Generator Sizing and Transient Response

Sizing a Generator Is Two Problems

A generator set has to do two different jobs, and a size that is correct for one can be wrong for the other. The first job is steady: carry the running load for hours or days without overheating or overloading the engine and alternator. The second job is transient: absorb a large step, usually the starting of the biggest motor, without the voltage and frequency falling far enough to drop out contactors, trip drives, or upset the UPS.

Many sets are sized only on the first job and then found to struggle on the second, because starting kVA can be several times the running kVA of the same motor. This guide walks through both checks in order, then runs one complete example. It sits alongside the demand load and motor starting guides, and the Generator Sizing and Step Load Calculator automates the sequence below.

Ratings: Standby, Prime, and Continuous

The same engine and alternator carry different kVA ratings depending on how many hours a year they run and how the load behaves. ISO 8528-1 defines continuous power, prime power, and limited-time running power, and manufacturers also publish an emergency standby rating. The definitions are not interchangeable, so the first question for any quoted number is which rating it is.

• Standby: for emergency use when the normal source fails. Annual running hours are limited, and manufacturers typically allow no overload at this rating
• Prime: for sites that rely on the set as the main source, or run it for long periods. Prime ratings are tied to an average loading limit over a 24-hour period (commonly 70% of the prime rating in the ISO definition), and an overload allowance is usually stated
• Continuous: for constant load at a steady level with no hour limit, usually the lowest kVA for the same frame

Confirm the exact definition and any limits in the manufacturer's rating data, because different brands phrase their ratings differently. A data center that uses on-site generation only for outages is a standby application. A site that runs gensets to avoid utility charges or as its primary supply is not, and needs the prime or continuous figure.

Step 1: Start From Demand Load and Power Factor

The input is the diversified running load in kW, built up with demand and diversity factors as described in Demand Load Calculation Basics: Demand Factors, Diversity and Design Load, not the sum of nameplates. Divide by the load power factor to get kVA. Generators are rated in kVA at a stated power factor, usually 0.8, so the alternator, not just the engine, sets what the set can deliver at the load's actual power factor.

• Use the right power factor: an inductive load at 0.85 needs more alternator kVA per kW than a resistive one, and a UPS with a poor input power factor can push this further
• Keep the motor list: the largest motor and its starting method are needed for Step 4, so record them separately
• Separate continuous from intermittent load: only the loads that run together at the worst moment count toward the steady figure

Step 2: Add Growth, a Loading Limit, and Site Derating

Three adjustments turn the running kVA into a required rating.

Future growth is a project decision. Document it as an explicit line, not a hidden margin. A maximum steady loading, commonly in the range of 70 to 80% for a single critical set, leaves room for transients, aging, and a load that turns out higher than estimated. It is a design choice, not a code value. Site derating accounts for altitude and ambient temperature: engines lose power in thin or hot air, and the manufacturer's rating table gives the correct figure for the engine and cooling system. Enter it as a percentage reduction or work from the derated kVA.

The required steady rating is then:

• Required kVA = running kVA × (1 + growth) ÷ maximum loading ÷ (1 − derating)

Light loading is the opposite problem. A diesel set that runs at a small fraction of its rating for long periods can foul and run poorly, so very large sets for small loads are not safe by default. Check the manufacturer's minimum load guidance and the exercise and load-bank requirements that apply to the application.

Step 3: Work Out the Motor Starting kVA

A motor draws its starting current at a very low power factor, and the generator has to supply it. Start with the full-load apparent power, which is the output kW divided by efficiency and power factor, then multiply by the starting current multiple for the starting method.

• Direct on line: locked-rotor current is commonly 5 to 8 times full-load current. Use the motor data sheet or the code letter, not a rule of thumb, for a large motor
• Star-delta: line current during the star period is roughly one third of the direct-on-line value, with a second current step at the changeover
• Autotransformer: with a 65% tap the line current is roughly 42% of the direct-on-line value, ignoring magnetizing current
• Soft starter: the current limit is set by the user, commonly around 2 to 4 times full-load current, at the cost of a longer start
• Variable frequency drive: starting current is close to full-load current, so the generator sees a gentle start, although the drive itself presents a non-linear load

The starting method is the biggest single lever on generator size, because it multiplies a number that appears directly in the dip calculation. The worked example below shows the effect.

Voltage Dip: The Series Reactance Model

When the motor starts, the generator sees a large, mostly reactive load. In the first cycles the alternator behaves like a voltage source behind its subtransient reactance X"d, and the starting motor behaves like a fixed apparent power S. With both on the generator's kVA base, the voltage at the motor terminals relative to the open-circuit voltage is the motor impedance divided by the sum of the motor and source impedances. That gives the standard estimate:

• Voltage dip = X"d × S ÷ (G + X"d × S), where S is the motor starting kVA and G is the online generator kVA after derating
• To limit the dip to d (as a fraction), the generator needs G = S × X"d × (1 − d) ÷ d

X"d is on the generator's own kVA base, and it is typically in the range of roughly 0.10 to 0.25 per unit, so always take it from the data sheet. A lower-reactance alternator is a common way to cut dip when the starting motor is fixed, usually at the cost of a larger alternator frame and higher fault level.

