The main objective of this article is to provide an efficient design approach of passive grid filters based on evolutionary search algorithms. The case study is conducted on a passive trap filter (one of modern high order passive filters that can provide better performance compared with first order scheme) to be incorporated with a three-phase grid-tied inverter fed from PV array and a battery bank, forming a typical PV grid integration system. The filter design approach is mainly based on particle swarm optimization technique (PSO) as one of the efficient evolutionary search algorithms meanwhile Genetic Algorithm (GA) has been utilized as well for comparison purposes. The developed PSO algorithm searches for the optimum numerical values of filter passive components that can optimize a customized objective function of multiple terms. Unlike single-term objective function, the adopted approach has the advantage of involving multiple items in the customized objective function, such as harmonic attenuation factor and the numerical values of passive filter elements to be minimized. Moreover, the evolutionary search-based design approach is characterized by offering several groups of solutions for the same optimization problem. Thus, the obtained solution is not unique. Consequently, the choice between those solutions depends on the system designers who can select the most convenient solution among several alternatives based on component availability in market, the overall cost of the filter and the desired upper limit of THD of grid currents. The obtained results of this study demonstrate also the capability of evolutionary search approach to get optimum values of filter components that achieve good harmonic attenuation and satisfy the related grid codes such as the IEEE standard 519. Compared with traditional filter design method, the obtained results indicate that the PSO technique achieves lower numerical values of filter parameters in the most of parameters, resulting in an overall reduction in the size and cost of the passive grid filter. In addition, the employed GA succeeded to provide satisfactory numerical values of trap filter as well. However, the PSO technique is better than GA in several terms such as lower final value of objective function and lower numerical values of filter elements and convergence to the optimal solution(s) irrespective the initial guess values.Matlab® has been employed as a simulation platform for PSO and GA algorithms. While the overall grid integration system has been modelled and studied using professional version of PSIM software®.
Read full abstract