When the utility or the solar input fails, the only question that matters is how long the battery will hold the site up. The basic runtime calculation is deceptively simple, usable capacity divided by the load, but the simple version tends to overpromise, because a battery does not deliver its full rated capacity when it is discharged quickly. This page shows the runtime formula and how to make the capacity figure honest, explains the Peukert effect that shrinks real runtime at higher discharge rates, and describes how a SCADA host uses live readings to predict remaining hold-up time so operators can act before a site drops offline.
Runtime Calculation in one line: Backup power runtime is estimated as the usable amp-hour capacity of the battery divided by the average load current it is supplying. Usable capacity is less than the nameplate rating because you should not fully drain the bank and because faster discharge yields less capacity, an effect described by the Peukert relationship. A SCADA host can improve on the static estimate by watching the actual load and battery state in real time to predict how much hold-up time remains before the site drops offline.
At its simplest, runtime is capacity divided by load. If a bank holds a certain number of amp-hours and the site draws a steady current, the runtime in hours is the amp-hours divided by the current in amps. The formula is exact in spirit but only as good as the two numbers you put into it, and both need care. The load has to be the true average current the site draws in backup, including everything still energized, and the capacity has to be the amount you can actually use, not the number printed on the label.
Usable capacity is smaller than nameplate for two reasons. First, you do not drain a battery to empty, both because deep discharge damages many battery chemistries and because you want a reserve margin, so the usable figure is only the portion between full and whatever depth of discharge the design allows. A bank rated at a hundred amp-hours that you only draw down partway to protect its life offers considerably less than a hundred amp-hours of usable backup. Second, capacity itself depends on how fast you take it out, which the nameplate rating usually assumes to be a slow, gentle discharge.
Getting the load number right is just as important and often harder, because the backup load is not always the same as the normal running load. Some equipment may be shed when backup begins, some may draw more as it works harder, and the average over a long outage can differ from the instantaneous draw at any moment. A realistic runtime estimate uses the average current the site will actually pull for the duration of the outage, not a snapshot, and it errs toward the higher end of expected draw so the estimate is conservative rather than optimistic.
The Peukert effect is the reason a battery delivers less than its rated capacity when it is discharged quickly. Battery capacity is rated at a specified discharge rate, typically a slow one spread over many hours. Pull the current out faster and the chemistry cannot keep up as efficiently, so the effective capacity drops and the battery reaches its cutoff sooner than the simple division would predict. The higher the discharge current relative to the rating, the more pronounced the shortfall, so a bank that comfortably lasts a long time at a light load can fall well short of the naive estimate under a heavy one.
The practical consequence is that you cannot take the nameplate amp-hours, divide by a large load, and trust the answer. At high discharge rates the real usable capacity is meaningfully less than the rating, so the runtime is less than the arithmetic suggests, sometimes by a wide margin. This matters most for sites that draw heavily in backup, because they are exactly the sites where the naive calculation is most optimistic and where an operator counting on that number could be caught out when the battery gives up early. The effect is stronger for some chemistries than others, so the size of the correction depends on the battery type as well as the discharge rate.
For a lightly loaded telemetry node the Peukert effect is small, because the discharge is gentle and close to the rating conditions, so the simple formula is fairly trustworthy. For a site running a substantial backup load the effect is large and cannot be ignored. The honest approach is to derate the usable capacity for the actual discharge rate the site will impose, rather than assuming the slow-rate rating holds. This is why two sites with identical batteries can have very different real runtimes: the one drawing harder gets less out of the same bank than the arithmetic promises.
A static runtime calculation done at design time is a planning estimate, but during an actual outage what operators need is a live prediction of how much time is left, and this is where a SCADA host earns its keep. Rather than relying on the paper figure, the host watches the real battery voltage, the real load current, and how fast the state is changing, and uses those to project when the bank will reach its cutoff. Because it is working from what is actually happening rather than from assumptions, the live prediction accounts for the true load and the battery's actual condition, including the Peukert shortfall showing up in the observed discharge rate.
The value of a live hold-up prediction is that it gives operators time to act while acting is still possible. A number that says the site has several hours left is a very different situation from one that says it has minutes, and knowing which one you are in decides whether to dispatch a generator, shed non-critical load to extend the runtime, or accept a controlled shutdown before an uncontrolled one. A prediction that updates as the load and battery change lets the operator watch the margin erode and intervene at the right moment rather than discovering the outcome when the site goes dark.
When this prediction lives in a cloud platform such as Merobix, it also becomes a fleet-wide view rather than a single gauge on a local panel. During a widespread outage the platform can show which sites have comfortable hold-up time and which are close to their cutoff, so limited field resources go first to the sites about to drop. Over time, comparing the predicted hold-up against what a bank actually delivered when it was tested also exposes batteries that are aging and no longer holding their rated backup, turning the runtime prediction from a one-time design number into an ongoing measure of whether each site can still ride through the outage it was built to survive.
Runtime in hours is the usable amp-hour capacity divided by the average load current in amps. The catch is that usable capacity is less than the nameplate rating, because you should not fully drain the bank and because faster discharge yields less capacity. So the formula is only as honest as the two numbers you feed it: a realistic usable capacity and the true average current the site draws during the outage.
The Peukert effect means a battery delivers less capacity when it is discharged quickly than when it is discharged slowly. Capacity is rated at a slow discharge rate, so pulling current out faster makes the effective capacity drop and the battery reach its cutoff sooner than the simple division predicts. The effect is small for a lightly loaded node close to the rating conditions but large for a site drawing heavily, where the naive runtime estimate can be far too optimistic.
A SCADA host watches the live battery voltage, the actual load current, and how fast the state is changing, then projects when the bank will reach its cutoff. Because it works from what is really happening rather than a paper figure, the prediction captures the true load and the battery's real condition, including the Peukert shortfall visible in the observed discharge rate. That live hold-up estimate lets operators dispatch a generator, shed load, or shut down cleanly before the site drops offline.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.