Research On Power Distribution Network Harmonic Detection Based On A Linear Neural Network Method

With some of the new technology is widely used in power systems, harmonicproblems become more serious. This not only causes serious harm to the security andeconomic operation of the power system, but also may affect the normal operation ofpower users, giving users unnecessary economic losses. Therefore, to take effectivemeasures harmonic detection has a very important practical significance.In this paper, the current harmonic detection method were analyzed first, andthen for the harmonic characteristics of the distribution network, the use of BP neuralnetwork and linear neural network based on a combination of methods to accuratelydetect harmonic distribution. The main idea is to use the subject of BP neural networkto construct the fundamental principles and methods of detection model, using thesteepest descent algorithm for the fundamental component of the distribution networkfor accurate testing. To detect the fundamental basis for the arguments, and then usinglinear neural network method for distribution network in all kinds of harmonicdetection. Because the whole harmonic and non-integer harmonics their differentcharacteristics, so this targeted linear neural networks are constructed and the wholeentire harmonic harmonic detection model non, and a detailed derivation of theimplementation process related algorithms. The last article in the absence of noise,respectively, of each harmonic detection phase and noisy environment for thesimulation and analysis of results.Experimental results show that, with a harmonic detection method combines BPneural network and linear neural network used in this paper has higher detectionaccuracy than the traditional FFT algorithm Hanning window up2to3orders ofmagnitude, and exhibits good noise immunity.