Artificial neural networks for power systems harmonic estimation

An artificial Neural Network model was developed and implemented for power system harmonics estimation. The model was given the name FNN, which stands for Fourier Neural Network. It was tested off-line under different conditions and was compared with FFT. The results of the offline tests indicate that the FNN has very high estimation accuracy. It has a recursive nature that makes it a candidate for real-time measurements. It also gave good results in a noisy environment where SNR is as low as 10 dB. The FNN model was implemented on a PC using a data acquisition board. The system was used for an on-line harmonic estimation study. The MN was able to estimate the harmonic components of voltage and current at various levels. The estimation results were compared with the data obtained using a FLUKE 41 harmonics meter. The comparison revealed that the ANN based harmonic estimation model performs similarly to industrial-approved meters.