Real-time estimation of the formation pressure can enhance the drilling operations in terms of non-productive time, costs, and safety. In the literature, the known methods for formation gradient prediction are derived from logging or a combination of selected drilling data and logging. The objective of this paper is to use random forest (RF) to develop a predictive model for real-time pore pressure gradient using surface drilling parameters. The used inputs were pump rate (PR), standpipe pressure (SPP), rate of penetration (ROP) and rotary speed (RS). A field data set was used to develop the prediction model. Another data set was used for validating the developed model. The proposed model estimated the pressure gradient with a correlation coefficient (R) of 0.99 and 0.98 for training and testing. The root mean squared error (RMSE) was around 0.01 and 0.014 psi/ft for training and testing. In addition, the average absolute percentage error (AAPE) was 0.96% and 1.66% for training and testing.
Formation pressure is the one exerted by the fluids contained inside the pores of formations. The normal pressure gradient ranges between 0.433 psi/ft and around 0.465 psi/ft (Bourgoyne et al., 1986). Abnormal pressure is used for any deviation from the normal pressure trend that can be either supernormal or subnormal (Mouchet, J.P., and Mitchell, 1989).
