Transformers are susceptible to insulation aging and failure under high temperature and overload conditions, which affects the normal operation of the power system. The winding hot spot temperature is a key factor in evaluating transformer load capacity, but it is challenging to measure directly. To accurately calculate the hot spot temperature, a new thermal circuit model is developed based on the thermoelectric analogy principle, considering the effects of external environmental factors, internal nonlinear thermal resistance, and oil viscosity. This model integrates traditional mechanism analysis with a data-driven method and employs an improved differential evolution algorithm, which uses population initialization and adaptive mutation factor based on prior knowledge. The fourth-order Runge-Kutta method is used to solve the differential equation of the thermal circuit model, with the 180 MV·A/220 kV transformer from a substation in southern China as a case study. The calculated hot spot temperature is compared with the measured values, and the determination coefficient of the proposed model reaches 0.84, showing an improvement of 14.60% and 5.53% over the IEEE and Susa models, respectively. The comparison demonstrates that the proposed mechanism-data fusion thermal circuit model offers significant advantages and enhances the accuracy of hot spot temperature calculation for oil-immersed transformer windings.