Coupled Drilling Optimization Based on Hybrid Physics-Machine Learning Models

Abstract Drilling operations account for a significant portion of well costs in the oil and gas industry. In the context of a cost-sensitive environment, there is an increased focus on drilling optimization. Traditional research has often targeted the optimization of the Rate of Penetration (ROP) or Mechanical Specific Energy (MSE) in isolation, overlooking the complex interactions between these parameters and drilling vibrations. Existing optimization models, which are either machine learning based or rely on classical physical models, exhibit limitations in predicting ROP, MSE, and vibrations accurately. This study aims to develop a comprehensive drilling optimization model based on physics-machine learning (physics-ML) model to simultaneously optimize ROP and MSE while mitigating drilling vibrations. This research entails processing an extensive dataset that includes weight-on-bit (WOB), rotary speed, flow rate, rock strength, vibration, ROP, torque, and MSE. The dataset is strategically divided, with 80% for model training and the remaining 20% reserved for prediction evaluation. ROP and surface torque models are then established based on domain concept and machine learning algorithms. The proposed physics-ML model, employing a residual approach, achieves an impressive R2 of 0.9836, underscoring its exceptional accuracy and interpretability. By integrating the ROP and torque models, a comprehensive MSE model is established. The HPSOGWO algorithm is subsequently applied to identify the most effective operational parameters, optimizing crucial control variables such as RPM, WOB, and flow rate. A novel target function is introduced, adeptly incorporating ROP and MSE with customized weights to balance their significance. The optimization results are remarkable, with ROP enhanced by 8% and MSE reduced by a staggering 26%. These outcomes not only demonstrate the practicality of the proposed optimization method but also highlight its superior performance in boosting drilling efficiency and curbing energy consumption.