Cummins Generators

The first argument on almost every project is the size of the set, and it usually starts from the wrong number. A generator's nameplate is given in kVA — apparent power — but your equipment does real work in kW. The two are linked by the power factor (PF), the ratio of useful power to the total power the alternator has to push. For diesel sets the industry convention is a lagging PF of 0.8, so an 800 kVA set is only an 640 kW machine. Specify in kW, confirm the PF of your actual load, then convert — never the other way round.

The reactive component matters because motors, transformers and fluorescent gear draw magnetising current that does no work but still heats the alternator windings and loads the engine. A site full of induction motors running at PF 0.65 will pull far more kVA than its kW figure suggests, which is why power-factor correction at the main board is often cheaper than a bigger genset.

Get this number wrong in either direction and you pay. Oversize the set and it spends its life lightly loaded, which — as we cover below — destroys diesel engines through wet stacking. Undersize it and the first big motor start collapses the voltage, drops out contactors and trips the main breaker. The right answer is rarely "round up to the next model"; it is a load study.

On most industrial and water-supply sites it is not the steady running load that sets the genset size — it is the worst-case motor start. A direct-on-line (DOL) induction motor draws a locked-rotor inrush of roughly six to eight times its full-load current for the first few seconds, at a miserable starting power factor of around 0.3. That transient is what the alternator must swallow without the voltage sagging far enough to drop out your contactors and controls.

So the real sizing question is the starting kVA (SkVA) of the largest motor that can start while everything else is already running. ISO 8528 classifies sets by how tightly they hold voltage and frequency during these steps (performance classes G1 to G4); a data centre or hospital wants G3, where the transient voltage dip is held to a few percent. Reciprocating-engine sets can typically accept a first step of 60–80% of their rating, but only if the alternator is sized for the inrush.

The fix is often not a bigger generator. Soft starters cut inrush to around 2–3× FLC, and variable-frequency drives (VFDs) bring it down near 1× while ramping — turning a brutal step load into a gentle one. On borehole and HVAC sites we routinely save a customer a whole frame size by staging starts and adding soft-starters rather than buying copper.

This is the single most common specification error we see, because most generators sold in the region are rated at the ISO reference of sea level and 25 °C — conditions that exist almost nowhere on the Kenyan plateau. A diesel engine breathes air; thinner high-altitude air carries less oxygen, so the engine makes less power. As a rule of thumb a naturally-aspirated engine loses about 3–4% of its output for every 300 m above 1,000 m, and roughly a further 1–2% for every 5–6 °C above the 25–40 °C reference. Turbocharged engines tolerate altitude better but are not immune.

Nairobi sits at about 1,795 m, Nakuru at 1,850 m, Eldoret at roughly 2,100 m. A "100 kVA" set delivered to a factory in Eldoret can be a genuine 80 kVA machine on a hot afternoon. Buyers then wonder why their "oversized" generator trips on overload every time the chillers and compressors coincide. The capacity was never there — it was de-rated away by physics the supplier ignored.

We always size against the site-corrected rating, not the brochure. That means knowing the altitude, the worst-case ambient (radiator air-on temperature, not shade temperature), and whether the canopy restricts airflow. For coastal sites — Mombasa, Kilifi, Diani — altitude is a non-issue but ambient heat and salt-laden humidity drive the derate and the corrosion-protection spec instead.

A generator does not have one power figure; it has several, and quoting the wrong one is how cheap sets are made to look competitive. ISO 8528 defines the operating ratings, and the gap between them is real money and real engine life. Pay for the duty you actually run, not the headline number.

For a building that draws grid power and only fires the genset during outages, Emergency Standby Power (ESP)is correct — full output for the duration of the outage, variable load, no sustained overload, limited annual hours. A telecom site or an off-grid lodge that runs the set as its main supply needs Prime (PRP), rated for unlimited hours at a variable load whose average stays around 70%. A process that demands constant full output around the clock needs Continuous (COP). Putting an ESP-rated set on prime duty voids the warranty and ages the engine fast.

Over a ten-year life, fuel — not the purchase price — is where most of the money goes. A modern diesel set burns roughly 0.25–0.30 litres of diesel per kWh generated at full load, but specific consumption climbs sharply at light load because the engine still has fixed friction and pumping losses to overcome. A set running at 25% load can burn nearly twice as much fuel per useful kWh as the same set at 80%. The curve below is why right-sizing and loading discipline matter more than any spec-sheet efficiency claim.

