Study on Deep-water Jet Drilling Parameter Optimization Based on Machine Learning

Abstract To enhance the efficiency and safety of shallow conductor jet drilling in deep-water environments, this study proposes a machine learning-based method for optimizing jet drilling parameters. The interquartile range (IQR) method was first employed to handle outliers in the collected dataset, and mutual information analysis was utilized to identify the key factors influencing the rate of penetration (ROP). A nonlinear mapping relationship between jetting parameters and ROP was constructed using the Gated Recurrent Unit (GRU), achieving a prediction accuracy of 92%. Building on this model, an intelligent optimization framework was developed by integrating an evolutionary algorithm. Weight on bit (WOB) and flow rate were selected as decision variables, with the maximization of ROP as the objective and bearing capacity constraints as limiting factors. This optimization method was successfully applied to a deepwater well in the South China Sea, resulting in a 22.6% increase in the average ROP. This research offers a novel and intelligent solution for improving the efficiency of deep-water jet drilling operations and serves as a practical reference for future applications in similar environments.