Applied methodology for temperature numerical evaluation on high current leads in power transformers

The majority of power transformer manufactures use their own proprietary design to build oil-immersed power transformers with high current busbar leads (HCBL) because there is no standardization or registered standard in the technical literature. Anyway, specialists certainly require a highly accurate temperature estimation regardless of the proprietary designs used to build these power transformers. It is well known that the temperature at the copper HCBL surface is mainly influenced by the Joule effect induced by the passing current, paper wrapping insulation characteristics, and temperature of the surrounding oil. In this context, we suggest a novel noninvasive technique for estimating the temperature of HCBL power transformers. Such methodology is simple and can be used to estimate the temperature for any power transformer despite designs strategies, becoming an algorithm that can be used in several commercial programs. The temperature is taken as a function of variables and measurements easily acquired from the transformer and, therefore, turns out to be an original and simple method. The methodology is based on the evaluation of the convection heat transfer coefficient h, which depends on the geometry factors and cooling fluid characteristics of the transformer. Since temperature levels are critically sensitive to h variations, Computer Fluid Dynamics (CFD)-based commercial software are a common-place tool for thermal analysis of power transformers as coefficient h must be calculated with very high precision and correctly evaluated. However, this solution have both high financial and computational burden. The skin and proximity effects are taken into account in the proposed methodology by using the Finite Element Method a<SIC>euro" FEM. Thus, we propose an efficient and simplified solution for temperature estimation. We will show that the solution is as accurate as the CFD programs, albeit, more economically and computationally viable. We validate the results by means of measurements made with Fiber Optic Temperature Sensors during a full-load test and show the estimation numerical accuracy for parameter h.