The objective of this study was to develop an efficient decision-supporting tool to optimize completion designs for development of unconventional reservoirs across multiple benches. The workflow is specifically designed to find the optimum combination of completion fluid volume and proppant amount that maximizes economic metrics (i.e., the revenue from incremental oil from a larger completion that exceeds the extra completion cost). The workflow was applied to a drilling spacing unit (DSU) in the Midland Basin to assist in completion design selection for field development planning.
A machine learning (ML) model was developed with R2 =0.80 to predict 2-year cumulative oil production (CUMO) in response to multiple input parameters, including completion design variables, geologic properties, a depletion factor for production degradation, well sequencing information (e.g., standalone, zipper, or parent-child), and well spacing. The 2-year CUMO is then extended to a 30-year forecast using automated DCA, which enables the quick calculation of economic metrics with detailed economic assumptions so that the process mirrors actual field development planning. A genetic algorithm then determines the optimum completion design that maximizes NPV. The workflow also provides production profiles and heat maps of completion fluid and proppant combinations to assess tradeoffs in oil recovery and NPV.
The workflow was applied to 17 wells in four benches in a Midland Basin DSU. The economic metrics and oil recovery at the base completion designs were compared with the output of existing tools and demonstrated consistency. The optimizer suggests that changes to the fluid and proppant loadings for four wells in three benches would result in a 38% NPV and 13% CUMO uplift. The results show the tool captures the effects of depletion and well spacing for wells landed in both the same and adjacent formations. Contour heat maps visually provide oil recovery, economic metrics, and corresponding costs at all completion design combinations within the optimization window. The maps allow for efficient business decision making and the maximization of capital allocation efficacy both with and without capital budget constraints.
The efficiency and accuracy of the workflow appears to be particularly well-suited for fast-paced multi-bench unconventional field development and accounts for the inherently multivariate nature of well design with bespoke recommendations. A single automated tool streamlines the entire workflow from completion design selection to well performance and economic evaluation. The results of this study point to the viability and power of this machine learning-based workflow, which may serve as a particularly useful supplement to existing tools for unconventional development, such as decline curve analysis and reservoir simulation.
