• Energy Research

Abstract Steam-Assisted Gravity Drainage (SAGD) recovery is strongly impacted by distributions of heterogeneous shale barriers, which impede the vertical growth and lateral spread of a steam chamber and potentially reduce oil production. Conventional reservoir heterogeneities characterization workflows that entail updating static reservoir models with dynamic flow data are quite time-consuming. Furthermore, certain assumptions are often needed to approximate the complex physical processes. This study proposes a workflow integrating artificial intelligence (AI) in a model selection framework that aims to identify associated shale heterogeneities in SAGD reservoir based on extracted features from production time-series data. A series of SAGD models based on typical Athabasca oil reservoir properties and operating conditions is constructed. After constructing the base homogeneous model, the shale barriers are assigned randomly by sampling their location, lateral extent, and thickness from several probability distributions, which are inferred from field data assembled from the public domain. Sensitivity analysis is carried out to identify and analyze features in the production response that are related to shale characteristics: whenever the steam chamber encounters a shale barrier, a drop in the production is observed; this drop continues until the steam chamber has advanced past the shale barrier, and the production would rise again. Several types of input feature extraction methods are introduced in this work: piecewise linear approximation (PLA), cubic spline interpolation (CSI), and discrete wavelet transform (DWT). Next, artificial neural network (ANN) is constructed to calibrate a relationship between the retrieved production pattern parameters (inputs) and the corresponding geologic parameters describing shale heterogeneities (outputs), which include some variables capturing the location, orientation or size of a particular shale barrier encountered by the steam chamber. The final model is implemented in a novel characterization workflow to infer shale heterogeneities from production profiles. A number of realistic applications are presented to illustrate its utility. The ANN models are validated using numerous synthetic models, where the exact shale distributions are known. The trained ANN models can reliably estimate the relevant shale parameters and the associated uncertainties, while accurately predicting the corresponding production responses. It is intended to extend the proposed method to construct the ANN models directly from well logs and production data. This work presents a preliminary attempt in correlating stochastic shale parameters with observable features in production time-series data using AI techniques. The proposed method facilitates the selection of an ensemble of reservoir models that are consistent with the production history; these models can be subjected to further history-matching for a precise final match. The proposed methodology does not intend to replace traditional simulation and history-matching workflows, but it rather offers a complementary tool for extracting additional information from field data and incorporating AI-based models into practical reservoir modeling workflows.

Summary The steam alternating solvent (SAS) process involves multiple cycles of steam and solvent (e.g., propane) injected into a horizontal well pair to produce heavy oil. These solvent-based methods entail a smaller environmental footprint with reduced water usage and greenhouse gas emissions. However, the lack of understanding regarding the influences of reservoir heterogeneities, such as shale barriers, remains a significant risk for field-scale predictions. Additionally, the proper design of the process is challenging because of the uncertain heterogeneity distribution and optimization of multiple conflicting objectives. This work develops a novel hybrid multiobjective optimization (MOO) workflow to search a set of Pareto-optimal operational parameters for the SAS process in heterogeneous reservoirs. A set of synthetic homogeneous 2D is constructed using data representative of the Cold Lake reservoir. Next, multiple heterogeneous models (realizations) are built to incorporate complex shale heterogeneities. The resultant set of SAS heterogeneous models is subjected to flow simulation. A detailed sensitivity analysis examines the impacts of shale barriers on SAS production. It is used to formulate a set of operational/decision parameters (i.e., solvent concentration and duration of solvent injection cycles) and the objective functions (cumulative steam/oil ratio and propane retention). The nondominated sorting genetic algorithm II (NSGA-II) is applied to search for the optimal decision parameters. Different formulations of an aggregated objective function, including average, minimum, and maximum, are used to capture the variability in objectives among the multiple realizations of the reservoir model. Finally, several proxy models are included in the hybrid workflow to evaluate the defined objective functions to reduce the computational cost. Results of the optimization workflow reveal that both the solvent concentration and duration of the solvent injection in the early cycles have significant impacts. It is recommended to inject solvent for longer periods during both the early and late SAS stages. It is also noted that cases with higher objective function values are observed with more heterogeneities. This work offers promising potential to derisk solvent-based technologies for heavy oil recovery by facilitating more robust field-scale decision-making.

Production forecast of steam-assisted gravity drainage (SAGD) in heterogeneous reservoir is important for reservoir management and optimization of development strategies for oil sand operations. In this work, artificial intelligence (AI) approaches are employed as a complementary tool for production forecast and pattern recognition of highly nonlinear relationships between system variables. Field data from more than 2000 wells are extracted from various publicly available sources. It consists of petrophysical log measurements, production and injection profiles. Analysis of a raw dataset of this magnitude for SAGD reservoirs has not been published in the literature, although a previous study presented a much smaller dataset. This paper attempts to discuss and address a number of the challenges encountered. After a detailed exploratory data analysis, a refined dataset encompassing ten different SAGD operating fields with 153 complete well pairs is assembled for prediction model construction. Artificial neural network (ANN) is employed to facilitate the production performance analysis by calibrating the reservoir heterogeneities and operating constraints with production performance. The impact of extrapolation of the petrophysical parameters from the nearby vertical well is assessed. As a result, an additional input attribute is introduced to capture the uncertainty in extrapolation, while a new output attribute is incorporated as a quantitative measure of the process efficiency. Data-mining algorithms including principal components analysis (PCA) and cluster analysis are applied to improve prediction quality and model robustness by removing data correlation and by identifying internal structures among the dataset, which are novel extensions to the previous SAGD analysis study. Finally, statistical analysis is conducted to study the uncertainties in the final ANN predictions. The modeling results are demonstrated to be both reliable and acceptable. This paper demonstrates the combination of AI-based approaches and data-mining analysis can facilitate practical field data analysis, which is often prone to uncertainties, errors, biases, and noises, with high reliability and feasibility. Considering that many important system variables are typically unavailable in the public domain and, hence, are missing in the dataset, this work illustrates how practical AI approaches can be tailored to construct models capable of predicting SAGD recovery performance from only log-derived and operational variables. It also demonstrates the potential of AI models in assisting conventional SAGD analysis.