Algorithm Research Of Trouble Diagnosis In Transformer Based On DGA

The power transformer as a kind of energy conversion device is at the core position in the modern power system, whose performance directly affects the safe and reliable operation of the entire power system. Therefore, it is necessary for the fault diagnosis of the power transformer to accurately master the operational status of the transformer, which will provide the most reliable guarantee for the security, stability and efficient operation of the power system.The analysis method of gas dissolved in the oil has been recognized as one of the most efficient ways in the studies of transformers’fault diagnosis. Compared with the traditional three-ratio method which does not have the integrity of the fault information, two algorithms of the nonlinear mapping neural network and the support vector machine with the establishment of a transformer’s fault diagnosis model were combined in this paper in order to improve the accuracy and reliability of the diagnosis. On the basis of a large number of references, the main research work of this paper is as follows:First of all, in view of the nonlinear approximation ability of the neural network, this paper takes the experimental data as the characteristic parameters to establish a internal fault diagnosis model of the three-tier network oil-filled transformer with the help of the genetic algorithm that can optimize the network parameters. And then the established model is simulated by the neural network toolbox of MATLAB and the results of simulation show that the application of the genetic algorithm to optimize the network in the fault diagnosis of transformers is feasible and effective.Secondly, a fault diagnosis model of the transformers based on the support vector machine was proposed in this paper. As the eigenvectors that can be collected when the transformers fail to work are limited, the training of the network that requires a large number of the training samples has some limitations and problems of local minimum and devilishly learning. Therefore, this paper introduces the support vector machine based on the small sample learning and combines it with DGA to establish a fault diagnosis model of transformers based on SVM. In addition, in order to further improve the diagnostic accuracy, the cross-validation function in the installation package is used to select the best paraments c and g, which can optimize the model. By contrast of the simulation results of the neural network transformers’fault diagnosis model, we can draw a conclusion that the support vector machine based on the CV way is used in the fault diagnosis of transformers,which can greatly improve the generalization ability of the diagnostic system and also shorten the diagnostic time, further demonstrating the pracitical value of the diagnostic system.