In short. Not every load deserves the same protection. Classifying loads by consequence of interruption and tolerable outage time decides which loads get UPS, which go on generator backup, which can be shed, and which can wait. Do the classification first; the architecture follows from it.
Why classify at all
Protecting every load to the highest standard is expensive and often unnecessary. Protecting too little risks safety, data or production. A classification turns a long load list into three or four tiers, each with a clear power-continuity requirement.
A practical three-tier model
| Class | Consequence of loss | Typical tolerable gap | Typical support |
|---|---|---|---|
| Mission-critical | Safety risk, data loss, process damage | None to a few cycles | UPS or battery, redundant paths, generator behind |
| Essential | Serious disruption or loss of service | Seconds to minutes | Standby generator with automatic transfer |
| Non-essential | Inconvenience or deferred production | Hours, or none needed | Shed during outage, restore later |
The terms vary between organisations and codes. What matters is that your site defines its own tiers and the gap each can tolerate.
How to classify a load
- State what happens if the load stops for one second, one minute and one hour.
- Note whether it can restart on its own or needs an operator or a staged sequence.
- Record whether regulation or a licence requires it to stay powered.
- Assign a class and record who approved it.
Examples to illustrate the thinking
- Servers, control systems and life-support equipment usually need uninterrupted supply.
- Emergency lighting, fire pumps, lifts and ventilation usually need generator backup within seconds.
- Process cooling or cold storage may tolerate minutes but not hours.
- Comfort cooling or general lighting is often shed during an outage.
These are illustrations. Your site’s tolerance may differ.
From class to design
- The total of mission-critical and essential loads sets the generator capacity. See capacity requirement.
- Mission-critical loads set the UPS rating and battery autonomy. See UPS autonomy.
- Non-essential loads become the shedding list. See load shedding.
- The classes decide which distribution sections need dual feeds and which can be single.
Worked illustration
A facility has 800 kW connected load. Classification shows 120 kW mission-critical, 280 kW essential and 400 kW non-essential. The generator is sized for 400 kW plus allowances for starting and growth, not 800 kW. The UPS covers 120 kW only. The non-essential 400 kW is shed on outage. This can reduce capital cost materially, but only if the classification was agreed, documented and tested.
Three classes of load, three expectations
| Class | Meaning | Typical backup | Typical interruption tolerated |
|---|---|---|---|
| Essential | Needed soon after an outage for safety or operation | Generator with ATS | 10 to 30 seconds |
| Critical | Interruption causes serious cost or risk | UPS plus generator | Zero to milliseconds |
| Mission-critical | Interruption is unacceptable at any time | 2N or distributed redundant, fully tested | Zero, including during maintenance |
Assigning classes
- List the loads and ask the operator to state, in time and money, what an interruption costs.
- Set the class by consequence, not by habit.
- Check whether the class needs different power for different seconds: a load may need UPS for the first minutes and generator beyond.
Examples (illustrative)
- Emergency lighting and fire pumps: essential, with legal start times set by applicable codes.
- Server rooms, control rooms and monitoring: critical.
- Operating theatres, trading, process control with safety consequence: mission-critical.
Cost scaling
Each step up in class adds equipment, space and maintenance. Designers often over-class everything; separating loads into classes can cut capital sharply because the generator and UPS carry only what needs them.
See critical-load classification and no-break, short-break and long-break.
Common mistakes
- Classifying by equipment type instead of consequence.
- Putting everything in the highest class “to be safe”.
- Forgetting auxiliaries such as cooling and controls that a critical load depends on.
- Letting the classification go stale as the site changes.
Documenting the classification
Keep a one-page register per load group: load name, class, tolerable gap, restart behaviour, dependencies and approver. The register becomes the input to generator sizing, UPS sizing, shedding logic and test plans, so it is worth keeping accurate. Review it after any plant change.
Dependencies that raise a load’s class
A load takes the class of the most critical thing that depends on it. Chilled-water pumps serving a data hall, ventilation for a hazardous area, or the control power for a critical process inherit the class of what they support. Trace dependencies for each critical load.
What to send us
Send your load list with the consequence and tolerable gap for each load. Send your application details. See also critical-load classification.
Questions engineers ask
Is “essential” the same as “critical”?
Not always. The words are used differently between standards and organisations. Define the tiers for your site by consequence and tolerable gap rather than relying on the label.
Who should approve the classification?
The operations or process owner, with safety and engineering input. It is a business and safety decision as much as a technical one.
Do essential loads need UPS?
Only if they cannot tolerate the generator start and transfer gap. Many essential loads accept a short break and do not need UPS.
How often should classification be reviewed?
Whenever loads or operations change, and as part of regular design reviews.
This guide gives general principles. Applicable codes and regulations for your facility type must be checked separately.