Auto-tuning the cost function weight factors in a model predictive controller for a matrix converter VAR compensator

This paper presents an auto-tuning technique for online selection of the cost function weight factors in model predictive control (MPC). The weight factors in the cost function with multiple control objectives directly affect the performance and robustness of the MPC. The proposed method in this paper determines the optimum weight factors of the cost function for each sampling time; the optimization of the weight factors is done based on the prediction of the absolute error of the optimization objective and the corresponding constraints. The application considered is a reactive power compensation technique using MPC of a direct matrix converter. This technique compensates lagging power factor loads using inductive energy storage elements instead of electrolytic capacitors (e-caps). The result demonstrates that the proposed auto-tuning approach of cost function weights makes the control algorithm robust to parameter variation and other uncertainties such as load variation. The proposed capacitor-less reactive power compensator based on auto tuned MPC cost function weight factor is implemented experimentally using dSpace DS1007.