Transformer oil temperature prediction based on PSO-SVR

Transformer oil temperature prediction is crucial to ensure power system stability and transformer selfmaintenance. In view of the limitations of traditional prediction models in terms of accuracy and real-time performance, as well as the advantages of deep learning techniques in dealing with complex datasets, this study proposes a particle swarm optimization support vector regression (PSO-SVR)-based transformer oil temperature prediction model, which significantly improves the prediction accuracy by optimizing the key parameters of SVR. In this paper, the theoretical foundations of SVR and PSO are first outlined, followed by a detailed description of the construction of the PSO-SVR model and the process of model training and validation with actual transformer oil temperature data. Through the iterative calculation of the PSO algorithm, when the penalty parameter is taken as 0.2201 and the radial kernel parameter is taken as 0.1864, the assessment indexes such as the mean square error of the model prediction results reach the lowest, which is much lower than the traditional SVR model, showing high prediction accuracy and good applicability, and providing a new technical way for intelligent management of the transformer oil temperature and the power system.