I heard a new that ARES Model, and i have no idea about it. It seems that only F call gas turbine with DLN 2.6+ configuration that is equipped with ARES Model.

Can someone give a brief introduction of ARES model.

What is it used for?

What are the input signals and output signals of that model?

Thanks in advance.
Neo
[email protected]

 

Neo,

Try this:

http://www.rvo.nl/sites/default/fil...be article in Turbomachinery Intl Oct2007.pdf

If the link doesn't work for you, try searching the archives of Turbomachinery International for 10 October 2007 for an article about GE Wide Wobbe control (also called OpFlex and a lot of other acronyms--they're wonderful that way).

Adaptive Real-time Engine Simulation, ARES, is part and parcel of MBC, Model-Based Control, that GE is touting as the next best thing since sliced bread in the area of gas turbine controls. While it's great in theory, it still needs some refinement--in particular, the ability to troubelshoot and understand turbine operation, specifically combustion problems.

What it boils down to is that it's like the computer in an automobile that controls the engine and fuel and emissions. The auto manufacturer controls how the engine operates and is about the only one who can determine if the engine is working properly or not because only they have the knowledge of how the computer was programmed and how they believe the engine should operate. (Yes; there are non-OEM devices for interrogating the computer, but they generally only produce codes that only the auto manufacturer can understand or interpret.) Essentially, it's a high-technology way of limiting third-parties--or even the owners and operators of GE-design turbines--from performing troubleshooting and tuning. By putting all of these parameters into a computer model and locking-out anyone who doesn't have the software and the passwords to look inside and understand what's going on the OEM is making it more and more difficult for anyone else to service the equipment--and guaranteeing the OEM lots and lots of profitable service work for many years (decades) to come.

Unfortunately, unlike the automotive industry, there aren't likely to be third-party computer diagnostic readers developed and sold for these turbine control systems--there just isn't a market for them, and no one wants to go to court to litigate access to the software. It's pretty crazy for someone to buy something like a multi-million USD turbine-generator and supposedly own the piece of equipment--but to be told that to troubleshoot or understand it the technology used to control it belongs to the OEM and only the OEM can use it. This OEM isn't the only one doing this, and, with remote access capabilities the OEMs can log in and have a look around to try to support operations or troubleshooting. But, that's going to come at an additional cost--and, still, what is really happening is that the owner, the person or entity that bought and paid for this power generation asset, doesn't really own it; they're really only "renting" the asset they've paid a lot of money for because they can't troubleshoot it or service it without someone from the OEM.

Hope this helps!

 

CSA:

Thanks again. Your comment is always appreciated.

I read the very good paper you recommended,and i have some questions.

1) "Under MBC, effectors-control outputs, such as Inlet Guide Vane (IGV) position, total fuel flow and fuel fractions are manipulated to close-loop on calculated operational boundaries, and the resultant effector positions are not constrained relative to each other. This natural <b>de-coupling</b> is fundamental to the MBC usability to robustly accommodate fuel quality variation."

What does that <b>de-coupling</b> mean?

2) As i know NOx and Co are not measured. Are they estimated by the ARES Model?

3) The main function of ARES model is to estimate important gas turbine cycle parameters that are not directly measured. And the Boundary Models are to generate limiting boundary based on cycle parameters estimated by ARES model.

And the control logic are the same with conventional control system. Is it right?

Hope i am clear.
Best regards!
Neo [email protected]

 

Neo,

1) Although I just recently received a printed copy of a GT Mark VIe running MBC I haven't really had the time to sit down and have a read. And, without any block help (if it even exists for MBC blocks) I don't know how much I'll learn from it. There's a bit of mumbo-jumbo in that document, as much a marketing document as a technical description it seems, and I'm not exactly sure what 'de-coupling' means or what exactly they are referring to. I <i>think</i> what they're trying to say (very poorly) is that outputs aren't just tied to feedbacks, that actual performance parameters are used by the model to determine what should be happening and adjustments are made to try effect (in a positive manner) what is happening so as to maximize performance. But, that's just my interpretation of the mumbo-jumbo, and while it may be close to what they're trying to say--I'm sure they would argue otherwise. I'm not sure what the 'de-coupling' has to do with that, other than instead of having some outputs impact others in undesired ways MBC allows them to operate independently of each other so as to improve performance. I would really have to review and analyze some examples of this "natural de-coupling" to be able to say much more than that.

2) As for NOx and CO, well it's my guess that some form of emissions feedback is naturally necessary to validate model observations. I think some of the newer OpFlex offerings and such actually do require emissions inputs for optimization and fuel flexibility. Emissions monitoring is becoming faster (which was the biggest problem in the past) as well as more reliable (which was the second-biggest problem in the past). I'm not sure without some fuel analysis inputs (say from a gas chromatograph) that too many emissions predictions could be made, or at least not very accurately. But, mathematicians and software engineers are doing some pretty marvelous things these days and the compute horsepower of the Mark VIe is pretty darned good.

