Reduce Acoustics, Emissions, and Fuel Spend with Automated Aeroderivative Gas Turbine Tuning

Webinar – Reduce Acoustics, Emissions, and Fuel Spend with Automated Gas Turbine Tuning

When your aeroderivative gas turbine falls out of its operability window, it takes more fuel and work to generate power — speeding up degradation and increasing heat rate. This webinar discusses GE Vernova’s Autonomous Tuning closed-loop solution, which uses an AI-powered digital twin model of the turbine to autonomously tune aeroderivative turbines for acoustics, emissions compliance, and ideal performance. Learn how Autonomous Tuning keeps turbines running efficiently using ambient temperature, fuel properties, and degradation.


--
TRANSCRIPT

Hi everyone, and welcome to our webinar, Reduce acoustics, Emissions and Fuel Spend with Automated Aero Derivative Gas Turbine Tuning. I'm Jackie Vineyard, Product Marketing Director at GE Fornova, and I'm thrilled you're here to join us today for a deep dive into a transformative technology called autonomous tuning. Autonomous tuning uses AI and machine learning, specifically neural networks, to dynamically tune turbines in near real time. Why does this matter? Seasonal manual tunings don't respond to changes in things like ambient temperature or fuel properties and cannot maintain optimal performance zones consistently. Because autonomous tuning uses advanced neural network technology, it can adapt to these variables like fuel types, ambient temperature, humidity, and even degradation.  
 
This in turn protects equipment, decreases heat rate and reduces emissions. Joining us today is Martha Saker who knows this technology inside and out. Martha brings 26 years of experience and has been instrumental in developing and deploying the solution. Martha passing the mic over to you to introduce yourself. Thank you, Jackie. Hello everyone and thank you for your time today. Yes, lots of experience, but in reality, I'm just lucky to be surrounded by a team of experts on machine learning, analytics and overall asset performance. As you mentioned, I had been part of the EG Bernova for the past 26 years.  
 
The first five or six years of my career I spent as a gas heavy duty controls engineer and that allowed me to spend a little bit time in the field servicing control systems and seeing first hand the challenges that our customers face. After that, I had different roles as product manager for all different areas of the power generation business and the past six years here with the software business developing solutions to optimize performance. And clearly, you bring a lot of depth and expertise, and we're honored to have you here with us. Today, we're going to do a deep dive into autonomous tuning and cover frequently asked questions by our customers.  
 
But before we dive into the FAQs and technology, let's go over some housekeeping items. Please submit questions at any time they come to you during this webinar. We don't want you to forget. We will leave time at the end to answer as many as we can. Contact information is available in the speakers section, and for questions we don't get to, we'll reach out after the webinar via e-mail. All right. First question, Martha, can you go into more detail about the challenges autonomous tuning is solving for? 
  
Sure, Jackie. We were trying to solve the issue of maintaining consistent performance for gas combustion. And yes, the tuning process is great. It involves a lot of domain knowledge from the history and the install base. And yes, it has the ability to bring the equipment back into that very small area where ideal performance is found. That a small area is determined by the boundaries, the physical boundaries of the equipment and also the overall surroundings of the process, the environment. Once the manual tuning process is completed, if degradation of temperature, humidity, any other variability drifts that space in any direction, the value of that manual tuning begins to degrade because it's a static process.  
 
So we were looking for a way to capture all that knowledge into a dynamic process that could be capable of following that process, whatever variability might take it, and bring it back to this ideal area of performance. That's what we found that machine learning, artificial intelligence has the ability to follow that variability, understand what's going on with the process, perform the changes that are necessary to bring it back into that ideal spot of performance. Yes. And having dynamic tuning that improves availability, I mean that's more important than ever today with the energy mix and increasing need and demand for fast and flexible power. And that's, that's incredible.  
Now let's dive into the frequently asked questioners by our customers. Our customers tend to be engineers. So all of you on this webinar, can you guess what the most common first question we get is? If you guessed how does it work? You got it right. So, Martha, how does it work? Thanks Jackie. Yes, only starts with a digital twin and this terminology just refers to a really fancy model of how your asset works and performs. We have this digital twins fully populated with all of the knowledge from the Burn Nova legacy.  
 
