Ten past four, and the inverter is beeping
The failure has a signature. Third grey day, no wind, ten past four in the morning, and the low-voltage alarm goes off in a cupboard. By the time anyone is properly awake the fridge has already shut down, the pressure pump will not start because the bank sags below the inverter's cut-out the moment the motor asks for its inrush, and the display is reporting a state of charge that was reading forty per cent at midnight. Nothing broke. The bank did exactly what the arithmetic behind it said it would do, and the arithmetic was assembled one afternoon from the labels on the backs of appliances.
Those labels are the problem. A watt figure moulded into a plastic housing describes an instant, usually the worst instant, and a bank is bought against a whole night. Between the two sits duty cycle, standby draw, ambient temperature, the conversion loss in the inverter and the plain fact that half the things drawing current in a van at three in the morning are not on anybody's list — the monitor, the gas alarm, the router, the inverter itself sitting idle so that a laptop charger stays live.
There are really two purchases hiding inside one number, and they fail differently. Energy is the hours between sundown and sunrise, multiplied by however many of them arrive with no charging in between. Power is a half-second in February when a submersible pump turns over. A bank can be generous on the first and hopeless on the second, and a specification quoted only in kilowatt-hours will never tell you which one you have bought.
Meter the night; do not total the stickers
Start with the compressor fridge, because it is the load that most often decides the answer and the one the label describes worst. Take a unit that pulls forty-five watts while the compressor is turning. In a well-insulated cabin in October it may run something like four tenths of the hour, which is eighteen watts averaged across the day and a shade over four hundred watt-hours. Park the same van in full August sun with the door opening every twenty minutes and the compressor runs most of the hour, and the same appliance with the same sticker is asking for double. Neither figure is wrong; the sticker simply was not answering the question.
The water pump behaves the opposite way and hides a different trap. A twelve-volt diaphragm pump drawing eight amps is a hundred-watt load, but it runs only while a tap is open, so its honest daily contribution is minutes and its energy is trivial. What is not trivial is a weeping fitting or a tired non-return valve. That pump then wakes every few minutes all night to restore pressure it keeps losing, and the bank empties into a puddle under the floor. It is one of the commonest reasons an owner reports that the battery went flat for no reason, and a shunt log finds it in an evening because the cycling shows up as a regular sawtooth at an hour nobody is drawing water.
Then the loads nobody counts. An inverter idling at twenty-five watts costs six hundred watt-hours a day, which on the figures above beats the fridge — and it costs that whether or not anything is plugged in, which is why a switched inverter and a separate small DC supply for the trickle loads is often worth more than another battery. A diesel heater takes a short heavy bite on every start for its glow plug, so a thermostat set to short-cycle costs far more than the same heat delivered in longer runs. Monitors, alarms and the battery management system itself sit under all of it at a couple of watts each, permanently.
None of that is knowable from a catalogue, so measure it. Fit a shunt-based monitor in the negative return, which counts amp-hours in and out rather than guessing state of charge from terminal voltage, and leave the system running normally for seventy-two hours in the season you actually care about. Read the amp-hours consumed at the same time each morning. Multiply by the nominal bank voltage and that is your daily load in watt-hours — the number every calculation downstream depends on, and usually not the number the spreadsheet produced.
- Fit the shunt on the battery negative, upstream of every load and every charge source, so nothing bypasses the count.
- Synchronise the monitor at a genuinely full charge, otherwise every percentage it reports afterwards is offset by whatever it was short at the start.
- Run three normal days and nights in the worst season available, with the heating, the fridge and the water in their real duty.
- Log amp-hours out at a fixed hour each morning, and note the ambient temperature alongside — the fridge figure moves with it.
- Isolate the inverter for one of those nights and log again. The difference is the idle draw, and it is the cheapest saving on the whole system.
- Convert to watt-hours at nominal voltage, then set the average overnight watts beside it: the first sizes the bank, the second sizes what has to survive until dawn.
Put the measured bank capacity against the average watts the night actually draws — not the peak, and not the sum of the labels — and see how far past dawn it reaches before anything is recharging it.
The total energy storage capacity of the battery bank.
The total wattage of everything running off the battery.
Inverter losses AND the battery usable depth of discharge, multiplied together.
Estimated runtime
5.67 hours
Real-world runtime varies with battery age, temperature, and how deeply it's safely discharged (especially for lead-acid batteries, which shouldn't be fully depleted) — check your specific battery manufacturer's guidance.
