Generalized Power Load Modeling Based on Dynamic RBF Neural Network

Abstract

Due to influence of distributed generationon synthetic load characteristics under new situation, a new method of generalized load modeling is proposed based on dynamic radial basis function (RBF). Differential changingprocessesof load power are described with dynamic RBF neural network, deeply revealing dynamic characteristics of generalized power load. The weights of the neural network are updated dynamically with state estimation error, and the weights of neurons not satisfying persistence of excitation are limited, making the parameters of dynamic RBF neural network model theoretically converge to optimality. Data of simulation platform and actual measurement are tested respectively, and results show effectiveness of the proposed method.