An evidence based hydraulic fracturing and production optimization procedure is presented using modern data science and data analysis techniques by leveraging publicly available data from the Midland and Delaware basins of the Permian. Results show a point of diminishing return has either been crossed or is being approached as Exploration and Production companies have intensified completions designs by tightening stage lengths, increasing number of clusters, increasing lateral lengths and more importantly escalating proppant and water volumes.
Multiple public data sources including FracFocus Chemical Disclosure Registry and the Texas Railroad Commission are combined, cleaned and processed for analysis with a unique workflow to showcase design optimization. Three critical completion parameters: proppant volume, lateral length and proppant loading have been addressed for select counties in the Midland and Delaware Basins of West Texas. The relationships between these three parameters and other variables affecting well performance are investigated and normalized using multivariate statistics techniques. These topics can directly help operators leverage data science methodologies to use historical data for data driven decisions that assist in optimizing completion designs while generating sound well economics.
This paper focuses on more recent completion designs (2013 onwards) where hydrocarbon production information reported is statistically significant while simultaneously focusing on the more relevant completion design solutions driving the ongoing shale revolution. In general, increasing proppant loading improves well performance, however the relationship is nonlinear, and operators need to be cognizant of the degree of nonlinearity associated with proppant volumes, potentially size and type. The approach described and presented here tackles the Midland and Delaware Basin of West Texas and can be applied to other basins as the data used is available and public.
This paper discusses a method that maximizes the use of public information while drilling down to specific completion parameters to assist in data-based decisions. The techniques used can be expanded to implement artificial learning in asset development.