Load shifting with a home battery: TOU arbitrage explained

Our time-of-use guide is about the rate plan: how to find yours, how to read it, and how to work out your spread. This guide is about the machine. What does the battery actually do during a load shift, what does it lose while doing it, and at what point does the arithmetic stop working?

The short version is that load shifting has one formula, and most explainers get it wrong by leaving out the losses. Below is the formula, then the same battery run against two real published tariffs, where it earns about 79 cents on one summer day and loses about eleven cents on a winter one.

Every rate figure here comes from a named utility tariff sheet with its effective date. Nothing here is a savings estimate for your house. It is arithmetic on published numbers, offered as a worked example.

The mechanic, in one paragraph

During the cheap window, the battery draws power from the grid and stores it. During the expensive window, your house runs off the battery instead of the grid, so the meter barely turns while the expensive price is in effect. Your consumption does not change. Only the moment of purchase changes. Each kWh you move that way avoids the peak price and instead costs you the off-peak price, and the difference is your margin.

Except that the battery does not give back everything you put in. That is the part that decides everything.

The one formula

A battery has a round-trip efficiency: the share of the energy you put in that you get back out. Call it E. To deliver one kWh to your house at peak, you must buy 1 divided by E kWh at off-peak. At 90 percent, that is about 1.111 kWh in for one kWh out.

So the margin per kWh delivered is:

peak price minus (off-peak price divided by round-trip efficiency)

And the breakeven condition is simply that the peak price has to exceed the off-peak price divided by efficiency. Rearranged into spread language, which is how most people think about it:

the spread must exceed the off-peak price times (1 divided by E, minus 1)

At 90 percent efficiency that multiplier is about 0.111. A 20 cent off-peak price needs a spread above about 2.2 cents to break even. A 40 cent off-peak price needs a spread above about 4.4 cents. Expensive off-peak power raises the bar, which is the counterintuitive part: a high off-peak rate is not neutral, it actively makes arbitrage harder.

Efficiency figures are not directly comparable across makers, because they publish different bases: AC to AC, DC to DC, CEC weighted, or solar round trip. Our round-trip efficiency explainer covers why, and our model pages record the basis alongside every figure. For load shifting you want the AC to AC number, since both the charge and the discharge cross the inverter.

PG&E Schedule E-TOU-C, worked through

PG&E's Electric Schedule E-TOU-C is the residential time-of-use plan with peak pricing from 4 p.m. to 9 p.m. every day. From the tariff sheet in PG&E's tariff book, read August 6, 2026, the total bundled energy rates are:

| Season | Peak (4 p.m. to 9 p.m.) | Off-peak |

| --- | --- | --- |

| Summer (June 1 to September 30) | $0.52240 | $0.39940 |

| Winter (October 1 to May 31) | $0.39757 | $0.36757 |

The same sheet applies a baseline credit of $0.08140 per kWh to baseline usage in both windows. Because it applies equally to peak and off-peak, it does not change the spread, so it drops out of the arbitrage arithmetic entirely.

A note on dating these, because it is easy to cite the wrong thing. These energy rates were set by PG&E Advice Letter 7846-E, submitted February 27, 2026 and effective March 1, 2026, and they are the rates in force on the date we read them. The rate sheet itself was re-issued on June 1, 2026 under Advice Letter 7921-E, which changed only the California Climate Credit line and left every energy rate untouched. So the sheet carries a June date while the prices carry a March one. The same applies to the EV2 sheet below.

Take a battery at 90 percent AC round-trip efficiency, which is the figure FranklinWH publishes for the aPower 2 on a grid to battery to load basis and the figure Enphase publishes for the IQ Battery 5P and 10C on an AC round trip basis.

Summer. Margin per kWh delivered is $0.52240 minus ($0.39940 divided by 0.9), which is $0.52240 minus $0.44378, or about $0.0786. Shift 10 kWh across the five-hour peak window and the arithmetic is: about $5.22 of peak charges avoided, about $4.44 spent buying the 11.1 kWh needed to deliver them, for a net of about $0.79 on the day.

