Tech Briefs: What got you started on this project?
Ben Ollis: In January 2018 I went to Puerto Rico after Hurricane Maria and talked to a lot of the folks there on the island about how we could help. We ended up doing a project for industrial scale microgrids. That project has since been completed, but I had established a connection to the island. One of my colleagues, Mike Ferrari, had volunteered some of his time in the town of Adjuntas, where they were installing a solar rooftop for energy resiliency. He introduced me to the community, where we saw an opportunity to take the research area that I focus on — microgrids — and apply it at the community level. The DOE Solar Energy Technologies Office (SETO) had called for microgrids to provide resiliency based primarily on solar energy, and we thought this fit the bill really well. We proposed a project and were lucky enough to win.
Tech Briefs: Roughly how big is the town?
Ollis: It’s about 18,000 people, not a huge town. It's in the central portion of the island, very mountainous, pretty remote. Some of the people there went months without power, I think some of them up to nine months with no power after Maria. It's an area that is definitely seeing issues with energy resiliency.
Tech Briefs: How many microgrids are there and how much of the town do they cover?
Ollis: There are two microgrids, which primarily cover the town square — the central part of the town — and the businesses located there. There are a variety of businesses: bakeries, pizzerias, hardware stores. There's a pharmacy, a church, a bank — basically the commerce center for that small town.
During Hurricane Maria, there was a lot of community gathering in that area. The pizzeria had its own generator and was able to feed people. The hardware store provided tools for people to do repairs on their homes.
So, our thought was that the businesses could set up their own cooperative to provide resiliency, which would then benefit the people living in the town. If another hurricane rolled through, they could continue to provide things like food, shelter, and water.
Each of the microgrids covers a certain portion of the businesses. There are seven or eight on one and about four or five on the other, totaling 13. It was designed that way based on the infrastructure layout — the distribution of loads. Some of the larger loads are on the smaller microgrid. So, for example, the bakery uses a lot of electrical energy to run ovens and coolers.
They were split up to have roughly about the same amounts of load and energy generation. Doing two also provides extra resiliency. If there was damage to one of the microgrids, the other could still be safely operated to provide some service.
Tech Briefs: Could you tell me about your orchestrator. What it does, how it interacts with the microgrids?
Ollis: In a microgrid, typically you have a controller that communicates with the assets — any solar, storage, generators, loads — anything that wants or needs to talk to their microgrids. Typically, the microgrid controller tells the battery when to charge or discharge. It coordinates the protection and a number of different functions.
But talking about microgrids coordinating with one another gets a little fuzzy. To my knowledge, I don't think there are any commercially available systems that can do coordination at a wide scale for microgrids, other than a full utility scale system — distributed energy resource management (DERM). And those can be expensive. A lot of utilities don’t have access to those, and even if they do, it depends on how much distributed energy there is, whether or not you're getting the full benefit of that type of system.
So, instead of going really big and trying to cover everything all at once, we designed this orchestrator to see what the microgrids can do to dispatch their assets and to coordinate them with one another. The orchestrator is designed to ensure that one microgrid is not pursuing a target or a use case or a set point that would be less resilient long term than if you had multiple microgrids.
So, for example, if you’re islanded [disconnected from the utility grid] for a long period of time, the battery energy in one of the microgrids might be running low — if it gets to zero you could black out that microgrid. However, the orchestrator could coordinate those two and share some energy so that both could stay up. The goal is to extend the life of the microgrids for as long as possible.
Tech Briefs: Are these both normally connected to the utility grid?
Ollis: Yes, their normal operation is to be grid-connected. What that results in is they have to have agreements with the local utility — purchase agreements for energy exports and imports — things like that. Although they’re grid-connected, they are designed in such a way that they can operate grid-independent for long periods of time, or under ideal conditions, indefinitely, although there are practical limits. Our goal is for them to operate for as long as possible following a disaster.
Tech Briefs: What kinds of practical limits are there?
Ollis: So, things like clear or cloudy days, maintenance, other things like that could cause the microgrid to go down.
Tech Briefs: Is the system up and running now?
Ollis: At this time, the microgrids are not yet fully installed, although the solar is in place. But the microgrid wiring and battery storage are still in progress. We're in year two of a three-year project, where year three is the demonstration space. We're still waiting on the energy-storage system. There’s a long lead time for batteries right now, so it will take some time before we can demonstrate the microgrid functionality.
Tech Briefs: Does the utility grid see the orchestrator as a single load?
Ollis: Yes, the orchestrator controls the two microgrids in such a way that the utility could view it as a single entity — in that sense, it sort of acts as an aggregator for the microgrids. So, from a utility perspective, instead of having to keep track of 10 solar inverters and two energy storage inverters and a bunch of switch gear and other things, there is just a single point they communicate with — the orchestrator.
Tech Briefs: How is the sunshine there, is it pretty constant?
Ollis: It's pretty good, they get a good bit of solar insolation [a measure of the solar energy that falls on a specified area over a set period of time]. Since they’re in the mountains, however, there is some patchy cloud cover, often in the afternoon. And it rains almost every day around 3 p.m. — you can almost set your watch to it. But morning to midafternoon, there's plenty enough to to run the microgrids.
Tech Briefs: It said in the description of your work that you investigated the particular needs of different users, such as the baker. Could that be done with AI?
