Artificial Intelligence is a major catalyst for change in the upstream sector and will provide a strong competitive advantage for early adopters. In this podcast we interview Professor Ertekin, Professor Emeritus and prior Department Head of the John and Willie Leone Family Department of Energy and Mineral Engineering, the George E. Trimble Chair in Earth and Mineral Sciences, at Pennsylvania State University to discuss his 4 decades of research in reservoir engineering.
Professor Ertekin shares his experience and research in deploying technologies like artificial intelligence and machine learning to enhance and accelerate to solve problems in reservoir engineering. He also identifies which technologies are the most mature and tested to provide high confidence intervals to those that wish to deploy them.
We also explore how companies and academia can work together to accelerate the deployment of these technologies to create predictable outcomes in the field. Ultimately, it is the innovators in the space that are willing to deploy these technologies that will gain a significant competitive advantage.
TRANSCRIPT BETA
Disclaimer: Our transcription software is in beta testing mode, using NLP and AI to convert our conversations from audio to text. We openly admit that at present, the output quality is poor, with numerous errors throughout (some of which are hilarious in nature), however, we would like to release this information should it be of value to any users.
With the help of time and money, we plan to increase the quality of this output as we build out sponsors that help support the growth of the show.
Adam Cohen: Hello Professor Ertekin and thanks so much for joining us here today. It's an honor to have you.
Professor Ertekin: It's my pleasure. Thanks Adam.
Adam Cohen: So I know you seem like a very humble man. So I'm gonna give you a quick introduction here for all of our listeners so that they can understand a little bit about your background in the extent of your experience. So today your head of the John and Willie Lyon family Department of energy and mineral engineering at Pennsylvania State University. you've been a professor at the school since 1978 and have had a tremendous amount of impact on students and the petroleum industry as a whole found a few quotes here on the internet from leading Executives speaking, very highly of you one is from Dave Stover who refers to you as a leader and innovator and and someone who is always inspired him in his career. They're also Martin Craighead who's CEO of Baker Hughes. He says that dr. Erdogan has enriched so many lives including mine his passion and knowledge for Reservoir modeling and Reservoir engineering helped lay the foundation for my future career and his wisdom and guidance helps me build on that Foundation. I think that's reflected strongly in what I found on your Personal bio here that for you you want to strengthen students thinking and their analytic powers and though you are demanding person. You you push your students towards success and that's that's something that you value in your relationship with your students. You know with that introduction. I think one more Point here that I'd like to make before we get more of a background from you as far as your career. In this space is the the recent article that you've released back in 2019. And in that article artificial intelligence applications and Reservoir engineering a status check you give a very detailed and thorough overview of how AI is is going to impact the space of reservoir engineering and I'm looking forward to getting into that conversation with you here today.
Professor Ertekin: All right. Thank you very much Adam. I also thank you very much for that. Very generous introduction. And yes, I've been at Penn State for a long time and I had the I was the lucky person having so many wonderful friends. I promised udents who work with me together over the course of years. We have the same in Turkish. We say that it's not possible to clap with one hand. So If you make some sound bites, I think we did them. And so that's what I have to I did them together with my shoes. underline this perspective as a fact. I am I left 12 engineer by education. I did my undergraduate studies. And also my Master's Degree at the middle the state clean diversity in Ankara Turkey and came to to Pennsylvania State University in 1975 is a PhD student. I work under the guidance of who was also very top who still is by the way that time he decided to move to Canada. And I they approached me and my wife that they would like to say we thought that the area Beaver was a wonderful place to raise a family and we stayed here and we've been here since then. Throughout my teaching and research career. I worked in the area of reservoir engineering and my phone bill training is Reservoir engineering. And so I taught courses in hard Computing classical Reservoir engineering and then mid-nineties 95 96 we stopped work in the application of artificial intelligence in Reservoir engineer and at that time when we started I want to combine or synthesize the conventional or hard Computing techniques with soft Computing techniques such that I could go and I could go and fix exploit the advantages offered by 2 is approaches and that's how we started. So it's been now almost 20 25 years we've been working in this artificial intelligence area is the technique is applied. So there's our voltage drops.
