Non-integer Lexicographic-optimization-based Sequential Model-Predictive Fault-Tolerant Control of T-Type Shunt Active Power Filter

Fault-tolerant control (FTC) strategies have been proposed to improve the reliability of power electronic systems. However, the FTC of the three-level T-type converter (3LT2C)-based shunt active power filter (SAPF) remains limited by redundant devices. To address this problem, an FTC strategy called sequential model predictive tolerant control (SMPTC) is combined with a non-integral lexicographic optimization (LO) algorithm for solving multi-objective optimization problems. First, the power circuit, space voltage vector diagram, harmonic extraction, and prediction models of the SAPF system are introduced. Second, based on power-switch fault analysis, an adaptive DC-bus voltage regulation method is proposed, which ensures that the candidate voltage vectors are fully utilized in the fault state. Third, an LO-SMPTC strategy is proposed by designing a sequential predictive controller that considers the neutral point (NP) voltage and the tracking of the SAPF output current. Compared with traditional model predictive control, the proposed method not only achieves NP voltage balance and excellent harmonic compensation but also eliminates the selection of weighting factors, thereby improving the control flexibility and time-efficiency. Finally, the effectiveness of the LO-SMPTC on the DC-bus voltage, NP voltage oscillations, grid current power quality, and pole voltage under different scenarios is demonstrated through experiments.