An adaptive neural network prediction for nonlinear parabolic distributed parameter system based on block-wise moving window technique

Abstract

This paper proposes an efficient adaptive artificial neural network (ANN) model for nonlinear parabolic distributed parameter systems (DPSs) with changes in operating condition. To obtain the complex spatiotemporal dynamics of DPS, the ANN model is updated via applying block-wise recursive formula. The improved group search optimization (IGSO) approach is proposed to optimize the connection weights and thresholds of the ANN to solve the problem of falling into the local optima. Meanwhile, when the number of the new data does not reach the threshold of the block-wise, the ANN does not need to update. And, the predictive output consists of the ANN model predictive output and the compensated output obtained from the real time predictive errors by recursive least squares method. The proposed method can effectively capture the slowly changing of process dynamics and decrease the computational cost. Simulations are presented to demonstrate the accuracy and effectiveness of the proposed methods.