A neural network and fuzzy logic approach for the control of an active power filter

A neural network and fuzzy logic approach for the control of an active power filter

Author: 
Hichem Laib and Abdelaziz Chaghi
Abstract: 

In this paper a three-phase shunt active filter is used to compensate current harmonics, reactive power in three-phase distribution network system. In order to improve its performances, a Neural Network and Fuzzy Logic approaches are used to control this device. The first controller called Adaptive Linear Neural Network (ADALINE) is used with the direct method in order to identify precisely the necessary currents to reduce the harmonics and to compensate reactive power. The neural network inputs are based on a decomposition of the measured currents. This decomposition is also based on the Fourier series analysis of the current signals and Least Mean Square (LMS) training algorithm to carry out the weights. In this case, three ADALINE are used to extract the fundamental component of the distorted line current directly from the axis (the three phase space). The second controller is the fuzzy logic controller, used to regulate the DC link capacitor voltage. This approach has the advantage to eliminate the PLL and Concordia, Park or Clark transformations method. Speed and accuracy of this approach results in improving the performance of the APF. All of the studies have been carried out through detail digital dynamic simulation using the MATLAB/Simulink Power System Toolbox.

Paper No: 
19