Backup System Annual Load Test
Purpose
A residential or small commercial backup system does not get a portable load bank, and it does not need one. The building is the load bank. What it needs is one scheduled outage a year in which the real house runs on the real source for long enough to prove the parts that a weekly no-load exercise never touches: the battery under crank, the fuel path under demand, the transfer back, and whether the load the system is being asked to carry has quietly grown since it was installed.
The value of the visit is the comparison, not the verdict. A set that passes at 45 percent of its rating one year and 54 percent the next is telling you something a pass or fail cannot, and it is telling you while there is still time to do something about it.
Scope
Covers the annual functional and loaded test of an installed residential or small commercial backup system under NEC Article 702, engine generator or battery energy storage, tested with the building's own connected load.
Does not cover the supplemental resistive load bank test in NFPA 110 Chapter 8, which applies to emergency and legally required standby systems rather than an optional standby set and which the generator load bank test SOP owns; commissioning of a new system, owned by the standby generator installation and automatic transfer switch commissioning SOPs; or engine overhaul and fuel polishing.
Roles and responsibilities
| Role | Owns | Hands off |
|---|---|---|
| Office | The window, the prior year's report and the parts likely needed | Gives the lead last year's report before mobilization, not after |
| Lead electrician | Every reading, the abort decision and the trend comparison | Hands this year's report with the trend lines to the office the same day |
| Customer | Confirming nothing critical rides through the window and naming who can abort | Signs the report including a failed or aborted result |
Procedure
Confirm the window, the abort authority and what cannot be interrupted. Establish that no medical device, no critical refrigeration and no work-from-home commitment rides on the outage window, and get the name of the person on site who can call it off. Acceptance: window confirmed in writing, named abort authority present or reachable, and a walk of the building noting anything on an uninterruptible supply. Wrong looks like a test starting because a scheduling email went unanswered; stop rule, no named authority means no test. Hazard: none, and it is the step that stops a test from becoming an incident report.
Read the record before touching hardware. Pull last year's report and read the controller's own event and run-hour log, which has been recording every weekly exercise since the last visit. Acceptance: run hours in, exercise count since last visit, every logged fault or alarm listed with its date, and last year's leg currents copied onto this year's sheet before any new reading is taken. Wrong looks like a technician who takes today's readings with nothing to compare them against; stop rule, a controller log showing repeated faults that were never reported goes to the office before the test proceeds, because a pattern is a different job from a service call. Hazard: none if the controller reads from the front, and it should.
Test the starting system as a system, under load, not by watching it start. For an engine set, measure open-circuit battery voltage, then measure voltage during an actual crank and compare it to the manual's minimum cranking voltage; check charger output and terminal condition. For a storage system, record reported usable energy against nameplate, cycle count and any module fault. Acceptance: open-circuit voltage, cranking voltage against the manual's stated minimum, and charging voltage all recorded; or for storage, usable energy as a percentage of nameplate. Wrong looks like a battery called good because the set started on a mild day, when starting current at freezing is substantially higher; stop rule, cranking voltage below the manual's minimum means the battery is replaced before the loaded test runs, not after. Hazard: a battery is a stored-energy source that vents hydrogen, so no metal tool bridges the terminals, rings and watches come off, eye protection goes on, and the set's controller is confirmed in OFF so nothing cranks with hands at the terminals.
Walk the fuel path, the exhaust and the clearances with disturbance in mind. For an engine set, check the fuel line, the regulator and the exhaust for holes, loose clamps and soot streaks, and confirm the discharge is still clear of doors, windows and intakes. For a storage system, confirm the manufacturer's ventilation and clearance dimensions are still met. Acceptance: exhaust intact, listed clearances re-measured and recorded against last year's figures, ventilation openings unobstructed. Wrong looks like a shed, a fence or a deck extension built up against the enclosure since the last visit; stop rule, a leaking exhaust inside or adjacent to occupied space means the set does not run today, the room is left, and the repair is quoted. Hazard: brushing out an enclosure or clearing leaf litter lifts dust and rodent debris into the air you are breathing, so that clearing is done with the set off, from upwind, wetted or vacuumed rather than blown, and a personal carbon monoxide monitor is worn for the whole test once the engine runs.
Run the loaded outage on the building's own load for a stated period, logging at fixed intervals. Open the utility means, let the system transfer on its own, and log every 10 minutes rather than when something looks interesting. Acceptance: current on each leg, voltage line to line, frequency where applicable, and for storage the state of charge, all logged at 10-minute spacing for a period agreed in advance and stated on the sheet; plus the largest motor started while on backup with no alarm logged. Wrong looks like a 5-minute run called a load test; stop rule, an alarm, a dropout or a reading trending toward a shutdown set point ends the run in a controlled abort rather than waiting for the equipment to protect itself. Hazard: this step runs an engine or a live inverter with the family in the building, so readings come from outside the guarded envelope, hearing protection stays on near a running set, and nobody opens a panel or a transfer switch while the system is carrying load.
