Transformer Fault Diagnosis Based On Genetic Neural Network Combined With LIBS

There are many power equipment in the power grid to support its normal operations,among them,the power transformer plays an important role in the whole power system.Once the power transformer fails,it will harm the normal operation of the entire power grid.So it is of great importance to figure out the fault condition of the transformer in time.In order to solve the large error caused by complicated operations procedure and the environmental factors in the routine chromatographic analysis,paper proposed apply laser induced breakdown spectroscopy(LIBS)techniques for detection of dissolved gas in transformer oil contains a variety of elements.First,a practical and effective transformer fault gases LIBS spectral data preprocessing method and model was set up,subsequent use of genetic networks,simulation experiment was carried out to simulate the type and content of the gas in the fault characteristic gases of the known fault types and to know the type of error.Through experiment we draw the conclusion that: LIBS technology can detect the spectrum map of C,H and O of the dissolved gas in the transformer oil contained in the transformer.By application of calibration curve method we can get the predicted concentration.Based on three ratio method,we establish a model of transformer fault diagnosis based on neural network RBF,the correct rate increased from 50% to 80%.In the course of the experiment,we found that there are also some problems within this model.When the sample data is inadequate,the error rate of the diagnosis result is very high.Based on the content of C,H and O,we establish the model of transformer fault diagnosis based on neural network SOM.During the experiment is to be found,whether by increasing the number of iterations or the number of training samples,the accuracy of diagnosis results is not able to be increased.Then we introduced the genetic algorithm and K-means clustering algorithms combined with SOM neural network,built the model of transformer fault diagnosis.And the correct rate of transformer fault diagnosis was increased from 50% to 87.50%.