A method for diagnosing the state of insulation paper in traction transformer based on FDS test and CS-DQ algorithm

Traction transformer is a vital equipment in high-speed railway. Insulation status decides its safety and reliability, and frequency domain dielectric spectrum (FDS) test is one of the most effective methods reflecting the changing of insulation status. For field applications, the following problems should be addressed: 1) how to obtain the result of paper insulation from the combined result of oil-paper insulation system; 2) how to discriminate the defects of insulation paper between aging and damp. In this paper, the first problem was transferred to a non-linear equation set, and a CS-DQ algorithm (cuckoo search algorithm optimized by the differential evolution algorithm and the quadratic interpolation method) was proposed to solve it. Then, the insulation states were discriminated by establishing a multi-class LS-SVM (least square support vector machines) model, in which CS-DQ algorithm was also used. Finally, a diagnostic approach for the insulation paper in traction transformer was proposed. The results in laboratory show that the pure result of insulation paper can be obtained and the insulation states can be discriminated effectively by using the proposed approach. Meanwhile the proposed CS-DQ algorithm has better performance than the conventional CS algorithm. The results of field testing also verify the proposed approach