Leak-Off Pressure Using Weakly Correlated Geospatial Information and Machine Learning Algorithms

Leak-off pressure (LOP) is a key parameter to determine the allowable weight of drilling mud in well and situ horizontal stress. The LOP test run frequently used by petroleum industry. If exceeds LOP, wellbore instability may occur, with hydraulic fracturing large losses formation. A reliable prediction required ensure safe economical operations. challenging because it affected usually complex earlier geological loading history, values their measurements can vary significantly geospatially. This paper investigates ability machine learning algorithms predict leak-off on basis geospatial information measurements. About 3000 data were collected from 1800 exploration wells offshore Norway. Three (the deep neural network (DNN), random forest (RF), support vector (SVM) algorithms) optimized three hyperparameter search methods grid search, randomized Bayesian search) compared multivariate regression analysis. algorithm needed fewer iterations than find an optimal combination hyperparameters. showed better performance linear when features inputs properly scaled. RF gave most promising results regardless scaling. not scaled, DNN SVM algorithms, even parameters, did provide improved scores analyses also that number points geographical setting much smaller other areas, accuracy reduces significantly.