Estimation of Drilling Rate of Penetration Using Hybrid Machine Learning Methods

Abstract

Abstract Rate of penetration(ROP) is one of the most important well drilling parameters, and its estimation and optimization is very important during well planning and reducing related costs. Meanwhile, the prediction of this parameter is challenging due to the complex interactions between the drill bit and the formation rock. In this study, different Machine Learning(ML) estimation techniques including Artificial Neural Networks(ANN), Random Forest(RF) and Least Squares Support Vector Machine (LSSVM) are hybridized with meta-heuristic algorithms, including Crow Search Algorithm(CSA), Genetic Algorithm(GA) and Particle Swarm Optimization(PSO) has been used to estimate ROP more accurately. The aforementioned meta-heuristic algorithms have been used to intelligently adjust hyper-parameters of estimation ML methods based on data. The results show that it will significantly improve the estimation performance. Among the models, RF-GA, RF-CSA and LSSVM-GA were recognized as the top three models, respectively. The value of the correlation coefficients between the estimated and the actual values of ROP in these models was 0.98, 0.974, and 0.972, respectively. Also, the mean square error (RMSE) values for these models were obtained 2.89, 3.25 and 3.37, respectively. Depth, mud weight and rotation speed are identified as the most influential parameters in the response of estimation models. The findings emphasize the effectiveness of combining ML methods with meta-heuristic algorithms to accurately estimate drilling penetration rates. The results provide valuable insights to optimize drilling operations, reduce costs and increase drilling performance in oil fields. The results of this study in the field of drilling optimization can be useful in engineering-based drilling decisions.