Predicting Drilling Rate of Penetration Using Artificial Neural Networks

Abstract

Abstract Oil well drilling processes are generally high-cost operations; however, these costs can be reduced by optimising drilling operations. In particular, it is vital to optimise the drilling rate of penetration during oil well drilling operations. Many parameters affect drilling penetration rate, and these have complex relationships with each other. The accurate prediction of the drilling rate of penetration has particular importance in the optimisation of all drilling parameters and the reduction of drilling costs. In this paper, a neural network model is developed to predict the rate of penetration for an Iraqi oil field as a function of well depth, drilling fluid inflow, bit rotation speed (RPM), weight on bit (WOB), standpipe pressure, and bit size. The data on which the network was trained was collected from one drilled oil well, and 3,939 data points were used to develop the new model. These were randomly divided into two parts, with 70% used for training and 30% used for testing. The results showed that the resulting neural network model offers high accuracy for predicting the drilling rate of penetration. The statistical analysis showed that the developed neural network model predicted the rate of penetration very high accuracy (correlation coefficient of 0.983 and average absolute error of just 7.78%). The new model can also be used to determine the optimum drilling parameters to obtain a desired rate of penetration.