Structural Optimization Model of Oil-Natural Air-Natural Transformer Radiator Based on Data-Model Hybrid-Driven
Chuan Zhang,Guoqiang Gao,Yujun Guo,Yijie Liu,Yicen Liu,Guangning Wu
DOI: https://doi.org/10.1016/j.applthermaleng.2024.125016
IF: 6.4
2024-01-01
Applied Thermal Engineering
Abstract:The power transformer is the central equipment of the power system. Establishing a model between radiator structural parameters and hot spot temperature (HST) is crucial for optimizing transformer cooling structures, enhancing transformer cooling capabilities, and ensuring the safe and stable operation of power systems. Existing models typically utilize data-driven methods. However, due to their lack of reasonable physical significance, existing models often suffer from issues such as insufficient generalization capability, high computational complexity, and poor interpretability, restricting their predictive accuracy and applicability. Therefore, this paper proposes a data-model hybrid-driven model (DMHDM) that integrates the radiator heat transfer physical model of the oil-natural air-natural (ONAN) transformer with HST data obtained from a finite element simulation model. Firstly, a theoretical analysis of the heat transfer process of the radiator fins is conducted, leading to the establishment of a low-fidelity physics model. Then, employing the full factorial design method and CFD model, sample data is obtained to develop a high-fidelity HST prediction model based on the data-driven method. Finally, a CFD model of a single-phase ONAN transformer was constructed. The DMHDM was compared with the traditional response surface methodology (RSM), and the sources of errors were analyzed. The results indicate that DMHDM provides better interpretability, improves accuracy by 20.5% within the sample set, enhances generalization capability by 98.6%, and reduces computational complexity by 92.5%. This study provides an efficient and feasible framework for establishing the relationship between structural parameters of ONAN transformer radiators and HST.