State estimation of 500 kV sulphur hexafluoride high-voltage CBs based on Bayesian probability and neural network

Abstract

Circuit breakers (CBs) are of vital importance for the stability of power systems, and a new two-stage hierarchical state estimation method is proposed for 500 kV sulphur hexafluoride CBs based on Bayesian probability and perceptron neural networks. On the basis of the samples collected from Yunnan Power Grid in China, a new indicator system is constructed by association rules. Bayesian probability is applied to measure the correlation between the individual indicators and comprehensive indicators at the same status level, to weigh the individual indicators. Also, an adaptive perceptron is improved to train the weights of comprehensive indicators in different operational conditions, to eliminate the influence of the imbalance problem of relative deterioration. Then, the operating state of equipment can be inferred according to the calculated comprehensive scores. Finally, taking the actual operating equipment as an example, the effectiveness of this proposed method is proved by sample tests and comparison with other existing linear methods.