Report 162 · Energy Storage
After Hurricane Helene washed out the substation feeding Hot Springs, North Carolina, a solar-and-battery microgrid kept the town's downtown lit for 143.5 hours. It is a genuine success story, and it usually gets told as "the battery powered the town for six days." That is not what happened, and the real version is more useful to anyone sizing backup power, from a utility down to a single house.
Hot Springs is a town of roughly 500 to 600 people in Madison County, fed by a single 10-mile distribution line that runs from a substation in Marshall through the Pisgah National Forest. Outages on that line were frequent, and a second line through the forest was judged too costly and too disruptive. So Duke Energy built a microgrid instead: a 2-megawatt (AC) solar plant and a lithium battery, with control software from Wärtsilä, placed into service in early 2023 at a reported cost of $14.5 million.
The design goal was modest and specific. According to the Southern Alliance for Clean Energy's account, the microgrid was designed to serve the town's entire load for 4 to 6 hours if the line went down, long enough for crews to fix most routine outages.
Helene's remnants hit on September 27, 2024. The Marshall substation was badly damaged by flooding, and SACE reports a temporary fix alone could take two weeks. Duke lost contact with the microgrid. When its crew reached Hot Springs on September 30, working with the mayor and the state transportation department to get through washed-out roads, they found the batteries depleted and downtown needing structural repairs before it could be re-energized.
By October 2 the batteries had recharged from the solar array and the microgrid began serving the town as an island. Jason Handley, who runs Duke's Distributed Energy Group, told WSOC-TV that from 10:30 a.m. on October 2 until a mobile substation was installed at 10 a.m. on October 8, "the downtown did not lose power," adding: "We powered it completely on solar and batteries." That window is the 143.5 hours in the Smart Electric Power Alliance case study, which says the outage could otherwise have lasted more than 262 hours.
So a system built for 4 to 6 hours ran for roughly 24 to 36 times its design duration (my arithmetic: 143.5 divided by 6 and by 4). How?
1. The load shrank. This is the part that rarely makes the retelling. Handley, quoted by SACE:
The damage in the town reduced the overall demand for power. That allowed us to use the solar and battery to deliver electricity to downtown Hot Springs all the time – and the rest of the town, all except for some overnight hours. We stretched the capabilities of the microgrid.
A flooded business does not draw power. The 4-to-6-hour design figure assumed the whole town's normal load. After the flood, the load was smaller, and Duke also asked the community to conserve energy to extend the microgrid's run, according to an Advanced Energy write-up based on interviews with Handley.
2. The battery was refilled every day. A battery's rated energy is only its runtime if nothing refills it. Here, a 2-megawatt solar array recharged it each day. The battery's real job was to carry the evening and night, not six days.
3. Not every circuit stayed on. Read Handley's quote again: downtown all the time, the rest of town "all except for some overnight hours." In a LinkedIn post quoted by Canary Media he put it this way: "Depending on solar output, we've also been able to bring on other load segments for periods of time." Handley described to WSOC how the team pushed further each day to maximize how long each load segment could stay on. WSOC's own summary is plain: the microgrid was not able to keep the power running continuously for everyone. Downtown got uninterrupted service; outlying segments were switched on when the sun made room for them.
That is not a knock on the project. It is the correct way to run an islanded system with a fixed battery and a variable source: decide what is critical, keep that on, and serve the rest when energy allows.
One detail is worth flagging because it trips up a lot of battery coverage. Duke's own 2023 announcement describes the battery as "a 4.4-megawatt lithium-based battery storage facility." Canary Media, WSOC and a January 2026 Idaho National Laboratory report describe it in megawatt-hours (Canary and INL write 4.4 megawatt-hours; WSOC writes "4.4 MW of on-site battery power"). Megawatts measure how fast a battery can deliver power; megawatt-hours measure how much energy it holds. They are different quantities, as I explained in Report 095, and runtime depends on the second one. I could not find a Duke primary source that states both.
If the battery holds 4.4 megawatt-hours, as Canary and INL report, a 4-to-6-hour design would imply the town's normal load averages very roughly 0.7 to 1.1 megawatts. That is my arithmetic, conditional on the megawatt-hour figure, not a number from Duke.
The honest version of this story includes the failures, and Duke has been open about them. In an April 2023 windstorm, the first real-world test, the microgrid did not island automatically; Duke had to intervene manually, and it carried the town for 12 of the 48 outage hours. Handley called it "a successful failure" that produced a long list of fixes.
Helene surfaced another one, in Handley's words to Advanced Energy: "The batteries need batteries." The uninterruptible power supplies that keep the microgrid's communications and controls alive had to be recharged when the crew arrived. Duke says it is adding a small generator for those auxiliary loads and plans to include that in future designs where long outages are expected. That is a detail every backup-power designer should steal: the control system is a load too, and if it dies first, the big battery is a brick.
The same arithmetic that governed Hot Springs governs a home battery. Rated capacity divided by your load gives a runtime, and that runtime moves by multiples depending on what you choose to keep on. Solar that can recharge the battery turns a fixed countdown into a daily cycle, weather permitting. And the controls, the inverter, the gateway, the communications, have to survive the whole event.
When someone quotes you "X hours of backup," the useful follow-up questions are the ones Hot Springs answers: at what load, with what recharging, and which circuits are on the list.
I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator. I do not own that company and earn nothing from this link. Full policy here.
The decisions that stretched Hot Springs (which loads stay on, when to let a segment back in, how much charge to hold for the night) are exactly the kind of decisions battery-cycling software makes. I find this case useful because it shows those decisions matter as much as the hardware. Nothing in this report is about that company's products; the microgrid is Duke Energy's.
The SEPA case study itself. The 143.5-hour and 262-hour figures come from the public summary page of the Smart Electric Power Alliance case study by Jared Leader (SEPA) and Jason Handley (Duke). The full report sits behind a registration form, which I did not submit, so I have not read it.
The 4-to-6-hour design figure comes from SACE's article, not from a Duke document I could open. It is consistent with the case being described as a stretch, but I have not seen Duke's design basis.
Load and solar data. No source I found publishes the town's load or the array's output during the event, so I cannot say how much of the stretch came from reduced demand versus shed segments versus solar. All three are described by Duke; their shares are not.
Population. Sources give "just over 500" (Duke, 2023), about 535 (Canary) and about 600 (SACE). I have used a range.
The Hot Springs microgrid is a real success, and the lesson is not that a small battery can run a town for six days. It is that runtime is a design choice. A system built for 4 to 6 hours at full load carried downtown for 143.5 hours because the load dropped, the sun refilled the battery every day, and the operators kept the critical segment on while rationing the rest. Ask any backup quote the same three questions, and you will know what it can actually do.
Scope note: this report explains how the Hot Springs microgrid outlasted its design duration, using the utility's announcement, an industry case-study summary, a national-laboratory report and contemporaneous reporting. The 24-to-36-times ratio and the implied 0.7-to-1.1-megawatt load are the author's arithmetic, labelled in the text. No system was measured and no product is evaluated or recommended. Disclosure: the author helps design AI battery-cycling systems for a veteran-owned energy-storage integrator linked in the body; he does not own it and receives nothing for the link.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.