Reciprocating Gensets: Diesel and Gas Engines

A piston engine burns fuel in cylinders, turns a crankshaft, and drives an alternator. The generator set includes more than those two machines: starting batteries or air system, cooling, lube oil, fuel system, controls, protective devices, exhaust, and often an enclosure. A medium-speed power-plant engine can be much larger than a compact standby diesel. Wärtsilä’s 31SG is an example of a gas-engine generating set; Cummins’ QSK95 is a diesel data-center set.

The key buying distinction is the delivered electrical package. A bare engine’s mechanical output is not the site’s usable electrical output. Alternator losses, radiator fans, fuel treatment, transformer losses, and site temperature all affect the result. Ask whether a listing covers an engine, an engine-generator pair, or an installed plant with switchgear and fuel infrastructure.

A genset can be one machine serving a small critical block or one module in a plant of synchronized units. Caterpillar’s paralleling explanation describes the second arrangement: each set synchronizes to a common bus and shares load. When a campus grows, the ability to add modules can be valuable, but the electrical bus and controls have to be designed for the intended plant size from the outset.

A diesel set brings its fuel from an on-site tank and can be designed for emergency starts. Its fuel independence is valuable when the utility fails and a pipeline interruption is part of the same incident. A natural-gas set receives pipeline gas or another gaseous fuel and can be specified for sustained generation if supply is firm. Cummins markets the HSK78G for prime and peaking applications; Caterpillar describes the CG260-16 with continuous natural-gas output. Those are distinct from a diesel standby proposal.

Neither fuel is a universal answer. A diesel plant needs storage, fuel-quality management, refill access, and emissions compliance for its duty. A gas plant needs a firm pressure and flow envelope, gas conditioning if required, and a contingency for curtailment. The site might combine a gas prime-power plant with a diesel standby tier. Caterpillar’s data-center gas conversion discussion gives examples of mixed diesel and gas configurations and explains why their load-acceptance behavior differs.

For a facility planning years of operation, start with the dispatch profile: emergency outage only, planned daily operation, or primary continuous power. The engine’s emissions category and fuel system follow that use. Selecting a “diesel model” or “gas model” before defining the duty can leave the site with equipment that cannot legally or technically run as intended.

There is no one genset size. Cummins’ HSK78G page lists continuous ratings in the 1.6–2.0 MW range. Caterpillar’s CG260-16 page lists approximately 4 MW at 60 Hz and a larger 50 Hz continuous figure. Wärtsilä’s 31SG publishes several 20-cylinder gas-engine configurations around 10–13 MW. These examples span a useful planning range, but they do not define a universal piston-engine ceiling.

Frequency, fuel, duty rating, and output boundary must travel with the MW number. A standby nameplate may exceed the same set’s sustained rating. An engine optimized for electrical efficiency can have a different output from a power-optimized variant; Wärtsilä’s 31SG table separates those configurations. A 50 Hz value may not transfer to a 60 Hz site.

For a 30 MW campus, a designer could meet nominal demand using many smaller units or fewer large engines. The first approach has more exhausts, fuel branches, breakers, and maintenance points; the second makes each outage a larger capacity event. Calculate firm capacity after one set is unavailable, including summer derating and auxiliaries. That plant-level calculation is more useful than dividing the IT load by the largest number on a product page.

A model number does not specify duty. Cummins’ QSK95 carries data-center rating language, while HSK78G is marketed for gas prime and peaking work; Caterpillar’s CG260-16 publishes continuous gas output. Record the duty before comparing the stated kW. The detailed ISO and OEM definitions belong in the generator-rating guide.

Piston engines often remain useful at varying loads, and a multi-unit plant can sequence sets to keep running engines nearer efficient operating points. Wärtsilä’s engine-versus-aero comparison publishes a specific dispatch comparison with a lower minimum stable load for its combustion-engine case than its aeroderivative case. It is vendor data, so use it as a model-specific illustration rather than a law of physics applying to every engine and turbine.

A simple plant example shows why sequencing matters. If four equal engines are installed and demand is half of the plant maximum, operators might run two nearer design load rather than four at half load. The actual choice depends on required spinning reserve, start time, maintenance state, emissions limits, and the risk of a large step when a unit trips. A hospital or data center may deliberately keep additional engines online for reliability even if fuel efficiency falls.

The energy model should therefore include the number of running units by hour, not just one full-load heat rate. Wärtsilä’s 31SG product data provides both output and efficiency for several configurations, but a campus dispatch study also needs load profiles, reserve rules, and emissions-compliant minimum loads. The answer may change during construction as IT occupancy grows.

Start time and load acceptance are separate tests. A set may reach rated speed quickly but be unable to accept the entire campus load in one step without excessive frequency or voltage disturbance. Wärtsilä lists 31SG start and load figures for defined configurations; Caterpillar’s 50 Hz G3520 Fast Response announcement describes a natural-gas product built for mission-critical applications. Neither statement proves every gas set is a drop-in replacement for every diesel emergency set.

The transfer system matters as much as the prime mover. Schneider Electric describes the UPS as the bridge while generators start and stabilize. In a mixed gas/diesel design, Caterpillar explains that a gas unit substituting for a diesel unit may need a slower UPS load walk-in to respect gas-engine load acceptance. This is an integrated controls problem, not a simple “seconds to full power” contest.