So from the start those models are already very capable of maintaining that performance. Then we add on top of this models the ability again like we started discussing, to introduce some dynamic response to variability and we do that through the addition of artificial intelligence, machine learning, neural networks. The way that works is to a three stage process for the purposes of autonomous tuning. The machine learning has three different stages. The first one we call learning mode, and in learning mode, the machine learning is just sitting there exploding the space of operation, how you're running the machine, what's going on with your environment, what's going on with your field, and accumulating all this knowledge about your process the same way an operator might.  
 
Once the learning mode has determined that there is sufficient variability to require updating updates or changes to the model, then it goes into what is called building mode. And building mode is where the machine learning is automatically, without any human intervention, updating the models to produce again the right combination of commands to be sent to the control system to nudge the asset back into that ideal son of performance. Once we have a fully updated new model that captures all that variability and all the necessary changes to correct it, we go into control mode. During control mode, artificial intelligence is exchanging information with your level one controller, your control system, to execute those changes.  
 
That is very important because it implies that our solution does not by itself applies any influence on the asset. Everything that we do, we do through the control system, which brings an additional layer of assurance in regards to risk. We are fully bound by the safety critical programming of the control system and therefore we cannot cause any harm to the equipment. It's quite incredible to have a technology today that can self update models in the real time. I I think I think that's incredible. There is a lot of hype though around artificial intelligence right now, driven by agentic AI.  
And as mentioned in their introduction, autonomous tuning uses the type of machine learning, specifically neural networks. You've hinted about the maturity a bit in the in the last two slides, but still, let's dive deeper into it. Our customers often ask us how mature is this type of artificial intelligence and how mature is autonomous tuning? Yes, and and and I get the question and, and I understand why everybody is curious. There is so much hype going on around AI, but I'd like to point out, you know, most of what we see in the press today is about a gentech AI, the form of AI that we use in here, the machine learning, the neural networks have been around for a very long time and specifically for us is a technology we have been using for over 25 years.  
 
We started this journey utilizing this technology as I previously outlined for coal combustion optimization. There is a whole long history of benefits over 200 installations worldwide that have demonstrated the maturity of that technology again throughout 25 years span. So I, I think that pretty much verifies that this the technology has been proven, it has consistently delivered great results and benefits for customers, millions of tons of emission saved and a lot of impact on the bottom line. Great. The follow up question we often get after explaining this proven technology is what kind of results can customers anticipate and are they consistent across the board?  
You know, they might say my region is cold, my region is hot or we use different fuel types. Will the results be consistent? Yes. And and of course that is variability. That is as much variability as there is variability in today operational missions of every power plant. So yes, I definitely understand why that would be a question that comes to mind. So maybe let's start with what is consistent. We are consistently seeing double digit improvements on emissions specifically on the CEO and Knox profiles. The double digit impact on Knox, some of you might already be thinking yes to that impact my ammonia definitely.  
 
So there are some additional savings there on ammonia as well as a byproduct of this optimization. Then also very consistent is the improvement on the heat rate, which at the end of the day is probably the most common impact to all of us on the fleet, which is the reduction on fuel utilization critical to those very tight budgets. And also associated with that heat rate reduction is a reduction on your CO2 footprint, which is critical to our green missions that many corporations have nowadays. Now, in regards to availability, there's also a very consistent pattern where we deliver an improvement that is key and that is the avoidance of unplanned manual tuning.  
 
So all these events where all of the sudden you have this drastic temperature change and you need to scramble to find a tuner that is gone. For most of our fleet, we have eliminated minimized the need for those events to effectively 0. Now yes, definitely the more you run, the more savings you will see. So yes, base loaders will accumulate higher savings in regards to fuel and others. But again, to summarize the missions, the availability and the heat rate improvements are there for all. I am here to emphasize the previous points. We have an example from the real world and this is captured from one of our installations.  
 
