New daily global solar irradiation estimation model based on automatic selection of input parameters using evolutionary artificial neural networks

Abstract

Assessment of solar potential over a location of interest is introduced as an important step for the successful planning of solar energy systems (photovoltaic or thermal). The aim of this paper is to develop an automatic selection of ANN inputs from available parameters set for the estimation of horizontal daily global solar irradiation in the city of Fez (Morocco). This selection is carried out using the evolutionary artificial neural networks (EANN) with different genetic operators including selection, crossover and mutation taking into account the coding of ANN as an individual in the pool of evolution. The study includes also an exploration of the effect of maximum generations, population size, cutting points positions/nature and the mutation percentage on the performance. In order to validate the efficiency of the proposed EANN model, comparative study is established using an implemented heuristic search and forward stepwise regression models. To compare the performances in term of estimation of the proposed model, three temperature-based and K Nearest Neighbors (K-NN) models were implemented and tested in our case study. Results show that the proposed model, not only achieves good solar irradiation estimation performance but also allows an automatic selection of the best input parameters for the ANN with lower population size and maximum generation ensuring good computation time.