Data-based Research on Predictive Control of Multiple Mismatched Model Parameters

The basis of the paper is that there are already some methods to accurately evaluate, test and diagnose the performance of the model predictive controller. And the result shows the reason of a bad performance of control system is because of model mismatch. There are much more complexity and variety in the problem of multiple mismatched parameters than single mismatched parameter, so we need consider more factors about it on the basis of the solution of single mismatched parameter. We propose a way of adjusting model parameters based on fuzzy rules when there are more than one mismatched parameters. The method is to adjust the step-size of parameters and get the adjustment rules on the basis of the changes of maximum overshoot, rising time and settling time. The last, verifying the method is effective by experiments.