Genetic algorithms-based gain optimization of a simple learning control for single-phase shunt active filters

The repetitive or learning based control has proven to achieve high steady-state performances for control systems. The iterative learning algorithm aims to accomplish zero tracking error without full knowledge of the system model. However, its dynamic behaviors were not satisfied in some conditions particularly under non-periodic disturbances since the control uses the information from the previous iteration to calculate the system input. This paper proposes the investigation of the use of genetic algorithm to optimize the learning gain of a simple proportional-type (P-type) learning control applied to current control for shunt active filters. The merit of the proposed control is its simplicity, potentially suitable for commercial active filters. All design concepts are verified and the results obtained in the simulation confirm the improvement in the dynamic responses during the transient condition while the harmonic control accuracy in steady-state can remain excellent with the proposed control system.