Yingchun Xu, Xiangmao Cheng, Peifeng Huang, Runjie Lin, Yangrui Lin, Jia Weng
View Full Paper
High voltage cables; Fault identification; Bagging; K-nearest neigh-bour algorithm; K-means
As a key transmission channel in the power grid, fault diagnosis of high-voltage cables directly affects the safety and reliable operation of the power system. However, traditional methods are susceptible to noise interference and have limited generalisation ability when pro- cessing high-dimensional cable monitoring data. A K-nearest neigh- bour cable fault diagnosis method based on a Bagging sampling strat- egy and adaptive sample weighting is proposed to address this issue. This method improves the ability to discriminate low-dimensional signals by embedding a learning process that preserves global and local structural features. It also uses a diverse ensemble of subclassi- fiers to improve adaptability to complex fault modes. Experimental verification showed that the proposed method achieved a recogni- tion accuracy of 95.7% on the training set and 93.7% on the test set, both of which were superior to the other two comparison methods. In terms of computational resources, this method was computationally efficient while maintaining diagnostic accuracy. It had an average memory usage of 24 MB and a single inference time of 2.3 s. This effectively reduced the risk of misjudging faults and made online de- ployment feasible. The research results can provide reliable technical support for the intelligent operation and maintenance of power sys- tems.
Important Links:
Go Back
