Research On Distribution Transformer Life Combination Prediction Based On Cross Entropy Theory

Distribution network is an important link connecting the power system with the user.Distribution transformers play a vital role in it, therefore improving their managementlevel as well as the reliability of their operation has important significance on the safeoperation of the grid. Besides increasing the reliability level in transformer design andmanufacturing process, using modern techniques and methods to predict its remaininglife then arrange the state overhaul reasonable to achieve maximum service life safety isalso very important. Distribution transformers life prediction model describes thedevelopment and variation of distribution transformers’ health status and operating life,and answers how long it can be used in current health status. It is the base of distributiontransformers’state maintenance.This paper has established a complete distribution transformers life predictionmodel, including distribution transformer health index model and residual life predictionmodel. In health index model, it digitally described the status of the distributiontransformer, finally the quantitative result is defined as the health index of distributiontransformer(HI).This paper, based on the transformer health index of calendar year, put forward akind of distribution transformer based on the theory of cross entropy life combinationforecast model. In residual life model, failure probability density function and crossentropy theory was introduced to dynamically determine the weight of the improved graytheory and markov prediction algorithm. It is more real-time, considering the life cycle ofthe distribution transformer failure probability density function of the shape parameterand scale parameter and comparing to the idealized view that the aging factor B in theentire life cycle of a given value. It is not only able to predict the remaining life of thetransformer, and can determine the health index variation with operating years, to achieveoperational status in real time forecasting of the transformer. Actual data verify theeffectiveness and reasonability of the model and it is more in line with the actualsituation compared to a single prediction method.