Data Drive Model of Permanent Magnet Synchronous Motor Direct Drive Pump Based on Improved Sparrow Search Algorithm-BPNN

Effective modeling of the structure and operating parameters versus the performance of electric coolant pumps is essential for its optimal design. However, the traditional finite element analysis is computationally expensive, and the theoretical model is limited to the separated motor part. In this study, a hybrid prediction model is proposed by combining an improved sparrow search algorithm with a back propagation neural network. The tent chaotic mapping and dynamic weighting methods are utilized to improve the search diversity and global search capability of the standard sparrow search algorithm. All experiments are repeated 30 times for taking the average value to ensure the validity of the experiments. The experimental results show that the RMSE of the proposed model for predicting output power is only 3.474w and the RMSE for predicting overall efficiency is only 0.611%, which can precisely predict the energy performance of electric cooling pumps under various operating conditions. The precision of the proposed model is higher than that of traditional regression models and back propagation neural networks, and the time cost is much lower than that of finite element analysis. Subsequent studies can use this model for multi-objective optimization of various servo-integrated pumps.