Adam Cohen: so that's pretty impressive that you know, right when I guess personal computers came on the market you you started using them right away to to use them in the models and applications that you were running for your and Reservoir engineering case studies and research
Professor Ertekin: One day this is correct. We'll close the the years that we started. We were doing most of our computational work using mainframe computers and personal computers were not available at that time and it was a much more difficult undertaking just running a complex simulation problem and The storage was limited computational time involved was much much larger speed was not there. So therefore I think myself and my friends my contemporaries we came through that grinder more or less and then when the electronic explosion took place starting late, I would say call it meet 80s late 80s, then we realize that the things that you want to implement could be implemented in a much easier way much faster way and in a much more effective way more importantly.
Adam Cohen: So it sounds like a lot of what you've done has gone hand-in-hand between Reservoir engineering from The Human Side and Reservoir engineering from the computer side. Would you say, you know kind of looking back was there? What did the way that the future moved forward from that from when you really started using that technology in the 80s in a more academic setting did you see the industry doing the same as far as running side-by-side with technology As Time Advanced or was there a Divergence away? What was the difference between the relationship I guess is my question in a more academic setting versus. Out in the actual business space where those similar where those different.
Professor Ertekin: Well III, I think it was quite similar writing quite a number of years ago when Reservoir simulation commercial Reservoir simulation techniques starts to appear in the Horizon and the Implement and there was some kind of education and and some colleagues were looking at the technique with a young boys. I should say and and then this continues for a relatively long period of time I still can get maybe about until maybe 15 years ago 20 years ago. I used to hear. Well, I really do not need these sophisticated models. I can still live with my material balance equation and then solve the problem using a simple tool. Black material balance for example, but I think if you look at the material don't balance technique, and also if you look at the reservoir simulation and sophisticated formulation, then I can take or anybody can take any Reservoir simulation equation and can collapse it into a much simpler form, which is mature adult a question, but during that process when collapsing it into it's very simple tool. You have to manage many assumptions and you need to simplify the problem. And as you simplify the problem, there's always a danger that you'll be moving away from the field conditions from the problem. Actually that you are trying to solve. So therefore it's a very thin ice that we have to walk very carefully if we are going to move in that direction. Action because it can really take us in a certain way that we solve a problem in a simple way but complex problem in a simple way, but it has nothing to do with the original problem that we are trying to solve. So it's someone said at one point in time complex every simple complex problem has a favorite every time. The complex problem as a simple solution and but they are all wrong. So that's something that we have to keep them. Keep that statement in our minds always.
Adam Cohen: And imagine that becomes more and more true as technology advances right as technology becomes more complex and more capable in its abilities to solve complex problems. It requires even deeper understanding of how the computations behind those Solutions are operating in relationship to the original problem. And I imagine, you know, just thinking this through in this was actually a conversation. I was having earlier today is you got to start with the problem first and then work your way backwards, but there's got to be a point when the the one who originates the problem in this case the engineer goes back and interacts with the data, but you know, maybe the engineer doesn't have that programming background to where they know how to communicate with was the computer. And what's the systems and then you end up in a pretty complex scenario where there may be a language barrier between the two that prevents growth. Is that is that would you say it's an accurate kind of summary of where things are at today and
Professor Ertekin: Exactly what I tried to say in the translation of the problem from the field to to a computer to a Computing machine device. We want to make sure that some important things are not lost in the translation so such that we can stay very close. The problem we know that the it's very difficult to be to replicate the entire problem with every demand in every Dimension which every single entity but still we should make every effort to stay as close as possible to the problem. There are sometimes Problems by the way, we really do not know exactly which mechanisms are playing which what are the Dynamics Is that are controlling governing the entire entire State of Affairs within the reservoir? So therefore there's something that we need to we need to be very careful. I only think that I always think that this is true for some other professions as well in Reservoir engineering. They are always working with a very huge jigsaw puzzle and just imagine a jigsaw puzzle with me. Be a hundred thousand maybe 1 million pieces of Twisted which hit many pieces and then and you are trying to assemble the puzzle. We are trying to put all these pieces together in a meaningful way, but it's a difficult task that but if you also think that Some of the pieces of the puzzle maybe out of 100,000 pieces maybe 5,000 of them are missing then the problem becomes even much more complicated but we are talking at that point. They are talking about what we call an ill posed problem what it means in simple words. Number of unknowns XE exceeds the number of equations that we need to solve the system altogether. So then it is the reservoir to assign responsibility. First of all to identify these pieces are missing and then Again, individual heads to create those pieces and then put them plug them in in the right place. So that's where the creativity of the human. It's okay to talk to engineer did you know scientists will come into the picture and if you don't have those that you mentioned the human dimension in place, you may have the most sophisticated computational devices systems and the we do not or whether we can subscribe our results. I mean ourselves to the results generate. G by D by dose device. So therefore again, it is something that we have to make sure that all the balances and checks are in place at every stage of the analysis and complications.
