Improving thermal-hydraulic modelling accuracy in CORE-Type transformers: a dynamic THNM approach.

Such thermal variations may accelerate the ageing of insulation materials, affecting the longterm reliability and lifespan of the transformer. Therefore, the development of methodologies capable of accurately informing about the temperature evolution within transformer windings, while remaining suitable for real-time operational use, has become a critical requirement for the energy industry today.

The mathematical models presented in IEC-60076-7 are widely employed to estimate hot-spot and top-oil temperatures under transient thermal conditions. However, accurate results from these models depend on prior calibration. In recent years, significant research efforts have focused on development of thermal hydraulic network models (THNM) for Power

Transformers. These models remain an attractive alternative to Computational Fluid Dynamic

(CFD) approaches due to their lower computational cost and reduced simulation time.

Additionally, CFD models often present challenges for real-time integration into transformer monitoring systems and in-service applications.

For THNMs to be effectively applied without prior calibration, they must describe in detail the hydraulic and thermal phenomena within the windings. Moreover, to support real-time applications, THNMs must be easily integrable in broader platforms and the simulation time of the dynamic THNM must be shorter than the real time transient processes, to enable performance prediction and serve as decision-support tool. The present work addresses the development and application of a proprietary dynamic and detailed THNM for core-type transformer windings, designated FluCORE Dynamic. The model is applicable to various geometric configurations and operating conditions, and it describes hydraulic and thermal behavior using algebraic, non-empirical equations. Although the spatial discretization is significantly reduced compared to CFD, the winding domain is divided into an optimal number of elements to ensure accuracy and numerical stability.

A winding composed of 54 discs and 6 radial passages, cooled with natural ester and operating under dynamic conditions, is analysed using both CFD and and the proposed THNM. The winding belongs to a full-scale experimental setup of a 15 MVA KDAF core-type three-phase power transformer subjected to heat-run tests. The model is first verified against CFD results and subsequently validated using experimental data. The simulated transient spans approximately five hours of real operation, while FluCORE Dynamic requires only 3% of the real-time duration. Compared to CFD simulations of the same winding, a 40% reduction in simulation time is achieved. The average absolute deviation is 2.8 °C for the hot-spot temperature and 1.4 °C for the top-oil temperature, confirming the accuracy of the proposed approach.

These results demonstrate that a detailed and dynamic THNM for core-type windings is both feasible and effective. FluCORE Dynamic combined reliable thermal predictions with significantly reduced computational requirements, representing an advance towards real-time monitoring solutions and decision-support tools, such as informed definition of loading conditions.