Power and Solar Energy Predictions Based on Neural Networks and Principal Component Analysis with Meteorological Parameters of Two Different Cities: Case of Diass and Taïba Ndiaye

The overuse of conventional resources based on fossil fuels increases the vulnerability of our environment. Faced with this effect, increasing the penetration rate of intermittent (non-polluting) energies in the electrical networks has become of paramount importance. However, this increase in the penetration rate allows on the one hand to improve the satisfaction of the producers and reduces the consumption of fossil fuels, on the other hand it is a major point of suffering for the non-smart electrical networks. In a dynamic of promoting intermittent energies while ensuring a permanent balance between consumption and production, the forecasting of these energies is an important lever. Hence, this paper studies artificial neural networks to predict the power and energy output of the Diass solar power plant in the short and medium term. Thus, the proposed approach consists in using not only the meteorological data of the city where the power plant is located, but also the data of a nearby city with a data acquisition station. The selection of the variables is done by principal component analysis (PCA). Moreover, our results were compared to the literature using only the meteorological data of the plant of implementation. The results obtained are more satisfactory with mean absolute and root mean square errors of 0.0223 KWh and 0.003 KWh respectively and a prediction accuracy of 94.57% in terms of energy and power. In terms of resource, it consumes more with simulation times varying between 1788 seconds and 2201 seconds.