The Harmonics Analysis Technology Of Power System Based On BP Artificial Neural Network

Due to the development of modern industry, nonlinear power supplies have been used pervasively, under this circumstance power harmonics are produced so excessively that users are suffering seriously from this problem, thus it is necessary to monitor and control this kind of harmonics.At the very beginning of this paper, possible harms caused by harmonics, as well as the significance for harmonics monitoring is given, next a comparison between several techniques for harmonic detection and measurement is made, based on which using BP neutral network to conduct harmonics analysis is chosen. According to the characteristics of power harmonics, a data set used for the training process of an improved BP neutral network is prepared, and a graphic software program developed in LabVIEW for the implementation of this algorithm is provided. After the data set for training is input into that program, numeric values for threshold and weight parameters of this network are obtained.Since the target network has been achieved through the training stage, analysis on an artificial designed data set and another set acquired from practical system are made respectively. This analysis indicates that designed neutral network in the paper satisfys high numeric precision and functions well for detection of harmonic components. Although relevant training and test have been only done on harmonics of odd times such as 3, 5, 7, 9 and other sequential orders such as 2,3,4,5 of base frequency of distorted non-sinuous periodical signals, the principal and methods are also suitable for signals consisting of much more harmonics. Meanwhile, two measurements on a signal from practical system are conducted, using the algorithms of BP neutral network and FFT respectively, a following comparison between the two achieved results shows output in fewer units on the advantage of high precision and speed that BP neutral network owns. At last, a front panel for parameters setting and the display of results produced by signal processing is designed in the LabVIEW software.