SDS 272: Data Science in Energy

Welcome back to the SuperDataScience podcast ladies and gentlemen today we’ve got yet another episode of data science in a specific industry. And the industry we’re looking at today is energy.

So the energy industry encompasses all the industries involved in the production and sale of energy including fuel extraction, manufacturing, refining and distribution. And check this out. Some experts even consider oil and gas exploration extraction to be the original big data industry. That’s how important data analysis has been in this space. And that’s because there’s large costs and large profits and where those things occur naturally businesses want to be competitive, they want to reduce costs, increase the profits and data science is exactly the tool that can help accomplish that. World energy expenditure is about 6 trillion US dollars per year. That’s trillion with a T and that comprise about 10% of global GDP making it the second after health care industry in terms of size and in many countries is actually the first in terms of expenditure. So let’s look at some use cases and examples of data science and artificial intelligence in the space of energy.

Number one, use case number one, predictive maintenance. As you can imagine, these oil rigs and massive oil plants, oil wells, they require a lot of machinery and any kind of breakdown at these sites can cause a lot of delays, a lot of opportunity costs and a lot of losses. So for instance, here’s an example. A day’s production at a small site is about a thousand barrels of oil and that represents $30,000 of revenue at current prices. So therefore, companies cannot afford to allow the machines to break down. They’d rather fix them beforehand, maintain them while they’re still operational to keep producing energy or keep extracting energy. An example here is General Electrics’ Predix platform helps oil and gas businesses create automated analytics models to facilitate predictive maintenance with machine learning. The Venezuelan oil company, Petro PR incorporated this platform to configure software trained personnel and establish functional guidelines to improve monitoring and failure prevention. This reduced the number of damage to field gauges, standardized data collection, and significantly reduced the number of urgent repair cases.

So there you go. That’s a data science in predictive maintenance. Actually we have in one of our courses in the Tableau Advanced course, there’s an example not for specifically an oil company, but in this example of using data science full predictive maintenance. And that is in mining. In that part where the coal is mined, it’s a little bit to the port. There’s machines that low like sort out that coal and load it onto ships. So we did an example of predictive maintenance there. So if you did that course and did that specific part of the course of predictive maintenance, you will know already the value of predictive maintenance. And here’s an actual application, real life in this space of energy.

All right example or use case number two of data science in energy is energy grid optimization. So here’s some stats. There’s about 58,000 power plants in the US alone, 58,000 and about 2.7 million miles of power lines. Again, just in the US and as you can imagine, they get old and they deteriorate. And of course they sometimes require replacing, but it takes time to get to the places where it need to be replaced or in fact if they’re not maintained on time, if the energy grid is… Here, we’re not talking about maintenance, we are talking more about energy grid optimization, meaning that if the energy grid is not used properly, if there’s too much load in certain parts of the grid where they shouldn’t be, that can cause huge blackouts and can cause damage to the infrastructure. For instance, there was a huge blackout in 2003 called the Northeast blackout that left 50 million people in the US without electricity for two whole days. So in some cases it was fixed earlier than two days, within a few hours, but many people were without electricity for two whole days.

And another complication on top of this for the grid is that now people are private users. They are generating their own power, whether it’s with solar or other means. Basically there’s power being generated not just by the power plants. And that power is going into the energy grid. And that’s a very complicated supply structure which can negatively affect the network. And so what data science can do is can help monitor and control this power network and balancing out the grid, making sure that there’s no threats or problems coming up. The example here is that the US department of Energy has invested $4.5 billion. That’s billion with a B in small smart grid infrastructure and installed over 15 million smart meters that monitor energy usage per device and alerts utilities of local blackouts.

It is estimated that while the total US energy demand is expected to increase 25% by 2050, 2, 0, 5, 0, this program will limit the rise in peak electricity load on the grid to only 1%. So you’ve got a 25% increase in demand on energy. But thanks to this program, the peak electricity load is showing an increase by 1%, which is fantastic. If they are, they manage to accomplish that. All right, so that was use case number two.

Use case number three, energy balancing. So the energy market has lots of users, lots of different players in it. And here I can actually speak from experience because in our consulting firm in Bluelife.ai, we’ve been working on a project in the space of energy balancing. And it is a really interesting space. There’s large distributors involved, there’s large suppliers involved, there’s large companies, corporate companies involved, and the factories and plants involved that require electricity, there’s consumers.

