It is very important for scientific drilling and reservoir protection to predict formation pressure exactly. Considering the complex origin of pressure in the carbonate formation, we propose a new method that combines PCA (Principal Component Analysis) with BP neural network to construct the formation pressure predictive system integrating the drilling parameters. Based on the measured data of the carbonate reservoir in Yuanba zone, Sichuan province, the procedure to construct predicting geo-pressure models is illustrated in detail. The result shows that the precision of the pressure in the carbonate formation is up to 90%.So this method can be used to predict the geo-pressure in the carbonate formation and is highly valuable in practice.