Optimized support vector regression for drillingrate of penetration estimation

Abstract In the petroleum industry, drilling optimizationinvolves the selection of operating conditions for achievingthe desired depth with the minimum expenditurewhile requirements of personal safety, environment protection,adequate information of penetrated formationsand productivity are fulfilled. Since drilling optimizationis highly dependent on the rate of penetration (ROP), estimationof this parameter is of great importance duringwell planning. In this research, a novel approach called‘optimized support vector regression’ is employed for makinga formulation between input variables and ROP. Algorithmsused for optimizing the support vector regressionare the genetic algorithm (GA) and the cuckoo search algorithm(CS). Optimization implementation improved thesupport vector regression performance by virtue of selectingproper values for its parameters. In order to evaluatethe ability of optimization algorithms in enhancing SVRperformance, their results were compared to the hybridof pattern search and grid search (HPG) which is conventionallyemployed for optimizing SVR. The results demonstratedthat the CS algorithm achieved further improvementon prediction accuracy of SVR compared to the GAand HPG as well. Moreover, the predictive model derivedfrom back propagation neural network (BPNN), which isthe traditional approach for estimating ROP, is selectedfor comparisons with CSSVR. The comparative results revealedthe superiority of CSSVR. This study inferred thatCSSVR is a viable option for precise estimation of ROP.