An Electromagnetic Torque Calibration Algorithm Based On Multiple Linear Regression Principle

Aiming at the problems of high cost and large amount of computation in real-time online calibration of PMSM’s electromagnetic torque, a multi-linear regression offline calibration algorithm based on speed partition was designed. First, it analyzes the reasons for the electromagnetic torque error. Then, based on the principle of multiple linear regression, an offline calibration model for electromagnetic torque in different speed ranges of the motor is established, and the model is solved using the linear least squares method. Secondly, the model was instantiated through motor bench tests, and the torque errors of calibration models with different elements and orders were compared. Taking into account factors such as the accuracy of electromagnetic torque calibration, the complexity of fitting equations, the computational ability and engineering implementation of hardware systems, different calibration models were selected at different speed ranges to complete motor torque calibration. Finally, the calibration algorithm was applied to the vector control system, further verifying the effectiveness of the algorithm. This algorithm does not rely on real-time parameter observation of the motor, and only designs calibration coefficients based on the motor speed segmentation, achieving electromagnetic torque calibration in the full speed range. The algorithm calculation is simple and easy to apply in engineering.