Adam Cohen: Yeah, a really common, you know phrase that to use throughout technology spaces garbage in garbage out. If you don't have good quality information to put into the system, then you're not going to get good quality output. And that's an interesting kind of concept. I think when you're talking about, you know, what you're talking about here is the oil and a conversation that's been happening a lot. Not recently with tech innovators in the oil and gas space and that is that the future of oil and gas may be more about tech companies running oil operations rather than oil it operations using Technologies. And the reason why I think it's relevant here is because the traditional Wildcat methodology while you know, while while has proven out, Out all of most of the oil that we have in this country, you know, when you're dealing with a hyper conservative Capital Market that doesn't want to spend in that those wildcatting Ventures. It really changes the landscape of what the oil and gas sector looks like and so I think it's it's interesting because a lot of what started oil and Us as an industry in America if you read the history, it's all about hunches on the lay of the land and you know hunches that there's oil deeper down. You just got drill deeper those kind of Concepts that are more human in nature and you look to the Future to today where you have a tremendous amount of data around fields that have already been proven out where technology makes a lot more sense and especially Ali in an environment like today where the capital is way more conservative than it has been before. so what you know to that point is there is there space for AI in true exploration wildcatting for Reservoir engineering how much data do we need to have enough pieces to put together that jigsaw puzzle or is it more of an in offsetting, you know development strategies or in fill type strategy situation or where you have a bunch of Cora logs to base your models off of
Professor Ertekin: Well again, we should see the AI applications is that other tools that the car in front of us as long as we use them in the right way we can try to expand the applicability of the technique from from the wild. It occasion from the exploration stage and later full development and that Maybe not maybe much more easily into impulse drilling programs. I mean, we need to make some critical decisions. So therefore again, as I said earlier a i self is going to be another tool in the engineers and geoscientists toolkit to make decisions. So therefore it's a decision-making tool and then we are using the machine to accelerate that decision and make sure that that decision is made in New York. Very large volume of data is the data becomes larger. Then the applicability of say I will become even much more effective. So that's something that we need to keep in mind as well typically typically in in our Reservoir engineering calculations use deterministic models something like a times B is equal to C and A represents the characteristics theater. Presents the project design parameters and T represents the response function a times B is equal to C. So therefore how do we make it population? So in order for this calculation to be accurate precise first of all, the characteristics the entries for a should be correct and then most probably we have Understanding about the project design parameters that you are implementing. So therefore most probably we are good with that design parameters as they are in place and then it's becomes simply multiplication. But just imagine if the entry certain entry in a are not correct then the response function which is typically either too. All rate or pressure. Okay certain location at a certain time as function function will be wrong if the response function wrong and if we take that response function, which is not correct and then try to use it in an economical analysis package there is going to take us towards the deep end. So that's four. have to be careful is that for AI can help us in assembling this large and categorizing these large sets of data and then sites that we can come up with much with a very large number of scenarios and try to create a number of different scenarios and then look at the results look at the confidence intervals and then hopefully maker Right decision at that point in time. So therefore it really strengthens the hand of the engineer or scientist who's making the analysis towards making the right decision.
Adam Cohen: So how much you know before we start moving into the really technical piece of this conversation, you know, we're an engineering background is probably pretty helpful to fully understand the the topic that will topics that we'll cover, you know from just a again a high level perspective here. How much of How much I guess Systems computer systems programming background does one need to work intelligently with the Advanced Technologies that are coming out right now. Is it something that you need a comprehensive background on in order to use today's Technologies or are there enough pre-built packages out there that it's a plug-in play like scenario that as long as you have good data. That you're confident in there's a platform there that you can plug into to generate results.
Professor Ertekin: I think in Artificial intelligence application, for example if he takes a special news networks as a told that we want to use then there are extensions of certain computational software. You can call that artificial neural network extension and then use but more importantly still before the Software takes the problems and then Grimes the numbers and that works with it and comes up we can resolve it is important that someone has to structure the topology or of the old architecture of the net and so at that point. It is very much important that you know, what are the input parameters? What are the output parameters? And...
Comentarios