Predicting the output of a PV plant

Energy market players (investors, power producers, grid operators, consumers, etc.) are facing potential challenges such as the growing demand for energy, new patterns of energy consumption, the integration of (intermittent) renewable energy sources into power grids and the evolution of power grids. This book investigates the possibility of predicting the production of a self-consuming photovoltaic installation by artificial neural networks. We cross-compared two neural network architectures (looped and unlooped) with respect to multivariate regression in order to have an efficient and reliable tool for predicting the production of a PV installation based on meteorological data (sunshine and ambient temperature). To do so, we used monitoring data of a plant over a 72-day period to build, train and test two neural network topologies (looped and unlooped) which are trained with the Levenberg-Marquardt algorithm.