Comparisons of Decentralized Model Predictive Control without Weighting Factors for Electrical Drive Systems

Model-based predictive control (MPC) has been widely spread in both academic and industry applications, due to its inherent merits of easy conduction, excellent dynamic response as well as adapive involvement of constraints. The optimal vector is selected via optimizing the error terms in the designed optimization function of MPC. However, the design of weighting factor is still a challenging task as various control objectives are coordinated in the cost function. In this paper, comprehensive comparisons of decentralized model-based predictive control without any weighting parameters for electrical drive systems are proposed. The comparisons not only evaluate the control performance but also the algorithm complexity. First, the novel construction of cost function in the presented decentralized MPC is described. According to the abovementioned concept, a complex MPC optimization problem is separated into a combination of simpler local problems, which can be solved by each sub-task. The initial optimization for each control objective are conducted, and then generate the optimal vector. The comparative results are implemented on a pair of 2.2 kW induction machine lab- constructed experimental platform. The proposed decentralized MPC methods are aiming to obtain the improvement of control performance for a large-scale control system with multiple parameters.