Hierarchical Detection and D-S Evidence Fusion-Based Insulation Diagnosis of Oil-Immersed Current Transformers

Abstract

Partial discharge monitoring of oil-immersed current transformers (CTs) is susceptible to electromagnetic interference, and single-parameter detection is biased due to incomplete information. To further improve the safety of the oil-immersed CT, an insulation diagnosis method combining hierarchical detection with Dempster–Shafer evidence (D-S) theory is proposed in this paper. A three-step diagnostic framework is constructed. At the first step, the liquid level is monitored since it changes with the gas generated by partial discharge. If an abnormality is identified in the first-step screening, features of partial discharge and oil chromatography are synchronously collected at the second step. At the third stage, the basic probability assignment (BPA) of evidence is constructed by adopting the Sigmoid membership function. Two types of diagnostic results, including normal state and discharge fault, are output through D-S evidence theory fusion. The unique contribution of this work lies in using simple oil level monitoring as a trigger for high-precision partial discharge (PD) and dissolved gas analysis (DGA) detection, which significantly reduces the diagnostic cost. Furthermore, by fusing the two evidence sources with D-S theory, the diagnostic reliability for oil-immersed CTs is improved. To verify the proposed method, tests are performed on a 110 kV oil-immersed CT. It is demonstrated that the rise in oil level is observable. Partial discharge signals are often mistaken for false pulses, while oil chromatography data has excellent anti-electromagnetic interference performance and can stably reflect the cumulative deterioration state of insulation. Restricting false partial discharge signals by oil chromatography data can reduce the misjudgment tendency caused by noise. After D-S evidence fusion, the fault confidence level is increased from 0.68 to 0.83, and the diagnostic uncertainty is reduced from 0.13 to 0.02, which effectively alleviates the ambiguity caused by the single partial discharge detection method. A cost analysis, compared with the PD and DGA fully installation scheme, further shows that the proposed scheme has significant cost advantage because the practical failure rate is far below the break-even value of 98.44%. The scheme proposed in this paper significantly improves the online insulation assessment accuracy for high-voltage current transformers via the hierarchical monitoring framework.