What buyers should know first: Critical facilities often have more connected load than the generator should accept at one time. Load shedding and staged starting help protect essential circuits and avoid unnecessary oversizing.

In practice

Hospitals, data centers, factories, and process sites should classify loads before requesting a generator quote. The supplier can then discuss generator capacity, ATS logic, synchronizing needs, and interface terminals with the electrical designer.

Yangzhou Tesla Power Equipment Co., Ltd. supplies diesel generator sets for standby and prime projects, including silent or soundproof packages, ATS options, 50Hz and 60Hz requirements, and engine-brand categories such as Cummins, Perkins, Volvo, MTU, and Mitsubishi. For a matched quotation, the purchaser should describe the site conditions instead of sending only a kW or kVA number.

How NFPA 110 classifies the load you are protecting

Before a load list means anything, someone has to say which standard the facility is designed to. In North American practice that is usually NFPA 110, Standard for Emergency and Standby Power Systems, and it sorts an emergency power supply system (EPSS) along three axes: Level, Class and Type. Getting those three wrong is the most common reason a shedding scheme is rejected at commissioning even though the generator itself is correctly sized.

Level is about consequence. NFPA 110 clause 4.4.1 defines Level 1 as a system whose failure would result in loss of human life or serious injuries; clause 4.4.2 defines Level 2 as carrying loads less critical to human life and safety. Life safety illumination, fire pumps and public safety communication are the examples given for Level 1; heating, refrigeration, sewage disposal and some industrial processes are given for Level 2. The standard does not itself decide which category a facility falls into - that is left to the authority having jurisdiction.

Type is the time allowed to restore power, and it drives shedding more than any other figure. NFPA 110 Table 4.1(a) lists Type U (basically uninterruptible, provided by a UPS), Type 10 at 10 seconds, Type 60 at 60 seconds, Type 120 at 120 seconds, and Type M for manual operation with no time limit.

Class is endurance: the number of hours the system must run at full rated output without refuelling. Clause 4.2 lists Class 0.083 (5 minutes), Class 0.25 (15 minutes), Class 2 (2 hours), Class 6 (6 hours), Class 48 (48 hours) and Class X for any other duration the application, code or user requires.

Source: NFPA 110 Table 4.1(a) and clause 4.2, as summarised in the Kohler / CK Power whitepaper "Understanding NFPA 110". Type and Class values are quoted from the standard itself; clause numbers as cited by that document. Checked 2026-10-08.

That 10-second allowance is tighter than it looks. A transfer switch normally waits 1 to 3 seconds before it moves, purely to ride through nuisance dips; on a Type 10 system with a 3-second delay, only about 7 seconds remain to start the set, transfer and pick up load. That is precisely why large motors are pushed into a deferred tier instead of being allowed to start together - a shedding plan is, in practice, a way of buying back seconds inside the Type window.

The fuel consequence follows from Class. NFPA 110 clause 7.9 requires the main fuel tank to hold a minimum of 133% of the fuel the system needs for its Class duration, so a Class 2 installation is not a two-hour tank, it is a tank sized for at least 133% of two hours of consumption. On-site fuel is therefore part of the shedding decision rather than a separate procurement line; the same argument is developed in our notes on fuel consumption data to collect before a quotation and on bunded fuel tank checks for generator projects.

Acceptance is fixed by the standard too: clause 7.13 requires on-site installation acceptance tests, including the two-hour load test at 7.13.4.3, before the EPSS is taken as having demonstrated that it can carry all emergency loads. A priority sequence that has only ever been tested on paper has not met that bar. The current edition is listed on NFPA 110 at nfpa.org.

Which information changes the generator scope?

What ISO 8528-1 says about the rating behind the plan

Shedding only saves money if the rating it lets you buy is the right rating. ISO 8528-1 defines four of them, and the difference between them is hours per year and load factor rather than marketing language.

• ESP (Emergency Standby Power) - for the duration of a utility failure. It is capped at 200 hours per year, and the average load factor over any 24-hour period must stay at or below 70% of the standby rating. When averaging a variable load profile, any power below 30% of ESP is counted as 30%.
• PRP (Prime Rated Power) - variable load for an unlimited number of hours per year (8,760 hours less service), again with a 24-hour average no higher than 70% of the prime rating. A 10% overload capability is available in accordance with ISO 3046, AS 2789, DIN 6271 and BS 5514.
• LTP (Limited-Time Power) - constant load for up to 500 hours per year. Above 500 hours the continuous rating applies instead.
• COP (Continuous Operating Power) - constant 100% load for 8,760 hours per year less service.

The practical consequence is that a shedding plan which keeps the set below its peak coincident load does not by itself justify an ESP rating, because ESP is also limited by annual hours and by the 24-hour average. If planned testing plus real outages will exceed 200 hours a year, the rating basis has to move to prime and the whole sizing conversation changes. Definitions checked against Cummins Power Topic #6240, "Understanding ISO 8528-1 Generator Set Ratings"; the standard itself is catalogued as ISO 8528-1:2018 at iso.org.

Whichever rating applies, name it explicitly in the enquiry. A quotation that only says "500kW" is not comparable between suppliers, because the same number at ESP and at PRP means a different engine, a different cooling package and a different price. Derating for site conditions is a separate step again - see altitude and temperature derating before the RFQ - and it should be confirmed in the factory acceptance test records before shipment rather than at handover. Tesla Power quotes against a named ISO 8528-1 rating for exactly this reason.

