Experimental analysis of forecasting solar irradiance with Echo state networks and simulating annealing

Abstract

The solar energy is a well alternative for covering the high electrical demand, and it starts to be integrated into the energetic grid infrastructure. High forecast accuracy can help in the management of industrial strategies.We present an approach that combines the potential of a Neural Network named Echo State Networks (ESN) and a wellknown optimisation technique named Simulating Annealing (SA). We use the SA technique for selecting the meteorological variables relevant in the forecasting task and the ESN as forecasting model. We present the results evaluating our approach on a public dataset.