Prognostics of transformer paper insulation using statistical particle filtering of on-line data

Abstract

The adoption of prognostics for critical assets has the potential to advance asset management in the power industry significantly. Whereas diagnostic techniques can identify the presence of incipient faults, prognostics aims to predict the future state of a given asset [1], [2]. Prognostics can therefore be used to estimate the remaining useful life (RUL) of the asset, and help plan maintenance while minimizing the risk of failure in service. Prognostics requires a good model of the process of deterioration, from inception through to failure [1]–[3]. Deterioration may be due to aging, as in the case of paper insulation, or it may be due to a fault. Regardless of the cause of deterioration, prognostics is useful only if the deterioration is slow enough that maintenance (repair or replacement) can be scheduled during the predicted RUL. Thus, prognostics is not superior to diagnostics if the deterioration is so rapid that failure cannot be prevented.