An Approach to Model Validation for Model Predictive Control Based on Dvurecenska’s Metric

In this paper, Dvurecenska’s validation metric is proposed as a criterion for selecting a suitable model to be used by a Dynamic Matrix Control algorithm. As part of the work developed in this paper, it is analyzed how this algorithm performs when using different models. Besides, a comparison between the validation of these model results is performed. Additionally to the above metrics, the Theil Inequality Coefficient index and the FIT index are employed too. This paper was developed on a simulated plant based on Clarke’s benchmark, which is controlled with a Dynamic Matrix Control. As a result, it can be seen that Dvurecenska’s metric accomplishes its objective and in some cases, gives better validation criteria than the other metrics analyzed.