- Usable energy
- 1,700 Wh
They open the calculator with your figures already in it
Battery Backup Runtime Calculator: 5.67 hours — shown in imperial, US market. The link sets both, so the result they see is the one on your screen.
What this calculation does not cover
- Assumes the load is constant for the whole discharge. Anything that cycles - a fridge compressor, a well pump, a heat pump - draws its full watts only part of the time, so a runtime worked out from summed nameplates comes out far shorter than the bank really delivers, while one worked out from an average hides the peaks. The inverter's own standby draw also continues whether or not anything is switched on, and it is not in the figure unless you add it to the load.
- Runtime is an energy answer and says nothing about whether the system can carry the load at all. Whether the inverter and the battery can supply the continuous watts, and the starting surge of a motor or compressor, is a separate power question this calculation does not touch.
- Takes the bank as fully charged with nothing recharging it. No solar, generator or grid contribution is in the figure, and a bank already part-discharged when the outage began gives proportionally less than the runtime shown.
- The efficiency figure is fixed for the whole run. Capacity lost to cold, capacity faded over the bank's life, and the extra loss a lead-acid bank takes when it is discharged fast all have to be built into that one number yourself - no correction for temperature, age or discharge rate is applied.
- This is not a standby-battery compliance calculation. Fire alarm, emergency lighting and similar life-safety batteries are sized by a prescribed multi-stage calculation with its own derating and recharge requirements, set by the governing standard and the panel manufacturer's own sheet, and a general runtime figure does not substitute for it.
A bank is five things and only one of them is cells
The quoted price covers the modules. The installation needs interconnects, busbars, an overcurrent device, a means of isolation, monitoring, a tray that restrains the mass and, for some chemistries, ventilation. Skipping any of them does not stop the system working on the first sunny afternoon, which is what makes them easy to leave out and expensive to add later, once the bank is wedged under a bed platform and the cables are cut to length.
How parallel batteries are wired decides whether you have one bank or a strong battery and some passengers. Take the main positive from one end of the row and the main negative from the other, or run equal-length leads from every module to a common pair of busbars. Wire both mains off the same end and the nearest battery sees the lowest resistance path, does a disproportionate share of every charge and discharge, and ages out first — after which the others follow quickly because they inherit its work. Terminal torque comes from the module datasheet, and it wants re-checking a few weeks in, because a loose terminal is a hot terminal long before it is an open circuit.
Protection is where lithium changes the conversation. A low-impedance bank can deliver an enormous fault current into a bolt dropped across two terminals, so the overcurrent device has to interrupt that current at DC, and the DC interrupting rating on the datasheet is the one that matters rather than the larger AC figure on the packaging. Class T fuses to UL 248-15 are specified for this duty for that reason. The fuse belongs as close to the positive terminal as the installation allows, with a separate switch for isolation, because a fuse is not a disconnecting means. In North America the requirements sit in the National Electrical Code, Articles 480 and 706 with Article 710 covering stand-alone systems; afloat it is ABYC E-10 and E-13; in a motor caravan the twelve-volt installation is covered by EN 1648-2.
Mass and gas finish the assembly. A lithium module is heavy enough to leave its shelf in a collision or a beam sea, so the restraint is structural rather than a strap that looks reassuring. Flooded and valve-regulated lead-acid cells evolve hydrogen on charge and need an enclosure ventilated on the basis set out in IEC 62485-2, which is a calculation and not a vent hole chosen by eye. Fixed installations at a dwelling also meet the siting and quantity rules of NFPA 855 where it is adopted, and the modules themselves are usually listed to UL 1973 with complete systems to UL 9540.
