Research On Temperature Prediction And Fault Warning For Electrical Equipment In Transformer Substation

Transformer Substations are an integral part of the power grid.With the development of China's economy and the surge in electricity consumption for daily production and production,the safety of electrical equipment has received more attention from the society.Therefore,to ensure the long-term stable operation of power equipment,scientifically optimize the layout of power equipment,and strengthen the temperature control and control of power equipment.The emergency warning plan has become the focus of daily grid maintenance management,and the temperature prediction and fault early warning research of electrical equipment in substations has become a proposition worthy of academic research.This thesis first studies the composition of the substation temperature measurement system,the factors affecting the temperature rise of electrical equipment,the types of thermal faults,and the characteristics of temperature data samples.Then starting from the basic data research,analyze the current temperature prediction and fault diagnosis principles and methods.Diversely combine the two methods of time series and neural network to design a device temperature prediction system and use three common neural networks.Models,that is,BP(Back Propagation)neural network model,radial basis neural network model and regression neural network model,are used to perform system performance prediction,and the optimal prediction method is selected through simulation.Based on this,the predicted temperature value is grasped,and the fault is diagnosed by the surface temperature measurement method,the similar comparison method and the contrast temperature difference method,and the alarm is quickly issued.Finally,according to the actual use needs of the substation,a set of core temperature early warning system for conventional equipment of the substation was developed and designed for Shaotong Power Supply Bureau 110 k V Nanzhai Substation Grid Company.This thesis uses MATLAB simulation experiments to confirm that the predictive ability of neural network models is much better than linear statistical models.And when the amount of sample data is large,the BP neural network model is the best prediction model.In terms of actual temperature prediction and early warning systems,the use of BP neural network prediction models can better manage the temperature of electrical equipment in substations,and provide scientific and efficient security for daily work in substations.