The model is deliberately simple. It ignores the load already on the set, the response of the voltage regulator and exciter, cable impedance, and the fact that a motor at reduced voltage develops less torque and takes longer to start. Those effects are real, and manufacturers' motor-starting programs include them. Use this estimate for early sizing and to compare starting methods, and use the manufacturer's analysis for the final selection.

The allowable dip is a project number. It is set by the most sensitive load on the bus: contactor holding voltage, drive undervoltage trips, UPS input windows, and lighting. A limit in the range of 10 to 25% is common, but the right figure is the one the sensitive load can actually tolerate, and that comes from the equipment, not from a rule.

Frequency Recovery and Block Loading

A load step also slows the engine until the governor responds, and the frequency dip and recovery time depend on the engine, its turbocharger, and the governor, not on the alternator. This is why engines have load acceptance limits: a set can usually take only part of its rating in one step, and the rest in following steps. ISO 8528-5 defines performance classes (G1 to G3, with a G4 class by agreement) that set limits for transient frequency and voltage deviation and recovery. Turbocharged diesels and gas engines generally accept smaller single steps than naturally aspirated or lightly loaded diesels, so ask for the manufacturer's step-load curve rather than assuming a number.

For emergency and critical systems, the load is usually staged: the transfer scheme brings loads on in blocks so that no step exceeds what the engine accepts. NFPA 110 sets the performance requirements for the emergency power supply system, including how quickly it must accept load, and the NEC articles for emergency (700), legally required standby (701), optional standby (702), and critical operations power (708) systems add the installation rules. Check the edition adopted by the jurisdiction.

Non-Linear Loads: UPS and Drives

A UPS rectifier or a drive front end draws current in pulses, and a generator has a higher source impedance than the utility. The result is voltage distortion at the generator terminals, which can upset the same electronics that created it. Three practical points:

• Rectifier type: UPS units with active (low-distortion) front ends and input filters are much easier on a generator than six-pulse rectifiers without them
• Soft ramp-up: many UPS units can delay or ramp the transfer from batteries to the generator input, so the load arrives gradually
• Alternator choice: a larger alternator or a lower subtransient reactance reduces distortion for a given non-linear load, which is why UPS-fed sets are sometimes specified with oversized alternators

See Harmonics & Power Quality: THD, Filters, and Resonance for the background. Whether a given UPS and generator combination is acceptable is a question for both manufacturers, not a calculator.

Paralleling and Redundancy

Large facilities usually use several smaller sets in parallel instead of one large one. Paralleled sets share kW through their governors and kVAr through their voltage regulators, and the paralleling switchgear synchronizes each set before it is connected. The benefits are redundancy, a smaller step per set, and the ability to run only the sets the load needs.

• N: the number of sets needed to carry the load. Each is sized for the total divided by N
• N+1: one extra identical set, so the load survives the loss of any one set. Size the dip and the loading for N sets online, because that is the post-failure case
• Starting kVA is shared: with identical sets the combined online rating is what resists the dip, so the formula above uses the total online kVA
• Fault current rises: paralleled sets add fault contribution, so check the switchgear ratings against the combined contribution, as discussed in Short Circuit Current Calculations: MVA Method and Impedance Method

Alternator fault current is a related point for protection. Without excitation support, the current collapses within a few cycles; with a permanent magnet generator or similar support, the alternator can sustain roughly three times its rated current for a few seconds, according to typical manufacturer data. Confirm the value for the actual machine before relying on it for protection coordination.

Data Center Considerations

For a data center, the generator is the last line behind the UPS, and the sizing conversation changes in three ways. First, the load is dominated by UPS input and mechanical cooling, so the non-linear and motor-start checks both matter. Second, the redundancy architecture sets the number of sets, so size from the electrical architecture first (see MV/LV Distribution Design: Feeder Sizing, Fault Levels, and Selectivity), not from a single kVA total. Third, the start-up sequence matters as much as the size: a facility that restarts cooling, UPS, and IT load in a defined order puts less stress on the set than one that reconnects everything at once.

