A novel fuzzy approach for winding paper insulation health monitoring of oil-immersed power transformer using non-destructive failure parameters

Power transformers, a significant and critical component of power system, require regular condition monitoring by the utilities for uninterrupted power supply. The kraft paper made of cellulose is used as the solid insulation of the transformer which undergoes thermal, electrical and environmental stresses causing its progressive ageing. It thereby reduces the physical strength of the paper and ultimately affects the transformer's remaining useful life (RUL). The current state (physical) of the cellulosic (solid) insulating paper is predicted in this article using the fuzzy inference system (FIS) based RUL assessment technique. Three non-destructive parameters, that is, 2-Furfuraldehyde (2-FAL), Carbon Monoxide (CO) and Carbon dioxide (CO₂) are used to determine the value of the Degree of Polymerisation (i.e. DP) which directly reflects the current physical state of the insulating material. Four states (Healthy, Poor, Very Poor, and Worst) of the insulating paper are identified based on the values of DP. Five FIS models have been developed separately with different membership functions and a number of fuzzy if-then rules are formulated to estimate the value of DP. Every model is tested with 30 test samples to validate the system accuracy. The physical condition of the insulation estimated using FIS built with trapezoidal membership function has been found to be considerably precise with an average absolute deviation of 0.33%. Hence, a comprehensive assessment of transformer's health is carried out with the prominent non-destructive parameters.