Abstract: Carbonate reservoir is characterized by strong heterogeneity and rapid production decline at the early stage, and the conventional production prediction methods can’t meet the production demands. The paper optimizes the conventional production prediction methods and discusses the basic models and matching methods of both exponential decline prediction and stretched exponential decline prediction. After the main factors influencing the production are identified, the producing wells are classified and an exponential decline model is used to match the production data of oil wells, thus the production prediction for producing wells during programming can be well guided. The stretched exponential decline model is applicable for the reservoirs with rapid production decline such as carbonate reservoirs. Taking the stretched exponential decline laws of producing wells in some similar reservoirs as references, the production of newly-drilled wells can be predicted. The combination of the 2 prediction methods can further improve the reliability of production prediction for carbonate reservoirs and guide the production practices of the similar reservoirs in the future