Online parameter estimation of permanent magnet synchronous machines by means of window LS optimization

Industrial applications, especially automotive ones should be robust and cheap. Both properties can be improved by using model based state estimation. Sensor cost can be reduced if some signal values are calculated from the other - already measured - signals or the robustness of the system can be increased by supervising the sensors by calculating their measurement value out of other signal values which are also measured. Robustness and redundancy is extremely important considering drive-by-wire technology, where the physical connection between steering wheel and the wheels of the vehicle is omitted. This paper reports advances in permanent magnet synchronous machine model identification. Continuously measuring machine input voltages, output currents speed and applying a least squares method for short time windows of the measurement signals, internal parameters of the machine can be online obtained. In the identification stage the model excitation signals are the current values and the speed of the machine and the response signals are the input voltages. Estimated parameters - especially the phase resistance - can be used to predict the machine temperature, letting the designers to supervise or omit the winding temperature sensor. Having a proper model of the machine also enables the engineers to supervise the current sensor and detect machine faults using the model based redundant information.