Forecasting Transformer Reliability

In this paper the concept of an integral transformer lifetime model is presented. The model provides the best possible prediction of future behaviour given the data available. It treats remaining life in terms of future failure probability, thereby giving better support to the decision taking process than a mere remaining life estimate. The core of the model is a generic description of ageing processes, coupled to a probabilistic approach. The approach presented utilizes various techniques to reduce the uncertainty that is inherent to modelling processes with incomplete knowledge of the operational history. One technique couples the process model to externally measurable quantities; another technique involves a sensitivity analysis, which shows what additional input data gives the most efficient way to improve accuracy. We will illustrate the approach by applying it to a well-known degradation process: thermal degradation of the transformer winding insulation.