ANN-based flux observer for the sensor-less control of a permanent magnet synchronous motor

The paper reports a neural network based flux observer for the sensor-less control of a three phase interior permanent magnet synchronous motor (IPMSM). The estimation of rotor position and speed at low speed range was achieved by extensive training of the ANN which is robust to variations in flux linkages. The ANN was trained extensively to overcome position- and speed-estimation errors at low speed due to the nonlinear behaviour of space vector PWM based voltage source inverter. The dynamic model of IPMSM was used. Temperature based variations in parameters have been accommodated in modeling of the machine and design of the observer. The proposed flux observer gives satisfactory performance in both the constant-torque and constant-power regions. The simulation and implementation results are shown which illustrate the effectiveness of the ANN based flux observer.