Method for judging abnormity of overhead line system based on autoregression and deep learning model

The invention discloses a method for judging abnormity of a contact network based on autoregression and a deep learning model. The method comprises six steps. According to the method, three different prediction methods including an empirical theoretical formula, a traditional prediction method and a deep learning method are adopted for data prediction, then measured data and prediction data obtained by the three prediction methods are used for drawing curves respectively, and then the obtained prediction curve of the B value and the measured curve are subjected to fitting comparison. And when the fitting degree of the actual measurement curve and the prediction curve is lower than a certain threshold value, judging that the overhead line system is abnormal, otherwise, judging that the overhead line system is normal. According to the method, manual inspection is not needed, the advantages of the theoretical formula, the differential integration moving average autoregression model and the convolutional neural n