Reservoir Software Developer: RTA Had Its Day, Hybrid Modeling Now the Way

Abstract

Rate transient analysis (RTA) is defined as the science of analyzing and forecasting production data. But in its application to ultratight reservoirs, it’s also been likened to an art form. There is no unified approach for reservoir engineers to follow in RTA, and as such, its solutions are nonunique. In other words, the accuracy of the analysis lies in the eye of the beholder. Ardent practitioners argue that none of this is a revelation, nor negates the value that RTA can deliver when used by a skilled reservoir engineer. Those closer to the middle of the spectrum will add that RTA is “good enough.” Another camp holds that the analysis method originally designed in the 1970s for forecasting in conventional reservoirs has reached its useful end in unconventional reservoirs. Among those turning the page on RTA, along with more traditional decline curve analysis (DCA), is a startup called Xecta Digital Labs. Founded in 2019 by a team of experienced oil and gas technologists, the firm has proposed a proprietary hybrid model as the way forward. It combines a reduced-physics approach with artificial intelligence and machine learning (AI/ML) techniques to issue daily estimates of well deliverability potential—a metric otherwise known as a transient productivity index (PI). Together—the model and the PI output—represent the core of Xecta’s software-as-a-service that was ultimately designed to guide major field development decisions. In framing the importance of using a reduced-physics approach, Sanjay Paranji, cofounder and CEO of Xecta, said, “The idea here is that in shale plays you have thousands of wells, and it’s very hard to routinely come up with a full-physics representation of the reservoir and all of the known calculations for well performance in a succinct and practical manner across all those wells.” But with a reduced-physics model, he explained that it’s possible to run constant simulations while “retaining the fidelity” of the first principles dictating much of reservoir and well behavior. The AI/ML methods add further constraints to improve accuracy along with offering Xecta the compute speed and scale it says is needed to generate recommendations across entire developments. On industry adoption, Xecta reports that its platform has been deployed to varying degrees by a handful of undisclosed US shale producers with major holdings covering thousands of wells in the Permian Basin and Bakken Shale. A key differentiator from existing technologies touted by the software maker is how it normalizes transient PIs for each well based on any significant changes to how they are operated, e.g., an increase in choke size, transitioning from one type of artificial-lift system to another, etc. Adding this context is what Xecta says enables its approach to cut through the noise found within tight-oil production data and surface the key drivers behind improved recovery rates. In terms of other inputs, the bare minimum required by the workflow are rate and surface pressure—data points that are collected daily from virtually all modern horizontal wells. Xecta adds that it employs a largely automated workflow that, once established, runs with zero human interpretation in the loop.