Reliability assessment based on BP neural network for relay protection system with a few failure data samples
Dai Zhihu · Dianli zidonghua shebei · 2014
A method based on BP neural network is proposed for the reliability assessment of relay protection system with a few failure data samples. The distribution models suitable for the reliability assessment of relay protections and their features are analyzed. The BP neural network is then trained by the small amount of original failure data sample to get the expanded data samples whose rules are similar to those of the original ones. The least square method is applied to estimate the distribution model parameters of expanded data samples. Case study shows that,the reliability assessment based on the expanded data samples has better effect and the proper empirical formula should be adopted according to the selected distribution model.