oil and gas well spacing and completions: untangling complex interactions with machine learning

After choosing where to acquire an asset, the biggest levers an operator has for development are choosing spacing and completions designs. This isn’t an easy problem, as just about everybody you ask will agree that completions and spacing interact. The crazy part is that engineering teams don’t always evaluate the two in tandem.

Completions results from parent wells or exterior wells do not extrapolate to interior wells. In stacked plays, even four- or six-well pads may be composed of mostly exterior wells. Consider the hypothetical “tight well spacing” test in the Williston Basin, below. It’s likely that Middle Bakken and Three Forks wells interact to some degree. Yet, they certainly do not show the interference that same-zone wells do.

Do you think large completions designs will perform the same on this four-well pad as they would on full row development?

Untangling Data with Machine Learning Methods

Fortunately, we now have enough data in most major unconventional plays to untangle this problem with empirical, machine learning-based methods. These tools are great at handling both complex variable interactions and nonlinear relationships. Additionally, they can handle multiple spacing inputs to properly describe & evaluate the above scenario. As opposed to just one “spacing” value, but distance to all the possible neighbors.

Below, we’re showing one of the outputs of our machine learning model for the Williston Basin. The model is trained on data from the North Dakota Industrial Commission (NDIC).

On the x-axis we have average in-zone lateral spacing. On the y-axis we have spacing impact on performance (720-day cumulative oil), with the dots colored by proppant intensity. The model has learned that different completion sizes will result in different spacing-related degradation. Larger jobs will cause more well interference, whereas smaller jobs will cause less.

Oil Well Spacing & Production Model

The impact of going from wide to tight spacing is quite severe for the largest jobs. This leads to a loss of over 30% production for the large jobs compared to a much more mild ~10% degradation with small jobs. Clearly, completions tests from parent wells and widely-spaced wells do not translate to full row development. (And infill drilling is another question entirely!)

Maximizing returns requires thoughtful design considerations — in many cases, the largest jobs or the tightest designs will crush returns. Machine learning software for oil and gas from Novi Labs provides engineers the tools to right-size designs for optimized development plans.

Ready to take your developments to the next level? 

Unlock the full potential of your oil and gas assets with Novi.

Get in touch with us and see how our cutting-edge solutions can help you achieve better results.

You can request a free demo here: novilabs.com/demo

______

Terms & meanings

Parent well

Parent wells are the first wells drilled in a unit to extract oil or gas. Parent wells are typically drilled to hold the lease, provide estimates of productivity, and generate a return for the operator. Based on parent well performance and offset data, operators then determine the full development of the unit. 

Complex variable interactions

Refers to the presence of multiple variables with non-linear relationships between them. This can lead to difficulty in using traditional methods like scaling factors. Machine learning algorithms have to take these complex interactions into account when making predictions. 

Nonlinear relationships

The relationship between variables where the output is not linearly proportional to the input. In other words, a change in the input variable does not result in a constant change in the output variable. Nonlinear relationships are often more complex and difficult to model compared to linear relationships, where the output is directly proportional to the input. Nonlinear relationships can be modeled using various machine learning techniques such as decision trees, random forests, support vector machines, and artificial neural networks. The choice of technique will depend on the specific problem and the type of nonlinear relationship present.

Stacked plays

Stacked plays refer to the presence of multiple hydrocarbon-bearing rock formations in the same area. This term is used in the oil and gas industry to describe areas where there are multiple potential reservoirs of oil or gas that are stacked on top of one another. In the Williston Basin, the Middle Bakken and Three Forks reservoirs are separated only by the thin Lower Bakken Shale. Stacked plays are attractive to operators as they offer multiple potential targets, which in some cases can be developed at different times. 

Ted Cross

Ted Cross is Director of Product Management at Novi, where he builds powerful machine learning data and analytics products to solve the toughest problems in unconventional oil and gas. Prior to Novi, Ted was a Senior Geologist with ConocoPhillips.