Fault Line Selection For Small Current Grounding System Based On GA-PNN And Comprehensive Hausdorff Distance

Distribution network is an important part of power network,and its safe and stable operation is related to people’s daily life and national economic construction.Small current grounding system is generally adopted in power distribution network in China,that is,isolated neutral system or neutral point via arc suppression coil grounding system.In the small current grounding system,single phase grounding faults occur frequently.With the continuous development of power system in China,the structure of distribution network is becoming more and more complex.It is difficult to judge the fault line quickly and accurately by single line selection scheme.How to find a fast and effective line selection scheme is particularly important.Based on the analysis of existing methods of fault line selection,the active component of zero sequence current and the fifth harmonic component are selected as fault characteristics.Meanwhile,the similarity measurement is applied to fault line selection,and the traditional Hausdorff distance is quoted then improved.A comprehensive line selection criterion based on comprehensive Hausdorff distance value,active component of zero sequence current and fifth harmonic component is proposed.The criterion combining transient and steady state characteristics,makes up for the shortcomings of a single method,and improves the accuracy and applicability of fault line selection.The simulation model of 10kV distribution network is built by using the software of Matlab/Simulink,and three kinds of fault characteristics are simulated.Through the measurement module in Simulink,the waveforms of three characteristic quantities are collected.By comparing the differences between fault lines and non-fault lines under each characteristic quantity,the feasibility of three feature quantities in fault line selection is verified.By setting different short-circuit distance,grounding resistance,fault voltage initial phase angle and fault lines,a large number of simulation experiments are carried out.According to experimental datas,calculate and record corresponding eigenvalues of comprehensive Hausdorff distance,active component of zero sequence current and fifth harmonic to prepare datas for training the neural network.Automatic line selection for single phase grounding fault in small current grounding system is realized by artificial neural network.Aiming at the problem of fault recognition and classification,this thesis adopts probabilistic neural network(PNN),which is widely used in pattern classification field,to abandon the traditional BP neural network.Considering the influence of smoothing factor on the performance of PNN network,this thesis uses genetic algorithm to optimize smoothing factor,and proposes a probabilistic neural network(GA-PNN)for fault line selection based on genetic algorithm to optimize smoothing factor.By sending experimental datas into PNN and GA-PNN respectively for training and testing,through comparative experiments,the fault diagnosis rate of GA-PNN network for training samples and testing samples is 100%,and the iteration time of genetic algorithm is very short,only needs four iterations to converge.Experiments prove that the line selection scheme in this thesis has the advantages of fast speed and high accuracy,and has certain practical application value.