The Future of Energy
(Fourth in a Series)
The coolest thing about Artificial Intelligence (AI) is that it can be layered on top of existing infrastructure. However, don’t picture it like putting a robot in front of a keyboard. That is most certainly not how it works. For one thing, it’s much more efficient than that!
You’ve seen evidence of AIs “typing” when you interact with a ChatBot online. Some are so cleverly designed that you can’t tell if it’s a person or a program.
Try the one at your local electrical utility to see if you can tell if it is a person or a ‘bot… The difference is that one of these ChatBots can carry on hundreds of simultaneous conversations with users. No more sitting on hold for 40 minutes, waiting for a person, to find out how long the power is going to be out! Just use your smartphone and get an instant answer from a ChatBot who is also talking to 200 other people wondering the same thing as you…
More particularly, you may remember a mellifluous female voice, affectionately named “Silicon Sally,” created by the telecommunications industry back in the 1980s in response to the need to automate telephone switching systems. True or not, people pictured Silicon Sally as a sophisticated speech-synthesis chip that could be plugged into just about anything that needed to speak, and sound convincingly like a human.
We’ve moved well beyond that, and you experience those advances on a daily basis with Siri or Lyra on your smartphones, Alexa (Echo Dot) in your home, and online animated characters that park themselves in the corner of your screen telling you about a website. Possibly even on your phone, though you may not have recognized an AI helping you if the need was uncomplicated. The voices sound very human, and the ability to cope with regional accents is getting better every day.
Keeping Customers Happy
For technophobes or simply dedicated phone users, new enhanced systems use voice-recognition and speech-synthesis. After recognizing the request for information about the power outage the conversation would go like this:
“What is the address experiencing the power outage?” to which the system makes a choice, based on the reply and what it knows about the current situation. It might be a simple “I’m sorry, but we don’t know the cause yet, but please feel free to call back, and we will update this information as soon as it becomes available” or “We are unaware of an outage at your location. Please let me connect you to one of our technicians to assist you further.”
In either case, the customer doesn’t need to know it’s an AI, but it lets them feel heard and, if the problem is new, connects them to a technician who isn’t harried with calls from customers, and who can deal with that customer immediately.
If you want to keep Federal or State regulators off your back, keeping customers happy is important. No complaints = no visits, however, that is not the only way that AI will help.
Consider the United States power distribution system. It’s a collection of almost 6,000 power plants and about 3 million miles of transmission lines. The three major grids are loosely associated (West, East, and Texas), able to shift power loads, if necessary, by covering shortfalls or accepting excess when the availability exceeds demand. That’s all well and good, and exactly how it should be.
The problems arise, however, when one power line has a ground-fault, and a series of cascading failures take out one whole grid, like the 2003 Northeast Blackout that left 50,000,000 people without power for two or more days. It’s estimated that with current technology, this will happen every 25 years. That is too often.
The government has mandated a “smart” energy grid before it happens again, and it will be driven by AI. With more than 15 million Smart Meters already widely installed, and with anonymized communication back and forth between customers and utilities, the regional grids now have instant feedback on the state of the grid, but it is a lot of information. Too much for a quick response by humans that are each focused on their generating plant, and their portion of the grid.
Add the complexity of residential and commercial green power generators intermittently feeding power back into the grid, from solar panels, wind generation, and biomass of many types. Sunny days, windy days, overcast days, windy nights—all of these cause significant fluctuations in the grid. Also, with estimates of a 25% increase in national demand by 2050, 30-year-old (average) power plants, and 40 year old (average) transformers could be under significant strain.
The AI would “know” all of these things, all of the time. According to estimates, a well-designed smart grid would see no more than a 1% increase in peak loads, despite higher demand. Yes, by 2050 those power plants and transformers would be 63 and 73 years old, respectively, but load balancing would keep them safe.
Problem: With 25,000 to 100,000 messages per second, covering 30-50 data points, it’s not hard to see why it represents information overload for humans monitoring a grid. By continuously (and remotely) altering power usage of customers in a way that doesn’t impact their operations, the demand could become predictable and steady.
That is accomplished by first, one grocery store running its refrigeration units, then another one follows, and as each finishes a cycle, another one can start. The same goes for air-conditioning units for office towers or any high-demand consumption. Participants are rewarded with lower costs, reliable service, and financial incentives.
