A test method of winding deformation excited by pseudorandom M-Sequences — Part I: Theory and simulation

A test method to detect winding deformation by using pseudorandom binary M-Sequence excitation signals is proposed. This broadband excitation signal and the corresponding response from the winding are combined to obtain the analytic expression of the winding's transfer function by using a Recursive Least Square method. The coefficients, poles, and zeros of the transfer function can be used to describe the winding status, and any mechanical changes in the winding over time can be therefore detected. The steps and algorithms involved in using the M-Sequence method are presented, and a comparative simulation between the M-Sequence method and Frequency Response Analysis (FRA) is carried out using an equivalent circuit for the winding. Results of the simulation confirm that the response sensitivity of the M-Sequence method is as good as FRA in a common frequency band, and is actually more sensitive for slight winding deformations of series capacitance. The correlation factor of the transfer function coefficients, used as example, can also diagnose slight deformation. Additionally, significant differences exist in the map of the poles and/or zeros before and after deformation. The study confirms the feasibility of using the M-Sequence method for detection of winding deformation in transformers. Transfer function can have a wider application than frequency response curves, then analysis and judgment based on the transfer function may offer more analysis tools. The theory and simulation study of the new method provide supports for subsequent development of its corresponding detecting system as well as experimental research.