueymard, FJ Santos-Alamillos, S Quesada-Ruiz, D Pozo-Vázquez. Bias induced by the AOD representation time scale in long-term solar radiation calculations. Part 2: Impact on long-term solar irradiance predictions. Solar Energy 135, 625-632. 2016
    • JA Ruiz-Arias, CA Gueymard, FJ Santos-Alamillos, D Pozo-Vázquez. Worldwide impact of aerosol’s time scale on the predicted long-term concentrating solar power potential. 2016. Scientific reports 6, 30546

     

     

  • Conferences:

     

    • Ruben Martin-Vazquez, Javier Huertas-Tato, Ricardo Aler and Ines M. Galvan. Studying the effect of measured solar power on evolutionary multi-objective prediction intervals. International Conference on Intelligent Data Engineering and Automated Learning IDEAL 2018, LNCS 11315
    • Francisco Rodriguez-Benitez, Javier Huertas-Tato, Clara Arbizu-Barrena, Ricardo Aler-Mur, Ines Galvan-Leon, and David Pozo-Vazquez. Evaluation of a short-term solar radiation ensemble forecasting system in the Iberian Peninsula. EMS Annual Meeting Abstracts. Vol. 15, EMS2018-272
    • Huertas-Tato, J., Aler, R., Rodriguez-Benitez, F. J., Arbizu-Barrena, C., Pozo-Vazquez, D., & Galvan, I. M. Predicting Global Irradiance Combining Forecasting Models Through Machine Learning. International Conference on Hybrid Artificial Intelligence Systems. Lecture Notes in Computer Science book series (volume 10870), pp. 622-633, Springer Cham. 2018
    • R. Martín-Vázquez, R. Aler and I.M. Galván. A Study on Feature Selection Methods for Wind Energy Prediction14th International Work-Conference on Artificial Neural Networks. June 2017
    • J. Huertas, J. Rodríguez-Benítez, D. Pozo R. Aler, Inés M. Galván. Genetic programming to extract features from the whole-sky camera for cloud type classification. International Conference on Renewable Energies and Power Quality. April 2017
    • C Arbizu-Barrena, JA Ruiz-Arias, FJ Rodríguez-Benítez, D Pozo-Vázquez, J. Tovar-Pescador. WRF advection and diffusion of MSG cloud estimates for short-term solar radiation forecasting.16th EMS / 11th ECAC (Trieste, Italia, 11/09/2016 – 16/09/2016)
    • FJ Rodríguez-Benítez, C Arbizu-Barrena, JA Ruiz-Arias, , D Pozo-Vázquez, J. Tovar-Pescador. Cloud nowcasting based on the combined use of MSG cloud estimates and the WRF NWP model. 16th EMS / 11th ECAC (Trieste, Italia, 11/09/2016 – 16/09/2016)