Research on Digital Analysis Method of Transformer Hot Spot Temperature Based on BP Neural Network Optimised by Genetic Algorithm

Abstract

As the core equipment of the transmission system, the high hot spot temperature of the transformer will accelerate the ageing of the transformer, which will lead to thermal faults. It has become an urgent task to introduce digital technology to monitor and analyse the hot spot temperature in real time. In this paper, the magnetic, current and thermal multi-field coupling numerical analysis method is used to establish the digital model of the transformer and carry out simulation analysis, and the transformer loss data and temperature distribution are obtained. Infrared thermal image thermometer, thermocouple temperature measuring device and optical fibre temperature sensor are used to monitor the thermal state characteristic parameters of the transformer. From the two dimensions of multi-source and heterogeneous, a multi-source heterogeneous information fusion method is proposed to extract, clean, and filter experimental data. Taking 100 sets of data as the sample data, the first 80 sets of data are used as training sets to construct a BP neural network model optimised by the genetic algorithm, and the last 20 sets of data are used as prediction sets to test the model, so as to predict the hot spot temperature. By comparing with other algorithms, it is found that the evaluation constraint index of GA-BP is the smallest, MRE is 0.0651 and RMSE is 0.2158. MSPE was 0.44%. Combined with finite element analysis and external measurement data acquisition, a digital twin platform is developed to realise real-time evaluation and analysis of the transformer operation status, which is of great significance to the operation and maintenance of transformer equipment.