An Optimal Control Method of IPM Motor Based on Multilayer Neural Network

The built-in permanent magnet (IPM) motor is a type of motor that positions permanent magnets within the rotor. It combines reluctance torque with magnet torque, making it extensively utilized in household appliances, industrial equipment, automobiles, and other domains. Neural networks are computational models with non-linear characteristics and potent learning capabilities, particularly adept at solving identification problems in complex, non-linear, uncertain, and ambiguous systems. During the network training process, the network progressively optimizes its performance through backward propagation and weight adjustment. This paper proposes a learning method that leverages a multi-layer neural network structure to optimize the efficiency of IPM motors, capitalizing on the characteristics of neural networks. The control method achieves optimal efficiency control by deriving and compensating the d-axis and q-axis current phase angles in real time so that the IPM motor can maintain efficient performance under various conditions. This method can effectively control the motor and improve control efficiency without relying on identification or estimation methods, even when the parameters of the controlled object change. The method's effectiveness is validated through comparative simulations with two traditional control techniques.