Abstract

Equivalent circulating density (ECD) is considered a very important parameter during drilling operations because when ECD is too high, it directly affects the reduction of the penetration rate, increasing the level of sticking of the drill string, and loss circulation, especially in the drilling section with narrow drilling mud windows, increasing pressure loss in the wellbore, and reducing the impact force capacity of the drill bit, leading to an increase in nonproductive time (NPT). ECD measurement must ensure high accuracy to avoid problems during drilling, such as well control. In practice, measuring ECD is very expensive, and other ways to predict ECD from mathematical methods provide low accuracy, which leads to high risks in drilling operations. The level of accuracy of ECD is questionable because there is still sometimes sticking drilling string or loss of circulation while drilling. The main objective of this study is to use machine learning techniques to predict the ECD model when drilling wells in the Oligocene formation of the White Tiger field based on surface drilling parameters without using downhole measurements. The study provides an ECD prediction model using artificial neural networks (ANN). Six surface drilling parameters collected in the field from drilling wells in the Oligocene Formation include the penetration rate (ROP), pump flow rate (Q), rotation speed (RPM), surface pump pressure (SPP), weight on the bit (WOB), drill string torque (T), and wide range of ECD. The model ECD was trained, tested, and validated to deliver highly accurate predictions for ECD. The results showed a strong overall prediction of ECD with a correlation coefficient (R) greater than 0.99. ECD model prediction is based on several drilling parameters of drilled wells with high accuracy by artificial neural networks (ANN), thereby minimizing risks in drilling operations of adjacent wells when adjusting surface drilling parameters so that ECD is always controlled lower than the fracture pressure.

This content is only available via PDF.
You can access this article if you purchase or spend a download.