Research For Fault Recognition Of Distribution Line Base On Clustering Artificial Immune Network

With the development of China’s power industry, distribution network constructionincreasingly large and complex, and also put forward higher requirements for power supplyreliability of distribution network, fault occurs, asked the staff can quickly and accuratelyjudge the fault location, fault isolation, and fault type identification is the basis of faultanalysis response, plays a vital the role of. But the research about the distribution networkfault type identification method is relatively less, this research still needs to be strengthened.In this paper, the specific fault type, in order to build a distribution network faultclassification model for the goal of fault classification, distribution line based on transientcomponent is studied. This paper mainly does the following work:Firstly, this paper briefly analyzes the neutral point grounding mode of the steady state,and line fault transient characteristics of power distribution system, extract the object selectedtransient signal for fault classification, using the wavelet coefficients statistical analysismethod, constructs the characteristics suitable for distribution network fault classification andrecognition to the amount.Secondly, based on the analysis of the artificial immune system and cluster analysistheory, from the immune clustering combined, using artificial immune network form tocomplete the clustering analysis, the distribution network fault types, designed a faultclassification model based on immune clustering: the ground voltage is greater than zero, inthe sample input, divided into the ground fault and non fault detection cycle into different,then the immune clustering and initial antibody extraction, antibody clustering core finaloutput of each fault type, complete the training samples, then simply comparing cluster coreand test samples of the Euclidean distance can output the identification results. Clusteringalgorithm based on aiNet to realize the identification of the model functions improved and,mainly for the hard threshold aiNet algorithm of too many problems, improvement for somegiven parameters self adaptive matching algorithm.Finally, in order to verify the validity identification method is established in this paper,the test model was set up by PSCAD IEEE34nodes,10kinds of fault was simulated byMATLAB, the correct rate of recognition test under different fault conditions, and in theneutral grounding mode and network topology changes, the method is analyzed adaptability.