In short. Critical power is the discipline of keeping essential loads running through utility failure, equipment failure and maintenance. It is a chain of sources, conversion equipment, distribution and controls, and the chain is only as strong as its weakest link. Resilience is designed into the architecture; it cannot be added later by buying a bigger generator.
What “critical power” actually covers
A generator alone is a source. A critical-power system adds everything needed for the load to see an uninterrupted, acceptable supply: transfer equipment, UPS or battery systems where no break can be tolerated, distribution that avoids shared failure points, controls that sequence the response, and a test and maintenance regime that proves it works.
The continuity chain
| Link | Typical equipment | Question to ask |
|---|---|---|
| Sources | Utility feeds, diesel or gas generators, BESS | What fails together? |
| Transfer | ATS, STS, bus-tie breakers | How long is the interruption? |
| Conditioning | UPS, stabilisers, filters | Which loads cannot tolerate any break or disturbance? |
| Distribution | Switchboards, busbars, cables, PDUs | Is there more than one path to the load? |
| Control and protection | PLC, relays, power management | Does it fail safe and stay selective? |
| Operation | Testing, maintenance, procedures | Has the whole chain been proven together? |
Four design questions that come first
- What must stay on? Classify loads before selecting equipment. See critical, essential and mission-critical power.
- How long can each load be without power? Zero, seconds, minutes or hours decide whether you need UPS, fast-start generators or manual recovery.
- What failure must the site survive? A single equipment failure, a failure during maintenance, a whole room, or a regional grid outage.
- How will you prove it? Acceptance and periodic testing are part of the design.
Redundancy is not the same as resilience
Redundancy adds spare equipment. Resilience is the system’s ability to continue and recover. Two generators on one fuel line, one control panel or one breaker share a failure mode, so they are not truly redundant. See single-point-of-failure elimination and common-cause failure.
Architecture families
- N: capacity for the load, no spare.
- N+1: one spare unit; survives one failure or one maintenance outage.
- 2N: two independent systems, each able to carry the full load.
- Distributed redundant: shared spare capacity across several systems.
Details are in N, N+1, N+2 and 2N architectures.
Where the gaps usually are
- A generator that is mechanically sound but not ready to start, because of batteries, fuel quality or a disabled controller.
- Source redundancy with no distribution redundancy behind it.
- Loads restarting all at once after an outage, tripping the generator again.
- Protection that clears a fault by dropping the whole bus.
- No integrated test of the full transfer and recovery sequence.
A short worked illustration
A site has a 500 kVA critical load and two 500 kVA generators. That reads as N+1. If both generators draw fuel from one day tank through one pump and one valve, a single fuel-path failure removes both. The real architecture is N for fuel. Adding a second fuel path, or a manual crossover, restores the intended resilience without changing the generators.
The continuity chain
Critical power is a chain: source, conversion, storage, switching, distribution and the load. Availability is set by the weakest link, not by the most expensive one.
| Link | Typical element | Typical failure |
|---|---|---|
| Source | Utility, generator | Outage, fail to start |
| Conversion | UPS rectifier and inverter | Fault, overload |
| Storage | Battery | Aged cells, low capacity |
| Switching | ATS, STS, breakers | Mechanical or control fault |
| Distribution | Switchboards, PDUs, cables | Overload, fault, human error |
| Support | Cooling, controls, fuel | Dependence on a single path |
Resilience is more than redundancy
- Redundancy: spare capacity and paths.
- Fault tolerance: a fault stays contained.
- Maintainability: equipment can be serviced without outage.
- Recoverability: restoration is fast and rehearsed.
- Observability: problems are visible before they become outages.
The failure question to ask of every component
Ask what happens to the load if this component fails now, and whether anyone would know. If the answer is an outage or silence, there is a gap.
Cost versus consequence
Match architecture to the cost of an hour of outage. A clinic and a trading floor deserve different designs. Over-building wastes capital, under-building shifts the cost to the day it fails. See the selection guide and source hierarchy.
Common mistakes
- Sizing sources first and classifying loads later.
- Counting spare units without checking shared dependencies.
- Accepting redundancy on paper without an integrated test.
Acceptance and proof
A critical-power design is accepted on evidence. Typical evidence includes factory and site tests of each component, a test of each transfer path, and an integrated test that removes the utility, starts the generator, transfers load and recovers, while recording the time and the voltage seen by the loads. Failure injection, such as a failed start or a tripped breaker, shows whether the redundancy works in practice. See integrated systems testing.
Lifecycle thinking
Resilience decays when batteries age, settings drift, fuel degrades or loads grow. Plan for maintenance windows that do not remove protection, spare parts, firmware control, and a design review whenever the load or supply changes. See maintenance and lifecycle resilience.
What to send us
Send your single-line diagram, load list with outage tolerance for each load, and the failure scenarios you must survive. Send your application details. See also source hierarchy and continuity chain.
Questions engineers ask
What is the difference between critical power and backup power?
Backup power usually means a generator that starts after an outage. Critical power includes the whole chain, including UPS, transfer, distribution, controls and testing, designed to meet a stated outage tolerance for specific loads.
Is N+1 enough?
It depends on what must be survived. N+1 handles one failure or one maintenance outage, but not a shared failure or a failure during maintenance. Sites that cannot accept that risk consider 2N or concurrent maintainability.
Do I need a UPS if I have a generator?
A generator takes seconds to start and take load. Loads that cannot tolerate that gap need UPS or battery support to bridge it.
How often should the whole system be tested?
Components are tested regularly, and the integrated sequence should be tested at acceptance and on a planned schedule agreed with operations, because that is what exposes interaction faults.
This guide gives general engineering principles. Site-specific design needs a review of your actual loads, supplies and risk tolerance.