Winter. Margin per kWh delivered is $0.39757 minus ($0.36757 divided by 0.9), which is $0.39757 minus $0.40841, or about minus $0.0108. The same 10 kWh shift avoids about $3.98 and costs about $4.08. It loses about eleven cents, before any consideration of wear.

That is the honest headline. On this plan, grid-charged load shifting is a modest summer proposition and a negative winter one, because the winter spread of three cents does not cover the round-trip losses on a 37 cent off-peak price.

If you repeated the summer cycle on every one of the 122 days in PG&E's summer season, the arithmetic totals about $95.92. We are stating that as arithmetic on the published rates, not as a forecast for your house: it assumes a full 10 kWh shift every single day, no backup reserve held, no cloudy interruptions, no rate change, and no value assigned to the wear.

The same battery on PG&E Schedule EV2

Now change nothing except the rate plan. PG&E's Schedule EV2, billed as EV2A, is its home charging plan. From the tariff book sheet, read August 6, 2026, on the same Advice Letter 7846-E rates effective March 1, 2026 as the schedule above, the total bundled energy rates are:

| Season | Peak | Part-peak | Off-peak |

| --- | --- | --- | --- |

| Summer (June 1 to September 30) | $0.53809 | $0.42760 | $0.22558 |

| Winter (October 1 to May 31) | $0.41099 | $0.39428 | $0.22558 |

The sheet defines peak as 4:00 p.m. to 9:00 p.m. every day including weekends and holidays, part-peak as 3:00 p.m. to 4:00 p.m. and 9:00 p.m. to midnight, and off-peak as all other hours, all year round.

Same battery, same 90 percent, same 10 kWh shift:

• Summer margin: $0.53809 minus ($0.22558 divided by 0.9), which is about $0.2875 per kWh delivered. On 10 kWh, about $2.87 on the day.
• Winter margin: $0.41099 minus $0.25064, which is about $0.1604 per kWh delivered. On 10 kWh, about $1.60 on the day.

Across the same 122 summer days, that arithmetic totals about $350.69, with all the same caveats.

The gap between the two examples is the whole point of this page. Identical hardware, identical behaviour, identical utility, on the same calendar day: about 79 cents against about $2.87. The battery did not get better. The off-peak price fell from about 40 cents to about 23 cents, and almost all of the difference came from that.

There is a footnote worth knowing here, because EV2 sounds like it requires a car. It does not, necessarily. The schedule's special conditions state that customers with battery storage may take service on the rate schedule on a pilot basis even without an electric vehicle, that participation is limited to 30,000 storage-only customers, that eligible customers must apply for interconnection and be granted permission to operate, and that installed storage capacity must be at least 0.05 percent of the previous twelve months of consumption for customers above 6,000 kWh a year, or at least 2 kWh below that. Other utilities set their own rules. The general lesson is that the rate plan is often a choice, and it is a bigger lever than the battery.

What round-trip efficiency does to the same rates

Holding the E-TOU-C summer rates constant and varying only efficiency shows how much the battery's own losses matter on a narrow spread:

| Round-trip efficiency | Margin per kWh delivered |

| --- | --- |

| 86 percent | about $0.058 |

| 88 percent | about $0.069 |

| 90 percent | about $0.079 |

| 93.8 percent | about $0.097 |

Those endpoints are not arbitrary. Across the batteries we track, 86 percent is the CEC weighted figure on the SunPower SunVault datasheet, 88 percent is Generac's system round trip for the PWRcell, 90 percent is the AC round trip published by FranklinWH and Enphase, and 93.8 percent is the figure on the Savant Power Storage 20 specification sheet, the highest among the all-in-one systems we track. Four DC battery modules in our set publish higher percentages, up to 98 percent on the Fortress eVault Max, but those are cell-level or module-level figures measured without a full inverter path, so they are not the number to run this arithmetic with. Compare bases, not percentages. On a wide spread the difference between them barely registers. On a narrow one it is most of the margin.

What the arithmetic above leaves out

Five things, and they all cut the same way.

