Prediction of chaos in permanent magnet synchronous motor using weighted delayed feedback reservoir computing with multi-head attention

Abstract Permanent Magnet Synchronous Motors (PMSM) exhibit chaotic behavior due to their complex nonlinear and time-varying characteristics, posing significant challenges to system stability. Predicting PMSM chaos is crucial for developing targeted control strategies in advance. To address this issue, this paper proposes a novel reservoir computing (RC) model, termed Weighted Delayed Feedback Reservoir Computing with Multi-Head Attention (WDF-MHA-RC), for predicting chaotic behavior in PMSM. Specifically, the reservoir computing framework first extracts the dynamic features of the PMSM system through a weighted delayed feedback mechanism. Then, a multi-head attention mechanism is introduced to enhance the model’s ability to capture complex dependencies in time-series data. The extracted features from the trained reservoir states and multi-head attention mechanism are used for nonlinear mapping to predict the future states of PMSM. To validate the effectiveness of the proposed method, numerical simulations are conducted on both single PMSM (SPMSM) and two coupled PMSMs The results demonstrate that WDF-MHA-RC accurately predicts the chaotic evolution trajectory of SPMSM and effectively captures the dynamic interactions between coupled motors in two coupled PMSMs Compared with traditional and related RC models, WDF-MHA-RC significantly improves prediction accuracy and provides deeper insights into the nonlinear dynamics of PMSM systems.