Time-delay prediction method based on improved genetic algorithm optimized echo state networks

Abstract

In a networked control system, the time-delay has random and nonlinear characteristics, make the stability of system is hard to ensure. It need the controller in system can accurately predict time-delay. So the precise time-delay prediction of networked control system is an important factor in ensuring the stability of the control system. The echo state networks has good predictive ability on nonlinear time series, it is suitable for predict the time-delay. But parameters of echo state networks learning algorithm has a great influence on the prediction accuracy. An improved genetic algorithm is proposed for parameters optimization of echo state networks. The simulation results show that the prediction accuracy of the predictive method in this paper is higher than the conventional predictive model such as auto regressive and moving average (ARMA) model, least square support vector machine (LSSVM) model and Elman neural network.