Asynchronous motor resistance thermal drift prediction and compensation strategy based on sparse search optimization of backpropagation (BP) neural network algorithm

This paper proposes a compensation strategy for thermal drift in motor resistance to enhance the control precision of induction motors. The compensation strategy primarily utilizes a sparse search algorithm (SSA) to adjust the initial parameters of a backpropagation (BP) neural network, establishing a refined thermal drift prediction and compensation model (SSA-BP). An 18 kW asynchronous motor is used as the test sample, with stator temperature and phase resistance data collected for model training and testing. The resistance prediction performance of the SSA-BP model is compared with that of BP and particle swarm optimization-BP models and the algorithm models are evaluated using metrics such as R², root mean square error, mean bias error, mean square error, and mean absolute percentage error. The comparison results demonstrate that the SSA-BP algorithm model exhibits higher prediction accuracy and generalization capability in the prediction of motor resistance thermal drift. Finally, by employing the SSA-BP thermal drift prediction compensation strategy in motor control, the tracking performance of d-axis and q-axis currents and speed feedback is superior to that before compensation, thereby optimizing the motor control performance.