The invention discloses a winding deformation intelligent identification technology based on a transfer function characteristic principal component and a neural network, belonging to the field of winding deformation fault detection. First, the zeros and poles of the current winding transfer function or frequency response are obtained, and the characteristic matrix is constructed from their frequencies and amplitudes, as well as their variations with respect to the reference winding. Secondly, the principal component of the eigenvector matrix is extracted by principal component analysis, and the principal component eigenvector is normalized. Finally, the BP neural network is constructed and trained, and the BP neural network is used to identify the current mechanical state of transformer windings. According to the idea of dimension reduction, the invention effectively extracts the principal component characteristic quantity of the winding state, and reduces the calculation amount and time. Compared with manual identification, the invention can well characterize the nonlinear relationship between the type, position and degree of winding deformation and the winding structure, overcomethe subjectivity and randomness of manual identification, and realize the intelligent identification of winding deformation.