Everything between the cells and the load
- Fuse and isolator — sized on the fault current the bank can deliver rather than the load it normally carries, and chosen on its DC interrupting rating rather than the AC figure printed larger
- Busbars and interconnects — the shared pair of bars every module lands on, sized by cross-section and current density so that no single link becomes the hot spot of the bank Busbar Ampacity Calculator (Current Density Method)
- Battery modules — the only part sized in kilowatt-hours, and the only one whose nameplate has to be discounted twice before it means anything at the socket Battery Bank Sizing Calculator
- Monitoring shunt — counts amp-hours in and out of the negative return, which is the only honest measurement of what a night costs and the only way to catch a pump cycling at three in the morning Battery Backup Runtime Calculator
- Tray, restraint and ventilation — carries a mass that must not move in a collision or a seaway, and on vented lead-acid cells it is also the enclosure the hydrogen ventilation calculation applies to
The label, the socket, and year eight
Two subtractions stand between the printed capacity and the energy that reaches an appliance, and they compound. The first is depth of discharge: lead-acid cycle life collapses if the bank is routinely taken much past half, so conventional practice holds it there and half the label is simply unavailable, while lithium iron phosphate datasheets permit most of the pack to be used. The second is round-trip efficiency, the energy lost putting a kilowatt-hour in and taking it out again, which is modest on lithium and considerably worse on flooded cells. Multiply the two and a lead-acid bank hands back a strikingly small fraction of the number on the case — which is the whole reason a lithium bank of half the rated size can do the same job.
Everything else that erodes capacity does so quietly. Cold takes a slice, and takes more of it from lead-acid than from lithium. Drawing fast takes another: a lead-acid capacity quoted at the twenty-hour rate is not what you get when the kettle is on, a behaviour described by Peukert's relation and largely absent from lithium. Age takes the rest, steadily, which is why a warranty is written to a fraction of nameplate after a stated number of cycles rather than to the nameplate itself. Read which fraction and after how many cycles, at what depth and what temperature, because those four conditions are what the guarantee actually promises and they are not the same across two products sold at the same price.
So size on usable energy, and then look at the bank you will own in year eight rather than the one arriving on the pallet. A system that exactly meets the load on commissioning day is a system that stops meeting it, and the moment it does is a winter evening. Testing on day one gives you the datum for that: charge fully, discharge at a known load to the manufacturer's cut-off, and record the amp-hours that came out. Repeat once a year. Two readings a year apart tell you more about the bank's future than any specification sheet.
| What removes it | Roughly what it costs | What decides the figure |
|---|---|---|
| Depth of discharge | Around half the nameplate on lead-acid; a small slice on lithium iron phosphate | The manufacturer's cycle-life curve — the choice is capacity now against cycles later |
| Round-trip efficiency | A few per cent on lithium, materially more on flooded cells | Datasheet charge and discharge efficiency at the current you actually use |
| Temperature | Falls away as the bank gets colder, and lead-acid falls faster | The capacity-against-temperature curve in the datasheet, read at your worst month |
| Discharge rate | A fast draw returns less than the twenty-hour rating promises | Peukert behaviour on lead-acid; largely a non-issue on lithium |
| Ageing | A steady loss to a warranty end-of-life figure below nameplate | Cycles, depth and temperature over the term, as written into the warranty |
| Inverter idle and conversion | A continuous background draw plus a cut on everything converted to AC | Inverter datasheet no-load current and efficiency curve at your typical load |
How many grey days are you buying
Days of autonomy is the most expensive input anyone types, because it multiplies the largest line on the quote. One day is a bank that gets you to tomorrow's sun. Three is a bank that shrugs off a wet weekend. Five is a bank that costs more than the rest of the system and spends most of its life half-empty of purpose, waiting for a week that comes round every few years.
Whatever number you choose has to be set against the site's worst month rather than its average. Solar resource at high latitude can differ between December and June by a factor that makes an annual mean actively misleading, and an off-grid system has nothing to fall back on when the mean turns out to have been generous. Take monthly irradiance for the actual coordinates from a named dataset — PVGIS from the European Commission's Joint Research Centre, or NREL's PVWatts — and read the worst month, at the tilt you will actually build. A steeper winter tilt trades away summer output you were going to spill anyway and recovers some of what December takes, which is usually the better bargain on a system with no export to sell.
Then check the other half, which is the half people forget: the array has to refill the bank as well as carry the day. A system where generation equals consumption on an average day never climbs back out of a deficit, because after three cloudy days the bank is deeply discharged and parity only holds it there. Recovery needs genuine surplus. Charge acceptance limits how fast that surplus can go in, too — a lead-acid bank tapers through absorption and the last stretch of the charge takes as long as the first two thirds, so a sunny afternoon that looks ample on paper can leave the bank at ninety per cent night after night, which is its own slow way of killing it.
