Vertical and horizontal inter-well communication in unconventional reservoirs remains a major uncertainty. This paper presents the results of geochemical analyses performed on several wells in the Bakken and Three Forks unconventional oil reservoirs. Geochemical analyses performed on oil extracted from core, oil sampled while drilling, and oil produced after stimulation indicate that the geochemical signatures of the Bakken and Three Forks Formations are different and unique to its respective stratigraphic units. Using unique geochemical signatures, this study developed a procedure for identifying the production of mixed oils and the relative contribution from each contributing startigraphic units. To further investigate vertical communication a detailed geologic model was constructed using core and outcrop data. The model was simulated and history matched to estimate contribution from adjacent layers. Various scenarios were simulated to understand the probability of communication. Analyses suggest that vertically adjacent layers contribute to production as predicted by the reservoir model and measured by the geochemical signature of the oil. This paper demonstrates (a) contribution from vertically adjacent formations can be significant, (b) geochemistry may be utilized to quantify vertical drainage, and (c) quantification of contribution from offset layers helped to constrain a reservoir simulation history match. Results from this study have facilitated the assessment of the degree of vertical communication across various flow units, which is the key to an efficient reservoir development.

It is widely accepted that unconventional resources hold enormous reserve potential. However, complex fluid flow physics and completion/stimulation practices pose a unique challenge in estimating reserves or making long-term production forecast for these unconventional reservoirs, as traditional methods are most often not applicable. This paper proposes the application of a probabilistic reservoir simulation workflow to provide realistic range of production forecasts with successful application in the Bakken unconventional tight oil reservoir. First, geomodels are constructed and ranked. Then, key static and dynamic uncertainty parameters are identified for the subsequent history-matching study (with each of the geomodel realizations) that provides not only production forecast for individual wells but also parameter ranges for experimental design (DoE) for field-level prediction. Then, DoE simulations are conducted to construct proxy equations that are used in Monte-Carlo simulations to generate Low-BTE-High response S-curves for the field-level models. Finally, based on these S-curve results, Low-BTE-High deterministic reservoir simulation models are constructed to generate corresponding long-term production forecast profiles for full-field development and optimization.

Performance prediction of wells producing from tight microdarcy formations is a daunting task. Complexities of geology (the presence/absence of naturally occurring fractures and contribution from different lithological layers), completion and fracture geometry complexities (multiple transverse and/or longitudinal fractures in long horizontal boreholes), and two-phase flow are impediments to simple performance forecasting. We demonstrate the use of various analytical and numerical tools to learn about both short- and long-term reservoir behaviors. These tools include (a) traditional decline-curve analysis (Arps formulation), (b) Valko's stretched-exponential method, (c) Ilk et al's power-law exponential method, (d) rate-transient and transient-PI analyses to ascertain the stimulated- reservoir volume, and (e) numerical simulation studies to gain insights into observed flow regimes. The benefits of collective use of analytical modeling tools in history-matching and forecasting both short- and long-term production performance of tight-oil reservoirs are demonstrated with the use of real and simulated data. Diagnosing natural fractures, quantifying stimulated-reservoir volume, and assessing reliability of future performance predictions, all became feasible by using an ensemble of analytical tools.