A proper acceptance test should stage utility loss, first engine start, dead-bus close, additional synchronization, UPS walk-in, cooling equipment restart, and failure of one intended starting unit. Record voltage and frequency at the load bus. The site’s critical equipment sees the whole sequence, not the best-case start time printed for an engine in isolation.

In the United States, stationary diesel and spark-ignited gas engines have different federal rule paths. EPA’s stationary-engine overview points readers to compression-ignition and spark-ignition standards. For an emergency stationary diesel engine subject to 40 CFR 60.4211(f), true emergency operation has no time limit under that provision, but maintenance and readiness testing are generally limited to 100 hours per calendar year; up to 50 of those hours may be used for defined non-emergency situations. The rule generally bars using those hours for peak shaving, subject to its narrow specified exception.

The gas-engine counterpart, 40 CFR 60.4243(d), has its own emergency-use language. An engine intended to carry routine prime load should not be presented to the permitting authority as an emergency-only machine. However, it is equally wrong to say every non-emergency engine must have one particular aftertreatment package. Compliance depends on fuel, model year, engine category, size, site permit, and applicable federal, state, and local rules.

Define the intended annual operating pattern before selecting a unit. A site planning to use a standby fleet as bridge power while waiting for a utility connection must evaluate non-emergency requirements at the start. A permit written for occasional outages is not proof that the same engine can operate every day. The data-center-rated generator set guide owns the rating and emergency-use distinctions in more detail.

Yes. An engine rejects heat through exhaust, coolant, and lubricating systems. If a facility has a steady thermal load, a heat-recovery package can turn some of that energy into hot water, steam, or another useful service. A manufacturing plant with process heat might gain more from this than a data hall without a suitable thermal customer. The engineering question is how much heat is useful at the same hours the engines run.

Gas and diesel engines have different fuel, emissions, and heat-recovery designs, so do not copy a CHP claim between them. Wärtsilä’s 31SG is a gas engine offered for power-plant applications; Cummins’ HSK78G is positioned for gas prime and peaking service. For either, the CHP project must provide a heat balance for the selected engine, exhaust conditions, cooling circuit, and thermal demand.

A standby data-center diesel that runs only during outages has little predictable annual heat for an adjacent facility to use. A gas engine running most hours can support a more stable heat-recovery case. But an HRSG, heat exchanger, pumps, and controls add failure modes. If lost heat demand causes the engine to trip or derate, the design must include bypass or rejection capacity so electricity remains available when the thermal system is down.

N+1 means the plant has enough installed capacity to serve the required load after one designated unit is unavailable, subject to the stated design basis. It does not simply mean buying one extra engine and ignoring the switchgear. Caterpillar’s paralleling paper explains common-bus synchronization and warns that central controls can themselves become a single point of failure. Caterpillar’s data-center design guide also discusses bus ampacity and short-circuit limits.

Take a plant with four identical sets needed at peak. A fifth set may create generator redundancy, but a single common switchgear bus, common fuel header, common cooling loop, or unprotected controls processor can still remove all five. The resilience study should draw fault zones and ask what happens when each shared component fails. Multiple buses or distributed controls may be justified by the load’s tolerance, but they also increase interconnection complexity.

The operating sequence should include maintenance. Can one unit be isolated, serviced, and returned without interrupting load? What happens if another unit trips during that interval? Is the spare already running or must it start? These are plant questions. The OEM’s engine reliability percentage, even if available, cannot answer whether the electrical topology survives the failure the customer actually cares about.

Specify the required net kW at the receiving bus, site ambient and elevation, frequency, fuel, expected operating hours, load profile, largest step load, and emissions jurisdiction. State whether the set is emergency standby, prime, continuous, or a data-center-specific rating. The rating vocabulary is explained in data-center-rated generator sets; a generic “3 MW generator” request omits the key duty constraint.

Next define scope: open skid or enclosure, radiator and exhaust arrangement, fuel storage or gas train, alternator voltage and power factor, switchgear interface, controls, remote monitoring, and factory or site tests. Compare Cummins QSK95, Caterpillar CG260-16, and Wärtsilä 31SG only after putting their differing fuels and ratings into the same request form. Their public numbers have different application contexts.

For an existing unit, request serials, operating hours, starts, overhaul record, load-bank tests, alternator insulation tests, maintenance logs, fuel-system condition, and preservation status. A buyer should confirm what is physically included. A Pantheon genset quote request can carry these requirements so Pantheon can confirm what its network can source; an owner can submit the asset details for possible placement.

One mistake is assuming diesel and gas sets accept identical step loads. Caterpillar’s mixed-fuel analysis explicitly changes UPS load walk-in when a gas set substitutes for diesel. Another is comparing a standby kW with a continuous kW; the number belongs to a duty, not simply to an engine displacement. A third is designing “N+1” engines behind one unexamined common bus. Caterpillar’s paralleling explanation makes that shared-control risk concrete.

A fourth mistake is taking a vendor’s published maximum efficiency as the whole-plant annual efficiency. A lightly loaded or reserve-heavy plant dispatches differently from a product test. A fifth is assuming a model’s fuel listing proves the receiving site can use that fuel; gas quality and supply pressure, or diesel storage and refill, may be the binding constraint.

Finally, do not treat emergency hours as an elective power supply. 40 CFR 60.4211(f) sharply distinguishes genuine emergencies from maintenance and limited other use. If the facility expects routine running, specify and permit that duty from the beginning. For the detailed rating and emergency-power question, use the data-center-rated generator set guide.