Run the arithmetic with the load you measured, the usable share your chemistry actually delivers, and the worst-month sun hours for your coordinates. The hours-of-autonomy result is the headline, but the line worth reading is the daily surplus or shortfall: it is what tells you whether the bank ever gets back to full, and it is the difference between a system that rides out a bad week and one that walks itself down over a fortnight.
Set the measured load against the bank and the array together, at the worst month's sun hours rather than the annual average, and watch the surplus line as much as the hours.
Rated capacity of the bank.
Depth of discharge times round-trip efficiency.
Average continuous draw.
Nameplate DC of the array.
Daily solar resource at your site.
Autonomy
25.5 hours
Generation exceeds consumption on an average day at the stated sun hours.
- Usable energy
- 12.75 kWh
- Days at this load
- 1.06 days
- Daily consumption
- 12 kWh
- Daily generation after losses
- 12.04 kWh
- Daily surplus or shortfall
- 0.04 kWh
They open the calculator with your figures already in it
Off-Grid Autonomy Calculator: 25.5 hours — shown in imperial, US market. The link sets both, so the result they see is the one on your screen.
What this calculation does not cover
- Peak sun hours vary by season by a factor of two or more. An off-grid system must be sized on the worst month, not the annual average.
- Assumes the load is constant. Real loads cycle, and a system that works on averages can still fail on a bad evening.
- Ignores charge-controller limits and the fact that a full battery cannot absorb surplus generation.
Energy is not power
A kilowatt-hour figure says nothing about what the bank can deliver in the next half second, and that is where the pump lives. A motor drawing its locked-rotor current on start-up asks for several times its running figure, briefly. The inverter has to carry that within its surge rating and for at least as long as the motor needs, and a soft-start module on a compressor or an air conditioner is very often cheaper than the inverter that would otherwise be required. Check the surge duration and not just the surge watts, because those two numbers are quoted together and only one of them is usually read.
On a lithium bank the real ceiling is frequently the battery management system rather than the cells. A drop-in module carries a continuous discharge limit in amps, and at twelve volts that limit turns into a surprisingly small number of watts — enough that two heavy appliances at once can trip the pack out even though its energy capacity is generous. When the BMS opens under load it does so instantly, and an inverter that loses its DC supply mid-cycle takes the whole cabin dark. Add the module limits up, confirm the parallel arrangement is actually sharing, and size the bank on that continuous rating as well as on kilowatt-hours.
System voltage is the lever that fixes most of this. The same power at twenty-four volts halves the current and at forty-eight volts quarters it, which takes cable cross-section, terminal heating and voltage sag down with it. A small van under a couple of kilowatts is comfortable at twelve volts and keeps the vehicle's own DC world simple; a cabin with a well pump and a workshop is far better served at forty-eight, where a two-kilowatt load is a modest current and the cable to the inverter stops being a structural component. Decide it before anything is bought, because it is the one choice that cannot be revised without replacing the inverter, the controller and most of the copper.
Cold
Most lithium iron phosphate datasheets forbid charging below freezing while permitting discharge well below it, and a management system enforcing that limit is the source of an entire genre of winter fault report: bright morning, healthy array, and no charge current at all. Nothing has failed. The pack is protecting itself from plating, and it will accept charge again when it warms. Heated modules solve it by drawing power from the bank to warm themselves, which is real energy that belongs in the winter load figure rather than being discovered afterwards.
Lead-acid does not lock out but loses capacity as it cools and needs its charge voltage compensated for temperature, using the coefficient the manufacturer publishes and a sensor at the battery rather than at the controller. Either chemistry argues for the same decision on site: put the bank somewhere thermally inside the building or the vehicle, not in the underfloor locker or the unheated crawlspace where it will spend the whole of the season it was bought for at its least capable temperature.
The fourth day is cheaper as fuel
Every extra day of autonomy adds a linear slab of the most expensive item in the system, and buys capacity that goes unused for most of the year. Past roughly three days the trade usually turns: a small generator, an engine-driven charger, or a DC-DC converter from a vehicle alternator covers the handful of genuinely dark stretches for a fraction of what the equivalent battery would cost, and it also covers the failure the battery cannot — a controller that dies in February. The bank then gets sized for the ordinary bad week rather than the exceptional one.
