Parameter Identification of Permanent Magnet Synchronous Motor Based on Long Short Term Memory Neural Network

The traditional BP neural network can identify the parameters of permanent magnet synchronous motor, but the identification accuracy of the algorithm is poor, and the dynamic identification is difficult to meet the needs. Based on the identification model of negative sequence current injection, this paper uses long short-term memory network ( LSTM ) to identify the four parameters of permanent magnet synchronous motor statically and dynamically, and discusses the influence of negative sequence current on the identification results. This method improves the identification accuracy and multi-machine identification accuracy by introducing historical data for state update and training. The simulation results show that the method has achieved good static and dynamic identification results, and has strong engineering practical significance.