| A disturbance predictor has been proposed earlier for a simplified Model Predictive Control (SMPC) algorithm. In this thesis, a closed-loop transfer function is derived for this disturbance predictor and its performance is analyzed. The analysis shows that the predictor offers an improvement in disturbance rejection. Based on this analysis, a relationship between the Dahlin's and the SMPC algorithms is provided. This relationship enables the selection of the tuning parameter of the SMPC algorithm by using a desired value of the closed-loop time constant. An optimization scheme is proposed for online determination of a tuning parameter employed in the disturbance predictor. The optimization is carried out by using historical data. The effect of the data length on the optimization is investigated. In addition, the effect of the disturbance predictor parameter on the control system stability is studied. The performance of the proposed predictor is presented on three example problems for non-stationary ARIMA disturbances that commonly occur in many industries. The applicability of the disturbance predictor to the DMC algorithm is also demonstrated. A comparison with the Generalized Analytical Predictor shows that the disturbance predictor provides improved control performance. Moreover, the disturbance predictor does not require knowledge of the disturbance models. |
