With the acceleration of global energy transformation, thermal risk issues in power grid project clusters under the context of Energy Internet have become prominent. This paper addresses this issue by constructing a multi-source data fusion thermal risk monitoring model. Using the COMSOL Multiphysics 6.0 platform, four 72 hours simulation scenarios were designed for a 110 kV power grid cluster in an industrial park, consisting of two main substations and eight 10 kV feeders. Simulation results show that in the extreme high temperature scenario (Scenario 3), the top oil temperature of the main transformer reached a peak of 89.3°C, a 22.7°C increase compared to the normal operating condition (Scenario 1), and the average cluster thermal risk index (TRIcluster) was 0.68. In the load sudden change scenario (Scenario 2), the cable joint temperature rose by 18.5°C within 15 minutes, with the maximum single-device thermal risk index (TRImax) being 0.91. In the equipment aging scenario (Scenario 4), the TRIcluster fluctuation increased by 42% compared to Scenario 1. The calculated values and simulation results agree 91.3%, validating the model's effectiveness. Finally, a cluster collaborative dynamic optimization strategy is proposed to enhance the thermal security defense capabilities of the Energy Internet.