Power time series data multi-label hidden danger classification method based on dimension conversion

The invention discloses an electric power time series data multi-label hidden danger classification method based on dimension conversion. The method comprises the following steps: (1) collecting one-dimensional time series data of an electrical fire; (2) preprocessing the collected one-dimensional time sequence data; (3) constructing a classification model which comprises a dimension conversion module, a semantic attention module, a graph convolution network module and a classification module; the working process of the model is as follows: the dimension conversion module converts preprocessed one-dimensional time sequence data into a two-dimensional tensor through Fourier transform, and extracts a feature map from the two-dimensional tensor; the semantic attention module decomposes the extracted feature map into a plurality of content-aware category representations; the graph convolutional network module enables the category representation to pass through a static graph convolutional network and a dynamic gr