What Type of Generator Set Is Best for a Data Center?

Data centers are among the most demanding generator applications in the world -- not because they require the most power (mining operations and industrial plants often draw more), but because the consequences of failure are immediate, visible, and commercially catastrophic. A data center that loses power loses customer data, breaks SLA commitments, and in a worst case, loses the trust of every customer whose workload was running at the moment of failure.

Standard commercial generators -- the same units used in hotels, offices, and hospitals -- are not the right answer for data center backup power. Not because they are inherently inferior, but because data centers impose specific technical requirements on generator type, start time, power quality, redundancy configuration, and fuel capacity that require deliberate specification choices. Get those choices right and the generator set performs exactly as required. Get them wrong and the generator becomes the weakest link in an otherwise well-designed facility.

This guide covers the full specification picture for data center generator sets -- from Tier classification requirements to alternator subtransient reactance to parallel control systems -- written for data center designers, facilities engineers, and the procurement teams who specify and source this equipment.

Why Data Centers Are Different: The Three Unique Demands

  ⚡  Demand 1: 10-Second Start and Transfer -- No Exceptions

  UPS systems provide bridge power during grid-to-generator transition. Most commercial UPS systems can sustain rated output for 5-20 minutes on battery, depending on battery bank size. But in the data center context, UPS battery autonomy is sized to cover the generator start-and-transfer window -- not extended outages. That window is typically defined as 10 seconds from mains failure detection to generator stable output. Every second of delay beyond 10 seconds consumes battery reserve. A generator that takes 25 seconds to reach stable output at a facility with a 10-minute UPS battery bank is consuming 25% of that battery reserve just to start -- leaving the facility with only 7.5 minutes of battery before the generator must take full load. Specify and verify cold-start time before commissioning.

  ⚡  Demand 2: Non-Linear Load Compatibility -- The Specification Most Buyers Miss

  Data center loads are dominated by switch-mode power supplies in servers, storage arrays, and network equipment, plus UPS rectifiers. These are non-linear loads -- they draw current in pulses rather than sinusoidally. Non-linear loads create harmonic currents (3rd, 5th, 7th harmonics) in the generator's output. These harmonic currents interact with the alternator's internal impedance (characterised by the subtransient reactance, X''d) to produce harmonic voltage distortion on the output. High voltage THD (Total Harmonic Distortion) degrades UPS performance, causes additional heating in server power supplies, and can trip sensitive protection systems. Standard commercial alternators with X''d of 0.16-0.20 per unit produce THD levels that exceed data center equipment tolerances under heavy non-linear load. Data center alternators must have X''d of 0.10-0.12 per unit.

  ⚡  Demand 3: Redundancy -- No Single Point of Failure

  A data center generator that fails leaves the facility dependent on UPS battery alone -- a finite resource. For Tier III and Tier IV data centers, no single component failure should be able to interrupt power to the IT load. This means the generator system must have redundancy: N+1 minimum (an additional generator that can carry the full load if one unit fails) or 2N (a complete duplicate generator system). Single-generator backup power is only acceptable for Tier I and Tier II facilities where some downtime is contractually accepted.

Uptime Institute Tier Classification: Generator Requirements at Each Level

The Uptime Institute Tier Standard is the globally recognised framework for data center infrastructure design. Generator requirements escalate at each Tier level.

  Tier I -- Basic Capacity

  Redundancy model: N (no redundancy) -- single path for power

  Generator set requirement: Single generator acceptable; standby rating may be used; no redundancy required

  Transfer time: No specified requirement -- UPS to generator; often 15-30 seconds acceptable

  Concurrent maintainability: Not required -- maintenance causes downtime

  Fault tolerance: None -- single failure causes outage

  Tier II -- Redundant Components

  Redundancy model: N+1 for most components

  Generator set requirement: Generator backup required; N+1 generator configuration recommended but not mandatory

  Transfer time: <15 seconds preferred

  Concurrent maintainability: Partial -- some concurrent maintenance possible

  Fault tolerance: Limited -- some component failures tolerated

  Tier III -- Concurrently Maintainable

  Redundancy model: N+1 minimum across all components; all active components maintainable without IT load impact

  Generator set requirement: N+1 generator mandatory -- minimum two generators, each rated for 100% of critical load independently

  Transfer time: <10 seconds -- specified and tested

  Concurrent maintainability: Yes -- all equipment maintainable without shutting down IT load

  Fault tolerance: Any single failure must not interrupt IT load

  Tier IV -- Fault Tolerant

  Redundancy model: 2N (fully redundant dual systems) -- two independent generator systems each at N+1

  Generator set requirement: 2N generator configuration -- two fully independent generator systems, each capable of carrying 100% of critical load

  Transfer time: <10 seconds -- specified, tested, and proven by fault simulation

  Concurrent maintainability: Yes -- full fault tolerance; no capacity reduction during maintenance

  Fault tolerance: Any single fault or maintenance event must not reduce IT capacity

  Tier III is the most commonly specified level for new commercial and enterprise data center construction globally. It requires N+1 generator redundancy as a minimum. A Tier III data center with a single generator -- regardless of that generator's size or specification quality -- does not meet Tier III requirements and cannot receive Tier III certification from Uptime Institute.

