Papier accepté pour '25th European Photovoltaic Solar Energy Conference and Exhibition

Un nouveau papier auquel j'ai collaboré vient d'être accepté au congrès suivant "25th European Photovoltaic Solar Energy Conference and Exhibition", à Valence, Espagne. 

Title of the paper : Optimization of an artificial neural network (ANN) dedicated to the daily global radiation and PV plant production forecasting using exogenous data 

Authors : Cyril Voyant, Marc Muselli, Christophe Paoli, Marie-Laure Nivet

Abstract :  In this paper, we present an application of Artificial Neural Networks (ANNs) in the renewable energy domain. We particularly look at the Multi-Layer Perceptron (MLP) network which has been the most used of ANNs architectures both in the renewable energy domain and in the time series forecasting. We have used a MLP and an ad-hoc time series preprocessing to develop a methodology for the daily prediction of global solar radiation on a horizontal surface and a 1.175 kWp PV plant production. Different forecasting methods are compared: a naïve forecaster like persistence, an ANN with preprocessing using only endogenous inputs and an ANN with preprocessing using endogenous and exogenous inputs (like temperature, pressure, nebulosity, insulation, wind speed and direction etc.). The endogenous case is easily computed: the use of Partial Auto Correlation Factor (PACF) allows to optimize the number of lag time to consider. For the exogenous variables, we have applied a Pearson correlation coefficient method to optimize the number of considered input neurons. Although intuitively the use of meteorological data in the input layer of the MLP can only increase the quality of prediction, the obtained results are relatively mixed. The use of exogenous data generates a decrease of nRMSE between 0.5% and 1% for the both studied locations. The absolute error (RMSE) is decreased by 52 Wh/m²/day in the simple endogenous case and 335 Wh/m²/day for the persistence forecast. The results are similar to the concrete case of a tilted PV wall (1.175 kWp), endogenous and exogenous data ANN inputs allow decreasing the nRMSE by 1% on a 6 months-cloudy period for the DC power production (January-June). Moreover, the use of exogenous data shows an interest only in cloudy period (winter season). In summer, endogenous data as inputs on a preprocessed ANN are sufficient. By comparison to a naïve forecaster as persistence, an ANN with endo and exogenous data improves the DC electrical power energy prediction by 9% (nRMSE). Next step of this work will drive to shorter horizons; power predictor of meteorological data should have a greater impact.

Keywords: Time Series Forecasting, Preprocessing, Artificial Neural Networks, PV Plant Energy Prediction.

Sources :  http://www.photovoltaic-conference.com