A New Predicting Method of Build-up Rate of Steering Tools Based on Kriging Surrogate Model

Abstract

Build-up rate is a key technical parameter of the steering tools in the drilling of complex directional wells, especially sidetracked wells, lateral wells and extended-reach horizontal wells. However, prediction of build-up rate is affected by various factors and very difficult to use explicit quantitative formula to portray its accuracy. So it is necessary to seek a scientific and efficient prediction method to be directly applied in drilling those wells, to improve wellbore trajectory control accuracy and speed, reduce development costs and improve economic efficiency. Based on hot methods of drilling engineering, from a regression analysis point of view, a novel method is proposed, which uses the Kriging surrogate model to build the predictive performance function to predict the build-up rate. In order to verify the effectiveness and superiority of the method, based on one field drilling data, Kriging surrogate model and commonly used regression models which are first-order polynomial regression models and radial basis function model, are built. Their three common performance indicators, i.e., root-mean-square error, maximum absolute error and average absolute error, are calculated and compared. The results reveal that, under different types of testing samples, the prediction performance of Kriging surrogate model is superior to the other two models in the three selected indicators. As to the computational process and prediction performance, the proposed method has better prediction performance, more robust prediction results, low computational complexity and more efficiency.