Groundwater Table Prediction Based on Improved PSO Algorithm and RBF Neural Network

Abstract

Groundwater table often shows complex nonlinear characteristic. Radial basis function (RBF) neural network is increasingly used to predict groundwater table. The traditional RBF training algorithm based on gradient descent optimization method can only obtain the partial/local optimums solution sometimes. Furthermore, man-made selecting the structure of RBF neural network has blindness and expends much time. In training RBF neural network, particle swarm optimization (PSO) algorithm is presented to optimize and automatically determine the parameters and structure of RBF neural network. In order to improve traditional PSO algorithm searching capacity, linear inertia weight and chaos variation operator are presented. Study case shows that, compared with back propagation (BP) or RBF neural network, the new prediction model based on PSO and RBF neural network can greatly increase the convergence speed and precision.