Introduction of upscaling methods derived from the simple averaging method and a comparison with the Backus method of upscaling

Alfataierge, Ezzedeen; Chesnokov, Evgeni M.; Gorb, Yuliya

doi:10.1190/segam2016-13278544.1

Cited by 2 publications

(1 citation statement)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The challenges become even more discouraging with a lack of rock and fluid data and insufficient production history. Researchers have tried to accelerate simulations through techniques such as upscaling [5,6], multiscaling [7][8][9], and streamline methods [10,11]. However, all of these speed-ups require a full physics-based model as a starting point for simplification.…”

Section: Introductionmentioning

confidence: 99%

Development of Decline Curve Analysis Parameters for Tight Oil Wells Using a Machine Learning Algorithm

Dong

Lee

et al. 2022

Geofluids

View full text Add to dashboard Cite

To obtain a reliable production forecast, one has to establish a geological model with well logs and seismic data. The geological model usually has to be upscaled using certain upscaling techniques. Then, a dynamic reservoir model is constructed with another dataset, including completion data, production data, fluid properties, and relative permeability curves. At last, the dynamic model needs to be validated by a history matching process. This approach is data-intensive, time-consuming, and often not rigorously accomplished due to the lack of skillset and time. In this study, 10,000 groups of reservoir/completion input data were generated by Latin hypercube sampling method, and then, 10,000 groups of output (oil rate and cumulative production data) were obtained by numerical simulation. Next, a machine learning technique was applied to establish a model between the input data and determining parameters of a decline curve analysis model by fitting the generated cumulative production rate. Overall coefficients of determination ( R 2 ) of the three Arps decline curve factors were 0.966, 0.990, and 0.945. The validation result shows that the production rate and cumulative production predicted by the proposed machine learning–decline curve analysis (ML-DCA) model agreed well with those simulated by reservoir simulation. As a result of the ML-DCA regression model, a complete understanding can be established of the impact of reservoir properties on the DCA model. The proposed ML-DCA model not only provides a quick and robust method for petroleum engineers to estimate production performance for unconventional reservoirs from reservoir and completion properties without full-field geocellular modeling but also can be used to optimize the completion and operation parameters for wells of interest.

show abstract