Two separate sizing problems: energy and power
A backup battery has to pass two independent tests, and failing either one leaves you in the dark. The energy test asks whether the bank stores enough kilowatt-hours to run your loads for as long as the outage lasts. The power test asks whether the inverter can supply the kilowatts those loads demand at any instant, including the surge when a motor starts. Capacity is measured in kWh, power in kW, and a product datasheet quotes both — a 13.5 kWh unit rated 5 kW continuous stores plenty for an evening but cannot start a 3-ton air conditioner and run a well pump at the same moment.
The energy calculation begins with a genuine load figure. The single most common error in residential storage is sizing to a guess. Your utility bill gives whole-house consumption, but backup panels carry only selected circuits, so you need the consumption of those circuits alone. Work it up appliance by appliance — the daily load audit calculator does exactly this — or read a monitoring device that meters the critical-load subpanel.
Then adjust for what the battery cannot give you. Nameplate capacity is not usable capacity: the battery management system reserves a fraction to protect cell life, and every kilowatt-hour that leaves the battery loses a few percent in the inverter on its way to your loads. A 13.5 kWh nameplate battery at 90% depth of discharge and 90% round-trip efficiency delivers 10.9 kWh to AC loads, not 13.5.
Installation in the United States is governed by the National Electrical Code Article 706, which covers energy storage systems, together with Article 705 for interconnection and Article 710 for stand-alone systems. NEC 2023 Article 706 sets the disconnect, working-space and listing requirements your inspector will check; it does not tell you what size to buy, which is what this calculator is for.
The formula, and why it has a floor in it
Start with the energy the battery must actually deliver. Over an outage of D days, your loads consume E_load × D, and any solar recharge gives back E_solar × D, so the battery supplies the difference. Divide by depth of discharge and efficiency to convert delivered energy into the nameplate capacity you have to buy.
The formula includes a floor of one full day of load, and the reason matters. If the array returns as much energy per day as the house consumes, the naive difference is zero and the arithmetic says you need no battery at all. That is false: solar produces during daylight and the house consumes around the clock, so the battery still has to carry every night. The floor makes the capacity never fall below one day's load, which is the minimum any solar-plus-storage system needs to bridge from sunset to sunrise. It is a conservative simplification — the true overnight requirement is the share of your daily energy used between sunset and sunrise, usually 40-60% — and erring on the high side here is cheap insurance.
Depth of discharge is set by the product, not by you. Lithium iron phosphate home batteries typically publish 90-100% usable capacity; older lead-acid banks are limited to 50% if you want any cycle life at all, which is why a lead-acid backup system needs roughly twice the nameplate capacity of a lithium one for the same job.
Round-trip efficiency in this calculator means energy delivered to AC loads divided by energy removed from the battery. Manufacturers quote 86-92% for AC-coupled residential systems. Do not enter the cell-level DC efficiency, which is higher and does not include the inverter.
Inverter sizing uses a different input entirely: the largest set of loads that can run simultaneously, plus a 25% design margin so the inverter is not operating continuously at its ceiling. Motor starting is checked separately against the inverter's surge rating, which is typically around one and a half to two times continuous for a few seconds — confirm the figure on the datasheet rather than assuming it.
Worked example: 12 kWh a day for a two-day outage
A household backs up a critical-loads panel carrying the refrigerator, the furnace fan, well pump, lighting, internet and a few outlets. Metering shows those circuits consume 12 kWh a day. They want to ride out two days with no solar recharge. The battery is a 13.5 kWh lithium unit with 90% usable depth of discharge, and the system delivers 90% round-trip efficiency. The largest simultaneous running load is 3.5 kW, and the well pump surges to 7 kW on start.
- Net daily draw. 12 kWh − 0 kWh of solar = 12 kWh a day.
- Energy the battery must deliver. 12 × 2 = 24 kWh, which is above the one-day floor of 12 kWh, so 24 kWh.
- Usable fraction of nameplate. 0.90 × 0.90 = 0.81. Only 81% of what you buy reaches the loads.
- Nominal capacity required. 24 ÷ 0.81 = 29.63 kWh.
- Module count. 29.63 ÷ 13.5 = 2.195, rounded up to 3 modules, giving 40.5 kWh nameplate.
- Energy actually deliverable. 40.5 × 0.81 = 32.81 kWh.
- Runtime. Average draw is 12 ÷ 24 = 0.5 kW, so 32.81 ÷ 0.5 = 65.6 hours, or 2.7 days.
- Inverter rating. 3.5 × 1.25 = 4.4 kW continuous. Check that the chosen inverter's surge rating covers the 7 kW pump start; at 4.4 kW continuous, a 1.6× surge capability is required.
The rounding up from 2.195 to 3 modules is significant here — it buys 0.7 extra days of runtime and adds a third of the capital cost. Two modules gives 27 kWh nameplate, 21.87 kWh deliverable and 43.7 hours, which covers the two-day target with about 10% to spare if the load estimate is accurate. Whether the third module is worth it depends entirely on how confident you are in the 12 kWh figure, which is the argument for metering the panel rather than estimating it.
