Estimated position error reduction of SMO-based sensorless control of IPMSM using variable notch filter

Abstract

Controlling IPMSM is categorized into two sections; an open-loop strategy in which the essence of shaft position information is not needed and a close-loop strategy in which shaft's position information is required and is subdivided into two methods: • Sensored • Sensorless In the sensorless method, there are three methods for rotor position estimation. The first one is fundamental excitation which includes adaptive and non-adaptive methods. Secondly, signal injection methods, and the last one is the help of artificial intelligence. Adaptive methods are categorized as model reference adaptive base systems and observer-based control, which is subdivided into Luenberger Observer, reduced order observer, sliding mode observer (SMO), and Kalman filter. Estimation error in SMO control is inevitable, and it is generated because of its sign function. In this work, a sliding mode observer is employed, and the error in this method is reduced by applying varying notch filters in the system. When a system's state is not measured directly, sliding mode observer control can be utilized to estimate the required dynamics. The main function of SMO is moving or sliding over a hyperplane to reach the desired or stable state and stays on it. With the help of these estimated state variables, it is possible to control our system. In the case of controlling IPMSM, the estimated state variables are the rotor's speed and position. A sign function is employed to determine the hyperplane. By multiplying the sign function of estimated value minus real value, the sliding surface is defined. If the sliding surface converges to zero, the error between the estimated and real value approaches to zero. The sign function, which is the main part of SMO, creates oscillations in the system and causes chattering in IPMSM. In order to overcome this obstacle, filtering the oscillated frequency is proposed. The range of the frequency which is being eliminated from the output of SMO should be narrow so that a notch filter is suggested to attenuate the oscillation frequency. In this case, in which the parameters of IPMSM have already been defined, the oscillated frequency for reference speed of 0.1 pu is obtained from the FFT analysis of the signal, which is 27.34 Hz. After simulating the proposed method in MATLAB®/Simulink® and filtering the aforementioned frequency from the estimated speed, a significant reduction in the first harmonic oscillation of the signal is achieved, and according to its FFT frequency xxi response, its amplitude is decreased from 0.14 to 0.05 when the motor is rotating at 0.1 pu. In conclusion, this method has reduced chattering in the system and attenuated the frequency, which leads to estimation error mitigation, and increases the robustness and accuracy of the position sensing of IPMSM.