This particular customer had the autonomous tuning running for for a little while and we perform what is called an on and off task. So for those of you that might not know what that is, is what you have an application running and you verify that you have equal conditions before and after you turn it off, you capture your readings and then you turn it back on. And what we see here then is that comparison on and off. The off is the basis schedule. So basic schedule is what's delivered to you through a manual tuning event, a brand new fuel schedule, and the bottom is your autonomous tuning results.  
 
And the first thing that jumps to you is a drastic reduction on your emissions footprint. Not only are we moving very far away to the left, away from your excursions and trips and all this undesirable behaviors what also greatly shrinking the overall emissions footprint. So for those of you that have any kind of credit schemes, this is good news. Now something that might not jump very clearly on the chart is the load. Why is there a load difference in here and why do that has to do with autonomous tuning? Yes, we do not have any impact whatsoever on load. But for this particular customer, what was happening is emissions were determining their production and that is a very undesirable behavior.  
 
Once the emissions were under control, they have total freedom to generate whatever they were called to generate, which of course had a great impact on their bottom line. Yeah, That is one of my favorite customer stories because I recall you telling me that they've had it now for five years, They've had autonomous tuning for five years, and since deployment, they haven't had to stop it from operating one time. And the customer is thrilled because they're generating more power while staying in compliance. Now definitely we'll have our best benefits. Yeah, yeah, absolutely. Now what about those who don't live in a mission strict region?  
 
Will they benefit from having autonomous tuning as well or no? Yes, definitely. As as I mentioned, the heat rate impact, the fuel savings, I'm still to encounter a power plant that does not have a limited budget. So I think everybody would welcome that. And of course the availability, whether you are a base loader that needs to run constantly and don't need surprises or you are back up to renewables and you must run when called to run, that availability can be critical. So the answer is yes. And the emissions are just the cherry on top.  
 
And also beyond that, well, yes, we help you meet those new strict emissions regulations. Many of you are encountering that as a change. You were not regulated and you might be today and you don't have a way to bridge that gap on those PPMS. So we can help you accomplish that and also accomplish your corporate sustainability targets, which can be very challenging, emphasizing again, this savings around fuel cost, seasonal weather impacts. I, I had many people comment on, hey, I used to be OK with my manual tuning events in March and October, but now it just happens to be December is a really hot month.  
 
So all this variability and weather changes, whether it is global warming order, we're not going to discuss that. But yes, there's definitely changing a wedding patterns and we can help you there. The need to maximize at least add some predictability to your availability is another point that we can support. And the overall load predictability. Again, having emissions dictate your megawatts is never a position you want to be on. For customers with air derivative gas turbines, you know everything you said so far. To be able to save money on fuel, reduce emissions, improve availability to this extent is very significant.  
 
But how does this translate to dollar amounts? What does the return investment look like using autonomous tuning? Can you give us an example? Yes, of course, Jackie, as I mentioned, there are multiple levels to the benefits here. So the story might be a slightly different from everybody. So we have a an example that captures a range and the range is based on the minimum amount of benefits, which for most customers might be just a hit rate and feel savings. And if you can compound that with additional levels like we have here. So your manual tuning avoidance that has a cost and that would increase that number your feel flexibility.  
 
So as we mentioned that BTU change can be easily corrected by autonomous tuning, which means you don't have to be as careful around where you source your fuel from. That could be also a source for additional savings. And if you are in a region where emissions have a credit value, then that could be a very substantial additional dollar amount on top of this range. So again, that's what we have, the bottom and the top. What I would like to highlight here is even at the bottom rate, at the bottom part of this range, you're still achieving a around a year return on your investment, which I think is pretty substantial.  
 
Yeah, absolutely. OK. At this point, we've covered the return on investment that you can expect, the challenges autonomous tuning is solving for how it works. The audience may be wondering what exactly is autonomous tuning? Is this just software or is this hardware plus software? And what are the risks, if any, to the equipment? Yes, definitely Jackie and it's it's both. So let's go there. There's hardware here and excuse me, and this also allows me to answer a couple questions that I saw here through glancing at the screen.  
 
