General regression neural network forecasting model based on PSO algorithm in water demand

Abstract

There is a complicated non-linear relationship between the factors and water demand. General regression neural network (GRNN) was adopted to model the non-linear relationship in the study. The prediction performance of GRNN can vary considerably depending on smoothing parameter. The optimal smoothing parameter is usually determined empirically based on trial-and-error. Particle swarm optimization (PSO) algorithm, to improve GRNN prediction performance, was employed to optimize GRNN and determine an optimal value of smoothing parameter. At the same time, linear inertia weight and chaos variation operator are presented to improve traditional PSO algorithm searching capacity. GRNN forecasting model based on PSO algorithm was used to water demand in Yellow River Basin. The result shows that, compared with Back propagation based on Genetic algorithm model and GRNN based on Genetic algorithm prediction model, the new prediction model is reasonable.