Chaos prediction of motor based on the integrated method of convolutional neural network and multi-reservoir echo state network

Abstract

Permanent magnet synchronous motor (PMSM) can exhibit chaotic behaviors detrimental to their regular operation in practical applications. To accurately predict the chaotic state of PMSM, this paper proposes a C-MRESN method based on the combination of convolutional neural network (CNN) and multi-reservoir echo state network (MRESN). The significant advantage of C-MRESN is that it combines the advantages of the two models, which can capture the complex temporal and spatial information from nonlinear time series and retain these features for prediction. In addition, this work uses the L-BFGS-B optimization algorithm to optimize the training process of C-MRESN and significantly improve the prediction accuracy of C-MRESN. By comparing the prediction experimental results with six other machine learning models, C-MRESN shows the minor prediction error and the most extended accurate prediction range. The root mean square error (MSE) of the 2000-step prediction results of C-MRESN for the three PMSM variables, [Formula: see text] and [Formula: see text] can reach [Formula: see text], [Formula: see text] and [Formula: see text], respectively. The experimental results substantiate that the C-MRESN is an effective and advanced method for the chaos prediction of PMSM.