In short. Start from what the site must tolerate, not from equipment. Outage tolerance decides whether UPS is needed; the failure to survive decides the redundancy; the need to maintain while live decides the paths; fuel and outage duration decide the energy source. This guide maps requirements to architectures and points to detailed pages.
Step 1: Outage tolerance
| Tolerance | Implication | Go to |
|---|---|---|
| None | UPS or battery bridging, STS or dual path | UPS architecture |
| Seconds | Standby generator with ATS | standby generators |
| Minutes to hours | Manual or staged recovery, rental backup | emergency rental power |
Step 2: Failure to survive
| Must survive | Usual architecture |
|---|---|
| Utility outage only | N generator with ATS |
| One equipment failure or one maintenance | N+1 |
| Failure during maintenance | N+2 or 2N |
| Loss of a whole system, room or path | 2N, separated |
| Regional or long outage | Fuel storage, resupply, island capability |
Step 3: Maintainability
If equipment must be serviced without shutting the load, add isolation, a second path and spare capacity. See concurrent maintainability.
Step 4: Energy and duration
- Short bridging: UPS or BESS.
- Hours to days: generators with fuel storage and a resupply plan.
- Reducing fuel or emissions: hybrid with BESS, or gas. See DG, UPS and BESS hybrid.
- Island or microgrid operation. See microgrid architecture.
Step 5: Constraints
- Space and ventilation, noise and emissions limits.
- Fuel storage rules and supply.
- Utility connection rules and fault levels.
- Budget and schedule.
- Existing equipment and brownfield limits. See brownfield modernisation.
Step 6: Verify
- Check single points of failure and common-cause failures.
- Run protection and fault-level studies for every configuration.
- Plan integrated testing.
Worked example of the thinking
A manufacturing plant has a 1.2 MW process with a 30-second ride-through, a 40 kW control and IT load with no tolerance, and an expected outage of up to 24 hours. The mapping: UPS for the 40 kW, with 10 minutes of battery; N+1 generators (three of 600 kW) for the process, with fuel for 24 hours plus margin; automatic transfer and staged restart for the large motors; manual shedding of comfort loads. The architecture came from tolerance, failure and duration, in that order. A different plant with the same load but a 2N requirement would end up with a different design.
Where to go next
From requirement to architecture
| Requirement | Architecture to consider |
|---|---|
| Short outage acceptable (tens of seconds), life-safety loads | Standby generator with ATS, readiness regime |
| No interruption for IT or control loads | UPS plus generator |
| Generator failure must not cause an outage | N+1 generators |
| UPS failure or maintenance must not cause an outage | N+1 modular UPS or two UPS systems |
| Any single path failure or maintenance with zero interruption | 2N or A/B distribution |
| Single-cord devices in A/B design | Rack STS |
| Long islanded operation | Multi-DG plant with fuel autonomy, microgrid controller |
| Cost reduction and resilience together | BESS with defined reserve |
| Old facility, cannot stop | Staged brownfield modernisation with rental bridging |
Questions that settle the design
- What does an hour of outage cost?
- Which loads tolerate which interruptions?
- How long must the site run without the grid?
- Can it be maintained without outage?
- Who operates it, and how well trained?
Principle
Choose the least complex architecture that meets the stated target, and prove it by test. Redundancy needs independent failure paths.
For help, use the application engineering enquiry. See resilience architecture.
Common mistakes
- Choosing equipment before stating tolerance and failure targets.
- Selecting the highest tier everywhere.
- Forgetting the operational and maintenance plan.
What to send us
Send your outage tolerance, loads, site constraints and risk targets. Send your application details.
Questions engineers ask
Where do I start if I do not know my requirement?
Begin with a load list and ask what happens to each load if power is lost for one second, a minute and an hour.
Is there one architecture that fits all sites?
No. The right design depends on tolerance, failure targets, duration and constraints.
Do I need 2N?
Only if you must survive the loss of a whole system or keep full redundancy during maintenance.
How do I check the final design?
By single-point-of-failure review, studies for protection and capacity, and integrated testing.
This guide gives general principles. Final design needs a site-specific engineering review.