A New Approach for Photovoltaic Power Prediction Based on Chaos Theory

Variability represents the main problem related to renewable energies. Their intermittent nature constitutes the greatest obstacle to their complete adoption. For this reason, and despite the efforts made in this field, renewable energies are not yet able to replace fossil fuels, hence the importance of prediction. This work proposes a new method of photovoltaic energy prediction, founded on dynamic behavior analysis. This approach is to use phase space reconstruction, to build the input of the neural network in order to take into account the dynamics of the system in the forecasting process. Then, to improve the precision, we introduce the wavelet transformation. We tested this approach on photovoltaic production of the Faculty of Science and Technology of Beni Mellal, Morocco. Finally, the comparison between predictions and actual observations confirmed the effectiveness of our approach.