Co-evolution Algorithm for Parameter Optimization of RBF Neural Networks for Rainfall-Runoff Forecasting

Abstract

In this paper, an effective co-evolution algorithm strategy is presented to optimize the parameters of Radial Basis Function Neural Networks by incorporating the metropolis process of Simulated Annealing (SA) into the movement mechanism and parallel processing of Particle swarm optimization (PSO), namely HPSOSA, for rainfall-runoff forecasting model. Firstly, this paper is constructed the co-evolutionary algorithm, the HPSOSA algorithm is combined the advantage of fast computation of PSO, and the advantage of SA searching in the direction of the global optimum solution, helping PSO jump out of local optima, avoiding into the local optimal solution early and leading to a good solution quality. Secondly, the co-evolution algorithm is used to optimize the structures and parameters of RBF neural networks, namely HPSOSA-RBFNN. Finally, The developed HPSOSA-RBFNN model is being applied for daily rainfall-runoff forecasting in Liuzhou of Guangxi, China. The performance of HPSOSA is compared to pure PSO in these Basis Function Neural Networks design problems, showing the co-evolution algorithm strategy is more effective global search ability and avoid falling into local solution. Experimental results reveal that the predictions using the proposed approach are consistently better than those obtained using the other methods presented in this study in terms of the same measurements. Therefore, the HPSOSA-RBFNN model is a promising alternative for rainfall-runoff forecasting tool.