Optimization of Drilling Parameters of Target Wells Based on Machine Learning and Data Analysis

The drilling process optimization can reduce cost and energy consumption by optimizing different metrics (ROP, TOB and MSE). It is crucial to establish the relationship between drilling parameters and metrics. Most of the traditional methods are to optimize the physical model established based on experience or formula derivation. However, the optimized parameters are not optimal due to the low prediction accuracy of these models. This paper proposes an optimization process for drilling parameters using machine learning and data analysis. Compared with traditional models, data-driven models with offset well data have higher accuracy. The input features used by the model include weight on bit (WOB), rotational speed (RPM), flow (FLW), lithology and depth, where rock mechanics parameters are not required. The operator can optimize different metrics according to project requirements and make judgments in different aspects based on the corresponding changes in the coupling model. By analyzing the optimization results, we believe that the inappropriate WOB and RPM are the main factors limiting the improvement of ROP, and the drilling efficiency can be further improved by optimizing the FLW. We used this method to optimize the drilling parameters of the target well. MSE optimization resulted in a decrement in MSE by 55%, a decrement in torque by 25% and an increment in ROP by 20%.