Skip to Main Content
This paper presents a complete strategy for harmonics identification and neural control of a three-phase voltage inverter used for the power active filtering. Based on the use of neural techniques, this approach of compensation is done in two stages. The first stage extracts the harmonic currents with the diphase currents method by using Adaline neural networks. The second stage injects the harmonic currents in the electrical supply network; it uses a control based on a PI-neural controller. Our approach is automatically able to adapt itself to any change of the non-linear load and thus to the harmonic currents generated. Furthermore, comparisons with hysteresis control and results of application of a conventional low-pass filter for harmonics identification are presented. The proposed neural compensation approach has been evaluated in simulations. The results show excellent behaviours and performances, as well as robustness and usefulness.