The power transformer is the most commonly used equipment in the power system, and winding deformation is one of the most frequent transformer faults. The short-circuit impedance method based on ΔU–I to construct the Lissajous figures is a newly proposed winding deformation detection method in recent years, but there are still problems such as difficulty in detecting the small deformation of the winding, and it is difficult to distinguish between different types of faults. In this paper, on the basis of the existing ΔU–I constructing Lissajous figures, a winding deformation detection method based on the least-squares fitting Lissajous track coefficient identification is proposed, which can effectively detect the winding deformation and determine the degree and type of faults. Firstly, the primary side current and voltage difference are used to construct the Lissajous figures, and the lower half of the Lissajous trajectory is excised to change the trajectory into a single-valued function; then the least-squares method is used for fitting, and the identification of winding deformation is carried out based on the coefficients of the power series of the fitted coefficients. The research results show that the method can effectively identify the small deformation of the winding and its degree of deformation, and distinguish the amplitude deformation and axial deformation, which has a certain engineering value.