A Weighting Factor Design Approach for FCS-MPC Techniques Based on PSO and K-Means Algorithm

A cost function with proper weighting factors (WFs) enables finite control set model predictive control (FCS-MPC) of power electronics to include multiple physical constraints and control objectives. On the other hand, lack of systematical WFs design and with the classical empirical trial-and-error adjustment will become time-consuming and tedious, particularly for complex system topology and multiple nonlinear control objectives. In this paper, a novel WFs design strategy based on particle swarm optimization (PSO) and k-means algorithm for the classical FCS-MPC techniques is proposed. A grid-tied three-level neutral point clamped (3L-NPC) converter is chosen as an explanatory case of verification. The fitness values of all control objectives (here, current tracking, DC-link voltage balancing, and quasi-fixed instantaneous switching frequency) are based on integral time-weighted absolute error (IATE). Because the control objectives have different natures (different physic units and numerical magnitudes), instead of globally optimal WFs, there would be Pareto (non-dominated) solution set during PSO operation. Then the k-means cluster algorithm is used to select typical WFs for efficient selection from Pareto solutions. Simulation data confirm the effectiveness of the proposed solution, which guarantees qualified WFs design and is simple in calculation.