It keeps prices down for all consumers, rewarding them when they lower demand, or allow remote control of high-demand equipment that can fluctuate without affecting their production (like air-conditioning). Also, it saves producers money because load-leveling means we could essentially do away with fossil-fueled Peaker Plants, eliminating that source of Green House Gases (GHG) entirely. Peaker plants are the most wasteful and costly aspect of energy generation for producers.
Customers with domestic solar arrays and battery storage capability could make their batteries available to the grid as well. When demand was low, they would charge. When demand was high, the utility could draw from hundreds of thousands of customers’ batteries, paying them for the privilege. While some limited number of people might regard this as an invasion of privacy, it is not significant, and getting paid for your storage capacity would decrease the payback time for your solar power investment, all while reducing GHG. Win-win-win!
The British are Coming!
This highly workable system is under development in the UK and helps keep the Peaker plants offline. It’s much cheaper to pay customers to use their extant batteries than to fire up a Peaker plant, for just 60 hours annually, or another for just a couple of hundred hours per year.
Keep in mind that an AI would monitor everything through deep learning, taught and managed by a neural network; it would know where a major problem had developed, why it had happened, and know how to shunt power around to prevent a cascade failure. It would keep the whole system in balance.
Something like the NE Blackout of 2003 would be virtually impossible to replicate. There is no mystery to it; when you can see everything and understand it in context, solutions are rather easy to come by.
Guardians of the Grid
Of course, there will always be people, foreign states, agencies, and so on that want to commit harm. Messing with a power grid could be highly effective at throwing a government into confusion. If you have millions without power, and some foreign state is performing illegal missile tests, the response to both situations could be slow and ineffective, precisely what the attackers had in mind.
Artificial Intelligence, properly programmed, can identify an electronic attack in milliseconds and stop it. Characteristic behaviors of backdoors, Trojans, worms, DoS, or DDoS will fail against an AI because the activity doesn’t get to run for minutes before a human notices or a Gatekeeper program is activated.
Where shall we drill?
AIs are helping us locate additional resources of natural gas and petroleum. Geologists are great and have been locating gas and oil for more than a century now, but letting them teach an AI, turning it into an Expert System (a program with the combined expertise of hundreds of top specialists) will allow the AI to pinpoint areas where there will be oil and gas. Dry wells are expensive and used to be necessary because of incomplete data.
Cleaner, More Energy, Less Maintenance
Siemens conducted a test in Asia that has now gone mainstream. By turning over control of a gas-powered turbine generator to an AI, they realized a decrease of 1/5th of their NO2 output, a significant GHG. General Electric is now employing an AI to manage wind generators to increase their energy production by 5% while decreasing maintenance by 20%. A German AI called EWeLiNE can utilize history, propensities of consumers, and weather predictions to calculate green energy output for the subsequent two days. This provides additional stability so producers can plan accordingly and not be forced into quick decisions that over or under-respond to situations, wasting energy.
New materials, new technologies, and AI will make ideal partners. After decades of R&D, we’re finally looking at commercial-scale battery-storage. This will significantly ease the strain on the grid. If your average requirements are 10 Gigawatts, and your peak is from 3 pm to 7 pm using 3 GW, with appropriately sized battery farms, your production can stay steady at 0.41 GW per hour, every hour, all day long.
Running anything at a steady pace is much more efficient, and requires a smaller investment in maintenance, than something which is constantly adjusted. Notwithstanding equipment failure, the AI could create remarkable savings by producing exactly what is needed with no shortfall or excess.
All the independent power generators don’t necessarily observe the other systems adjacent to them, or their impact upon them. Cooperation is an essential component to making sure our power grid is reliable.
One overarching AI, monitoring everything, could make sure every part of the grid was fully powered and that every single bit of green energy was used. No plant with significant GHG production would be chosen when a cleaner alternative was available. That is how we get to clean, cheaper to produce, and cheaper to purchase energy—with the help and direction of Artificial Intelligence that can analyze mountains of data quickly and efficiently.
If you want to learn more, we provide webinars to show you where the future is going and how you can get there. Don’t shortchange yourself. Join us today!