It's the 'Shazam' for the oil and gas industry: the technology that identifies the sounds in a production well. That's the premise behind an innovation allowing BP's experts to listen for, and pinpoint, problematic sand in business operations. Meet the scientist behind the algorithm...
What does sand sound like? Pradyumna (Prad) Thiruvenkatanathan knows, but isn’t telling. Inspired by apps like Shazam that recognize the song or genre of music, he, along with his BP colleague, Tommy Langnes, also from Upstream Technology, has figured out how to distinguish the acoustic ‘signature’ or ‘fingerprint’ of sand entering a well from other sounds, like flowing oil. So valuable is this technology that it’s protected by four pending patent applications. “Nobody else in the industry can do this,” says Prad. “That gives us a very big advantage.”
Sand isn’t music to the oil industry’s ears. It’s an unwelcome contaminant that enters production wells from weak reservoir rock, causing serious damage to equipment. But, knowing exactly where sand is entering is difficult and, historically, the industry has had to rely on ‘choking’ the whole well - restricting its flow rate - which reduces the sand entry but also oil and gas production.
Sand is a particular challenge at BP’s facilities in Azerbaijan and, in late-2014, the business asked the Upstream Technology team for help. “So, we surveyed the market,” says Prad, “and found DAS.”
'Listening' to oil wells
The Distributed Acoustic Sensing (DAS) monitoring system transforms a fibre optic cable installed in the well into a distributed array of microphones. “This lets you ‘listen’ to your well as you produce from it,” says Prad. “You do that by shooting light along the cable and measuring changes in the optical reflection, what’s called backscatter. The technology, however, comes with big data.” A typical cable running about 10,000 metres (or 33,000 feet) down a well would produce 100 million individual data points per second at typical DAS sampling rates, with data acquired from every single one of those metres. That’s one terabyte of data every hour -equivalent to downloading 1,000 Netflix movies every single hour.
There's an app for that: how digital innovation helps BP listen underground
What’s more, no one knew exactly what sand movement in a well sounded like. So, how do you distinguish sand from oil from background noise? The answer lay in the techniques used in the music industry, similar to those used in the music recognition app Shazam. Prad designed an algorithm that takes the DAS data and pattern-matches different acoustic signals. “Instead of listening for guitar music, I was trying to listen for sand,” he says. To identify which signal was sand, the Upstream Technology team designed a ‘flow-loop’ experiment with optical fibre installed. Varying concentrations of sand were injected into the test set-up, along with other fluids. Prad and the team then pattern-matched the signals to identify the acoustic ‘fingerprint’ of sand under varying flow conditions using DAS. He then wrote an algorithm to extract this ‘sand noise’ in near real time. “I must admit, writing the algorithm was a bit complex,” he says.
Putting the tech to the test
Once the different acoustic signatures were identified and the algorithm written, it was time to test the technology in a real well. The results were almost immediate - the processed DAS signals appeared to show BP exactly where sand was entering the well. A second test was conducted six months later with an improved algorithm and new dashboards that allowed the Azerbaijan team to monitor the well in real time. In fact, remarkably, it takes just five seconds from the moment sand enters a well for it to show up on an engineer’s computer screen anywhere in the world.
“This second test gave the team the confidence to go into the well and fix the problems,” says Prad. “Once the identified zone was remediated, the amount of sand entering the well dropped by 90% and we’d added around 2,500 barrels of oil per day production from just one well.”
That was just the start. Within 12 months of developing the system, the Azerbaijan team had completed 50 surveys. Prad’s algorithm has the potential to add thousands more barrels of oil to BP’s daily production around the world.
“The ability to see sand moving in real time has made a huge difference,” he says. “It allows us to see exactly where inside a well we need to remediate and, in a lot of instances, we can now make adjustments to increase production without that expensive intervention. It has been a great team effort to take this from a concept to a product that is now running the algorithm to generate real business value. It’s exciting to see work I’m doing have a real impact on the business.”