How to prepare the request

• Prepare a load list with running kW, starting current, and priority level.
• Separate emergency, essential, and optional circuits before sizing the generator.
• Confirm whether load shedding is controlled by ATS, PLC, BMS, or manual procedure.
• Ask the supplier which signals are included in the generator panel scope.
• Include staged-start and load-shed logic in the site acceptance test.

A worked shedding calculation you can check

Here is a three-tier scheme whose numbers a reader can re-add. The site has 700kW of connected load: 300kW of essential circuits, 200kW that can wait for voltage and frequency to settle, and 200kW that should never run on generator power at all.

Ratings used here (200, 300, 500 and 700kW) are taken from the approved power segments for this site, so they are comparable with the published set ratings rather than invented for the example.

Add the tiers as they would be connected with no shedding at all and the worst-case coincident demand is 300 + 200 + 200 = 700kW, which pushes the purchase to the next rating above it, an 800kW set. Apply the shed and the stage instead and the peak the generator ever has to carry is 300 + 200 = 500kW, because the rejected 200kW is blocked while the set is on load and the deferred 200kW only arrives once the essential tier is stable. A 500kW rating covers the site - 300kW of rating that does not have to be bought, housed, cooled or fuelled.

Now check that against the rating rules instead of stopping at the arithmetic. At ESP the permissible 24-hour average is 70% of 500kW = 350kW. The essential tier alone is 300kW, so even a long outage with tier 1 carrying continuously stays below the permissible average, and the deferred tier only lifts that average while it is actually running. The scheme is therefore defensible against the standard, not merely cheaper.

One caution on the same numbers: the 200 hours per year ESP cap still applies. Planned monthly testing, a long commissioning period and real outages consume that allowance faster than most buyers expect, and once it is gone the set is being operated outside its rating basis. Run-time hours belong in the same document as the load schedule.

Finally, the shed has to be executed by something. The usual options are a shunt trip on the downstream feeder, a PLC or BMS sequence, or a dry-contact output from the generator controller, and the boundary between them is worth agreeing early. Our notes cover dry contact and alarm interfaces for SCADA and BMS, and load bank testing before handover is where the staged sequence should actually be proven.

Which configuration mistakes cause the most rework?

• Sizing the generator from total connected load without defining which loads run together.
• Leaving HVAC or pump restart delays to be decided during commissioning.
• Assuming the generator supplier controls every downstream breaker.
• Testing the generator alone but not the load priority sequence.

Related equipment to compare

For a larger packaged set, review the Cummins QSB5.9 open frame diesel generator options. For another configuration path, compare it with the Cummins KTAA19-G6A 500kW diesel generator set. Related planning references include load bank testing before handover and step load acceptance checks.

Which documents should you request before approval?

• Load schedule with priority levels
• Single-line diagram
• ATS or switchgear control description
• Signal list for generator panel interface
• Commissioning sequence for staged loads

Before requesting a quotation

Send the load schedule, starting method for large motors, required restart delays, and control responsibility. The generator quotation can then address rating, alternator, controller, ATS, and signal scope more accurately. The rest of the buyer-side checks are collected under help and advice, and Tesla Power will review a tiered load schedule before it goes to tender if you send it with the enquiry.

Frequently asked questions

Does load shedding reduce generator size?

It can, if non-essential loads are blocked or delayed and the remaining load profile is technically acceptable.

Who designs the load shedding logic?

The site electrical designer usually defines the load priority, while the generator supplier confirms available control interfaces.

Is load shedding only for data centers?

No. It is also useful for hospitals, factories, water systems, mining sites, and any facility with mixed critical and non-critical loads.

What should be tested on site?

Test the start sequence, delayed loads, rejected loads, alarms, and recovery after utility power returns.

How fast must a generator pick up load under NFPA 110?

NFPA 110 Table 4.1(a) sets the Type by restoration time: Type 10 means 10 seconds, Type 60 means 60 seconds and Type 120 means 120 seconds. Ten seconds is the working default for Level 1 emergency power, and because transfer switches normally wait 1 to 3 seconds to avoid nuisance starts, the set may have as little as 7 seconds to start, transfer and accept load.

What does NFPA 110 Class 2 mean for the fuel tank?

Class is the number of hours the system must run at full rated output without refuelling, and Class 2 means 2 hours. It does not mean a two-hour tank: clause 7.9 requires the main tank to hold a minimum of 133% of the fuel the Class duration needs, so the tank is sized against roughly 2.7 hours of consumption.

Can load shedding let us buy a smaller generator set?

Yes, and the arithmetic is checkable. With 700kW connected as 300kW essential, 200kW deferred and 200kW rejected, unshed demand is 700kW and the next rating up is 800kW. Shed the rejected tier and stage the deferred one and the peak is 500kW, so a 500kW set covers the site.

Which ISO 8528-1 rating should a standby shedding plan use?

Emergency Standby Power (ESP) is the normal basis, but it is capped at 200 hours per year and limited to a 24-hour average load factor of 70% of the rating. If planned testing plus real outages will exceed 200 hours a year, the basis should move to Prime Rated Power (PRP), which allows unlimited annual hours at the same 70% average.

Who signs off the load shedding sequence?

The authority having jurisdiction approves the installation, and NFPA 110 clause 7.13 requires on-site installation acceptance tests including the two-hour load test at 7.13.4.3. The generator supplier can confirm which control outputs exist, but the priority list and the pass criteria belong to the site electrical designer and the AHJ.

Does load shedding need its own breaker control?

Something has to execute it. The common options are a shunt trip on the downstream feeder, a PLC or BMS sequence, or a dry-contact output from the generator controller. Whichever is chosen, agree in the enquiry which party supplies the signal and which supplies the breaker, because that boundary is where most handover disputes start.