3) As far as the sequencing of auxiliaries (solenoids; fans; pumps; etc.) and gas turbine operation (starting/acceleration; synchronization; speed/load control; shutdown; cooldown; water-washing; etc.)--I'm sure that's all relatively unchanged, if it's been changed at all. It's just that, as with automobile engine control being computerized, the OEM is trying to optimize gas turbine operation and parts life--while at the same time keeping a tighter control on parts and service by restricting access (just like the automobile companies). Everyone wants to emulate all the high-technology companies in patenting as much as possible--even if they don't ever intend to use it in the short term. They can get paid if someone else wants to use the technology even if they don't. And, it can add up to some serious money. (As a shareholder, I'm not averse to that. As a technician, I have some issues with these practices as they relate to being able to provide help to Customers with questions or problems.)

Hope this helps!

By the way, MBC isn't just for F-class turbines, any more!

 

CSA:

Thank you again for your patient comment, and it really helps.

I heard the term ARES Model just several days ago, and i think if i had chance in the future i would like to go deeper and ask more questions.That's enough for now

Now i am reading CSP of one GE 9E gas turbine with MARKVIe. It has some difference from the one i read before (6B as you know, i read that under your guidance), for it has DLN combustor burning natural gas, and it has primary frequency control and some other function that former 6B unit doesn't have.

I wish i would have chance to talk about it latter and learn from your comment. And take it as a chance to continue our journey. I know there is a long way to go.

What's your opinion?

Best regards!
Neo

 

Neo,

Unless I have an opportunity to work on a unit with MBC-based control, I'm not going to be of much help. And, as I'm hearing it's a "work in progress"--meaning it's in development. And, if I know GE they will keep tweaking this model because they are still learning the capabilities of model-based control and how to apply mathematical models to turbine control, and so it's going to be a while before it settles down into a refined, defined product. And, even then--there are going to be some (Belfort?) who will tweak because they can.

If you want to open specific threads for some discussion, we can do that. Let's try to follow the same process of writing a sentence for any Boolean blocks (ladder rungs by any other name), meaning that you will write the sentence as you believe it works. And then we can correct or clarify and embellish.

There's a vicious rumor that control.com may allow graphics some time in the near future, and when that happens I'll probably post my servo description. It just doesn't work without drawings and graphics. I may post it to the controlwiki regardless, though I don't think most readers now about the controlwiki. (Look under the 'Communities' tab on the Menu bar.) It's not as well-used as it should be; hopefully that will change.

 

Neo, 3 1/2 years after your post I'm going to take a crack at explaining it. I've been lurking on here for years and this is my first post.

First off, just a shout out to CSA. He has very succinctly described many concepts that the manufacturer's own documentation has fallen miles short on.

> Thanks again. Your comment is always appreciated.
>
> I read the very good paper you recommended,and i have
>some questions.
>
>1) "Under MBC, effectors-control outputs, such as Inlet
>Guide Vane (IGV) position, total fuel flow and fuel
>fractions are manipulated to close-loop on calculated
>operational boundaries, and the resultant effector positions
>are not constrained relative to each other. This natural
><b>de-coupling</b> is fundamental to the MBC usability to
>robustly accommodate fuel quality variation."
>
> What does that <b>de-coupling</b> mean?
>
>2) As i know NOx and Co are not measured. Are they estimated
>by the ARES Model?
>
>3) The main function of ARES model is to estimate important
>gas turbine cycle parameters that are not directly measured.
>And the Boundary Models are to generate limiting boundary
>based on cycle parameters estimated by ARES model.
>
>And the control logic are the same with conventional control
>system. Is it right?

So the question I'm going to try and answer is "what does decoupling mean?" This decoupling is a concept that they demonstrate with a little picture in GEH-6810 (newest is Rev C I believe). They don't explain how it really works. They just trust that you will understand what they mean and since, as CSA said, a lot of this is written by and aimed at higher level marketing types and 9 times out of 10 nobody stops to question it. I've heard people try and explain what this "decoupling" means many times and usually they just confuse the issue even more. I apologize in advance if I end up doing the same thing.

Let's start by taking as an example a 7FA.03. Commissioned in 2002 and never upgraded because, well, we don't have the money. This unit has DLN 2.6 combustion system but no Model Based Control. No ARES. We'll call this our "Legacy Machine". And our Legacy Machine is trucking along happily at 80% of base load. We got to this 80% of base either by preselecting it or by just starting from Spinning Reserve and pressing the INC Speed/Load button. Either way it doesn't matter for this argument (two very different governor responses I know but we can talk about that another day). And we are also very fortunate at our site that every day is an ISO day (59 F, 60% RH, 14.7 psi). Now, at this particular load point, the IGV is open but not fully open. What controls the IGV position?

If we dig into the Speedtronic's control sequence program we will find a block called CSRGVV3. And if we study what this block does by poring over the Block Help we will discover that it basically controls the position of the IGV based upon the measured exhaust temperature (TTXM) and one of two control curves (TTRX and TTRXGV). It's a bit more complicated than that really but that's the crux of it. So now if I decide to go into the control room and press the Speed/Load INC button because the boss wants more MW, what happens? Since the speed reference increases Speedtronic will call for more fuel (more FSR) which will in turn raise TTXM. Since TTXM increased there is now less of a difference between TTXM (measured exhaust temperature) and TTRXGV (our boundary) so the control loop sees an error and compensates by opening the IGV further. So in a nutshell, FSR and IGV are “coupled” via exhaust temperature and it just so happens that that is exactly what the graphic shows in the GEH 6810.

