Residual Lifetime Evaluation of Power Transformer Insulation Based on PSO-Wiener Model

Abstract

Power transformers play an important role in the operation of power system. Prediction of transformer insulation lifetime reasonably can improve the reliability of power grid and bring in economic benefits. In this paper, the transformer life correlation index (LCI) is constructed as its degradation data by fusing the multi-dimensional data including dissolved gas and polymerization degree (DP) in oil instead of existing single parameter. Then Bayesian updating and Maximum Expectation (EM) algorithm are used to update the parameters of Wiener model. To solve the problem that uncertain initial parameters of Wiener model bring stochastic error to life prediction, Particle Swarm Optimization (PSO) algorithm is proposed. The combined model is efficient to get the optimal solution of initial parameters to improve the prediction accuracy of remaining useful life of power transformers. A 500 kV power transformer in the field is taken as the case. The minimum loss between actual degradation path and predicted trajectory are compared and evaluated. Finally, its predicted total life is 37.96 years, which is in close to the general service life of transformers. Therefore, the PSO-Wiener model is effective for the practical application in the field.