We host this application on Prem and the reason why we do that is because this is a closed loop application that interfaces every two seconds which is almost near time with the control system. That means being on Prem is imperative. There is no cloud version of the solution. Currently. The hardware is the box that we use to host the application. We call it a control server and edge device. There's so many fancy names for just a really big box with a lot of resources. And yes, that's what we use to host our application.  
 
That is the hardware component, the rest is all software. So as I previously mentioned, that is a digital twin that is hosted on our artificial intelligence platform, which is called Process Link. That's where the learning building control happens. Then we interface all of that infrastructure, which is called the Level 2, with your level 1 infrastructure, the level 1 infrastructure being your control system. We are control system agnostic, which is another big advantage of residing in that additional layer or level 2 because it allow us to interface with any control system out there and that is very important. We understand power plants have legacy equipment, have different types and preferences on controllers, on the CSS.  
 
So we do accommodate that. Also you will notice on this architecture that there is no direct connection between our Level 2 infrastructure and the asset, the turbine itself, which means there is basically no risk from our application in any possible case of scenario to exceed the safety critical boundaries. They are built into the control system who are fully bound by that control system, and that is again to limit the risk and to ensure seamless operation with the programming of that control system. Yeah, customers love to hear that because autonomous tuning, you know, as you cover, not only provides a significant return, return on investment as you showed, but it also is completely safe for the gas turbine.  
 
But what about adoption? Is it hard for customers to adopt? You know, we have customers who are concerned about their artificial or the artificial intelligence we bring with their old legacy systems. Is is that a problem? It could be a problem, but we have been very deliberate again about being agnostic and trying to work with whatever control system might be at potentially at any of the plants with equipment that we service. And the way that we do that again is through having the application hosted on that edge server which allow us to connect with a variety of legacy systems through off the shelf protocols or PCE or others depending on the generation of that legacy equipment.  
 
And so far we have not encountered any issues around connecting with the systems and being able to have its seamless execution through those systems in yes. Another big concern around, Oh well, how difficult it is to implement machine learning, artificial intelligence solution is the training. And everybody gets a little concerned. I don't have data scientists and stuff. So how am I going to adopt this solution? Who's going to be managing this solution? And the good news is you don't have to. Nobody has to manage this is close loop fully automatic. It runs in the background and you're interfacing with. It is as basic as a on and off button and some parameters that show you what we do and in general.  
 
So yes, this might be a little bit of a joke, but it's really applicable. You just set the application and you forget about it. There's really no need to become a data scientist. I don't remember seeing this guy, but apparently I looked him up and apparently he was a legendary infomercial pitchman. And I've definitely heard the memorable and famous catch phrase just set it and forget it. But yeah, that's hilarious. Next question, what about deployment? Do customers have to shut down or shut off operation of the gas turbine to deploy autonomous tuning? That's definitely another plus from the solution and as compared to many other choices that you might have in order to improve your emissions profile, most of them involving major hardware upgrades to your engine.  
 
We do not require an update, an outage. So in general, we can install the software in a matter of hours. All we need is a window for you to allow us to connect at level 2 to your control system. So a water wash, a weaken outage, any type of a small plant event will be a sufficient window for us to Commission this software. No outage required, no associated downtime. That's again another benefit we're very proud of. Got it. No outage necessary to deploy. What about cybersecurity?  
Can you speak to that? Yes. And that brings up a very important point that I might have escaped. So it's good that we're talking about cyber. Our level 2 is connected, that control server is connected to GE. We do some monitoring especially at the beginning. Well, while the first phases of that learning mode are happening to ensure that your artificial intelligence is running as suspected and the changes from to the models are being implemented as suspected, that connection of course cannot bring or introduce any risk into your enterprise.  
 
And for that reason, we offer incremental security measures to ensure or minimize the risk associated. We have the simple and the complex. So this is based on your preference around technology. The simple 1 is just interrupting that connection. If those cables don't touch, then there's no connection and you are the full administrator as when you want to eat to have access to that level to exclusively, we won't have access to anything else. Then if you want something more sophisticated, maybe you are under their state regulations and standards. They are very strict and you have to have something with very high compliance thresholds.  
 
