2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Quantifying the Impact of Well Spacing on Bakken Production: A Multivariate Study


The Middle Bakken formation of the Williston Basin has been commercially developed for the last fifteen years. Initial horizontal drilling programs were designed as single one-mile laterals per 640-acre drilling unit. As technology continued to develop, the play saw increasing numbers of more closely spaced wells with longer laterals and more intense completions. This not only increased per-well production but also encouraged economic development of poorer quality reservoir. This study was undertaken to isolate, understand, and quantify the impact of well spacing on production in Mountrail County, North Dakota, but additionally it provided insights into how other variables also control production. Estimated ultimate recovery (EUR) of oil was selected as the production metric for two reasons: 1) oil represents most of the value from Bakken wells; and 2) a long-term metric helps avoid potential bias in favor of closely spaced wells; that is, since the more closely spaced wells generally came later in the play’s development and were thus completed with higher proppant loads, they usually resulted in higher initial production than pre-existing widely spaced wells - but not necessarily higher long-term production. Hierarchical clustering analysis was used to classify 381 wells into four different reservoir quality classes. Next, multivariate linear and non-linear regression was used to build a class-dependent model predicting oil EUR from what were found to be the most dominant variables: water cut, lateral length, resistivity, porosity, GOR, proppant per foot, and well spacing. The model’s correlation coefficient is 0.902; but do the mathematically-derived relationships of the predictor variables to the response variable make physical sense? This question is answered through inspection of the optimal transform plots generated by the regression algorithm (Alternating Conditional Expectation), which validates the data relationships and allows quantification of well spacing’s impact on production. In high-quality reservoir, spacing closer than 3500 feet decreases EUR, while in low-quality reservoir wells can be spaced as closely as 1500 feet before EUR is negatively impacted. Several model realizations were then calculated, each normalized for engineering parameters. These show the economic feasibility of downspacing mediocre reservoir to eight wells per section and of completing wells in poor reservoir with 2500 pounds of proppant.