APPLICATION OF PRINCIPAL COMPONENT ANALYSIS FOR POWER TRANSFORMER ASSET MANAGEMENT

Measurements from insulating liquid are utilised to indicate the health of power transformer insulation systems. Due to various dynamics within the transformers, assessing the transformer health through condition monitoring data and their risk profile require input from transformer experts. This paper investigates the application of Principal Component Analysis (PCA) as an automated tool in ranking transformers based on their condition monitoring data with minimum expert input. Oil test data obtained from a transformer fleet over a five-year span was used for the analysis. For each transformer, a single value for each measurement was obtained by using the 90^th percentile data from the five-year measurement span. Furthermore, multiple imputation was applied for transformers with missing data. The rankings obtained from PCA was compared against the ranking developed based on expert knowledge. Results indicated that PCA has potential to identify transformers with poor health condition with minimum expert input.