Prediction of line current harmonics in an AC-DC converter using neural networks

A circuit topology for an AC-DC single-phase single-stage half-controlled converter with reduced line current harmonics is presented. It is shown that these harmonics are reduced below limits imposed by relevant international standards governing the generation and injection of harmonics into the distribution systems by equipment which draws current less than 16 A. The paper also demonstrates the use of a feedforward neural network (NN) for the prediction of the line current harmonics. The neural network predicted results compared favorably with the simulation results and hence were also within the limits of the harmonic regulations.