Accurate modeling of heat transfer devices is important for airborne electronic equipment cabin thermal prediction and thermal management. The current thermal models are mostly lumped model based on Thermal Network Model (TNM). The least square method is often used to compute and identify its parameters. Because of the principle of lumped parameter, thermal network model cannot correctly characterize the nonlinear temperature change process in the cabin, and the prediction accuracy is poor. Recently, neural network has gradually become a major research direction of heat transfer process modeling due to its powerful learning ability and data approximation performance. In order to achieve more accurate online thermal modeling of electronic equipment cabin, a sliding time window method based on Random Vector Function Link (RVFL) is proposed. By training the measured temperature in the electronic equipment cabin, the sliding window RVFLNN is built to predict the temperature of the equipment in the subsequent window. When the accuracy of this method cannot meet the requirement, the model is quickly updated according to the data acquired in real time. The real data experiments verify the effectiveness of this method as well as fast modeling speed.