Now add solar. Suppose the array can reliably return 6 kWh a day even in poor weather. The net draw halves to 6 kWh a day, the three-day requirement becomes 18 ÷ 0.81 = 22.2 kWh, two modules suffice, and the runtime with that recharge stretches to 87 hours. Solar changes the answer more than any other single input, which is why an outage-resilience conversation should always start with whether the array is there.
How to read the result
Compare the runtime against your target, not against the marketing claim. The runtime figure assumes your load draws evenly across the day. Real households draw more in the evening, so the battery may hit its floor before the arithmetic says it should, and then recover overnight when demand falls. Treat the runtime as a good average, not a guarantee for any particular hour.
The inverter rating is often the binding constraint, not the capacity. If your continuous requirement exceeds what one module's inverter supplies, you are buying a second module for its power rating rather than for its energy, and the runtime will be longer than you needed. That is a legitimate design outcome, but you should know it is what you are paying for.
Cutting the load is almost always cheaper than adding capacity. Removing electric water heating, EV charging and central air conditioning from the backed-up panel typically halves the daily energy on a critical-loads design. At current installed prices, an hour of electrician time moving a circuit competes well with a battery module.
The solar recharge figure must be a bad-weather figure. Grid outages correlate with storms. Entering your annual-average production overstates what the array will actually give you during the event that matters. Use something near your worst month's daily average, or run a cloudy-day case in the solar output calculator and use that.
The runtime with solar is capped at 720 hours in this display. When the recharge equals or exceeds the load, the battery cycles rather than depleting, and the true limit is a run of consecutive days too dark to recharge — a quantity this model does not know. Read the cap as a signal that the constraint has moved from capacity to weather.
Typical critical-load energy for a backed-up subpanel
| Load | Running power | Hours or cycles per day | Energy per day (kWh) |
|---|---|---|---|
| Refrigerator (modern, 20 cu ft) | 150 W running | 8 h equivalent run time | 1.2 |
| Chest freezer | 120 W running | 7 h equivalent | 0.8 |
| LED lighting, whole house | 200 W | 5 h | 1.0 |
| Gas furnace blower and controls | 500 W | 6 h | 3.0 |
| Well pump, 1 hp | 1,000 W | 1 h | 1.0 |
| Internet router, modem, laptop | 100 W | 16 h | 1.6 |
| Television and receiver | 150 W | 4 h | 0.6 |
| Microwave, kettle, small kitchen | 1,200 W | 0.75 h | 0.9 |
| Sump pump (wet season) | 800 W | 1.5 h | 1.2 |
| Typical total | — | — | 11.3 |
Running powers are representative nameplate figures for common residential equipment; your appliances will differ and the labels on them are the authority. Refrigerators and freezers cycle, so the hours column is equivalent full-power run time rather than hours the appliance is plugged in.
Sizing mistakes that show up during the outage
- Sizing to nameplate capacity instead of usable capacity. A 13.5 kWh battery at 90% depth of discharge and 90% efficiency delivers 10.9 kWh. Ignoring both factors overstates runtime by about 24%.
- Using whole-house consumption for a critical-loads panel. The backed-up circuits typically consume a third to a half of the whole house. Using the bill figure buys two or three times the battery you need.
- Forgetting motor starting current. A well pump or air-conditioning compressor draws several times its running current for a second or two. If the inverter cannot supply it, the pump does not start and the inverter faults, whatever the state of charge.
- Assuming the array recharges the battery during an outage. Grid-tied inverters shut down without the grid unless the system is explicitly designed for islanded operation. Confirm your equipment supports it, and confirm the array is connected on the backed-up side.
- Entering annual-average solar production. Outages happen in storms. Use a poor-weather figure or the requirement is met only on the days you did not need it.
- Ignoring cold. Lithium batteries cannot charge below freezing without integral heating, and heating draws from the battery. An unconditioned garage in a northern climate is a real derating.
- Backing up the electric water heater. A single tank can consume more energy in a day than every other critical circuit combined. Take it off the panel and heat water on the grid when it returns.
Where a battery sits among the alternatives
Against a generator, a battery wins on noise, maintenance, instant transfer and indoor air quality, and loses on multi-day outages where a fuel supply keeps producing energy after the battery is flat. A standby generator with a fuel line is the more capable choice for a week-long ice storm; a battery is the better choice for the two-hour and one-day outages that make up the overwhelming majority of events. Some households install both and use the battery as the transfer device.
Against a bigger array, extra modules do nothing during a night-time outage and everything during a multi-day one that happens to be sunny. Storage and generation are complements, not substitutes: the array extends the battery's runtime, and the battery lets the array serve loads at all.
Against demand-charge or time-of-use arbitrage, note that backup sizing and economic sizing give different answers. A battery bought to shift energy from midday to the evening peak is sized to the evening load for a few hours, not to a multi-day outage. If you want both, size for backup and check the arbitrage value separately — under net billing tariffs the daily cycling value can be the larger part of the case, which the net metering savings calculator helps you quantify.
If you are building a system with no grid at all, this is the wrong calculator. Off-grid banks are sized to days of autonomy at a specified bank voltage with a temperature derate, and the design revolves around amp-hours rather than kilowatt-hours — start with the off-grid battery bank amp-hour calculator.