Intelligence First: The eSmart Systems Blog

Artificial Intelligence: How to take advantage of new technology in the energy industry

Written by Forsetlund, Henriette Næss | 15. November 2018

Chief Analytics Officer at eSmart Systems, Davide Roverso, recently sat down with Torgeir Micaelsen in the Norwegian podcast LØRN.TECH to talk about how to utilize big data and artificial intelligence in the energy industry. 

Artificial intelligence and machine learning are techniques that help us intelligently manage big data, and these technologies could be a major asset for the utility industry. The LØRN.TECH podcast invites you to learn more about technology through inspiring episodes about technologies like blockchain, drones, VR / AR, artificial intelligence and big data.

Listen to the episode with Davide Roverso here. Alternatively, just continue reading a transcript of the conversation below.   

#51: «Virtual machines»

Torgeir Micaelsen: We'll do this episode in English. Davide Roverso, Welcome. Obviously, you’re not from Norway originally.

Davide Roverso: No, I’m from Italy. I have been in Norway for a very long time.

TM: Exactly. And you work in a company called eSmart Systems. And you do have a lot of clients in Norway, but you also work internationally. So, could you start by telling us a little bit about the company and where you work today?

DR: Yes, eSmart Systems is a software company, its not a start-up anymore, now it’s a scale-up. We are about 75 people based in Halden. But we have offices in Oslo, in London and we are expanding a lot in the US. And also, in Denmark and Germany. And we develop mostly cloud-based software for energy companies, mostly electrical utilities, so we work in the power and utility sector. The core of the company is a bunch of people that have worked in this domain for a very long time, for thirty years. And about five years ago they started hearing about all these new technologies, about Big Data and AI and they took a trip to the US to Silicon Valley where they were a few months, and they come back home with the idea of how can we apply this new technology to this domain of power and utilities. So that’s how the company started. And since the beginning, we have had a very big focus on both big data and artificial intelligence. And kind of that’s my role in the company. I am Chief Analytics Officer, so I’m responsible for what we do around analytics and AI.

TM: I can say it, so you don’t have to say it yourself. You’re an academic?

DR: Yeah, originally.

TM: Exactly, so you actually know a lot about this phenomenon called Big Data, Artificial Intelligence, not just the guys running around in conferences pretending to know…

DR: Yeah, nowadays everybody is talking about machine learning, AI and big data. Not too many have a long experience in that area. And a lot of people rebrand themselves as data scientists when they have a different type of background. Because, of course, there is a huge demand for this competence, and also a lot of people are actually learning new skills, and nowadays it’s much easier than, I mean you don’t need to take a Ph.D. in machine learning to be able to do data science. So, a lot of people follow courses on the Internet, and that is very helpful. And there are a lot of very bright young guys, and some of them are with us, so we are very proud of our team.

TM: You should be, you should be. You said that you have, that eSmart Systems have, obviously, lot of experience in data analytics, big data, artificial intelligence, but you have domain knowledge about the energy sector in Norway. Is that still one of the largest client bases for the company?

DR: Yeah. It is. In eSmart Systems, we think about three domains that we need to have competence in. One is the data engineering part, about big data, cloud systems and that kind of thing. The other part is more on the AI and machine learning, and the third one is the domain knowledge on the area of which we apply this technology: the power and utilities. So, we have a lot of people with that kind of background as well. We think the combination of these three is the core of what we do and is our advantage in this area.

You might also want to read: Towards Industrial-Grade Machine Learning

TM: So, let’s talk about big data, and continue to talk about the energy sector. I think everyone understands that a grid that transports electricity - a power and utility station or something like that installations - creates data. But what is actually the data? Can you give people some examples? When you collect data from the grid, what do we collect?

DR: Well, before we go into that, I think one interesting fact is that many people don’t think about it that way, but the power grid is actually the biggest machine ever made. The most complex machine ever made. You have hundreds of thousands of kilometers and hundreds of thousands of components and everything need to be synchronized together for you and me to charge or phones or watch TV.

