Effective Seismic Inversion Through Microsoft Azure Cloud
Cloud computing has gained a significant foothold in the oil and gas industry, promising IT cost reductions, scalability opportunities, and collaboration improvements. Even among geoscientists, the cloud offers considerable efficiency opportunities as recently developed seismic inversion technologies can leverage the cloud’s high computing power for improved performance and reduced costs.
Cloud Computing: A Short Introduction
Simply put, cloud computing is a catch-all phrase for everything from data processing and data storage to software on servers made available via the Internet. Microsoft describes it as the delivery of computing services, whether that is servers, storage, databases, networking, software, analytics, or intelligence, over the Internet.
The cloud has considerable advantages when it comes to cost, time, and flexibility. Compared to traditional on-premise solutions, cloud computing offers significant cost-saving opportunities as its users only pay for the actual processing power and information storage they require. Efficiency improves as the need for underlying infrastructure cost and maintenance is reduced. Moreover, flexibility is enhanced as users can scale the cloud to their needs, customize applications, and access it from anywhere in the world.
These benefits have not gone unnoticed by the oil and gas industry. According to the Accenture and Microsoft survey “2017 Upstream Oil and Gas Digital Trends”, 45.7 percent of the surveyed companies are already investing in the cloud, and 40.1 percent are planning on doing so over the next couple of years.
BP, for instance, has selected Microsoft Azure as part of their global computing strategy, moving their proprietary data lake to Microsoft’s cloud platform for rapid data analysis, improved insights, and enhanced decision-making. ExxonMobil has turned to Microsoft’s cloud technology to collect real-time oilfield data, make faster drilling decisions, prioritize personnel deployments, detect leaks, and monitor greenhouse gas emissions. Likewise, Chevron has partnered up with Microsoft to connect many of its million sensors to the cloud.
Along with the advent of the newly developed inversion technique, Blueback ODiSI (One Dimensional Stochastic Inversion), Cegal’s geoscientists can now leverage the high computing power of Microsoft’s Azure cloud for improved and time-efficient seismic inversion.
ODiSI: Changing the Game of Seismic Inversion
For the uninitiated, ODiSI is a transparent and innovative approach to seismic inversion. Initially developed by BP and commercialized by Cegal, ODiSI captures all prior knowledge about a given field, prospect, or reservoir and enables the user to create thousands of pseudo-wells to build a robust and comprehensive model of the subsurface. ODiSI solves the problem of Seismic Inversion by brute force. Treating each trace location of the seismic independently, ODiSI composes a picture of your entire reservoir, one trace at a time and independent of the previous trace.
ODiSI uses all the known information about the target interval to build thousands of realizations of that trace within a Bayesian framework. These realizations are known as pseudo-wells; a pseudo-well is based upon a lithology column created from bed thickness statistics, a geological model and individual facies-based rock physics models. From the thousands of pseudo-wells ODiSI creates at each trace location, these are sampled in an accept/reject approach against the seismic data. The 100 best-matched, based on lowest RMSE, are then selected and this becomes the resolution for that seismic trace. ODiSI moves to the next trace and repeats the process. Over time you have a complete picture of the reservoir target as well as 17 associated petrophysical curves (Vp, Vs, Density, Vshale, etc.) for each trace location and pseudo-well. It is these synthetics which allow direct comparison between the seismic and well information. Uniquely, no statistical spatial control is imprinted, and all spatial continuity is directly inferred from the seismic.
Traditional seismic inversion techniques, such as deterministic and stochastic inversion, are limited by the number of well penetrations you have to refine the input model. Additionally, these techniques assume that your understanding of the spatial relationship is well known or can be modeled effectively. ODiSI solves this problem by combining the best of both techniques to provide a robust estimate of your reservoir properties.
While there is a large amount of calculation power required for such a calculation, the problem is highly parallelable as the calculation is one-dimensional. The Blueback ODiSI code is extremely efficient. What’s more, ODiSI can now rely on the high computing power of the cloud to considerably reduce the seismic inversion processing time.
Seismic Inversion in the Cloud
Seismic inversion is normally a time-consuming and expensive transformation requiring specialized expertise and significant computing power. Today, however, we can implement large volumes of data through ODiSI parameterization for rock physics work and use the high-computing power of the Microsoft Azure cloud to speed up the inversion run significantly.
The seismic information is broken into “blobs” composed of 64 traces; these traces are then run through the ODiSI algorithm within the cloud and then recompiled within the ODiSI program. The principle advantage is that each “blob” can be run on a separate machine within the cloud, dramatically improving the run times. It becomes a function of how many resources you wish to throw at the problem. Effectively, the process can be thought as taking a pane of glass, smashing it into thousands of pieces, understanding each piece in the cloud and then seamlessly stitching it back together.
Taking this to a real-world example, Cegal has worked on seismic problems in excess of 1200km2; at 25m bin spacing and more than one second of vertical thickness. Running this through ODiSI requires the creation of billions of pseudo wells and hundreds of billions of corresponding elastic curves. Such a problem would take months to run on a standard workstation but can be run in days on the cloud. Proving that any inversion problem becomes nothing more than a function of scalability. Furthermore, this can be done extremely cost-effectively utilizing the power of the cloud.
In summary, Blueback ODiSI not only provides an efficient method for seismic inversion, but the technique can dramatically improve understanding of your reservoir. Resultantly, the end user has more time and great products to spend more time extracting understanding and knowledge from their seismic data, ultimately adding significant value.
