Summary In this paper, we propose a machine–learning methodology using domain–knowledge constraints for well–data integration, prior/expert–knowledge incorporation, and sweet–spot identification. Such methodology enables the analysis of the effects of the main variables involved in production prediction and the evaluation of what–if scenarios of production prediction within geological zones. This methodology will allow streamlining the process of data integration, analytics and machine learning for better decisions, saving time, and helping geologists and reservoir and completion engineers in the task of sweet–spot identification and completion design. We tested the proposed methodology with production, completions, and petrophysical data from a field within a geological target zone. For instance, using local Kriging, we estimated gamma ray features from gamma ray measurements from vertical and horizontal wells, and we integrated those features into the production– and completion–well data, generating an integrated data set for machine–learning modeling. Besides the usual black–box machine–learning models, we used generalized additive models (GAMs) and shape–constraint additive models (SCAMs) for predictive modeling. Those models permit the incorporation of prior/expert knowledge in terms of interaction terms and mathematical constraints on the shape of the effect of the covariates, such as petrophysical and completion parameters, resulting in greater accuracy and interpretability of the predicted production vs. classical black–box machine–learning modeling. We also defined hypothetical what–if scenarios of oil production, such as by estimating the empirical distributions of production estimates using hypothetical settings for completions within the region of interest.
Ambiguity in a geophysical interpretation problem is the possibility of accepting more than one solution caused either by solution nonuniqueness or instability. Nonuniqueness is related to the existence of more than one solution regardless of the precision of observations. On the other hand, instability is related to the acceptance of different solutions producing data fittings within the expected observational errors. We studied the ambiguity in the inversion of well‐logging data using a method based on the analysis of a finite number of acceptable solutions, which are ordered, in the solution space, according to their contributions to the overall ambiguity. The analysis of the parameter variations along these ordered solutions provides an objective way to characterize the most ambiguous parameters. Because this analysis is based on the geometry of an ambiguity region, empirically estimated by a finite number of alternative solutions, it is possible to analyze the ambiguity due not only to errors in the observations, but also to discrepancies between the interpretation model and the true geology. Moreover, the analysis can be applied even in the case of a nonlinear interpretation model. The analysis was performed with recorded data, and compared with the analysis using singular value decomposition, leading to comparable results. Following the determination of the most ambiguous parameters, a reparameterization is possible by grouping these parameters into a single parameter leading to a simpler interpretation model and, therefore, to a drastic reduction in the ambiguity.
We present a new methodology for modeling the cumulative oil and gas production of horizontal wells in a shale play given two types of well completion parameters: lateral length and proppant intensity, we also consider the location of the wells and the shut-in days of wells. We show experimental results evaluating the predictive accuracy of this methodology on hold-out data and compare it to standard data-driven estimation procedures. Our approach is based on the use of Generalized Additive Models (GAMs). The main advantage of using GAMs is that we can easily obtain effect plots that allow us to quantify the effect of each completion parameter and the location of the well on the production. Furthermore, it is possible to explicitly model the interaction between variables and impose shape constraints based on physical knowledge. We have implemented and tested this methodology in R and compared its predictive accuracy against a variety of standard data-driven modeling procedures available in the Caret package, using leave-one-out cross validation. We present experimental results using data from a shale play with 152 horizontal wells that have cumulative production for more than 12 months. Our experiments show that using GAMs in most cases leads to better predictive accuracy than the standard data-driven estimation procedures available in Caret, possibly because they allow us to explicitly model the interaction between input variables that are known to be are physically related and to impose shape constraints based on known physical relationships between input variables and target variables. We conclude that it is possible to accurately estimate the effect of the completion parameters on the cumulative production when we ensure that the models obey physical constraints, i.e., that the cumulative production is a monotonic increasing function of lateral length and of proppant intensity. We also obtain effect plots showing the "sweet-spot" effect, i.e., the relationship between the location (x, y) in the field and the cumulative production. The novelty of this hybrid data-driven and knowledge-driven methodology is to allow the quantification of the effect of completion parameters and location on well production through the combination, in a principled way, of data available from existing wells with prior/expert knowledge regarding known physical constraints.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.