Prediction of Rate of Penetration (ROP) in Petroleum Drilling Operations using Optimization Algorithms

Abstract

In drilling operations, by choosing the proper tools and also incorporating more accurate and reliable parameters, this operation can be performed in less time and cost manner. Among drilling parameters, Rate of Penetration (ROP) is viewed as the main parameter in drilling operation evaluation. Field data investigations can be considered the most fruitful approaches to evaluate drilling performance, or ROP, as well as development of predictive models although laboratory tests and experimental formulas are vastly used to identify the drilling problems. In this research, intelligent modeling was used to predict the penetration rate of drilling operations through analyses of an established comprehensive data base from drilling operations in one of Iranian oilfields, Shadegan oilfield, in which novel artificial intelligence techniques such as Grey Wolf Optimization (GWO), Particle Swarm Optimization (PSO), and Grasshopper Optimization Algorithm (GOA) were applied. Since the database includes 400 data, these techniques were utilized due to their effectiveness on a large set of data. In this research, using drilling data compiled from Shadegan oilfield, a precise model was developed to predict the ROP. Results showed that determination coefficient (R²) and Root mean squared error (RMSE) parameters for Particle Swarm Optimization (PSO) are found to be as R²=0.977 and RMSE=0.036, for Grey Wolf Optimization (GWO) R²=0.996 and RMSE=0.014, for Grasshopper Optimization Algorithm (GOA) R²=0.999 and RMSE=0.003, respectively. Ultimately, it can be concluded that all predictive models lead to acceptable results but GOA yields more precise and realistic outcome.