• Load banks: plan for commissioning and periodic testing with resistive and reactive load banks, because the real load may not exist on day one
• Fuel and runtime: the rating is only half the question; tank size sets how long the set can run
• Gas-fired and other prime-power units: sets that run as the main source behave differently from standby diesels in rating, load acceptance, and maintenance, so use the prime rating and the manufacturer's step-load data

Worked Example: 1,200 kW Plant With a 250 kW Motor

Given: running load 1,200 kW at 0.85 power factor with all loads on, 950 kW already running when the largest motor starts, 10% growth, a maximum steady loading of 80%, no site derating, a 250 kW motor at 95% efficiency and 0.88 full-load power factor, X"d of 0.18 per unit, and an allowable voltage dip of 10%. One set carries the load (N = 1).

• Running load: 1,200 ÷ 0.85 = 1,412 kVA
• Steady requirement: 1,412 × 1.10 ÷ 0.80 = 1,941 kVA
• Motor full-load kVA: 250 ÷ 0.95 ÷ 0.88 = 299 kVA

Case A, direct on line at 6 times:

• Starting kVA: 299 × 6 = 1,794 kVA, and the total at the instant of the start is 950 ÷ 0.85 + 1,794 = 2,912 kVA
• Required for the dip limit: 1,794 × 0.18 × (1 − 0.10) ÷ 0.10 = 2,907 kVA
• Governing: the dip, at 2,907 kVA, is well above the steady 1,941 kVA
• Standard size: 3,000 kVA, which is 2,400 kW at 0.8, and steady loading is 47%
• Check: dip = 0.18 × 1,794 ÷ (3,000 + 0.18 × 1,794) = 9.7%, inside the 10% limit

Case B, soft starter limited to 3 times:

• Starting kVA: 299 × 3 = 897 kVA
• Required for the dip limit: 897 × 0.18 × 0.90 ÷ 0.10 = 1,453 kVA, below the steady 1,941 kVA
• Governing: the steady load, so the standard size is 2,000 kVA (1,600 kW) and steady loading is 71%
• Check: dip = 0.18 × 897 ÷ (2,000 + 0.18 × 897) = 7.5%

Changing the starting method on one motor removes 1,000 kVA from the set. A drive (about 1.1 times) would cut the starting kVA further, to about 329 kVA and a dip of about 2.9%, but the steady load already governs at 2,000 kVA. That is the general pattern: starting method matters until the steady load takes over.

Case C, two paralleled sets (N = 2, DOL): each set needs 2,907 ÷ 2 = 1,453 kVA, so each is a 1,500 kVA unit. With N+1, three 1,500 kVA sets are installed. The dip with two sets online is the same 9.7%, because the combined online rating is the same 3,000 kVA, and the loss of one set still leaves two online.

These are screening numbers. The final selection needs the manufacturer's rating tables, X"d, and motor-start analysis for the exact set and starting method.

Where Standards Diverge

The sizing method is the same under NEC and IEC practice, and only the installation and performance rules differ. In North America, NFPA 110 covers emergency and standby power system performance and the NEC covers the installation and the load classes (700, 701, 702, and 708). Under IEC practice, ISO 8528 (all parts) covers the set and its performance classes, IEC 60034-1 covers the alternator rating, and IEC 60364-5-55 covers low-voltage generating sets, with IEC 60364-5-56 covering safety services. Frequency (50 or 60 Hz) changes the engine speed and the generator data, but not the kVA arithmetic here. The IEEE Orange Book (IEEE Std 446) is an older reference on emergency and standby power that is still widely cited, but check its current status before specifying it.

Practical Checklist

• 1. Build the running load from demand and diversity factors, in kW and kVA
• 2. Apply growth, a maximum loading, and the manufacturer's site derating
• 3. Identify the rating basis: standby, prime, or continuous, and confirm the definition
• 4. Find the largest motor, its full-load kVA, and the starting multiple for the starting method
• 5. Estimate the voltage dip, with X"d from the data sheet and a dip limit from the most sensitive load
• 6. Take the larger of the steady and dip requirements and round up to a standard size, per set if paralleling
• 7. Check engine load acceptance, UPS and drive compatibility, and fault contribution with the manufacturer
• 8. Plan commissioning load banks, fuel runtime, and the restart sequence

Summary

• A generator has a steady-load requirement and a transient requirement, and the larger of the two sets the size
• Voltage dip is estimated as X"d S ÷ (G + X"d S), so lower reactance, a larger set, or a gentler starter all reduce it
• Starting method is the cheapest lever: in the example, a soft starter on one motor removed 1,000 kVA from the set
• Frequency recovery and block loading depend on the engine, so use manufacturer step-load data
• UPS and drive loads are a compatibility question for the manufacturers, not just a kVA question
• Paralleling spreads the step and adds redundancy, but raises fault level and needs paralleling switchgear

Disclaimer: This guide is for general information and preliminary reference only. Engineering judgment is required. Verify methods, values, and code requirements against the edition adopted for your project, the project specification, and manufacturer data before use in design documents.

Found an error or an outdated reference? Let us know at [email protected].