Ollis: Certain aspects of it could — things like trying to learn the behavior of the load over a particular time period. So, if the baker every Tuesday at 5 a.m. fires up their oven to cook the bread for that day and it's very consistent, you could pick up on that and consider it. But we're really dealing with the load on an aggregate scale. We see what's at the meter, so we see the total load for the building. It's not like we could learn exactly what's happening, we could just learn in general what the patterns of consumption are.
An advantage of talking with people is to learn why they do what they do. If we then wanted to ask them to change their future behaviors, it would help us figure out which ones were changeable while still keeping within their business model. Things like what's the minimum amount of load they could run at in an emergency and still provide their critical service to the community. For those sorts of things, you have to get boots on the ground and talk with people to understand what to do.
When you're running as a microgrid, you don't know when your bulk energy supply is going to come back. So, it definitely pays to be frugal with your energy consumption. Your behaviors don't just affect you, they affect all the other businesses that are fed by the microgrids — it's your neighbor and the other people in the community. That’s a bit of a mindset shift that has to happen for people when they get into situations where they are running in microgrid for an undetermined amount of time.
Tech Briefs: It seems like you'd need some humans managing the system, educating people, things like that. Are there any arrangements being made for that?
Ollis: We do some partnering with the environmental organization Casa Pueblo, which is based in Adjuntas. They do a lot of community outreach and even have some public spaces where they offer free solar-based electricity for charging electronics. They are also sending out messages about how important it is for people to change their mindsets about using energy — to realize that we need to operate with resiliency in mind as opposed to just our normal day-to-day habits.
We also have partners at the University of Puerto Rico, Mayaguez, who have been doing a lot of public education in the town and visit often to talk with people in the community about this.
Tech Briefs: Are you thinking of generalizing this for use in other places?
Ollis: Yes, absolutely. We designed the orchestrator concept in such a way that it's not specific only for this use case. We tried to make it such that if you have two or three or four or 10 or 12 or 15 microgrids, no matter how many, the orchestrator can accommodate it. It doesn't really matter much where the microgrids are installed, as long as you have the correct information to support them. For example, information about the loads, estimates of solar insolation, and estimates of wind and wind capacity in that area if wind is part of it.
So, as long as you have the right inputs into the orchestrator, it should be agnostic as to the location.
Tech Briefs: Could you tell me one more time the kinds of things the orchestrator organizes.
Ollis: The orchestrator is really looking into the future to coordinate the microgrids in such a way that it provides the most resiliency for the most people who are islanded most.
It's trying to make sure that if there's excess energy in one microgrid, it gets routed to another.
It's predicting the solar insolation for the next few days; it’s predicting the load for the next few days — it’s trying to predict out over the next two to five days what the expected operation of the microgrids will be.
As an example, if we have access to weather reports and we see that we're expecting low solar insolation two days from now, we can take steps well ahead of time to try and conserve energy because we know we might require shifting some load, reducing load, maybe storing some extra power in the batteries to make sure we can ride through that period of low solar insolation and continue to operate the microgrid.
So, managing energy, managing interconnections between microgrids, and managing loads is what the orchestrator is looking at.
Tech Briefs: Would someone have to input that current weather forecast?
Ollis: For the most part it can run on its own, completely automated. As long as it can communicate with a weather service, it can get a forecast for 10 days.
But, since in an emergency situation there's no guarantee you have an Internet connection, there’s also a local repository of historical data that we can use to make predictions of future load and solar power.
There definitely is the option for humans in the loop. For this particular installation, it won't be 100 percent hands off. There will be certain situations where we will need a human to take a look and do some action. However, we try to automate as much as possible, so you don't need someone constantly monitoring the microgrid.
Tech Briefs: How is the weather information automatically input into the system to affect control?
Ollis: If you have an Internet connection you can just go out to the public weather API and pull down a 10-day weather forecast for any location, store it locally, and feed that directly into your solar-power forecast. Also, temperature is going to be very highly correlated with your HVAC load for a given building.
So, the weather forecast is one of our best friends when we talk about extended microgrid planning. We can take a 10-day forecast and feed it into a model of the microgrid that exists on the controller, run the model for the duration of that data, and then have an algorithm that can use the model to make control choices at a certain point in time based on how we've coded it. So, either charge or discharge the battery, or cut the load, or improve the load.
There are a few different methods we use. For example, it could be based on weights throughout the system that fluctuate throughout the day under different conditions. There are others that are price-based — there are quite a few options.
The model then informs the controller as to what it should do to maintain the resiliency of the system for the desired duration. We can run it for a week and the algorithm will work on the model and say this is the best solution we could come up with to operate the system for this entire week. For example, you should charge the battery one time, discharge it one time. Basically — these are the control actions that you should take over a long period.
Tech Briefs: Is there anything you’d like to add?
Ollis: To sum up, the orchestrator is all about trying to provide a highly resilient system for an area that’s seen a lot of reliability issues and natural disasters. If we can get this to work and get it out there, I think it could make a difference — so we're really excited about it.
Postscript: Hurricane Fiona made landfall in southwest Puerto Rico on September 18. According to the National Weather Service, the town of Adjuntas received 22 inches of rain on that day alone. Rivers overflowed and there were landslides on roads that cut off access and communication to rural areas for many days. The good news is that the installed solar panels were not damaged, and the Casa Pueblo organization has been donating thousands of solar lamps to residents. Casa Pueblo director Arturo Massol-Deya also reports that his organization’s own small microgrid, which was installed before this project had begun, has been available for residents to charge phones and medical equipment.