An improved artificial neural network for a direct-power control based on instantaneous power-ripple minimization of the shunt active-power filter.

The paper presents a method for a direct-power control (DPC) based on Artificial Neural Networks (ANN) applied to a shunt active-power filter (SAPF). The aim is to improve the performance of conventional controls. SAPF is one of the most advanced pollution control solutions. DPC is a high-performance control for PWM converters based on the instantaneous-power theory. However, the presented control has some drawbacks, such as the presence of ripples in the current. To improve the performance of the system to be controlled, artificial neural networks are applied to the conventional control. To achieve the objective, DPC-ANN is combined with conventional DPC using MATLAB/Simulink, Simulation results show a very satisfactory performance. [ABSTRACT FROM AUTHOR]