Turning Pump Jacks Into Robots Using Edge Computing

54BAAD72 06C6 44D6 991E 79FC94238215 e1557071344665

Turning Pump Jacks Into Robots Using Edge Computing

Oil and gas companies have so many assets running far far away from civilization it’s a wonder that more are not getting an automation makeover using edge computing. 

Edge Computing?

Pull up Google Earth on your device and zero in on Odessa, Texas, heart of the Permian oil basin revolution in the US. Pan in any direction and all you can see are well pads laid out in a neat checkerboard pattern. But what you don’t see are the trappings of modern civilization — big cities. All that oil infrastructure is far away from where people actually want to live because it’s hot, dry, and inhospitable.

Finding highly skilled talent in places like this is a stretch — limited school choices, long commutes, and isolated living. Oil companies find it similarly difficult to find the talent they need in parts of rural Saskatchewan and Alberta, the Bakken basin found in North Dakota, the shale basins in northern BC, and the vast stretches of Texas.  

But that oil equipment still needs to be supervised. Wells, considered a pretty sensitive asset, are typically connected to a traditional SCADA system that captures a number of variables about well performance and feeds that back to a control room somewhere. Even the term SCADA belies its bias — Supervisory Control and Data Acquisition. It’s about pulling data from a controlled asset during operations, but not to analyze that data and take action.

When SCADA architectures were first evolved (circa 1960 and later), the notion that you could place a computer at the site of the well (at the edge as opposed to the center), was unheard of. It was too costly.   

For the enormously profitable and huge producing wells, it makes solid economic sense to allocate a team of production engineers to take all that data from the SCADA systems, build excel models, figure out to tune well performance, and do that every day. A single off-shore platform will be richly endowed with human talent.

But the vast majority of on-shore wells are in fact very low production — just a few barrels per day — and their productivity is in constant decline as the wells slowly drain the reservoirs. These wells receive little analytic attention because they’re marginal. The general rule is to turn them on and leave them to run. Anecdotally, as much as 85% of the on-shore wells in North America are run this way. 

I liken them to planes on auto pilot, except they fly exactly the same way without regard to the load they carry, the weather they encounter, the cost of fuel, the travel distance, and the negative impacts on the equipment. 

It’s not at all uncommon to find oil companies with large fleets of producing but marginal wells. Big oil companies trim their portfolios by selling off older wells as they age and decline. It makes little sense tie up capital in wells that do not generate the same returns as the best wells. Other oil companies buy these assets and in the process, build up their portfolios. They apply their smarts to these wells and extend their productive lives.

In time, the new owners too are forced to trim the portfolio, and the cycle continues, with wells constantly changing hands until finally, the well is abandoned when no one thinks they can make a buck with it.

This scenario illustrates some of the prevailing orthodoxies of the oil and gas industry:

  • Low volume, marginally economic wells do not warrant or cannot afford much analytic attention.
  • Technology costs are too high to equip remote marginal assets with anything other than SCADA
  • Run marginal wells hard.
  • Abandon or sell wells when they are uneconomic in your business model.

 

I suppose when oil prices are strong and everyone is making money, this situation is acceptable. But in this new era, of low prices forever, hot competition, more demanding regulation, there’s pressure to do better. And new technologies are finally making it possible to revisit these industry orthodoxies and do things differently. 

Optimizing Oil Equipment

One of the key ways to keep a well productive is to supply it with artificial lift, or a pump that pulls fluids and gas from the well. Once the natural pressure from the earth decreases, pumps become necessary, which is why they are so common. These pumps are a regular sight throughout oil production country (there’s up to a million of them), and look like a horse’s head bobbing up and down if the horse was eating grass and not lifting oil to the surface. 

The rocking motion pushes a shiny rod (the sucker), attached to the horse head down the well to a set of valves that open and close to let fluids flow in to be lifted up. The reason the contraption is so large is because of the weight of all that steel rod, the weight of the liquids being lifted, the depth of the well and the size of the well bore. 