The Right Generator Type: Open Frame vs Containerised

Data center generators are almost universally specified as open frame (no canopy enclosure), installed in purpose-built generator rooms or external enclosures that form part of the building infrastructure. This differs from hotel, hospital, and commercial building installations where silent canopy generators are typically specified.

Why open frame is preferred for data centers:  Data center generator rooms are designed by mechanical and electrical engineers as part of the building infrastructure. The room provides weather protection, acoustic attenuation, ventilation, and security -- all the functions that a canopy provides on a standalone commercial generator. Adding a canopy to a generator in a properly designed generator room adds cost without benefit and complicates maintenance access. Open frame generators also provide easier integration of paralleling bus bars, protection relay panels, and synchronising equipment in multiple-generator configurations.

When containerised is specified:  Modular data centers and edge data center deployments sometimes use containerised generator sets -- generators in acoustic-lined shipping containers that can be positioned adjacent to the modular data hall. This approach suits rapid deployment and site flexibility. Container generators achieve 70-75 dB(A) at 1 metre -- adequate for most commercial sites -- and provide IP weather protection without a dedicated building structure.

The Complete Data Center Generator Specification

Every item in the following specification has a reason related to data center operating conditions. Each is non-negotiable for Tier III and Tier IV facilities.

Alternator Specification Deep Dive: The Non-Linear Load Challenge

The alternator specification is where most data center generator quotations fall short. Standard commercial alternators are adequate for resistive and motor loads -- hotels, factories, hospitals. They are not optimised for the predominantly non-linear load profile of a data center. Understanding the key parameters allows you to specify and verify correctly.

  ✔  Subtransient Reactance (X''d): The Most Important Alternator Parameter

  X''d (subtransient reactance, in per unit) is the internal electrical impedance of the alternator winding during the first few cycles after a load change. Lower X''d means lower voltage distortion when harmonic currents flow in the alternator windings.

  Standard commercial alternator: X''d = 0.14-0.20 pu
  Data center specification: X''d = 0.10-0.12 pu

  To achieve X''d of 0.10-0.12, the alternator winding requires more copper -- a physically larger alternator at the same kVA rating. This adds 15-25% to the alternator cost but is non-negotiable for facilities with predominant non-linear loads. When specifying, write 'subtransient reactance not greater than 0.12 pu' in the alternator specification and request confirmation from the alternator manufacturer's published datasheet for the specific model.

  ✔  PMG Excitation: Why It Matters for Non-Linear Loads

  Standard alternators use shunt excitation -- the AVR draws its power from the alternator's own output terminals. With predominantly non-linear loads, the harmonic distortion on the output terminals can interfere with the AVR sensing circuit, causing voltage instability or oscillation.

  Permanent Magnet Generator (PMG) excitation provides an isolated, clean power source for the AVR that is independent of the main output. The PMG is a small permanent magnet generator mounted on the alternator shaft. Because its output is independent of the main alternator output harmonics, the AVR receives a clean sensing signal regardless of the distortion level on the main output.

  For data centers: specify PMG excitation where available (Stamford P7 series, Leroy Somer R448 AVR configuration). If PMG is not available, specify AREP (Auxiliary Rectified Exciter Power) as the minimum standard.

  THD verification: request the alternator manufacturer's test data for THDv (Total Harmonic Distortion, voltage) at rated non-linear load -- not at rated resistive load. Standard alternators achieve 2-3% THDv with resistive loads but 8-15% with non-linear loads. Data center UPS systems typically require <5% THDv at the generator output. Always test with a non-linear load profile during factory load bank testing.

N+1 and 2N Parallel Generator Systems: Sizing and Configuration

Most Tier III and all Tier IV data centers require multiple generators in parallel. The paralleling configuration determines both the redundancy capability and the fuel efficiency of the installation.

N+1 configuration (Tier III minimum):  Two or more generators connected to a common bus bar, with one generator beyond the minimum needed to carry the full load. Example: a data center with 1,000kW critical load requires N+1 with two generators each rated 1,100kW prime -- either unit can carry the full 1,000kW load independently with headroom. When both run, each operates at approximately 45% load -- consider running one at 90% load and keeping the second on hot standby, starting it automatically when demand requires.

