![]() | Only 14 pages are availabe for public view |
Abstract Harmonics becomes one of the most significant problems affected power quality of distribution systems. The increasing use of power electronics and distributed generators in nowadays networks producing different types of harmonics in turn. The severity of harmonics is lied in its effects as it causes many problems of loads fed from distorted network, it leads to improper operation, causing additional losses, overheating and overloading; it can also cause errors in power metering and protection devices. To mitigate harmonics, an accurate method needed to evaluate the harmonic levels in the networks. Also it is important to identify the source of harmonics whether it is caused by the load or supplied from the source. The techniques used in this issue are divided into three main techniques: frequency domain, time domain and artificial intelligence techniques. In this work a study to evaluate harmonic level is provided by modeling the nonlinearity of loads in micro grid network using one of artificial intelligence techniques. Nonlinear AutoRegressive with eXogenous input (NARX) neural network is trained using field measurements to model loads in a petroleum field micro grid which contains variety of loads. After verification of this technique, a shunt active power filter is designed to mitigate the nonlinear load harmonics. This filter is designed using a NARX with backpropagation training algorithm. The instantaneous reactive power algorithm is integrated within the neural network to extract the dominant harmonics. |