I came across a great post describing ways to talk about machine learning, and one of the analogies the author brought up was a washing machine. “Washing machines are robots, but they’re not ‘intelligent’. They don’t know what water or clothes are. Moreover, they’re not general purpose even in the narrow domain of washing – you can’t put dishes in a washing machine, nor clothes in a dishwasher (or rather, you can, but you won’t get the result you want). They’re just another kind of automation, no different conceptually to a conveyor belt or a pick-and-place machine.”
I’ve been inspired by this more accessible way of talking about technology lately, and I wrote about this in a blog post a month or so ago advocating to let technology do the blocking and tackling for you so you can be the QB and score touchdowns. Having lived that life in the field, I always felt like Ops was blocking, playing quarterback, coaching, and even selling tickets and concessions sometimes with everything we had to do in a given day. There was always too much to do, and not enough time to do it in, and I felt like the technology that existed was not keeping up with the rest of the world.
That’s why the washing machine metaphor is a good one when it comes to PLC automation vs. optimization. I don’t have time to wash my own clothes or like to wash clothes. A washing machine washes clothes certainly better than it used to in the ‘50s, and it’s a huge step forward from using a board and hand scrubbing. But the problem I’m solving isn’t just washing clothes. It’s doing laundry.
Laundry is a more complex problem that a washing machine alone cannot solve without a significant amount of my time. Think about all of the tasks that we have become numb to: sorting towels from dress shirts, separating out the whites to bleach, moving the load to the dryer or hanging on a rack, folding clothes, and finally putting them in the right drawer or hanging them up. You get where I’m going. Yes, the washing machine has made strides, but it hasn’t been able to solve the complex problem of doing laundry.
This is the same thing with PLC-based controllers in regards to optimization. Optimization is a complex multi-dimensional problem, which is why setting and forgetting any PLC is a suboptimal strategy.
A pumpoff controller protects against pumpoff–it’s in the name. A flow computer RTU auto-tuning algorithm protects against well instability and loading up. Similarly, a washing machine has presets, warnings, and sensors to make washing clothes more efficient. These features are helpful, but alone do not tackle the many tasks required to finish a load of laundry.
For decades, producers have needed employees to play the role of operators, techs, and engineers. Even with optimization, I think we all have become numb to the difficulty of doing this on a single well-let alone on hundreds of wells. Production data is extremely noisy with multiple streams and it changes daily by definition of a decline curve. Additionally, it can be impacted by external forces such as dynamic mechanical systems and imprecise measurements/calibrations on the surface, and scale in the formation or communication with other wells in the subsurface.
Those are a lot of factors…kind of like the many tasks of laundry. Numerous studies across industries have shown that people working together with AI improve business outcomes for their companies. Given the complexity of optimization, it requires a combination of the best of both worlds. AI can help humans scale their abilities across several tasks. It is able to implement consistency day in and day out, remove process dispersion, and eliminate a ton of noise out of the system. This removal of noise and burden allows us to focus on tasks where creativity and complex thought are needed to optimize other aspects of the system: wells that need better designs, wells that aren’t operating on the bottom valve, or optimizing plunger types for wells.