Motor Parameter Estimation Based On Improved Kalman Filter Algorithm

Brushless DC motor is widely used in industrial equipment for long lifetime,simple control and reliable operation,e.g.,industrial robot,numerical control machine tool.The reasonable motor control mode can greatly exert the motor function while saving the driving electric energy.At present,there are many controllers used to control brushless DC motor,such as various types of single-chip microcomputer,DSP,FPGA and so on,and most of them adopt the control mode of PWM and PID.However,no matter what kind of controller and what kind of control mode,the prerequisite for accurate motor control depends on the accurate motor parameters,in order to ensure the control accuracy and improve the motor performance,it is necessary to identify the motor parameters in engineering practice.Therefore,the parameter identification of motor has always been a hot topic in motor research.In this article,a series of improvements are made on the basis of the extended Kalman filter algorithm,and it is applied to the on-line identification of brushless DC motor parameters.The main contents are as follows:Firstly,the basic structure and mathematical model of brushless DC motor are introduced,and the augmented continuous state space model is established by adding the motor parameters into system states,and then the identification model for extended Kalman filter algorithm is obtained by discretization and linearization.Secondly,in view of the large estimation error of the extended Kalman filter algorithm,the linear correction is carried out on the estimation results.Two linear correction methods are introduced: First,the correction is obtained by the gradient descent method based on the Armijo-Goldstein criterion,and the other is obtained by the weighted least square method.The effectiveness of the two methods is verified in Matlab/Simulink environment.Finally,for the fact that the statistical characteristics of noise are unknown or the accuracy is poor,the quality of filtering will be directly reduced,and even lead to divergence of filtering results.This method updates the process noise covariance matrix in real time according to the input and output data of the system,and takes the state estimates by traditional extended Kalman filter as onestep estimation for calculation of covariance matrix.