157 / 2018-08-25 17:49:24

Research on Transformer Fault Diagnosis Based on BP Neural Network Improved by Association Rules

Transformer Fault Diagnosis；BP Neural Network；Association Rules

The transformer is the core equipment of the power system. The safe operation of the transformer is not only related to the operating cost and operational safety of the power system, but also to the normal operation of various industries in society and the safety of life and property of the people. Due to the continuous expansion of the power system, the incidence of power equipment failures has gradually increased, which puts higher requirements on the fault diagnosis of transformers. Through fault diagnosis technology, faults can be detected in advance during the operation of the transformer, and measures can be taken in time to reduce the possibility of accidents.
Fault diagnosis of transformers can be regarded as a process of pattern recognition. After a long period of research and development, a variety of algorithms have been cited in transformer fault diagnosis. Current transformer fault diagnosis methods include ratio method, support vector machine (SVM), artificial neural network (ANN) and other algorithms. Most of the above algorithms use a single or a small number of fault characteristics for analysis and calculation when diagnosing and predicting transformer faults. There is a lack of common influences on multiple faults and possible links between faults, resulting in low accuracy of prediction results.
In order to improve the accuracy of fault diagnosis and obtain higher prediction accuracy, it is necessary to consider not only the various characteristics such as oil gas and partial discharge, but also the degree of correlation between the fault type and each feature quantity and the relationship between each feature quantity. Correlation. Therefore, an association rule analysis method based on multiple transformer fault characteristics is proposed. The Apriori algorithm of association rules reveals the association rules between the high-frequency items in the feature quantity data set. The Apriori algorithm can effectively explore the degree of association and confidence of different feature quantities, and apply it as a weight value to the prediction link.
In this paper, BP neural network algorithm based on association rules is adopted. BP neural network is a multi-layer feedforward network. In the training process, the correlation degree and confidence between different feature quantities of association rules are used as network weights. The sample is trained, and the connection relationship between the layers is adjusted according to the method of error back propagation. Through the repeated forward training and reverse transmission, the weight training steps are adjusted to finally obtain the training result.
The BP neural network algorithm based on association rules used in this paper considers the multiple factors affecting transformer faults. The Apriori algorithm is used to mine the correlation degree and confidence between different feature quantities, and the influence of various influencing factors on fault prediction is obtained. Value, the BP neural network algorithm is improved. The training results obtained by the improved BP neural network algorithm can effectively improve the accuracy of transformer fault diagnosis, improve the prediction accuracy of transformer fault diagnosis, and have a positive effect on the safe and stable operation of the transformer.