Optimization of a passive harmonic filter based on the neural-genetic algorithm with fuzzy logic for a steel manufacturing plant

This report presents a method which combines an artificial neural network and a genetic algorithm (ANNGA) in the planning of large-scale passive harmonic filters (PHF) for a multi-bus system under abundant harmonic current sources. The objective is to minimize the cost of the filter, its power loss, the total demand distortion of harmonic currents and the total harmonic distortion of voltages at each bus, simultaneously. First, a Taguchi experiment was used to perform an efficient experimental design and analyze the robustness of the PHF. Following, the results from the Taguchi experiment were used as the learning data for an artificial neural network (ANN) model that could predict the parameters at discrete levels. Finally, a genetic algorithm was applied to obtain a robust PHF setting of the parameters with continuous variables. Besides, to deal with the multiple objectives of the addressed problem, the membership function from fuzzy logic theory is also adopted in the algorithm for determining the weight of each single objective considered. A search for an optimal solution was applied to the harmonic problems in a steel plant. The results showed that the performance of the harmonic filters was significantly improved when compared with the original design.