TM: Exactly. In my old life, in politics, we developed an expression in how to visualize how big it is. So we created; if the grid is like the highway the electrical energy. And a complex highway, because as you say.. How many kilometers did you say? Thousands of kilometers only in Norway.

DR: Yeah, there are 300 000 kilometers just in Norway.

TM: Sorry, I interrupted. Continue.

DR: No, no, but going to the data part. You know, the grid is changing and the last years a lot. You know we are going more towards the smart grid. So originally, power distribution was a centralized thing. You have big power plants, and you transmit energy over long distances and then distribute it to the lower voltages to the houses and so far, but this was a one-way thing, and not much data was collected at that time. So, a lot of this electrical utilities were kind of operating the line because they were not measuring much of what was going on, especially the lower voltage grid.

TM: So a small, lower voltage grid is kind of dumb.   

DR: Yeah, a lot of people in Norway have experience. Most of the people are getting smart meters at home. Then suddenly electrical companies have instruments in every house so they can directly measure with higher frequency consumption and all other parameters. So, there is much more data than they are able to handle. And also, there are many more opportunities to be able to operate the grid in a more optimal way. Because the grid is changing, not just because we are instrumenting it, but also because we have what we call distributed energy resources. People are installing solar panels. There are windmills, there are electric cars, that need charging. It’s a much more complex and flexible grid and that you can not operate blindly, so you need data. So that’s one big source of data now. Nowadays there are many more measuring points, much more data that comes in to be analyzed. So that is one area we work in, using machine learning to make predictions of what kind of load we can expect in a certain area or certain transformer station and so forth.

Another big area we are working on is more related to inspections. We said that the power grid is the biggest machine ever built, but it was built quite a long time ago.

TM: So, it’s getting old. It needs new investments.

DR: Yes, a lot of the infrastructure is quite old. Some is a hundred years old. And if you look at the number of outages, I have some figures from the US - just in the 80’s they had about five major outages every year. Nowadays they are above 200 a year. So, there is an increased need for inspection and maintenance. And in that area, we are using AI to analyze the big amount of data the electrical components gather when they inspect power lines. For example, they fly helicopters, and they take pictures of the grid, so this generates a huge amount of data. Big utilities like Hafslund here in Oslo collect hundreds of thousands of images every year. So that’s one example of kind of biggish data, not very huge data but hundreds of thousands images that need to be processed, so they have people that day in and day out look at the pictures and try to find problems and faults. We try to apply AI to this to automate as much as possible of this process, so we use all the latest technologies that we have today of image recognition. Instead of recognizing faces we recognize isolators and recognize problems.

TM: And when you say you apply AI, is it like you give up your own kind of algorithms and AI technology or do you reuse others, from all the biggest platform companies?

DR: We do both. We stand on the shoulders of giants. So, we use everything that is available, and now there is a huge speed in development. Every week we see some new exciting development and see opportunities of using it in our domain, but of course, all of what we can find of open source or things that are published needs to be adapted and improved to able to have an optimal application in our domain. But one of the biggest advantages, well, one big factor is not necessarily the AI technology or the model that we are using. It is the data that is one of the most important things. Recently we competed in the US with different companies such as IBM, Intel and some big partner of defense contractors in the US doing exactly this. Recognizing faults on images, and we came out on top of that competition. So that we are very proud of. But the reason for that was not that we use methods different from others, but because we had a longer history, we had more data to build our models on. We have been working on this with Norwegian utilities for three years now, and we are able to combine data from many different customers and that is a huge advantage. It’s not only the technique – the machine learning algorithm – but its also good quality data.

TM: Good quality data sets from different sources combined, then you get…

DR: Yes, a good quality data set trumps any good algorithm.

 

Listen to the full episode here.

With technologies like machine learning and artificial intelligence the energy industry has a great opportunity to analyze their data and find out exactly which areas they need to invest or do maintenance. It also brings huge opportunities for optimization of the grid and saving investment costs.

Read also our guide to efficient power grid operations for the digital age. 

Download for free:
Next Gen Utility Infrastructure Operation