Backup power: deciding what genuinely has to stay running

Start from what a disorderly stop costs

Losing supply mid-cycle is not uniformly expensive. A machining centre stops and restarts. A furnace charge, a batch part way through a reaction, material setting in a die or a mould, a coating in progress, a fermentation, a refrigerated store of finished product and a controlled environment losing pressure all have costs that are specific and sometimes very large. Work through the plant asking what is spoiled, what is damaged and what recovery involves. That analysis produces the load list, the required duration and the tolerable transfer behaviour, in that order, and without it any sizing exercise is guesswork.

Riding through and restarting are different problems

Some loads must not experience any interruption at all: controllers holding process state, safety-related systems, servers, instrumentation. Others merely need supply restored within a reasonable time so the plant can be brought back in an orderly way. Others again need only enough support to reach a safe stopping point rather than to continue producing. Those three needs are met by different equipment and different sizing, and conflating them produces a system that is oversized for the plant and still fails to protect the loads that mattered. Classify every load into one of the three before specifying anything.

Life-safety provision is not a commercial decision

Emergency lighting, alarm and detection systems, means of escape provision, and any equipment a fire strategy relies on are subject to statutory requirements wherever the site operates, with prescribed duration, testing and record-keeping. Those are obligations, not options to be traded against production priorities, and they are typically kept separate from any production standby arrangement. Where a process presents its own hazards on loss of power — controlled ventilation, containment, cooling of a reaction — the safe-state requirement must be determined by the engineers who assessed that hazard, and it takes precedence over output considerations.

Untested standby is a story, not a capability

Standby equipment fails when called because it has been sitting idle. Flat or degraded batteries, stale fuel, a control system that was left in the wrong position after maintenance, a transfer arrangement that has never operated under real load, and a generator that starts happily with nothing connected but stalls with the plant on it are all routine findings. A testing regime that runs the equipment under genuine load, at intervals, with results recorded and defects corrected, is what turns installed hardware into available capability. Budget for the fuel, the disruption and the engineering time this consumes.

Alternatives to owning generation

Owning standby plant is not the only answer and sometimes not the best. Options worth comparing include a second supply from a different part of the network where geography allows, arrangements for rapid hire of mobile generation, moving critical computing off site, redesigning the process so an interruption is survivable, or simply accepting the risk where the exposure is genuinely small. The comparison should include capital, the maintenance and testing burden, fuel storage obligations, space, noise and planning constraints, set against a realistic view of how often supply is actually lost at that location.

Frequently asked questions

How long should backup power last?

Duration follows purpose. Supporting a controlled shutdown needs only enough time to bring the process to a safe state and secure the plant. Continuing production through an outage needs however long outages realistically last at that location, plus a margin, and that implies fuel storage and refuelling arrangements. Life-safety systems have durations set by regulation rather than by you. Establish the requirement separately for each category, because a single duration applied across everything will be wrong in both directions.

Should we back up the whole factory or just critical equipment?

Selective provision is almost always the better economics, but it has to be designed into the electrical distribution so that supported loads can be separated from the rest. Retrofitting that separation into an existing installation is often the largest part of the cost, which is why the question is worth settling during construction. Whole-site provision makes sense mainly where the process cannot usefully run in part, or where the supply is unreliable enough that continued operation is the normal expectation.

How often should standby equipment be tested?

Frequently enough that a defect is found by you rather than by an outage, and in a manner that reflects real conditions. Running a set off load proves it starts; it does not prove the transfer arrangement works or that the machine will hold the plant. Periodic testing under genuine or simulated load, with recorded results and corrective action tracked, is what gives confidence. Requirements for life-safety installations are set by regulation, so check what applies at your site and follow that as a minimum.