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

The substation is an important part of the electricity transmission system in the power grid. With the development of social economy and the increase of electric equipment used in our life and industry, more attention has been paid to the safety of the power system. Ensuring stable and safe running for the electric equipment becomes a key problem. It is a crucial step of the power grid management to construct and optimize the electric equipment temperature prediction and fault early warning management system.At the beginning of this thesis, the structure of the passive and wireless temperature measurement system, the principle of electric equipment temperature rise, the categories of equipment failure, and the characteristic of the temperature data are investigated. Next, the methods for temperature prediction and fault early warning for electrical equipment in the substation are analyzed. Taking the impact factors such as equipment, environment temperature and operating load into consideration, the equipment temperature is predicted from existing temperature data by using both time sequence and neural network in multiple dimensions and multiple levels. Then, the prediction performance of three frequently-used neural networks(Back Propagation Neural Networks, Radical Basis Function Neural Networks and Generalized Regression Neural Networks) under the background of this article are tested and the optimal prediction model is selected. The equipment temperature difference and temperature rise value in the period ahead can be acquired according to the predicted value. One or two methods including the surface temperature judgment method, the similar comparative judgment method, and the relative temperature difference judgment method could be used to diagnose failures and trigger alarm devices. Finally, a substation primary equipment key point temperature early warning system is designed and developed for Zhao Tong 110 kv substation of the Nan Zhai substation power grid company according to its real need.The MATLAB simulation shows that more information makes the prediction more accurate and the neural network model performs better than other statistical models. At the same time, BPNN is the optimal model when the sample set is large while GRNN is the optimal one when there is not too much sample data. Using BPNN prediction model in real temperature monitoring and early warning system provides strong support and security for the safety management of the substation electrical equipment.