PV power forecasting improvement by means of a selective ensemble approach

Research dealing with renewable energy sources is focusing on the possibility to forecast the daily output power profile of a given power system. In particular, photovoltaic has gained more interest because it is relatively easy to install but the power output shows high level of variability, strongly depending on meteorological conditions. The present work is devoted to improve the accuracy of the photovoltaic power prediction. For this reason an Hybrid Artificial Neural Network based system is used to the scope and in order to reduce the difference between the forecast and the daily measured power profile, a novel criterion based on night hours forecast is proposed. The implemented methodologies have been applied and validated throughout the work, emphasizing their partial contribution. The simulations and results are obtained on the basis of a photovoltaic module recordings of the year 2017 in Solar Tech Lab in Milano, Italy. Weather forecasts and real hourly photovoltaic power data are made available to the readers.

[1]  Francesco Grimaccia,et al.  Hybrid model analysis and validation for PV energy production forecasting , 2014, 2014 International Joint Conference on Neural Networks (IJCNN).

[2]  Sonia Leva,et al.  Physical and hybrid methods comparison for the day ahead PV output power forecast , 2017 .

[3]  Marco Mussetta,et al.  PV Module Fault Diagnosis Based on Microconverters and Day-Ahead Forecast , 2019, IEEE Transactions on Industrial Electronics.

[4]  R. Urraca,et al.  Review of photovoltaic power forecasting , 2016 .

[5]  Alessandro Niccolai,et al.  Computational Intelligence Techniques Applied to the Day Ahead PV Output Power Forecast: PHANN, SNO and Mixed , 2018, Energies.

[6]  Lars Kai Hansen,et al.  Neural Network Ensembles , 1990, IEEE Trans. Pattern Anal. Mach. Intell..

[7]  Marco Mussetta,et al.  Intelligent Approach to Improve Genetic Programming Based Intra-Day Solar Forecasting Models , 2018, 2018 IEEE Congress on Evolutionary Computation (CEC).

[8]  Aie.  World Energy Outlook 2017 , 2017 .

[9]  Igor V. Tetko,et al.  Neural network studies, 1. Comparison of overfitting and overtraining , 1995, J. Chem. Inf. Comput. Sci..

[10]  M. Brenna,et al.  Dynamic analysis of a new network topology for high power grid connected PV systems , 2010, IEEE PES General Meeting.

[11]  James L. McClelland,et al.  Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations , 1986 .

[12]  Carlos F.M. Coimbra,et al.  History and trends in solar irradiance and PV power forecasting: A preliminary assessment and review using text mining , 2018, Solar Energy.

[13]  M. Milligan,et al.  Integrating Variable Renewable Energy: Challenges and Solutions , 2013 .

[14]  Francesco Grimaccia,et al.  Validation of ANN Training Approaches for Day-Ahead Photovoltaic Forecasts , 2018, 2018 International Joint Conference on Neural Networks (IJCNN).

[15]  A. Mellit,et al.  Day-Ahead Photovoltaic Forecasting: A Comparison of the Most Effective Techniques , 2019, Energies.

[16]  Francesco Grimaccia,et al.  Artificial Intelligence Forecast of PV Plant Production for Integration in Smart Energy Systems , 2012 .

[17]  Marco Mussetta,et al.  Metaheuristic Algorithm for Photovoltaic Parameters: Comparative Study and Prediction with a Firefly Algorithm , 2018 .