APPLICATION OF PSO-LSSVM IN PREDICTION AND ANALYSIS OF SLOW DRILLING (RATE OF PENETRATION)

Abstract

The application of artificial intelligence to predict the accuracy of the rate of penetration (ROP) is very important in optimizing drilling parameters and increasing ROP during drilling. Slow drilling refers to a rate of penetration (ROP), not at the desired level. Drilling an oil well is usually expensive, but this cost can be reduced by carrying out optimal operations. In this paper, the model used is PSO-LSSVM to predict penetration rate. This requires drilling data sequentially and predicting ROP continuously and has a higher success rate in predicting ROP, especially Hole Depth, weight on bit (WOB), Bit Rotation per minute (RPM), Torque, Hook Load, and Standpipe Pressure. The trained data was collected from one drilled oil well, and 7,553 data were used to create the model and were divided into 70% trainset and 30% test-set. The results show that the PSO-LSSVM model has a high level of accuracy in predicting drilling penetration rates. The statistical evaluation shows that the developed PSO-LSSVM model has a high level of accuracy (MAE = 20.10, MSE = 757.9, RMSE= 27.53, and R2 = 0.83) and also in this study PSO which is used in conjunction with SVM in the ROP prediction, the optimum value (best pos): -0.0429913, 0.00350291, and the optimum value (best cost): 0.00186. The results show that the LSSVM optimized with PSO has stronger search and convergence capabilities and higher prediction accuracy for ROP prediction in well X.

.