Now we move from the "Legacy Machine" to what I'll call the "ARES Machine". It's a 7FA.04. Commissioned in 2016 that has the hot gas path upgrade, the ARES "digital twin" of the 7F engine running inside the Speedtronic sequencer and all of the features that sales has touted to achieve "operational flexibility" or OpFlex as they've branded it. We've located it to run right side by side with our "Legacy Machine". And it just so happens that we are running the "ARES Machine" at 80% of base and of course under the exact same conditions as the "Legacy Machine". Now, at this particular load point, the IGV is open but not fully open. What controls IGV position?

If we dig into the Speedtronic's control sequence program (and we can actually do that now thanks to something they call Open Epic...ARES itself is still locked down though), we see that now instead of just 2 boundaries for IGV there are over 20! They are all being calculated simultaneously in real time but only one of them is being controlled to. For simplification, let's say that our unit is being controlled to the boundary called TFIRE_MAX (Max firing temperature). That boundary may in fact not be realistic for 80% of base load but hang with me. If we tear into the control sequence program, we find that the IGV reference for this boundary is coming out of a PI control block whose input and feedback is something called TFIRE. What is TFIRE? And also, where does TFIRE come from? TFIRE is the ARES estimate of the first stage firing temperature. It's also an encoded index and you won't find it to be in degrees F but likely to be around 95 or so for this example. What's significant about this is that we aren't waiting for a feedback from a group of thermocouples. It's almost like a feed-forward. So ARES is going to drive the IGV to where it thinks it needs to be based upon what it thinks the 1st stage firing temperature is. So if I go back to the PEEC and press the INC Speed/Load button here, ARES is going to know from a whole plethora of both hardwired inputs and historical data that has been programmed into it what the TFIRE is going to be and it's going to move the IGV accordingly. It ain't waiting for no stinking feedback. IGV and FSR are now "decoupled".

Make sense?

 

As I reread what I wrote, I realized that I was a bit unclear in my description of the control loop driving the IGV reference. I said that both the reference and feedback to the PI regulator was TFIRE. How can a regulator regulate if both the reference and the feedback are the same? My bad. Let's look at it again.

A more likely scenario for our 80% of baseload example, and one that's a bit easier to describe, is if we are running with the IGV being controlled by the TFIRE_TG (firing temperature target boundary) instead of TFIRE_MAX. Remember too that there are around 20 possible boundaries that the IGV effector can control to. Sometimes that's a bit hard to get your brain around (apart from MIN and MAX I think there were two in the legacy controls: TTRX and TTRXGV). The logic to "auctioneer" it is hard to follow. In this case, Tfire_Target (not real signal name) is the reference to the PI controller and TFIRE (from the ARES digital twin) is the feedback. The Tfire_Target comes from a lookup table...no they didn’t disappear with the advent of MBC and ARES...with measured CPR and CTIM as its inputs. The lookup table is a bit more complicated than most of the linear interpolators that we are used to. This one is a double analog interpolator (DALIP00 block) that finds a value for Tfire_Target (indexed. no units) for two measured inputs: CPR (ratio) and CTIM (degrees F). So an example that's not too far off would be CPR = 14.6, CTIM = 59, Tfire_Target = 92.6. Tfire_Target gets pumped into the PI block's reference (with some minor offsets) and TFIRE (from ARES model) goes to the PI block's feedback. The output from the PI block is added to a small feed-forward offset from ARES and IS the IGV reference.

That's the best way I can explain how the effectors are not coupled.

 

Carlo_Muro_,

I know this is an old thread, and I don't know how I missed your last two posts; sorry!

Anyway, these are very good explanations. I think I want to add something that is maybe not obvious to some--and that is, I believe, the 7FA.04 units have multiple stages of variable axial compressor stator vanes that the 7FA.03 unit might not have. And even if the .03 units have multiple stages of variable axial compressor stator vanes, they are probably different--as are the IGVs than those in the newer .04 units. That's a pretty SWAG (Scientific Wild-Arsed Guess) on my part, but probably not too far off.

I believe ARES is all about wringing as much power out of every operating point as is mechanically possible. To do this, they have added multiple sensors not in use on units without ARES, and, of course, the accompanying algorithms in the application code (Control Sequence Program (CSP) is so ... old-school and out-dated!).

And, I don't believe us lay people are ever going to get a lot of understanding of how it works and how it is supposed to work. That's all IP (Intellectual Property), and we will only probably ever get to learn more about it on a need-to-know basis--so, if we are working around or on the unit when there is some kind of problem, or failure, and things kind of "slip out" during review and analysis, and "some" explanation of what happened or should have happened.

Anyway, sorry for missing these responses!