Retransfer, cool down, return to automatic, and re-prove the protective functions. Restore the utility, let the retransfer and cool-down run at their set delays, then confirm the system is back in AUTO before anything else. Acceptance: retransfer and cool-down times measured against the settings on file, controller confirmed showing AUTO and photographed, and a GFCI or AFCI device on a critical circuit tripped and reset on its own test button. Wrong looks like a hot shutdown straight off load, or a system left in MANUAL after the visit; stop rule, a system that will not return to AUTO is red-tagged and the customer is told in the same visit that they currently have no backup. Hazard: this is the step that puts fuel, energy and motion back with everyone present, so deadfronts are secured before the transfer, everyone stands clear of the set, and the protective-function check is the verification that a year of vibration and weather has not undone anything.
Write the report as a comparison and set next year's date. Put this year's numbers beside last year's, state the direction of every difference in words, and give every deficiency an owner and a date. Acceptance: a side-by-side table of the key readings, a stated trend for load and for battery or state of health, deficiencies with owners, next visit booked, customer signature. Wrong looks like a report that says passed; stop rule, a deficiency with no owner does not close. Hazard: none, and this is the section a customer reads when deciding whether the annual visit is worth booking again.
The record this produces
One annual report per system, filed and given to the customer, built to sit beside last year's.
Header: date, technician, make, model, serial, rating, run hours in and out, exercise count since last visit, ambient temperature. Body: the step 3 battery or state-of-health readings; the step 4 clearance measurements; the step 5 interval log at 10-minute spacing with the motor-start observation; the step 6 retransfer and cool-down times against the settings on file. Footer: the side-by-side comparison table, the trend statement, deficiencies with owners and dates, next visit date, signature.
The next technician reads the comparison table first. A single year's readings say whether the system works; two years say whether it is going to.
Worked pass: 22 kW set, third annual visit
Step 2: run hours in at 61.4, exercise count 52 since the last visit, controller log showing two low-battery warnings in December that were never reported. Last year's leg currents copied onto the sheet: 38 A on L1, 44 A on L2.
Step 3 fails. Open-circuit battery voltage reads 12.4 V. During an actual crank, voltage drops to 8.9 V against the manual's stated minimum cranking voltage of 9.6 V, so it is 0.7 V under. The set started anyway, on a mild afternoon, which is exactly the reading that gets a marginal battery signed off. The stop rule fires: the battery is replaced before the loaded test, and the December warnings in the log are now explained rather than filed. The replacement battery cranks at 10.8 V, clearing the 9.6 V minimum by 1.2 V, and charging voltage reads inside the manufacturer's band.
Step 4: exhaust intact. Clearance to the nearest openable window measures 7 ft 2 in against 7 ft 4 in last year, because a new hose reel was mounted on the wall; still above the manufacturer's 5 ft listed distance and recorded as moved.
Step 5: the loaded run goes 60 minutes at 10-minute intervals. Leg currents settle at 46 A on L1 and 52 A on L2. That is (46 plus 52) times 120, or 11,760 W, against the set's 22 kW rating, which is 53.5 percent. Last year the same reading was (38 plus 44) times 120, or 9,840 W, which was 45 percent. The largest compressor started with no alarm logged.
Step 6: retransfer at 10 minutes 3 seconds against the 10 minutes on file, cool-down 5 minutes 1 second, AUTO photographed. A GFCI on a critical circuit tripped and reset on its test button.
Step 7: the report states the load on the set rose from 9,840 W to 11,760 W, a growth of 1,920 W or 19.5 percent in one year, and that the remaining headroom of 22,000 minus 11,760, or 10,240 W, would be consumed in roughly five years if that rate held, which it may not. The customer is told that before they add the workshop circuit they mentioned, not after. Deficiencies: none open, since the battery was replaced on the visit.
References
- NEC (NFPA 70) Article 702 for optional standby systems; confirm the adopted edition
- NFPA 110 Chapter 8, in the edition your authority having jurisdiction has adopted, which governs testing for emergency and legally required standby systems and is named here to mark what this procedure is not
- Manufacturer's operation manual for the set or storage system: minimum cranking voltage, charging voltage band, clearances, cool-down period, and alarm and shutdown set points
- See related: the generator load bank test SOP for systems under NFPA 110, the automatic transfer switch commissioning SOP for the settings this test measures against, and the critical load panel design SOP for the tally the load trend is compared with