Optimal design method for LLCL filters based on NSGA-III

Abstract

The LLCL filter is usually used as a switching ripple suppressor since it can attenuate switching-frequency current harmonics much better than an LCL filter using lower total inductance and capacitance. However, it is more difficult to design LLCL parameters. In addition, it has a number of initial design constraints: the fundamental reactive power, the resonant frequency fres, etc. are coupled and always contradictory, which means that designing the parameters is a Many-objective optimization problem (MaOP). The non-dominated sorting genetic algorithm-III (NSGA-III) does well in solving this kind of problem. In recent studies, the proposed methods only provide a range of parameters. Thus, obtaining certain parameters relies on experience, and using the boundary value cannot be proved optimal. However, using original NSGA-III is somewhat time-consuming for achieving specific parameters. To deal with this problem, this paper proposes a novel optimal design method for an LLCL filter based on NSGA-III with the handling of constraints. The proposed method gives a set of specific parameters and achieves a high computing efficiency. The proposed method is verified through simulations and a grid-connected inverter system based on a virtual instrument to show the effectiveness of the proposed method.