Sizing a generator for site loads

Scope. This method organises declared load information. It does not select a generator, design earthing/protection, model an alternator’s transient response or authorise a location.

Start with location: fuel generators and carbon monoxide

Carbon monoxide (CO) has no colour, taste or smell. HSE warns that petrol or diesel equipment can produce dangerous CO concentrations in enclosed or poorly ventilated places, including trenches. Moving a generator nearer a doorway, adding an ordinary fan or relying on smell is not a sizing control. [HSE]

For an open-air proposal, still consider exhaust direction, nearby openings, access, refuelling, fire precautions, weather, security and noise. Those are task controls rather than variables in the kVA formula.

Record the location decision with the load sheet.

Include the proposed operating hours, nearby occupied areas and the person responsible for maintaining the exclusion and exhaust arrangement.

Build a running-load schedule

List equipment that can draw power at the same time. Use electrical input kW, rated VA or current/voltage data that can be tied to the actual item. Mechanical output is not a substitute. Record power factor (PF), voltage, phase and frequency, and identify loads that cycle automatically.

A credible schedule includes support loads: transformer losses where known, extraction, pumps, chargers, lighting and controls. It also separates mutually exclusive work from genuinely concurrent work. “Probably not together” is too vague; state the operating rule that prevents concurrency.

• Identity: equipment or load group and rating-plate reference.
• Supply: voltage, phase and frequency.
• Running demand: electrical kW or kVA and PF where conversion is needed.
• Start behaviour: manufacturer start kVA, current or multiplier, and duration/basis.
• Sequence: what is already running when this item starts.
• Duty: continuous, cyclic or intermittent, without assuming intermittent means negligible.

Cummins’ sizing guidance describes load characteristics, starting requirements, duty and site conditions as inputs to generator-set selection. This is manufacturer guidance, not a universal selection table. [MFR]

Model starting demand in the real sequence

Starting demand can exceed running demand, but there is no defensible universal “three times” rule. Use a manufacturer-stated start kVA/current or multiplier for the particular load. If several loads may start together, model that event explicitly or change the control sequence; the Job Lab’s simple route models one selected start while other scheduled loads continue running.

Record why the selected item is the governing event. A large running load may not create the largest step, while a smaller motor started across the line may matter more. Variable-speed drives, welders, lifts, compressors, electronic supplies and cyclic loads can require more detailed treatment than a multiplier.

The Cummins application manual discusses transient performance, motor starting, nonlinear loads, load steps and environmental influences. Its figures describe Cummins’ engineering method and must not be turned into a universal product threshold. [MFR]

Running, peak and planning formulas

For each load stated in kW and PF, running apparent load is Sᵢ = Pᵢ ÷ PFᵢ. The concurrent total is Srun = ΣSᵢ. For one selected starting item j, the simple event is Sj = (Pj × Mj ÷ PFj) + Σ(Pi ÷ PFi) for every other running item. [ARITH]

An entered margin is then shown as Splan = Sj × (1 + margin ÷ 100). The margin is Job Lab/user policy, not law and not a substitute for derating, load-step acceptance or verified reserve. [POLICY] Comparisons use unrounded numbers; displayed kVA is rounded to three decimal places.

Keep three lines in the record: running kVA, the declared start event, and the margin-adjusted planning figure. Calling all three “generator size” loses the reason for the calculation.

Worked example

Load A is 1.6 kW at PF 0.9 and has a manufacturer-declared 2.5× start multiplier. Load B is 0.8 kW at PF 0.8 and is already running. They operate on the same declared voltage, phase and frequency.

The 20% margin is an entered assumption. The 6.533 kVA display is not rounded to a catalogue size and does not establish that a generator can accept the step. Carry the load list, start basis and site conditions into the manufacturer/supplier check.

What the arithmetic does not model

A scalar kVA total does not predict voltage dip, frequency dip, recovery time, alternator excitation or engine response. It also does not account automatically for harmonic/nonlinear current, regenerative loads, unbalanced phases, repeated load cycling or a sequence that controls large steps.

Generator ratings can be stated on different duty bases. Continuous, prime, standby and other rating terms are not interchangeable. Obtain the basis and conditions for the actual set. Temperature, altitude, fuel, ventilation and ancillary loads may derate available output; use the manufacturer’s data for the proposed environment. [MFR]

Also confirm voltage, frequency, phase, neutral/earthing arrangement, protective devices and distribution. A transformer or long extension lead downstream introduces its own limits and voltage drop. Use the lead voltage-drop tool only with verified cable data, and read the 110 V site-power method.

Load quality matters as well as magnitude. Rectifiers, variable-speed drives and switched supplies can distort current, while welding and cyclic plant can impose abrupt repeated steps. A single power-factor value and one multiplier do not capture harmonic heating, phase imbalance or recovery between events. Put these load types in the manufacturer enquiry rather than forcing them through the simple model.

Write the proposed operating sequence in ordinary language: “extractor and lighting running; saw starts; compressor locked out until saw stops”. This lets a reviewer test the concurrency assumption. If the sequence depends on operator memory, state the control that enforces it. The schedule should also show what happens after a trip or automatic restart and whether a safety-critical load must remain energised.

Finally, distinguish nameplate demand from measured evidence. Measurement can help validate a repeated setup, but it must capture the relevant start event with appropriate equipment and competence. A clamp reading observed after startup does not establish the peak. Preserve instrument, method and conditions with any measured value.

Stop and confirm with the manufacturer or competent person

The generator source group records the source keys, dates and registry version. Methodology explains full-precision comparison, fail-closed behaviour and change control.

This guide is a load-planning aid. It cannot confirm a generator, electrical arrangement or location. Follow the site assessment, manufacturer instructions and competent electrical design.