Warranty consumption. Cycling daily for a rate plan is a much heavier duty than sitting charged for outages, and every warranty has a throughput or cycle limit alongside the years. At one full cycle a day, from the datasheets we record: FranklinWH's 15 years or 60 MWh works out to 4,000 full cycles on a 15 kWh aPower 2, which is about 11 years at a cycle a day, so on a hard daily-cycling duty the throughput cap arrives before the 15 year term does. Shift less than a full battery a day and the term binds instead, which is the general shape of these caps: the calendar governs light use and the throughput governs heavy use. Enphase's 15 years or 6,000 cycles is about 16 years of daily cycles, so there the term binds first either way. Savant's 10 years, 73 MWh or 3,650 cycles works out to exactly 10 years at a cycle a day, so they expire together. Generac's PWRcell module warranty of 10 years or 7.56 MWh per module is 2,520 full module cycles at the published 3.0 kWh each, which is about seven years at a cycle a day, so on that model the throughput cap arrives before the term does. Our warranty guide covers how to read these.

Backup reserve. Any capacity you hold back for outages is capacity you are not shifting. Generac publishes this trade explicitly for its program participants: it says the most common backup reserve setting is 20 percent, and that lowering it toward zero increases what the battery can contribute. Twenty percent reserve on a 13.5 kWh battery is roughly 2.7 kWh you are not arbitraging.

Days you cannot fully cycle. The arithmetic assumes a full shift every day. Real households do not consume a full battery's worth inside every peak window, and a battery that only displaces 6 kWh on a given evening earns six tenths of the margin.

Whether you are allowed to charge from the grid. This is not universal. Rhode Island Energy's ConnectedSolutions battery program page, read August 6, 2026, states that the battery storage system must be charged from an on-site renewable resource such as solar to participate. If you are enrolled in a program with that condition, grid arbitrage and program income are not simply additive. Ask your utility whether grid charging affects your solar tariff or your program eligibility before you configure anything.

Power is rarely the constraint, capacity is. PG&E's peak window is five hours. Spreading 13.5 kWh evenly across five hours is about 2.7 kW, well inside the continuous rating of every whole-home battery we track. For load shifting, the number that limits you is usable kWh, not kW. Continuous power matters for backup, not for arbitrage.

If you have solar, the arithmetic is different

Everything above assumes the battery charges from the grid, so the cost of a stored kWh is the off-peak retail price.

With solar, the cost of a stored kWh is not a price you pay. It is the export compensation you gave up by keeping that kWh at home. Under California's net billing tariff, exports are valued closer to the utility's avoided cost rather than at retail, which is why our NEM 3.0 explainer and the fuller NEM 3.0 guide frame the battery's job as self-consumption rather than arbitrage. When forgone export credit is low, solar charging is cheaper than any off-peak rate, and the margin per shifted kWh gets much wider than the examples above.

The practical implication: a solar owner should run this arithmetic with their export rate in the charging-cost slot, not the off-peak retail rate. And note that California net billing customers are moved onto specific electrification rate plans, so the E-TOU-C and EV2 numbers here are illustrations of the method rather than the plan you would be on.

How to run this for your own house

1. Get your tariff sheet, not a summary. Your utility publishes the schedule with an effective date. Take the peak and off-peak totals from it.

2. Get your battery's AC round-trip efficiency and its basis. The model pages in our comparison set record both.

3. Run the breakeven test. Is peak greater than off-peak divided by efficiency? If not, stop. Grid arbitrage does not work on that plan at that efficiency, no matter how large the battery is.

4. Multiply the margin by what you can realistically shift, after backup reserve, on the days you actually have peak consumption to displace.

5. Check what a daily cycle does to your warranty, using whichever of the year, throughput, or cycle limits binds first.

6. Ask about grid-charging restrictions on your tariff and on any program you are enrolled in.

Then put the result into the Worth It calculator alongside your backup value and any state or utility program payments, because on most plans in most states, load shifting is a contributor to the case for a battery rather than the whole case. Our guide to when a battery is not worth it covers the plans where it contributes nothing at all.

One last note on 2026 money. None of the arithmetic above is softened by a federal credit. The Section 25D residential clean energy credit expired on December 31, 2025, so a homeowner buying a battery outright this year gets zero federal cash, and the only surviving federal route is Section 48E through a lease or power purchase agreement where a company owns the hardware. That makes the daily margin above carry more of the argument than it did two years ago, which is exactly why it is worth calculating honestly rather than optimistically.