So there’s lots and lots of players and basically electricity needs to be moved around. And the goal of what countries are after is to make sure that electricity or you know, like local governments what they’re after is that they want to make sure that there’s enough electricity for everybody because the electricity, it’s not just like for us as users, it seems like you put it into… you switch on electricity, you put in your cattle into the, the socket in the wall and there it happens. But electricity needs to be moved around between power stations, between locations. Even between jurisdictions sometimes and therefore somebody has to balance all that out. And there’s huge departments, huge facilities, lots of people working in this space in order to balance out the supply in the amount of electricity, in different locations, at different points in time.

And they require data science in order to predict, right? You cannot balance it out on the spot. You need to move with electricity beforehand or might not be just like this. It might be gas that you’re moving. You need to move it beforehand in order for it to be there where it needs to be. And in order to do that, you need to predict what is the demand going to be. Otherwise they just won’t be enough power, enough energy in that part of the country or part of the continent in order for it to meet the supply. So there’s a very, very interesting space, heavily data science and artificial intelligence driven because of the amount of costs that that can save and efficiency that can add.

I cannot speak to the example of Bluelife.ai because that’s all confidential. However, I can share another example that we found, which is a Swedish energy company, Vattenfall worked with data analytics company SAS to develop an energy balancing calculation software with the ability to forecast, simulate and optimize their power plant auction offers. They’re reporting 5 to 15% success increase in the field of balancing energy. And if you put that into context or like you put the numbers to it and we start talking about trillions of dollars or billions of dollars to make it more realistic, you know, 10% of $1 billion is $100 million. That’s the numbers there or the cost savings and potential for increasing efficiency is incredible. So even 5 to 15% success, increase is huge. So yeah, another massive use case of data science and AI in the space of energy.

All right, example or use case number four, exploration. So back in the early days, the way you’d explore for oil and gas is you’d look for seeps near the surface where the oil or gas naturally bubbled. That’s not the case anymore. There is… It’s so much more sophisticated now because simply we’ve already used up all or many of those sources where it was that simple. Now very complex geological analysis has to come into play and test wells have to be created, which cost over a million dollars at times. The mistakes are quite expensive in this space. That’s why the more we can apply data science and artificial intelligence to find these locations, the more efficient this industry becomes. And we don’t have like a specific example here with numbers, but in general, to date 3D seismic data has been the industry’s most impactful scientific breakthrough. This data vastly improves the picture of the earth’s subsurface and removes the need to drill a multimillion dollar hole. And with very little data, you can already assess what is in that rock, and you kind of explore it and save some costs. Okay, so that’s use case number four, exploration.

And use case number five renewable energy. Renewable energy sources include sunshine, wind, tides and others. And they’re quite heavily dependent on climate. In terms of the market itself up to a trillion dollars will be invested in green energy projects in the coming decades. And by 2030 renewable energy will comprise almost 30% of the US generating capacity. And this is where data science can come into help. Data scientists can find and work on datasets, such as historic regional wind speed, direction and intensity of sunshine, tide and currents, terrain models and so on. In order to impact the performance of these renewable sources, if it’s going to be 30% of the US generating capacity just alone in the US, how important is it going to be to make it very efficient and get the most out of it? And here we’ve got an example, IBM’s HyRef, H, Y, R, E, F, can predict incoming weather patterns and calculate wind turbine performance from 15 minute intervals up to 30 days in advance with 92% accuracy.

How crazy is that? 92% accuracy, 30 days in advance at 15 minute intervals. This can increase the amount of renewable power generation that can actually be used by 10%. In 2017 that was enough to power an additional 14,000 homes. And that’s all thanks to data science. So there’s a couple of ways of looking at the energy industry. One way is to look at it as something that’s potentially some people might consider energy as evil as like we’re abusing the plant. But on other hand projects like this where you are using renewable energy in order to help, with data science, to help an additional 14,000 homes, I think that’s very admirable. And that’s a very noble thing to do. So you can always find applications of data science that are for good and for the right cause in any industry that you look into. And that’s one of the examples in the energy industry.

So there you have it, ladies and gentlemen, that’s data science and artificial intelligence in the energy industry. If you’d like to get the links for this research, you can find them at www.superdatascience.com/272. That’s www.superdatascience.com/272. There you’ll also find the transcript for this episode if you’d like to check it out. 

If you know anybody in the space of energy, whether it’s a data scientist, a manager, a business owner, and entrepreneur, feel free to forward them this episode and share the love, spread the word about the data science field that they’re interested in. And thank you so much for being here today. I look forward to seeing you back here next time. Until then, happy analyzing.