Transformer winding deformation fault identification method

A transformer winding deformation fault identification method comprises the following steps that a transformer winding lumped parameter model is established, and normal winding amplitude-frequency data and phase-frequency data are obtained; adjusting model circuit parameters to obtain amplitude and phase data of the winding in different fault states; obtaining new characteristic indexes from the obtained amplitudes and phases through a moving window calculation method, and establishing a new characteristic sequence; obtaining a cosine value of a polar coordinate through an arc cosine of a new feature sequence normalization result, performing conversion to obtain a Gramb angle difference field numerical value, and generating a corresponding thermodynamic diagram according to the Gramb angle difference field numerical value to obtain a required image data set; and establishing a convolutional neural network (CNN) model, training the image data set, finally carrying out fault identification and classification, a