Torque Estimation of Permanent Magnet Synchronous Motor (PMSM) Using 1D Convolutional Neural Network

Traditionally, different current and voltage sensors are used to find out the values of differential flux and they are then used to calculate the value of torque. This dependency on the accuracy of sensors, their limited range, and the increment in cost and size of the drive set off the torque calculation to a larger process. To reduce this burden, this paper proposes the estimation of torque values based on the sensor's measured values of direct electrical quantities, instead of contemplating through complex mathematical equations. The model gets trained based on the permanent magnet synchronous motor (PMSM) dataset provided to it and converges accurately. This paper proposed a 1D convolutional neural network (CNN) for torque and temperature prediction using a PMSM. To evaluate the performance of the proposed model R2 score and adjusted R2 score values were evaluated. The performance of the proposed model was found to be more accurate as compared to the other existing methods.