For solving the problem of the weak classification due to the difficulty in artificially selecting feature during the power quality classification the traditional machine learning

a kind of classification method of highly comparable time series analysis（HCTSA）with combination of BP neural network is proposed in this paper

which can not only extract a large number of features

but also filter features without the need of artificial selection feature indicators. Firstly

a large number of algorithm databases are used to extract thousands of interpretable features from time series;then

linear classifiers are used to select forward features and the features with the largest amount of information in the class structure are selected as the optimal features. Finally

the optimal feature is used to extract the features of disturbance samples again and the BP neural network is used for classification. It is shown by the simulation result that the classification accuracy of hctsa-bp method is 97.3%

which is higher than the traditional wavelet-bp method by 8.7%.