Fast non-recursive extraction of individual harmonics using artificial neural networks

Non-linear loads, which cause harmonic distortion, are increasingly being used in electrical power systems. This is causing a major concern and real-time harmonic monitoring in electrical power systems has become important. Many applications such as harmonic monitoring and filter design need techniques for the fast extraction of individual harmonic components. The time limitations and computational complexity associated with conventional techniques make it appealing to investigate alternative techniques for harmonic extraction. A technique based on artificial neural networks (ANNs) for the fast extraction of individual harmonic components is presented. It uses the non-linear mapping capabilities of ANNs to accurately estimate the individual harmonic components of a distorted signal. The proposed algorithm is implemented on a real-time hardware platform and tested. Results show that the ANN-based harmonic extractor is significantly faster and less computationally complex than conventional techniques.