Physics Informed Neural Network Induction Motor Equivalent Circuit Parameter Estimation with Only Electrical Measurements

The induction motor is used in many applications across industry. It is a preferred choice because of low cost and high reliability. To control the induction motor, it is critical to know the induction motor equivalent circuit parameters. Knowing these accurately allows for more precise and efficient controllers. The accuracy constraints on these estimates is compounded in situations where sensorless control is being attempted. These parameters are normally identified by performing no load and blocked rotor tests and using various measurements to calculate the equivalent circuit parameters. There is much interest in finding these parameters without such tests because it is not always feasible to perform these tests. In this work, a physics informed neural network (PINN) is used to estimate these parameters from measured stator voltage and current with the rotor locked. The proposed method solves the inverse problem to estimate the desired parameters with high accuracy.