We also offer a zero trust command broker based solution that allows you to monitor every second we're connected, record every one of those sessions and even automate the access. So for example, you may decide I want GE to have access for half an hour on Tuesdays afternoon and then you also can come here and set it and forget it. All right, now for a question that is top of mind for many of us, you know, all driven by the height of artificial intelligence, is is often the question, how much will artificial intelligence replace people? Since autonomous tuning is closed looped, you know, does it replace the job?  
 
Does it replace a person? I get this question because yes, it's all the hype of artificial intelligence. And yeah, I'm, I'm, I'm really sorry for those fields that legitimately might have this concern, but I think anyone that has been in a control room or in a command center will laugh at this sustainment. There's just no shortage of work and activity in this environment. The pure cognitive overload of an operator is just cannot be understated. And aside from everything the control system and DCS are throwing at you for decision making and alarms and so for you also have maintenance task on the side.  
 
So no, definitely no, not ever. All we're doing is contributing a little bit to alleviate that cognitive overload by automating a small piece of that process where we are very effective and where there is enough predictability for us to model and manage without overwhelming the human. So again, all we attempt to do here is free up a tiny bit of productivity for you to take care of more important things. Wonderful. Next kind of tough customer question for you is there are a handful of options out there for automatic tuning. You know what?  
 
What makes autonomous tuning stand out? Yes, there's, there's definitely options and 1st and all well, manual tuning. We're not here to knock manual tuning. Manual tuning is is very effective. Again, the the only issue there is how do you repeat a process that requires human intervention without having to bring the human. So that's that's where we trying to supplement a static process with a dynamic one. There's also some physics based approaches out there and physics based is great. Everything that is programmed into the equipment by default is physics based, is gathering all that knowledge and understanding of what's going on with the boundaries in the development of that equipment and using that to optimize it.  
 
But even that pretty quickly will reach those limitations where again, the human intervention is needed to update those physics based models. Then there are a handful of machine learning solutions out there, mostly provided by those who are really good on the machine learning space, data scientist. This kind of people, yes, they have great knowledge on the artificial intelligence side, but we supplement that artificial intelligence knowledge with equipment knowledge. So they burn Nova Legacy and the hundreds or thousands of installed base out there, they give us information that is captured on that base model or your digital twin. So again you are accumulating here multiple sources of optimization and value to give you that ideal solution.  
 
Then also the fact that we are agnostic, so our understanding that many of you might have not chosen AG control system, that's very sad, but we understand it. So we will accommodate you and we'll connect again with whatever control system and this yes, might have been your choice. And lastly, this solution is synergistic with whatever might be on your hardware. So if you're thinking well, planning to do a combustor overhaul 2 years from now, would I still get value from the solution once I have the latest and greatest hardware? And the answer is yes.  
 
We aim to improve whatever hardware behavior we are managing, whether it is old hardware or new hardware. So you will always get a level of benefits on your emissions, on your hit rate and the avoidance of your manual tuning, regardless of how new your hardware might be. Well, let's see how this plays out in the real world. Can you cover examples of how autonomous tuning has helped customers meet regulation compliance as well as reducing fuel spend and vibrations? Yeah, sure. Jackie, we have a couple examples here just for the sake of time and you guys will see towards the end we give you a slide that talks about where to get additional information.  
 
You will see a bunch of use cases there, a very particularly like this one because it's, it's a little odd. It highlights the versatility of autonomous tuning as a solution and how it adds resilience to the combustion process. So this one is from a customer who was struggling greatly with their field quality. They were getting a wide range of BTU's, they didn't have control over the sourcing of their field and what this was causing was a lot of dynamics issues, high acoustics, blowouts, trips, multiple excursions requiring constant visits from PAS from donors. So huge effect on their availability and reliability.  
 
