Drilling Data Based Approach for Equivalent Circulation Density Prediction While Drilling

The current research presents a machine learning application toward predicting the equivalent circulating density (ECD) in real-time while drilling operations. Maintaining high accuracy in estimating equivalent circulating density (ECD) is crucial due to the potential well control issues that can arise from inaccurate ECD values. Inaccurate ECD can lead to problems such as fluid losses, formation fracturing, kicks, and blowouts. Therefore, precise ECD estimation plays a vital role in preventing these issues and ensuring the safety and stability of drilling operations. The fourth industrial revolution provided high technology-based solutions to serve the oil and gas industry and machine learning is considered one of the best tools in research and development. The developed model within this study utilized the drilling hydraulic data in terms of initial mud weight, mud pumping standpipe pressure, and pumping rate to be fed to the random forest model to predict the ECD with good accuracy. A data set of 5, 140 data points was used through a comprehensive data analytics workflow to enhance data quality that greatly impacted the model accuracy. The obtained results showed the outperformance of the developed model to predict ECD with a correlation coefficient (R) of 0.99 between the actual and predicted ECD data with a low mean square error of 0.04. The developed model will add more to the drilling automation process toward safe and efficient operations through providing high accurate ECD values while drilling.