As the old say, the devil lies in details. Writing a global control is no means an easy task, as

1. You need to recognize the opportunities exist in the first place.

2. You need a global controller that can aggregate and optimize for a global solution (and the global solution might not necessarily simply to maximize the aggregate throughput, but there might be other factors into account), which may involve some algorithmic design (in some cases, you need to design new algorithms).

3. You need to justify that global controller gives you a superior solution compared to locally greedy solution. As in this article, a global solution gives you about 3% improvement compared to the local controller, and the local controller algorithm is substantially easier to write.

Background: in my previous job at Meta, I wrote such a global control algorithm for controlling the rate of data going in and out each data center. It involved some really interesting algorithmic design.

I think the real thing that's difficult is-- every installation is different, in the geometry of the turbines, turbine sizes, terrain shape, etc.

Even if you have a perfect implementation of this, and you don't need to deploy new networks, etc, and you put in a lot of NRE to make this easy deploy... how much engineering effort is still needed to start squeezing 1-2% out of a wind farm?

It has been done earlier; multiple times, by every serious manufacturer. :-]

The real difficulty lies in:

1) Noise in the on-turbine wind speed and direction measurements and/or robustly (see point #2) operating LIDAR or met masts in front of the farm to try to avoid said measurement noise.

2) Actually arriving at a robust, operational in real-world conditions, fully closed-loop control system. A commercial wind farm has to operate 24/7 for 25 years without a bunch of engineers and scientists babysitting it, which is what is likely to end up happening if the cool control system relies on offline simulation results, topographical data, and/or human-supervised calibration & tuning.

It's not all doom and gloom: Ongoing improvements in sensor price/quality will probably make these kind of global control systems more and more practically feasible in the future.

Having off-site control of configuration and sensor information seems desirable and obtainable regardless. And once that is in place it is just a commercial decision. If spending x on cloud modelling delivers a multiple of x then its a simple decision.

It's a 1.2% overall improvement. That's the kind of number where it's worth doing in general, and worth using if it's been developed, but not nearly worth the development effort and headache for a wind farm operator trying to invent it on their own.

considering it amounts to 1.2% over a period, I wouldn't say "obvious improvement"

local maximizing presumably already deals with the effects from upstream turbines, global (also presumably) only adds consideration for downstream turbines

I suspect actually interfacing with the windmills is the hard part. Especially if you want to support many different types of devices.

Local control, and no need for a global coordinator, might be so much simpler as to be worth losing some efficiency / not going into integration hell.

All the wind turbines in any given windfarm come from the same manufacture, and they all have monitoring and control from the utility's main offices in whatever city (at the very least they have enough control to prevent damage when the weather gets bad, I don't know what else they can do). As such all that is needed is a software update from the manufacture to give the central office more control.

The utility might own wind turbines from different manufactures, but they are not in the same wind farm. I suppose there might be places where two wind farms border each other that you want different devices, but for the most part you can optimize each wind farm individually with no need to worry about manufactures.

I don't know that any of that is true.

Farms can have different makes and models. I also think that there is a lot of on board control.

Sure they can, but the reality how wind farms are built is they won't buy more than one make on any given farm. They place an order from some factory to deliver, getting a into the factory order stream is a large part of the process. Of course it does take a couple years and so there is plenty of opportunity for the factory to change internals in the middle, but the manufacturer is locked in before the first turbine goes up.

Shipping is also rather expensive, so your are probably going with the nearest factory for everything. (each blade needs a semi with an oversize load permit, and "chase cars" both in front and behind with the right lights and signs)

Of course if a turbine fails in 5 years (I don't know what the warranty is, so I'm going to use 5 years) they might replace it with one from someone else, but that isn't common.

I don't know how much is onboard controls. However someone is feeding the weather instructions in, and when demand is low someone is telling a few to shut down. They also do remote monitoring for issues that maintenance needs to fix. That connection just needs an upgrade, along with some new software for the onboard controls and it can be done offboard. (this may not be easy, but compared to a turbine it is cheap)

Im guessing you are thinking of very different farms than I am.

At the extreme, I am familiar with farms that have been expanded over 25 years and have a huge variety of builds.

I am also aware of a number of farms that have at least 2 different sized turbines.

I count an expansion as just putting up a new wind farm next to a previous one, which happens often. At least in my experience the expansions often get a different model (visually different), but I count that as a new wind farm.

I'm reasonably sure that all wind turbines in my area come from Siemans. (they have a factory in my state, any other make would be shipped in from a considerable distance)

If the turbines are close enough to impact the air flow to the other ones then they are effectively the same for the purposes of the cfd model in the parent post

A typical wind farm will have 1 or 2 types of turbine.