Simplified Model Predictive Current Control for PMSM Drives Based on Bayesian Inference

Abstract:
In order to solve the problem of the dependence of the traditional model predictive current control method on motor parameters, this paper proposes a simplified model predictive current control method based on Bayesian inference, which makes the predictive model structure simpler by reducing the motor parameters in the predictive model. Firstly, the d-q-axis current prediction model is simplified and the q-axis current prediction model is incrementally processed to eliminate the influence of the magnetic chain parameters, so that the d-q-axis current prediction model contains only the unique motor parameter - inductance. Then, Bayesian inference is used to identify the parameters of the inductor, and the likelihood function is established with the d-axis current error as the criterion, and the posterior distribution of the inductor is sampled by the Metropolis-Hastings sampling algorithm, and the inductor sample is selected to be accepted or rejected based on the likelihood of the inductor samples, which generates a randomized sequence of inductor samples. After the sample burn-in period, the inductor sample sequence can converge and form a Markov chain with a smooth distribution, and the expectation value of the inductor parameters in Bayesian inference can be calculated by intercepting the valid inductor samples in the smooth distribution, which completes the accurate identification of the inductor parameters. Simulation results show that the control method proposed in this paper significantly reduces the dependence of the prediction model on the motor parameters, effectively improves the robustness of the whole control system, and achieves good control results.