Frequency response analysis (FRA) is an effective method for detecting internal winding deformation faults in transformers. However

in traditional FRA methods

only the amplitude response is typically used to interpret FRA characteristics

while the importance of phase response has not been adequately emphasized. To address the above issues

this paper proposes a transformer winding deformation fault diagnosis method based on frequency response reconstruction and digital image processing. This method firstly integrates the frequency

amplitude and phase elements into a three-dimensional coordinate graph and applies image processing techniques for texture analysis feature extraction. Subsequently

the fault type classification is achieved using the triangular coordinate statistical graph of image sum square ratio error (ISSRE) index trigonometric graph

and parameterized threshold indicators for winding fault type discrimination are summarized. Then

based on the polar coordinate frequency response characteristics

a quantitative indicator

namely

image sum square max-min ratio error (ISSMMRE)

is proposed to identify the degree of winding faults. Finally

the experiments with various faults were conducted on a custom-designed power transformer

demonstrating the effectiveness and feasibility of the proposed method

and providing a new idea and method for the improvement and automation of FRA method.