Development of GRBFN with global structure for PV generation output forecasting

Abstract

This paper presents a new method for forecasting of PV generation output. The output of PV systems is significantly affected by the weather conditions. As a result, forecasting of PV systems generation output is one of the most difficult time series forecasting. However, power system operators require more accurate prediction model to deal with power system operation such as economic load dispatching, unit commitment, etc. The proposed method makes use of a hybrid intelligent system that consists of Generalized Radial Basis Function Network (GRBFN), Deterministic Annealing (DA), and Evolutionary Particle Swarm Optimization (EPSO). GRBFN is one of artificial neural networks (ANNs) that provide good performance with complicated nonlinear time series. DA is used for determining the center and width of radial basis functions in GRBFN. EPSO is useful for optimizing weights between neurons in GRBFN to improve the performance from a standpoint of global optimization. Also, this paper applies the weight decay method to the cost function to avoid overfitting for learning data of nonlinear complicated data. The proposed method is successfully applied to real data of the PV system in Japan.