Translate that into shillings and the case for treating a generator as backup, not base-load becomes obvious. At a pump price around KSh 165/litre, diesel generation lands near KSh 40–55 per kWh once you include oil and maintenance — well above commercial grid tariffs. The sets that wreck a budget are the ones quietly running as primary supply because nobody trusts the changeover, or because a solar-plus-storage layer was never added to carry the daytime base load.

This is the conversation we have with every serious B2B client: the cheapest kWh is the one you do not generate on diesel. A correctly sized genset for genuine outage cover, paired with solar PV and (where the tariff justifies it) battery storage, routinely cuts a site's diesel bill by 40–70% while keeping the resilience.

A diesel engine is designed to run hot and worked. Keep it below about 30% load for long periods and combustion temperatures never rise enough to burn the fuel completely. Unburnt diesel and soot then condense in the cylinders, exhaust and turbocharger — the black, oily weep at the exhaust that gives wet stacking its name. Left unchecked it glazes the bores, fouls injectors and valves, and progressively strangles the engine.

This is the hidden cost of the "buy big to be safe" instinct. The oversized standby set that idles at 15% during every short outage is slowly killing itself. The remedy is either to size honestly, or — where load genuinely varies — to commission an annual resistive load-bank test that runs the set to 80–100% for an hour to burn off the deposits and prove it can carry full rated load. NFPA 110 builds this into its monthly/annual test regime for exactly this reason.

Modern sites are full of non-linear loads — VFDs, UPS rectifiers, LED drivers, IT power supplies — that draw current in sharp pulses rather than clean sine waves. Those pulses inject harmonic currents back into the alternator, distorting the voltage waveform (measured as total harmonic distortion, THD), overheating the windings and confusing voltage regulators. IEEE 519 recommends keeping voltage THD below 5% at the point of common coupling; a poorly matched genset on a heavy VFD load can drift well past that.

Two things protect against it. First, a permanent-magnet generator (PMG) excitation with a good digital AVR sustains field current even when the waveform is distorted, so the set holds voltage during step loads and fault clearing. Second, where the non-linear fraction is high, the alternator is deliberately oversized — a common rule is to rate the alternator at 1.5–2× the connected UPS or VFD load so the harmonic heating stays within the insulation class. Skipping this is why some sets "work on the bench" but nuisance-trip once the real IT and drive load is connected.

On data-centre, hospital and broadcast work we specify the alternator and the UPS together, because the two have to cooperate during transfer. A UPS with an aggressive input filter or a low-quality rectifier can fight the genset; the cure is matched ratings, a PMG alternator, and sometimes a passive harmonic filter or an active front-end UPS.

A standby generator earns its keep on the day the grid fails — and the cruellest statistic in our trade is how many sets fail to start on exactly that day. The leading cause is not the engine at all: it is a flat or sulphated starting battery, followed by stale fuel and a discharged or faulty charger. Reliability is a maintenance outcome, not a purchase decision.

We schedule service by running hours or by calendar, whichever comes first, and we log every visit. Oil and filter changes at 250–500 hours or annually; coolant condition and concentration checked each visit; fuel polished or treated to stop diesel-bug growth in tanks that sit full for months; valve lash and injectors on the manufacturer's longer interval; battery tested under load — voltage alone lies. Above all, the set is exercised under loadregularly, not just idled, for the wet-stacking reasons above.

The number that matters to a facility manager is availability — the probability the set runs when called. It is driven by how often it fails (mean time between failures, MTBF) and how fast it is repaired (mean time to repair, MTTR). A genset with a long MTBF but a three-day parts wait can have worse real-world availability than a humbler set backed by local spares and a four-hour SLA. This is why we hold fast-moving Cummins and FG Wilson parts locally and contract response times rather than selling iron and disappearing.

When two quotes differ by a few hundred thousand shillings, the gap almost always reappears — multiplied — in fuel, parts and downtime over the life of the machine. The honest comparison is total cost of ownership: purchase and installation, plus fuel over the expected running hours, plus scheduled maintenance, plus the cost of the outages the set fails to cover. A generator that is 15% cheaper but burns 10% more fuel and waits days for parts is the expensive option on any horizon longer than a year.

We put this number in front of clients deliberately, because it reframes the decision from price to value — and because it is where a properly engineered solution (right size, right rating, local parts, an SLA, and a solar layer to shave the fuel) pays for itself. Ask any supplier for the ten-year TCO, not just the invoice. If they cannot produce it, they have not engineered your power — they have just sold you a box.