An effective principal component regression method for transformer life management based on indirect parameters

Transformer life management has gained remarkable recognition due to its crucial role for safe operation of the power grid. Since transformer life is dominated by insulating state of its oil-paper insulation, the DP value has been traditionally regarded as the primary indicator. However, the DP sampling procedure is destructible and inconvenient. In this paper, we present an effective principal component regression method to assess the insulating state of oil-paper through indirectly estimating the degree of polymerization based on the aging characteristic parameter in oil. Thermal aging experiments were firstly conducted on palm oil and mineral oil impregnated paper, respectively. After aging experiments, aging characteristic parameters of oil were tested, including moisture, acidity, 2-furan, surface tension, dissolved gas-in-oil analysis, methanol, and ethanol. Principal component regression method was then performed to find the principal component and build a relationship between the degree of polymerization and the aging characteristic parameters of oil. After computation, the estimation of degree of polymerization value can be obtained with a high goodness of fitting, which is 0.92 and 0.78 for mineral oil and palm oil impregnated paper respectively. The estimation could be improved by data processing.