Sensorless closed-loop speed estimation of BLDC motor drive using physics-informed NN algorithm

Brushless Direct Current (BLDC) motors are predominantly used in Electric Vehicle (EV) applications because of their higher efficiency, higher torque-to-weight ratio, lower maintenance requirements and capability to operate in a wide speed range. However, conventional sensor-based and traditional sensorless estimation methods experiences inaccurate performance under high-speed operating conditions because of parameter variations and nonlinear system dynamics. To overcome these limitations, this article implemented an improved sensorless Extended Kalman Filter (EKF)-based closed-loop speed estimation algorithm using Machine Learning (ML) based Physics-Informed Neural Network (PINN) algorithm specifically designed for high-speed applications. Performance of the implemented sensorless speed estimation scheme is experimentally validated and compared with conventional algorithms using the LAUNCHXL-F28379D development board. The experimental results indicates that the implemented method has a maximum steady-state error of 7 at 500 rpm and 8.95 at 1000 rpm between the actual motor speed and the reference speed. Furthermore, during higher operating speeds of 1500 rpm and 1600 rpm, the motor precisely tracks the reference speed.