Electromagnetic Resistance Model Based Maximum Efficiency Control for IPMSM with Reduced Experimental Cost

Maximum efficiency (ME) control with look-up table (LUT) is commonly utilized in drive applications to achieve energy conservation, due to its fast dynamic response. However, existing methods to obtain ME LUTs need precise machine characteristics or parameters across the entire operating range, causing extensive experimental costs. Unfortunately, limited effort has been made to reduce such cost. Furthermore, the traditional iron-loss resistance model usually encounters rank-deficiency issues during parameter calibration, preventing the unique determination of parameters. To address these challenges, this paper develops an optimized ME LUT control. It leverages an improved electromagnetic resistance model (ERM), integrated with neural networks (NNs) to describe variable model parameters, for calculating ME LUTs. The ERM addresses the rank-deficiency problem by integrating copper and iron losses into a unified resistance. Besides, a selection criterion for operating conditions is proposed to collect ERM parameters for training NNs. The describing accuracy of the NNs, trained with different amounts of variable parameters, can be estimated using a derived boundary. By striking a balance between the amount of training data and the accuracy of NNs, the proposed approach can decrease the experimental costs while slightly compromising the performance of calculated ME LUTs. The proposed method has been validated experimentally.