Output PV power prediction using an Artificial Neural Network in Casablanca, Morocco

Optimal use of renewable energy requires its characterization and prediction to size detectors or estimate the potential of power plants [20-21]. In terms of prediction, electricity suppliers are interested in different horizons to manage power plants and predict their production [1-2]. This paper proposes a model for predicting the output power in photovoltaic (PV) panels installed on the rooftop of the Ben m'sik faculty at Hassan II University, Casablanca, Morocco, and this model is based on a multilayer perceptron (MLP) model. In this work, different combinations of weather variables were used to develop this model and for validate the proposed model results different practical measurement methods are used, such as mean square error (MSE), mean absolute error (MAE), correlation (R) and coefficient of determination (R2). The determination coefficient of the proposed model is 0.98501 with an RMSE value of 30.663. The proposed model was tested on new data, the results showed that the model works with a good preferment and that the prediction quality depends on the time of year with a determination coefficient of 0.9972, 0.9856, 0.9487 and 0.9942 for summer, autumn, winter and spring respectively.