Application of adaptive Kalman filtering to system identification of brushless DC motor

To restrain the effect of variable measurement noise and to acquire the accurate model of a brushless DC motor, the identification method for the motor based on adaptive Kalman filtering algorithm was proposed. By computing the maximum likelihood estimation of the innovation variance and using it to modify the filter gain, the influence of variable measurement noise could be restrained and the parameters could be estimated accurately. In this way, the identification accuracy was improved. Experiments show that the adaptive Kalman filtering algorithm can follow the change of actual measurement noise accurately and get smooth estimation results. Compared with the recursive least square algorithm which is widely used in system identification at present, the root mean square value of output error is reduced by 73.5% under the variable measurement noise. The identification results can describe well the system behavior, and offer the same response with the real system. The algorithm is easy to apply to the engineering practice.