Study On DGA Data Analysis And Fault Diagnosis Of Oil - Immersed Transformer

Transformer achieves transmission and substation engineering as a hub device in power system, and its running state directly affects the security, reliability and stability of all the power system. However,various types of insolation failure due to aging and othe factors can not be completely avoided when the transformer is in the long running process. Hence we need to identify and exclude early latent fault of transformer so that the economic losses caused be reduced by the accident, and the supply rate of the power system be improved,also more convenient for people’s production and life through early diagnosis and repair can be achieved. Analysis of failure mechanism of oil-immersed transformers to find that, oil dissolved gas data as an important basis can be used for diagnosis, which also can be combined with artificial intelligence methods to achieve the type of fault diagnosis and prediction. The main contents of this paper are as follows:1、the oil dissolved gas’s generation mechanism and the dissolution process of the transformers be analyzed, and a study to find the correspondence between the main internal fault types and oil gas components by transformers be made. H2, CH4, C2H6, C2H4 and C2H2, these five characteristics of gas content can be used as parameters in diagnosis to determine the fault. Fuzzy C-means clustering algorithm be used for a cluster analysis with 102 fault samples with definite conclusions of transformers. The accuracy rate of this classification is 82.35%. To effectively prevent plunging local optimum circumstances, the genetic algorithm and the simulated annealing algorithm are used to improve its search performance. The improved performance of the algorithm is demonstrated by 400 sample points, and then SAGA-FCM algorithm is used to cluster analysis with the 102 fault samples above. The classification accuracy rate is more 4.9 percentage points than FCM algorithm and cluster centers of the 6 typical fault types is got.2、A transformer fault diagnosis model is established based on the theory of grey correlation entropy. In the model, reference sequences are obtained by using the SAGA-FCM algorithm to the cluster centers of the 6 typical fault types in the second chapter. The comparison sequence uses three ratio method to encode the incomplete 8 fault samples, and the grey correlation entropy transformer fault diagnosis based on the SAGA-FCM cluster center based on the cluster center is carried out, and the 8 samples are all correct.However,there are 3 samples in the diagnosis of grey correlation entropy method using a single reference sequence, and the reference sequence of grey relational entropy in SAGA-FCM cluster center is found. It not only enhances the correlation of the comparison sequence and the reference sequence, but also improves the accuracy of the fault diagnosis, also the problem of the three ratio code have been solved.3、Based on the historical data of the time variation of dissolved gas in oil, the content of dissolved gas in oil is predicted with GM (1,1) prediction model and polynomial regression model. In order to stabilize the prediction accuracy of two models for different characteristics of the gas, the root mean square error is used to carry out the weighted average of the two prediction models. The prediction accuracy of the weighted combined model is judged, and the gas change in the later period is reasonably predicted. The results are analyzed and the timeliness of the forecast model is explained.