Prediction of vertical well inclination angle based on stacking ensemble learning

Well deviation is a common technical challenge in vertical well drilling operations. To accurately predict the Inclination angle in a certain oilfield in the Xinjiang work area, a Stacking-based ensemble learning method was established using historical drilling data from this work area. This method integrates Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbours (KNN) algorithms through a Stacking ensemble strategy. Genetic algorithms were employed to optimise the parameters of each base model. The study resulted in a prediction method for the Inclination angle suitable for this oilfield. Field test results show that the optimised Stacking-based learning model has the best prediction effect, with a 95.3% hit rate in predicting well inclination angle ± 0.01°, higher prediction accuracy than both single-base learners and traditional Stacking ensemble learning models. This method provides a new approach for predicting the Inclination angle and optimising construction parameters in this oilfield, thereby improving the efficiency of vertical well operations.