Remaining Useful Life of Oil-Immersed Transformers: A Combined CFD and Surrogate Model
Please login to view abstract download link
The lifespan of oil-immersed power transformers depends heavily on the ``hot spot temperature" (HST) within the insulation. High HST can accelerate aging and reduce the transformer's life, hence the importance of accurate HST assessment. Among the methods used to determine the HST, a convective model based on Computational Fluid Dynamics (CFD) and Heat Transfer is widely employed for its accuracy. However, this model is computationally expensive, making real-time monitoring challenging, especially in scenarios with rapid fluctuations in ambient temperature and load. This paper proposes a novel approach to predict HST by employing an efficient surrogate model based on a model-order reduction technique called sparse Proper Generalized Decomposition \cite{ibanez2018multidimensional}. The methodology uses a detailed 3D CFD model, implemented with OpenFOAM \cite{OpenFOAM}, to generate accurate data for various parametric scenarios. This data is then used to build a surrogate model capable of real-time HST prediction. Additionally, the approach predicts the remaining useful life of power transformers by leveraging real operational data over several years and adhering to the international standard IEC 60076-7. The results demonstrate that this approach effectively adapts to variations in load and environmental conditions, providing accurate real-time temperature predictions for transformer monitoring.