Real-time Evaluation Model Construction of GIS Conductor Thermal State Based on PSO-BP Neural Network

The thermal state of the conductors inside the GIS is an important indicator to assess its safety, but due to the closed nature of the GIS, it is often difficult to achieve a relatively fast monitoring. At present, multi-physics field simulation with the help of finite element method can realize the rapid monitoring of the thermal condition inside GIS, but some of its input parameters are difficult to obtain, and often can only” explore the causes from the results”. Meanwhile, it is difficult to guarantee the real-time performance. In this paper, we propose and construct a real-time evaluation model of conductor heating condition based on multi-physics field simulation and a BP neural network optimized by particle swarm algorithm. With the multi-physics field simulation model, parameters such as current, conductor resistance, gas pressure, and shell temperature rise are extracted as inputs, and contact resistance and conductor temperature rise are used as outputs to train the PSO- BP neural network to build the inverse model of GIS conductor heating state. The inversion system is highly accurate and real-time, reducing the input parameters that are difficult to obtain, and effectively realizing the improvement and enhancement from “exploring the causes from the results” to “deriving the results from the causes”.