Operating the pumps efficiently is trickier than it sounds. The weight of the liquids can vary as the fluids coming to the surface might be a shifting concoction of gas, water and oil. The volumes might be off as the well slowly peters out. The rod might pull up progressively less and less volume of anything. The cost of fuel or electricity to run the motor to move the pump varies. Even the value of the product being lifted might be too low to justify bringing it to the surface.

Pumps might be working too hard for too little, which causes unnecessary wear and tear on the pump, consumes costly and emissions-laden fuel, and ruins the economics of the well. Or pumps might not be working hard enough, which reduces their revenue. Or the pumps experience mishaps and upsets (water slugs on a gas well, or reductions in lubrication, or pressure problems).  

Engineers know all this, of course. They’ve been running these wells for years. The efficiency formula are well understood. The mishaps, upsets, and issues reveal themselves in the data. The actions to be taken are documented. 

The problem is they haven’t the time to gather and analyse the data constantly as the conditions vary for the hundreds of thousands of marginal wells. There are too few production engineers for the sheer number of wells. Add the remote location of the wells, the lack of real time data and the complexity of the analysis, and voila: a recipe for artificial intelligence at scale. 

Digital To The Rescue

How might new technologies be configured to tackle this problem of making pumps more efficient? Enter the power of digital.

Internet of things

Place an actual computer at the pump site to supplement the data being backhauled via SCADA to a control room. Run it using the power already to the pump. This is the edge device. 

Artificial intelligence

The computer runs an on board artificial intelligence engine that constantly takes data readings from the pump, interprets the data, and adjusts the performance of the pump to optimise production. The pump becomes a self running, self optimizing robot.

Cloud computing

Move data and software to and from the AI engine using a cloud computing environment rather than SCADA. Capture pump data for reporting, and share AI generated insights across all similar AI engines. Use the cloud to send software updates to the AI engine. Now it looks like smart phone and an App Store.

Modern telecoms and security

Equip the computer and site with the latest encryption code, light duty satellite uplinks, patch appliers, app managers and other tools to give it autonomy.

 

The Solution

This kind of solution delivers the goods. It brings human-level expertise as codified in the AI engine to those wells that get little to no attention now. Human talent is ultimately better utilized. By reducing over pumping, operating costs fall  because of lower maintenance costs, and lower fuel costs. Production increases to the optimal level, and stays at the optimal level responding much better to reservoir changes. Revenues from the well are optimised (not necessarily maximised).

The wells when they are sold can fetch a higher price as they are more valuable (or, said another way, growth companies can bid slightly more for wells that could be equipped with these digital smarts). Infrastructure utilization (flow lines, batteries), improves which helps returns to shareholders. 

This technology also gets better over time as the AI engine ingests more and more data from across more and more wells. Best of all, the AI engines can be moved to another well once the economics are thoroughly depleted. 

Fact or Fiction

For regular readers of this series, you might recall my nifty Framework for thinking about digital.

  • Digital is all about data
  • Internet of things generates the data
  • AI interprets the data
  • Robots apply the data
  • Cloud stores the data

You can read the full story in this article about the Framework. The pump rod solution is a perfect illustration of how these digital tools can be creatively combined to solve real world problems. Better still, the solution is already in the market and delivered by Ambyint, a leader in improving artificial lift systems in oil and gas using digital. After just a few weeks on the job, customers report that Ambyint’s engines deliver up to 5% gains in production, and up to a 10% reduction in operating cost. On a 100 well site producing 25 barrels per day per well, that’s like having an extra 5 wells (which could be several million dollars in avoided capital spend).

The smart oil and gas operator should be actively converting its dumb horse iron into artificially intelligent autonomous production units, or risk being left behind in this horse race.

 

Much thanks to Alex Robart and Brian Arnst for sharing their insights with me.

*****

Check out my new book, ‘Bits, Bytes, and Barrels: The Digital Transformation of Oil and Gas’, available on Amazon and other on-line bookshops.

Mobile: ☎️ +1(587)830-6900
email: 📧 geoff@geoffreycann.com
website: 🖥 geoffreycann.com
LinkedIn: 🔵 www.linkedin.com/in/digitalstrategyoilgas

No Comments

Post A Comment