Neural Network-Based MTPA Control Strategy for IPMSMs under Temperature Variations

Abstract

This paper proposes a control algorithm for the Maximum Torque per Ampere (MTPA) operation of Interior Permanent Magnet Synchronous Motors (IPMSMs) that consider temperature variations using an Artificial Neural Network (ANN). As the temperature increases, the residual magnetic flux density of the permanent magnets decreases, leading to a reduction in the magnitude of the magnetic flux. Furthermore, even at the same temperature, the d-q axis fluxes vary depending on the current operating points due to magnetic flux saturation. These nonlinear d-q axis flux variations result in nonlinear torque variations at current operating points. ANNs are well-suited for modeling nonlinear correlations and have been widely applied across various fields. Hence, ANN is applied to MTPA operation, effectively representing the nonlinear relationship between the d axis flux and MTPA operating point variations. The proposed method utilizes a Frequency-Adaptive Observer (FAO) to estimate the d-axis flux under the present temperature. Using the ANN, the MTPA operating point is adjusted by accounting for the d-axis flux variations caused by temperature changes and magnetic flux saturation. The proposed algorithm was validated through experiments on an 11 kW IPMSM.