A Method for Hot Spot Temperature Monitoring of Oil-Immersed Transformers Combining Physical Simulation and Intelligent Neural Network

The hot spot temperature (HST) of oil-immersed power transformers is closely related to their service life, so it is critical to accurately assess the HST. Currently, the existing methods for calculating the HST are not accurate, especially in scenarios where the ambient temperature and load change rapidly. In this paper, a novel HST monitoring method is proposed by combining the advantages of multi-physics field simulation and intelligent neural networks. With the established transformer high-fidelity simulation model, this method obtains plausible sample data under different scenarios of ambient temperatures and rapid rise and fall of loads. Then, based on the streamline feature analysis technique, the ambient temperature, load factor, and typical area temperature of the tank shell are extracted as feature variables, and a neural network-based HST evaluation model is established to achieve dynamic monitoring of HST. A transformer with windings configured with distributed optical fiber is analyzed as an example. The findings demonstrate that the approach presented in this paper is capable of adapting to variations in ambient temperature and load factor, while accurately computing the HST. The mean square error between the HST curve evaluated by the method and the measured curve is 0.94°C, which is significantly better than the guideline empirical formula and the thermal circuit model.