Health Index Analysis of Power Transformer with Incomplete Paper Condition Data

Abstract

Transformer is a vital equipment in electrical power system that can degrade faster or slower than its designated life. In order to recognize the vulnerability of a transformer in a fleet, Health Index is commonly used. Conventional Health Index approach require all the data to be available in order to obtain accurate condition of a transformer. However, frequently incomplete data such as furfural is often faced by asset manager. This paper demonstrated the use of seven models to substitute unavailable furfural. Health Indices of 200 transformers with complete data were calculated, and compared to the alternative models. Multiple imputation approaches to predict paper condition of transformer using Multiple Linear Regression (MLR) and ANFIS (Adaptive Neuro-Fuzzy Inference System) had better agreement than other approaches shown by higher coefficient correlation with complete Health Index, as much as 0.959 and 0.960 respectively.