In general, they adopted autonomous tuning without introducing any changes. Again, that the fuel source was not something they could directly manage. And after the installation of autonomous tuning, this problem was pretty much solved. They didn't have any subsequent trips. So it for over that 12 month period, 0 trips on acoustics, no blowouts, all that was solved. They saw the common benefit here on the double digit production, on the emissions footprint and all of that without the constant PA and tuner deserts, but for also the same period of time. Credible.  
 
And then we have a second example here. We wanted to use this example for multiple reasons. So the first one is it highlights what we can do for a very regulated environment. Many of you as I previously mentioned might have been for many years in a non regulated environment and you find that not being the case anymore. Now you have limitations on your PPMS or you might have them all along. It's just that your equipment degradation is now reducing your threshold on being able to achieve those goals. This particular customer is in a highly regulated environment with a very strict credit scheme and they were having a lot of emission issues to the point that as in previous slides where we were discussing that ugly case scenario where regulations are the term where emissions and regulation are deciding what your MW output is and not your demand.  
This particular customer and in the summer prior to the installation of autonomous tuning had to shut down prior to the end of the season. They run out of credits. After the installation of Autonomous Tuning, they were able to run not only through their High Peak season, but also through the end of the year, which gave them some incremental revenue that year and thereafter. So again, impacts go beyond just the emissions. It also reflects on the bottom line. Fantastic. Well, thanks for sharing those customer examples. That wraps up our presentation.  
 
So now we'll open the floor to questions. Let's see what has come in. Yes, I guess see a few questions here, so let me scroll to the bottom. Yes, there is a really good question around where the emissions abatement technology can be used. And yes, it's very specific to the LED today. And the reason for that of course is, is the water injection brings a totally different flavor to your combustion. It changes how emissions are managed and somewhat reduces the impact from autonomous tuning. So we're concentrating today on the DLE, which is where we can give you the most benefit.  
 
Then there is another question, does this solution eliminates the need for initial tuning done during commissioning? Great question. The answer is no. As you guys previously, if you follow some, some of my random questions, random comments, I mentioned the, the combustion schedules. We need a robust combustion schedule as the seed of the boundaries that we manipulate in the control system, that commissioning process tuning goal or or main deliverable is that base combustion schedule. And yes, it is needed. And thereafter, that's what we would interface for with to deliver our optimization.  
 
It's also a great question to point out that at any point where you might perform a major hardware change, so anything that will affect the accuracy of that combustion schedule overall, we would need to have it updated by a tuner. So keep that in mind. Avoiding the avoidance of manual tuning is referring to your ongoing maintenance manual tuning events or maybe your unplanned tuning events related to weather or field quality or other, not necessarily that initial one or those associated with major hardware overhauls. And let's see what other questions for the LM6000 example, can you provide internal engineering detail design and validation reviews for this software solution?  
 
That is a great question. I'm not sure all that information can be shared, so let me put a pin on that one and we'll get back to you. If I'm allowed to share that information, I'll be happy to do so. Let's see. What other questions do we have? I don't have a GE Bernova Aero derivative gas turbine. We have a Mitsubishi turbine. Will this product work on my Mitsubishi gas turbine? Great question. So let's let's divide it in two parts.  
 
The technology would work. So this approach to optimize combustion through the combination of a digital twin and machine learning for variability corrections is universal to all types of combustion. As you guys heard me mention at the beginning, this is started in the cold space and then we expanded that as an application to the aero derivatives and its orders that we would like to address in the future. Mitsubishi will be an specific application that we will need to do some productization and gather some information around the specific boundaries for that turbine in order to introduce a solution. We won't rule it out.  
 
We definitely want to expand the product for many other types of combustion. We have a wide road map. We just don't have an specific date for Mitsubishi, but we'll keep you guys updated. And any. Words, Jackie. OK. OK. If if we didn't get to your question, we will follow up in the next few days. If you're interested in running your own numbers through the autonomous tuning ROI calculator to see if it's a good fit for your company, feel free to reach out to me.  
 
As a reminder, my contact information is in the speakers section and, and also you can see additional information to learn more in the Resource Center. And thank you so much for joining us today. Thank you for your time. Thank you.