Study on Prediction of Top Oil Temperature for Transformers Based on Bayesian Network Model

Ran Li,Ya Yuan Yi

doi:10.1051/matecconf/20166301027

Abstract

The top oil temperature for transformer has a great influence on transformer’s operational life and load capacity, therefore, it is important to predict the top oil temperature. On the basis of analyzing and summarizing the main impacts on the top oil temperature, an idea is proposed to predict the top oil temperature by means of Bayesian network, and Bayesian network model is established. The model takes active power, reactive power, load current, ambient temperature and previous time oil temperature as its quantitative indicators, and trains the sample data to find out the probability distribution between various factors. The model is verified according to data collected from the transformer of SSZ11-50kV/220. The results show that the relative error between predictive value and measured value is small, which can be accepted completely in engineering. Therefore, Bayesian network is reasonable and can be widely applied to forecast the top oil temperature.

Highlights

With the rapid development of science and technology, electricity that an emerging green energy has been widely used in various fields which range from the national defense science and technology to the people's livelihood
Affecting the top oil temperature,and both of those are interactive.,Bayesian networks can be used as a effective method of analyzing and forecasting the top oil temperature.After careful analysis and consideration,the paper choose 5 factors,namely active power,reactive power,load current, ambient temperature around transformer and the previous time oil temperature,and build Bayesian network model according to the 5 parameters.The model has been tested,and obtain predictive data.When predictive datais compared withthemonitor data,it is found that the predictive data is close to monitor data.And when Bayesian network model is compared withBP model,the Bayesian network model outperforms obviously theBP model consistently in relative error.,Bayesian network model is appropriate to predict transformer’s top oil temperature
Bayesian networkprovides an effective method with regard to the causal information presentation.In addition, it is one of the most effective theoretical models in the areas of expression and reasoning of uncertain knowledge currently[12].Bayesian network is a directed acyclic graph (DAG), that is made up of nodes and directed edges.Nodes represent random variables,and directed edges between nodes represent the relationship between the nodes

Summary

Introduction

With the rapid development of science and technology, electricity that an emerging green energy has been widely used in various fields which range from the national defense science and technology to the people's livelihood. The large demand for electricity lead to the power system developing along the direction[1] of large capacity, high voltage, large grid , power grid,and smart grid,and the development is more and more rapid, which requires power supply and reliability of the power system to achieve a higher level.In power systems,the transformer plays an very important role[2,3] in economic transmission, flexible allocation, safely use of electricity.Winding hot spot temperature is an important performance index of the transformer, and it is closely related to the operation life and the load capacity of transformers.Once transformer’s windings get overheated[3],the internal insulation will decreaseand ultimatelyaffect the normal operation of the transformer.,it is difficult to measure windings hot spot temperature[4].In practice, top oil temperature are usually used insteading of windings hot spot temperature to scale whether the transformer operate normally. Predicting the top oil temperature timely and taking protective measures in advance notonly to ensure the stable operation of the transformer,and extendedlife. Network model according to the 5 parameters.The model has been tested,and obtain predictive data.When predictive datais compared withthemonitor data,it is found that the predictive data is close to monitor data.And when Bayesian network model is compared withBP model,the Bayesian network model outperforms obviously theBP model consistently in relative error.,Bayesian network model is appropriate to predict transformer’s top oil temperature

The Bayesian network

Bayesian network inference

Prediction of top oil temperature by Bayesian network

Conclusions