What decides the generator hours is the charger, not the generator. A set with plenty of output feeding a modest charger returns energy only as fast as that charger will push it, and on lead-acid the absorption taper means the last part of the charge is slow no matter what is connected. Plan on running to the point where charge current has fallen off rather than to full, let solar do the finishing, and give the engine a loaded run rather than an idle one — light-loaded running is what glazes a diesel set. Sizing the fuel that has to be on site for a dark week is arithmetic on the burn rate at the load you will actually apply, which is not the figure at the top of the spec sheet.
The safety side is not negotiable and is where off-grid installations get hurt. Exhaust goes outdoors, never under an awning, into a partly open garage or near an opening window, and a carbon monoxide alarm is mandatory equipment rather than a nicety — NFPA 1192 covers the recreational-vehicle case and NFPA 37 governs the installation of stationary engines. Fuel is stored in approved containers away from the living space, and refuelling happens with the engine stopped and cool.
Work out how many hours a tank actually gives at the load you will run the set at, then decide how much fuel a dark week needs standing on site before winter starts.
The generator's fuel tank size.
Check the manufacturer's fuel consumption chart at your expected load percentage.
Estimated runtime
6.67 hours
Actual runtime varies with load percentage, fuel type, temperature, and altitude — check your specific generator's published fuel consumption chart at your actual expected load for a more precise figure.
They open the calculator with your figures already in it
Generator Fuel Runtime Calculator: 6.67 hours — shown in imperial, US market. The link sets both, so the result they see is the one on your screen.
What this calculation does not cover
- Treats the burn rate as constant for the whole run. Real load moves as pumps, fridges and heaters cycle, so this holds only if the rate you entered is the average across the run rather than the figure at one load point.
- Assumes every drop in the tank reaches the engine. Fuel pickups sit above the tank floor, low-fuel cut-out and surging arrive before a tank is dry, and a propane or LPG vessel is filled to a working fraction of its nameplate size — usable fuel is always less than stated capacity.
- No derate for site or engine condition. Altitude, intake air temperature, fuel type and a worn or poorly serviced engine all move consumption away from the manufacturer's chart, which is measured on a new set under test conditions.
- This is not a sizing or load calculation. It does not check that the generator can carry the load you intend to put on it, and it cannot tell whether the burn rate you entered matches that load — read the rate off the consumption chart at the load percentage you will actually run.
- Ignores the stops a long run needs. Oil-level checks and oil-change intervals on portable sets arrive well before a large tank empties, and refuelling means shutting the engine down and letting it cool. The hours here are fuel-limited, not the hours the machine will run unattended.
Write it on the lid
The single most useful thing to leave behind is a label on the enclosure carrying the numbers this exercise produced: measured daily watt-hours and the month they were measured in, the usable share assumed for the chemistry, the days of autonomy the bank was sized to, the fuse type and its DC rating, the charge voltages and current limits taken from the module datasheet, and the day-one capacity test result with its date. Nobody remembers any of it two years later, and the next person to touch the system is otherwise starting from the labels on the backs of appliances again.
One last discipline, because it undoes more banks than cold does. Do not add new modules to an aged bank. The old cells set the pace, the new ones are dragged into their pattern, and you have paid for capacity that the weakest string will not let you use. When the bank is due, replace it as a set — and take the last measured load figure into that purchase, because by then the fridge, the pump and everything else will have told you far more accurately than the first calculation ever could what a night in that building costs.
Before the battery order goes in
Six numbers decide this purchase. Get them onto one page before anyone is asked for a quote, because the last of them fixes the cable, the fusing and the inverter, and cannot be revised once the copper is cut.
- Measured daily amp-hours, in the worst month — From a shunt over seventy-two hours of normal use, converted at nominal voltage. Not a total of nameplate ratings.
- Average overnight watts, separately — What has to survive from last light to first charge. It sizes the night; the daily figure sizes the bank.
- Worst-month peak sun hours for the coordinates — From PVGIS or PVWatts at the tilt you will actually build. The annual average is the number that leaves you dark in December.
- Chemistry and the usable share you are assuming — Depth of discharge multiplied by round-trip efficiency, both from the datasheet for the model being bought.
- The largest starting load and its inrush — Pump or compressor locked-rotor current, against the inverter's surge rating and duration and the BMS continuous limit.
- System voltage, decided and written down — Twelve, twenty-four or forty-eight. It sets every cable, the fusing and the inverter, and it cannot be revised cheaply.
Opens the calculators above on one screen with the dimensions from this article already filled in. Quantities only — this site publishes no price list, because local prices vary too much to publish honestly.