An Earth-to-Air Heat Exchanger (ETAHE) uses the massive thermal storage capacity of the ground to dampen ambient air temperature oscillations by delivering outdoor air to the indoors through a horizontally buried duct. Owing to their low airflow resistance, large cross-sectional area ETAHEs have been found more energy efficient than the conventional small duct ETAHEs, especially when integrated in hybrid ventilated buildings. However, the lack of available methods for determining the surface heat convection has made accurate energy simulation and design difficult.  A field investigation in a large ETAHE was carried out at beginning of this research. Detailed airflow and heat transfer data were collected to analyze the heat convection process and to verify computational fluid dynamics (CFD) simulations. In the CFD model, a two-layer turbulence model was used to ensure accuracy in resolving flow information in the near-wall region. The modeling method was verified by comparing its results with measured data. The results indicated that the entrance and buoyancy effects in ETAHEs caused surface heat convection distribution to be highly non-uniform, and its intensity is significantly higher than that in conventional small duct ETAHEs. Using the CFD model, sensitivity of heat convection intensity to various design parameters was analyzed, and six parameters were identified as being influential. A large number of CFD simulations were then performed to find mathematical relations between the design parameters and local heat convection rate. Based on these cases, an Artificial Neural Network based Heat Convection (ANN-HC) algorithm was developed. It predicts local Nusselt numbers on ETAHE surfaces, and the results are in good agreement with CFD predictions. A thermal simulation model of ETAHEs was developed to solve three-dimensional unsteady conductive heat transfer in the ground surrounding ETAHEs, and the ANN-HC algorithm is coupled with the model to provide local heat convection boundary conditions at duct surfaces. A case study showed that the new model can properly simulate the interactions between an ETAHE and its environment, and the results are more accurate than the existing simulation models. It will be a useful tool for designers to predict and analyze ETAHE performance in order to obtain optimal design solutions.