AbstractThe failure of electrical equipment directly or indirectly caused by overheating has become one of the main reasons for equipment accidents. The real‐time condition monitoring method of electrical equipment based on digital twin (DT) has received extensive attention and is considered as a technology with great engineering values and excellent application prospects. However, the current calculation of DT mostly relies on the traditional finite element method (CFEM). This method becomes less computationally efficient as the size of the DT increases, especially for electrical equipment with high complexity. It is difficult to meet the requirements real‐time calculation in DT. Therefore, starting from the algorithm optimization and parallel architecture, based on the weighted residual method and stabilised conforming nodal integration, a novel discrete method of electrothermal coupling equation is proposed and the data storage structure of the algorithm from the bottom layer is redesigned, and the corresponding GPU and multi‐core CPU parallel framework are proposed. Finally, taking the high voltage bushing as an example, the correctness of the method in this paper is verified by the CFEM code and commercial software ABAQUS. Under the same grid and accuracy requirement, the calculation time is shortened by at least five times than ABAQUS. And the method is easy to extend to other types of multi‐physics calculations.