Take AIM!

Thanks to Artificial Intelligence, we can now improve well trajectory options while reducing manual work. Let’s find out more about AIM!

Planning a well is an incredibly complex task and it can take years from planning until the drill bit gets to work. Part of this planning phase is figuring out what trajectory your well is going to take – basically how to get from point A to B – and figuring out how to balance all the different tradeoffs you have to make in order to get the best well.

What qualifies as the “best” well depends on who you ask. A drilling and well engineer will say the best well is cheap with low risk, while a reservoir engineer will say the best well is the one that creates the most value. And the geoscientist will say the best well is one that avoids geohazards.

“This results in well planning being a series of trade-offs between several factors, for example risk, value and cost. As a result, we get a lot of iterations in the well planning process. If everyone’s working in their siloed tools, these iterations take a lot of time,” Bj?rn Peter Tj?stheim, AIM project manager, explains.

The goal of AIM (Artificial Intelligence for Maturation) is to auto-generate all possible well trajectories, truly changing the well planning process. AIM was developed in collaboration with WhiteSpace Solutions and is already in use on 11 of our fields, with more to come.

Instead of engineers manually creating a handful well trajectories and iterating on these, they will be spending their time and competence on selecting the best one from the options that AIM provides.

“The main value of AIM is in what we call non-human solutions – solutions that you wouldn’t think of in a manual process. This isn’t necessarily because the algorithm is smarter than our engineers, but because it considers all possible solutions. This is similar to how chess computers improve chess by coming up with new and surprising moves,” Bj?rn says.

Lowering the user threshold

One of the unique aspects to AIM is how it’s not just a domain-specific tool, it has users from both our Subsurface and Drilling & Well (D&W) domains. D&W will provide rules for well geometry, length of section, total length of well, dogleg, anti-collision and more. Subsurface will provide rules for targets, geohazards, over-pressured zones, colling zones, major faults and more.

“The trajectories AIM creates adheres to both Subsurface and D&W rules, so they don’t have to go into their separate software to adjust the time tube trajectory further. This saves us a lot of time. Another benefit is that it has a very low user threshold, compared to other related software like Compass or DSG,” Bj?rn explains.

So, what does the AIM workflow look like? Users first pick their fields, then look at input parameters and can change rule settings if needed. The next step is to pick your scenario: for example, you can have 5 starting points, known as slots, and three different targets you want to drill towards.

“Then, you select the factors and options you want to emphasize, hit compute and AIM will generate all possible slot-target combinations with several wells going between the different combinations. Computing time can take several hours for complex scenarios and mere minutes for simpler scenarios,” Bj?rn explains.

After computing is done, you can move to the visual inspection where you can also filter down towards the options that are most interesting to you as the user. “Then, you can ask AIM to optimize for you, based on cost, risk and value, before you can export the trajectories to Compass or DSG and work on them further.”

Avoiding collisions

With your filters and options selected and computing done, it’s time to review the trajectories AIM gives you. Which can be in the thousands, depending on your selection of options.

“No one likes to get 4,000 trajectories to choose between, so it’s incredibly important that we provide an interface where the user can quickly boil it down to what well trajectory options are the most important to them,” Bj?rn explains.

They do this by ensuring that the interface is customizable to each field.

“We tailor each version of AIM to the field it will be used on, so in effect we have 11-12 different applications. Traditionally, similar software has a one-size-fits-all approach, but with AIM we want to give our users what they want and need,” Bj?rn says.

When planning a well you also have to take existing wells into account. Naturally, you want to avoid collisions with them.

“What we do is import the existing wells with an uncertainty ellipsis around each well and for the wells we generate we must include uncertainty ellipses. Based on these we can then create safety factors and in that way ensure we avoid anti-collision,” Bj?rn explains.

“Technically, it’s not that complex to describe but implementing it is quite hard and now it’s a significant contribution to our compute time. We’re actively working on reducing the compute time going into anti-collision while still giving users maximum flexibility,” he adds.

Collaboration is the name of the game

Bj?rn explains that the success of AIM is due to the close collaboration between Equinor and WhiteSpace, and the fact that the algorithm was initially demonstrated using our own Volve datasets.

“We’re a small team on both sides of the table, so working both openly and closely together has been key for us. We share our data, and they share their code, and this open attitude has been important in making sure AIM is successful,” Bj?rn says.

While AIM is currently being used to plan well trajectories, there are far more use cases available. The technology and algorithm itself are quite generic, even though the latter has been tweaked extensively, and can easily be used for anything to generate paths from A to B.

“As long as we can describe rules we can test and implement. We can use AIM to test the well concepts of the future, simply by implementing a set of rules and evaluate the design trade-offs,” Bj?rn explains.

“We could potentially also use AIM to optimize offshore wind farms. Finding the right placement for wind turbine anchors is not that different from what AIM is doing now. It’s all about balancing cost, risk and value,” Bj?rn says.