Saltar al contenido

3D oil

The latest technology in seismic imaging makes it possible to perform large-scale scans of the substructure and obtain virtual models of its composition. This makes it possible to find hydrocarbons in places that until now would have been unthinkable, such as in the sea bed or in areas with complex geological structures. Only a few oil companies have the necessary technology to achieve this, in a sector where the pace of research is so relentless that technology that is currently being developed will be obsolete in two years time.
Petróleo en 3D
It is estimated that in the deep and ultra-deep waters of the Gulf of Mexico and the pre-salt area off Brazil contain 100,000 million barrels of oil. However, this is beneath thousands of metres of water and deep layers of salt. Every well drilled in these areas costs between $175 and $200 million. Therefore, obtaining the best images of the substructure is key for oil companies when it comes to risk assessment. This is how the Kaleidoscope project came about; to develop advanced techniques to process these seismic images in extreme conditions.

Kaleidoscope: Looking at the substructure with the Hubble

Underwater and sub-salt exploration for hydrocarbons is a technological and engineering challenge that is comparable to the challenges of the aerospace industry. To explain this, Francisco Ortigosa, Director of Geophysics at Repsol, compares astrophysical research to geophysical research: "The sky is the same for everybody. However you can get closer to the sky using binoculars or get a closer look thanks to the Hubble". The same is true in geophysical exploration, "the Earth is the same for everyone and it depends on the technology you use to look at the substructure".
To rise to this challenge, Repsol launched the Kaleidoscope project, 4 years ago, focusing on developing the latest generation of mathematical algorithms and the most recent advances in supercomputing to obtain high resolution seismic images, "which enables us to view the substructure in areas that would previously have been impossible". The results obtained allow the company to process data 15 times faster than their competitors.
In the Kaleidoscope project, there is a multidisciplinary team of geophysicists, geologists, mathematicians and Information Systems specialists and it is the result of collaboration with private companies such as IBM and public research centres such as The National Supercomputing Centre, The Spanish National Research Council, (CSIC) and the University of Stanford (US). The Kaleidoscope project has received worldwide recognition as one of the most innovative projects in this field and the applications for the oil industry are 100% Repsol technology.

Large scans of the inside of the Earth

Exploration based on a seismic image focuses on obtaining three-dimensional images of the outer layers of the Earth's crust. To explain this, Ortigosa makes a further comparison: "We perform large scans of the inside of the Earth". To do so, vessels drag sensors called hydrophones that pick up sound waves generated by large compressed air canons. These waves are reflected and refracted as they pass through the different layers of the substructure and the time required for their journey is recorded. This data undergoes computerised processing that uses complex mathematical algorithms to reconstruct the geological profile of this layer of the Earth's crust. 

However, while in medicine scans can be performed in real time, a seismic campaign can require between 10-20 Terabytes of data and to process this data a supercomputer like the one Repsol built in its Houston centre (US) that has a calculating capacity of 128 Teraflops (128 billion operations per second) that is equivalent to 10,000 PCs.

The algorithms used in seismic imaging are not universal. They vary according to the geological conditions in each place in the world and algorithms used for hydrocarbon exploration in the North Sea for instance, cannot be used in the Brazilian pre-salt area. Experts spend years studying this and they specialise in the variations that exist in different areas of the planet.
The Kaleidoscope project came about in response to an additional difficulty that exists in the waters of the Gulf of Mexico: the presence of deep layers of salt that absorb a high percentage of the field of waves emitted and make it very difficult to reconstruct the seismic image of these areas. The mathematical equations that have the capacity to process this information required supercomputers with such high calculation capacities that it was thought impossible.
"The major innovation of Kaleidoscope", explains Ortigosa, "was to conduct research simultaneously on both hardware and software. Rather than trying to achieve approximations in algorithms we decided to process the complete image, without taking any short cuts, to obtain high fidelity images of the substructure and to achieve this, it was necessary to conduct parallel research in hardware".
The mathematical equations that have the capacity to process this information required supercomputers with such high calculation capacities that it was thought impossible

Virtual reality for Oil Exploration

Realidad virtual
In developing this capacity of supercomputing, IBM played a fundamental role, as did the Barcelona Supercomputing Centre (BSC). In the view of Francesc Subirada, Associate Director of BSC, it's all about being able "to see in a virtual world what we believe we will be able to see in the real world".
To create these virtual models, high performance computing is required: supercomputers composed of thousands of processors that work in a network and whose calculation capacity is multiplied since they rely on each other for help to perform operations. The BSC has one of the largest supercomputers in Europe, Mare Nostrum, and having access to this equipment was key in developing the Kaleidoscope project.
"Other players in the industry, that do not have access to such supercomputers have to test what they are constructing with small sets of data", says Ortigosa. "Whereas we are able to process large volumes of real data and see the quality of the algorithms that we are generating very quickly". Collaboration between Repsol and the BSC goes back to 2007 and has been extended recently following the creation of a Joint Research Centre in Barcelona.
In an effort to virtually reconstruct the substructure, Kaleidoscope researchers came across an unexpected ally. A microchip created for a Playstation 3 video console was able to be used in seismic imaging applications. "These microchips" continues Ortigosa, "have been designed to create models of physical phenomena that appear real. To do this they have an enormous computing capacity". Following a complex reprogramming process, a supercomputer based on these chips in Repsol's centre in Houston was built successfully. However research in this field is moving so quickly that now, just three years after the discovery, this supercomputer is already obsolete and has just been replaced by another more powerful one that works with video accelerators.

Supercomputers of the future

With current mathematical knowledge it may be possible to create higher quality virtual models of the substructure however the processing of the seismic imaging requires so much supercomputing capacity that to process it "we would need equipment that is a million times more powerful than current equipment" sums up Francesc Subirada. Continuous progress in information technology anticipates this future and "in 2020 we will probably see supercomputers that will be 10,000 times more powerful than Mare Nostrum".
The challenge is also to build sustainable supercomputers. "At this point in time, we could make computers with a Petaflop (one thousand trillion operations per second) using normal processors. The problem is that, in addition, we would need to build a power station to run it and more importantly to cool it down", explains Ortigosa.
In an industry such as the oil industry where a 25% chance of success in drilling is considered high, any advancement in this field is decisive: "A seismic imaging project currently takes between 6-8 months. If we could increase supercomputing to do it in just one month, this would be a tremendous step forward. We are working on it" concludes Francisco Ortigosa.
We would need equipment that is a million times more powerful than current equipment to achieve higher quality virtual models
Last updated: January 2012