The Robust Harmonic filter Design Using Artificial Neural Network and Taguchi Method

A systematic approach to achieving optimal discrete-value harmonic filters using artificial neural network and Taguchi method was developed for a multi-bus system under abundant harmonic current sources. In this simulation, the Taguchi method was used first to perform an efficient experimental design and analyze the robustness of the harmonic filters. Then an artificial neural network (ANN) was used to minimize the variation and make the harmonic filters less sensitive to variations. The L₃₆ orthogonal array was used as the learning data for the ANN to construct an artificial neural network model that could predict the parameters at discrete levels. In the design of harmonic filters, the outer array used to present the noise factors effect of loading uncertainty and changing of system impedances. The searching for optimal solution is applied to the harmonic problems in a steel plant, where both AC and DC arc furnaces are used and a static VAr compensator (SVC) is installed. The results showed the harmonic filters performance was significantly improved compared with the original design