The Pattern Recognition Method And Application For Grounding Fault Line Selection In Non-effectively Grounded System Based On Probabilistic Neural Network

The non-effectively grounded system mainly adapts the way of isolated neutral,neutral grounding by the arc extinction and neutral grounding with resistance in ourcountry of low voltage distribution network. Once there is a failure with this kind ofsystem, the faulted current is too small to be detected, for the current can not find itsloop. This way can effectively protect the distribution network working stability andsustainability, but there are so many problems in the faulted line selection. At presentthe small current grounding faulted line selection method is mainly divided into threetypes: one is to make use of the faulted steady state features, the second is to use thefaulted transient features, the third way is to inject special signal into the system to dothe faulted line selection. As the fields of electronics, communication, computer andintelligent control have been making such rapided development of technology, peoplebegan to explore the use of more advanced methods and computer equipments inorder to get a better effect. Artificial intelligence, neural network and fuzzyrecognition are applied to the emerging technologies in the faulted line selection.Probabilistic neural network is a kind of neural network which can be used forpattern classification, so in the fields of machinery, materials, environmentalengineering and economic it has been used effectively. In this article, firstly the threekinds of faulted characteristics(wavelet energy of the sequence current, activecomponent, and fifth harmonic component of the distribution network) must be got,then make fusion of them through the probabilistic neural network, in this way wecan predict the faulted line selection.By making a large number of simulations in MATLAB and comparing it withsome other ways(probabilistic neural network with single characteristic and the BPneural network method for faulted line selection), it is verified that this method notonly has good versatility and anti-interference ability, but also owns the followinggood traits, they are high accuracy, operating simply and quickly, rich contain offaulted data, and easy to expand the knowledge base, etc.