Moment Inertia Identification Method Research For Permanentmagnet Synchronous Motor

The permanent magnet synchronous motor with high power density, small size, high reliability was widely used in high performance control field. Such as Industrial Robots, CNC Numerical Control Machine and Hybrid Electric Vehicle, and so on. Among them, the moment of inertia is a very important parameter for permanent magnet synchronous motor. Usually, if the load changes, the inertia would also change. If the moment of inertia has great changes, but the servo motor control parameters are still the same, it would lead servo motor’s operating conditions deteriorate, even it will cause the control system became unstable. So it is necessary to identify the moment of inertia of the motor system accurately, and adjust the servo system controller according to the identification value of inertia. Only in this way can the performance of servo system be improved and will the system of immunity and robustness be enhanced.Due to the existing identification algorithm, such as reduction method, artificial track method, belong to off-line identification. Although they are simple, the precision of the identification result was low, and the identification time was long. The established online identification algorithm, such as least square method, the model reference adaptive algorithm, etc., while some scholars are studying them, but they still can not get satisfactory results. So the online identification is still in the research phase and it is worth research further. This paper first introduces the mathematical model of permanent magnet synchronous motor and the basic principle of vector control; Second, it studies deeply the principle of integral method, the improved least square method and model reference adaptive identification algorithm, and build the motor control system simulation model in the Matlab, using different algorithms which based on the model simulation to work to get the recognition results. Then, analyzing different identification algorithm results, and improve the identification algorithms according to the identification results. Simulate again and draw the conclusion comparing with previous results.Aiming at the adding the forgetting factor into the traditional least squares identification of data saturation phenomenon, and the simulation results shows that using the forgetting factor least square method can effectively prevent data saturation and get the good identification time. Second, based on the larger error phenomenon traditional principle of model reference adaptive identification, adding the average weighted recursive filtering algorithm, the simulation results shows that after adding the filtering algorithm, it can not only reduce the error of identification, but also can have a comprise between the identification error and recognition time. Finally, according to the results of the moment of inertia of the motor, applying the parameter self-tuning function, the controller can automatically adjust the PI parameters when the inertia changes, then improve the system robustness and real-time control.