ABSTRACT: This paper presents the comparison of three different reference current extraction methods based on Adaline neural networks for controlling three-phase shunt active power filters. The objective of the active power filter is to eliminate harmonics and reactive power. The filter acts as a current source, which is connected in parallel with a nonlinear load and controlled to generate the required compensation currents. The reference current extraction methods to be used for the control of the active power filter are 3-phase currents method, instantaneous power method and two-phase (DQ) currents method, respectively. In all of these methods, Adaline neural networks functioning as an adaptive noise canceller to extract the reference current from the load current, which will be injected to the mains by shunt active power filter. Under different operation conditions the simulation results of three-phase shunt active power filter that is controlled by all of the three methods proposed in this paper are presented and their operation performances are compared.
Keywords: Shunt active power filters, artificial neural networks, Adaline, reference current extraction, adaptive noise canceller, harmonics and reactive power compensation.