Hindcasting production in four shale gas basins using a physics-based approach

In this work I attempt to characterize uncertainties in production forecasting. The analysis focuses on how decline analysis results, that depend on the production profile of older wells, change as the population ages. Also I investigate systematic and random errors that affect EURs. This work is a retrospective on production forecasting performed in the Marcellus, Fayetteville, Haynesville and Barnett shale gas resource plays. Publicly available and subscription-based production data on 34,000 shale gas wells were forecast multiple times over a six-year period. We perform a hindcasting analysis on these wells, where production has been forecast using a physics-based decline curve analysis (DCA) and other published approaches. This allows us to test the accuracy of the decline methods and assess reasons for errors in this analysis. I also review data quality and its effect on production forecasting. Using this analysis, we find that there are a number of systematic and random errors that affect ultimate recovery estimates. Production forecasts for wells in which the production has been back allocated from lease level to well level show regression to the mean over time. This has an effect on the errors associated with individual well forecasts in these fields. Also, fields where few wells have entered boundary-dominated flow (BDF) show higher uncertainty for field-wide production forecasts. A proper accounting for individual well and population uncertainties is necessary for sampling data and risk assessment. Understanding the sources and magnitudes of errors and uncertainties in EUR values for unconventional gas wells allows operators to account for these in determining economic outcomes and financial planning of wells.