Puerto Rico: how existing home batteries became emergency supply

Puerto Rico’s thermal fleet is old and often offline. The island has more than 6,500 MW of installed generation, but only about 3,200 MW is typically usable. Peak demand in August 2026 reached about 3,234 MW, the highest in four years (Metro Puerto Rico). LUMA counted 17 generation-shortfall events in fiscal 2026; the North American planning standard is about one event in 10 years (LUMA).

When supply falls short, the grid operator cuts power to circuits in rotation, known as load shedding. LUMA forecast up to 154 hours of load shedding across 36 days for July 2024 to June 2025 (LUMA). In July 2026 alone, there were 17 load-shedding events (LUMA August 2026 filing).

Each fix has a price. FEMA bought 14 temporary generators, about 340 MW, for about $306.6 million. That is roughly $902 per kW before fuel or operation, with 90% paid by federal funds (White House fact sheet). An emergency contract for 400 MW of temporary generation costs 22.4¢ per kWh (IEEFA). LUMA’s own analysis says about 800 MW of firm temporary generation would be needed to bring shortfalls near the industry standard (Metro Puerto Rico).

Doing nothing has a cost too, mostly borne by customers. Puerto Rico customers lose power for at least 27 hours a year, and close to 200 hours in bad years (pv magazine USA). Puerto Rico publishes no official value of lost load. The Energy Bureau’s judgment that avoided load shedding is worth about twice CBES incentives is the closest benchmark (PREB).

CBES did not pay anyone to buy a battery. It paid owners to share part of batteries they already had. At the end of 2025, Puerto Rico had more than 171,000 behind-the-meter batteries totaling 2,864 MWh, alongside 1,456 MW of rooftop solar (pv magazine USA). Four forces built that fleet:

Before CBES, a battery earned its owner nothing on the bill. Net metering already credits solar exports at retail, and the residential energy charge is flat, so shifting solar into the evening has no value (LUMA tariff). The battery’s value was backup. For companies that lease the systems, the batteries earned only the monthly lease or power purchase payment. One Sunrun customer pays $173.64 a month for 17 panels and a battery (Sherwood News). CBES added a new revenue stream for both, from hardware already installed.

LUMA does not dispatch batteries directly. Approved aggregators do: Sunrun, Tesla, sonnen, Fortress Power, Flip Energy, SunStrong Management and Virtual Peaker, which runs the platform for Enphase systems (Energy Changemakers). Sunnova’s customers were paused after its bankruptcy (LUMA filing). Enphase does not contract with LUMA directly; it takes part through Virtual Peaker (Virtual Peaker).

That control is what made auto-enrollment possible. LUMA limited it to aggregators whose existing customer agreements already allowed them to enroll batteries (LUMA program proposal, LUMA responses). Each aggregator reconciled its customer data with LUMA, sent a 30-day opt-out notice, then configured and tested the batteries (joint aggregator motion). Auto-enrolled batteries defaulted to sharing 20% of their capacity and keeping 80% as backup (joint aggregator comments). Fewer than 7% of Sunrun’s auto-enrolled customers opted out (joint aggregator motion).

1. The shortfall is forecast. Events are called when available generation is forecast to come within about 200 MW of load shedding (Solar and Energy Storage Association).
2. LUMA calls the event and notifies aggregators. Events last about four hours (LUMA).
3. Batteries discharge. Aggregator software discharges each enrolled battery down to its reserve. The energy powers the home first and is exported to the grid beyond that (LUMA). This is active discharge, not just a pause in charging.
4. Owners can opt out of any event. There is no penalty for skipping (LUMA).
5. Aggregators report telemetry and are paid per kWh. LUMA pays $1.25 per kWh delivered, and aggregators pay customers on their own schedule. Sunrun pays every six months (Sunrun).

Both labels are partly right. By its architecture, CBES is a virtual power plant (VPP): thousands of batteries aggregated, dispatched by software with two-way communication, exporting energy to the grid and settled on telemetry. The Department of Energy’s VPP definition fits (DOE Liftoff). LUMA files it as a demand response program because of how it is used: only in emergencies, called by the utility, paid per kWh with no capacity commitment (LUMA).

For program designers, the lesson is that auto-enrollment, not the pay level, moved the numbers. The $1.25 rate was unchanged from the pilot. On June 30, 2025, no customer was auto-enrolled; by September 30, 67,161 were, alongside 13,843 who had opted in (LUMA).

The second default mattered as much as the first. The average reserve of 73% is largely the 80% reserve that auto-enrolled batteries started with, not a choice each owner made (joint aggregator comments). Sunrun noted that enrolled homes hold about 400 MW of four-hour capacity, while the default shares only a fifth of it. In 2026, PREB approved a higher default share, which raised expected delivery from about 50 MW to about 80 MW (Energy-Storage.news). Participation was also lower than planned: 61% of customers responded to a typical event, against the 85% LUMA assumed (LUMA).

The dispatch record is real. On July 9, 2025, about 70,000 batteries delivered 48 MW against a shortfall of about 50 MW (Utility Dive, Sunrun). The program dispatched 49 times from June to October 2025, averaging 33 MW (Microgrid Knowledge). An analyst’s review of LUMA reports counts about 10 occasions from June to December 2025 when CBES avoided or reduced load shedding (9 Millones). Load shedding still happened when the shortfall exceeded what batteries could deliver (joint aggregator motion). Tesla and LUMA won a 2026 Peak Load Management Alliance award for the program (PLMA).

An average participant delivers about 3.1 kWh per event, roughly a fifth of a typical 15.6 kWh battery. At 48–58 events a year, that earns about $150–$182. Sunrun told customers to expect at least about $200 per battery in 2025 (Sunrun). That is a rebate, not a payback: it would take decades to recover a battery’s cost on CBES pay alone.

For the grid and its ratepayers, CBES is a success as emergency capacity. It needed no capital, no permits and no construction. It came online in weeks, and it costs about $263 per kW of event power a year. Generators avoid fuel limits and owner opt-outs, but cost about $902 per kW just to buy. PREB concluded that the economic benefit of avoided load shedding is likely “on the order of, twice as large as the cost of the incentives, if not more” (PREB).

It is not a replacement for power plants. Even at 75 MW, CBES covers about a tenth of the deficit. There is also a feedback loop on the bill: when LUMA asked for its record October 2026 rate, it listed greater use of customer batteries among the cost drivers (NBC News). The worse the grid, the more CBES pays out.

So the case study is less about economics justifying battery adoption and more about a utility recognizing an asset already in place. Households bought batteries for resilience. LUMA spotted a few hundred megawatts sitting idle in its territory and bought a slice of it through a traditional emergency program, for less than the alternatives.

The Puerto Rico model needs three things: a large fleet of home batteries already under aggregator contract, frequent real shortages, and a regulator willing to approve opt-out enrollment. Few places have all three. The Clean Energy States Alliance reviewed the program’s lessons for other states (CESA).

Related: how per-kWh event pay flows in the per-kWh money trail.