2N configuration (Tier IV):  Two completely independent generator systems (A-side and B-side), each at N+1. A 2,000kW critical load requires two independent systems, each of N+1 capacity (e.g., A-side: 2 x 1,100kW; B-side: 2 x 1,100kW). Total installed capacity: 4,400kW for a 2,000kW load. Each system feeds separate UPS modules that independently supply separate PDU branches to the IT load. Any single generator or complete generator system failure cannot interrupt IT power.

  ⚡  Parallel system technical requirements

  Auto-synchroniser: each generator must have an auto-synchroniser that matches voltage, frequency, and phase angle before closing the parallel breaker. Incorrect synchronising causes severe mechanical and electrical damage to both machines. Isochronous load sharing: when running in parallel, all generators must share load equally. Isochronous load sharing governors (electronic) adjust engine fuel delivery to maintain equal load distribution. Mechanical governors cannot share load accurately in parallel. Bus protection relay: a protection relay panel monitors the common bus for fault conditions (short circuit, earth fault, reverse power) and trips individual generators without affecting the others. Specify a dedicated bus protection relay panel -- not just individual generator protection.

Fuel System Design for Data Center Autonomy

Data center fuel system design is driven by the target autonomy -- how long the facility can operate on generator power without fuel resupply. Target autonomy is typically set by the data center operator's SLA obligations and local regulatory requirements.

  ⚠  Fuel polishing is mandatory for large data center fuel storage

  Data center bulk diesel fuel storage tanks accumulate water contamination and microbial growth over time. Unlike industrial or commercial generators that run frequently and draw fresh fuel, data center standby generators may not run for months between tests. Fuel that sits without circulation in a large tank degrades significantly. Specify a continuous fuel polishing system (circulation through water separator and particle filter) for any tank above 5,000 litres. Fuel quality failure at the moment of a real grid outage -- when the generator is actually needed -- is the worst possible time to discover contaminated fuel.

Recommended Engine Platforms for Data Center Applications

Data Center Generator Commissioning: What Must Be Tested

Commissioning a data center generator set is not the same as commissioning a commercial generator. The stakes are higher and the test protocol must verify performance under the specific conditions the facility will experience in operation.

· Cold start test:  om ambient -- record actual time from mains failure signal to generator at rated voltage/frequency; must be <10 seconds; test minimum three consecutive cold starts

· Step load transient test:  0% step loads; record voltage dip and recovery time at each step; verify voltage returns within ±3% within 3 seconds

· Non-linear load test:  inear load for minimum 4 hours; record THDv at 30-minute intervals; verify <5% at all measurement points

· Parallel operation test:   auto-synchronisation; test load sharing accuracy (each generator within ±5% of equal share at all load levels); test bus fault protection trip and recovery

· Fuel system test:  ery failure: run to sub-base tank empty; verify low fuel alarm before shutdown; verify restart sequence

· Remote monitoring test:  verify Modbus data feed to BMS/DCIM; test all alarm conditions

· Commissioning certificate:  eport with all measured values, test dates, and technician qualifications

Leading Power Data Center Generator Sets

We supply generator sets for data center applications from 200kW to 2,500kW. Our data center specification is configured to meet the requirements described in this guide.

· Engine: Cummins QSK series or Volvo Penta TAD series -- electronic isochronous governor; cold start <10 seconds; block heater standard

· Alternator: Stamford PI or UCI series with low-reactance winding -- X''d 0.10-0.12 available; PMG excitation on request; THDv <5% with non-linear load confirmed by test

· Control: ComAp InteliGen NT -- Ethernet, Modbus TCP/IP, SNMP; auto-synchroniser; isochronous load sharing; bus protection relay panel

· Dual battery starting: independent primary and standby battery banks with automatic test and charge monitoring

· Factory test: 4-hour load bank test at 100% rated output; cold start timing test; transient response test; non-linear load THD measurement -- all documented

· Parallel systems: we supply complete parallel packages (auto-synchroniser, load sharing, bus bar, protection relay panel) for N+1 and 2N configurations

· Range: 200kW-2,500kW in data center configuration; parallel packages to any total capacity

· CE certified; ISO 8528 compliant; DCIM/BMS integration documentation provided

  ✔  Data center project support from Leading Power

  For data center projects, we provide a complete technical support package beyond the standard quotation: generator sizing calculation based on IT load, UPS efficiency, and cooling load; N+1 or 2N configuration design with parallel control system specification; fuel system sizing for your target autonomy; site-specific derating for altitude and ambient temperature; and factory-witnessed load bank test with data center-representative non-linear test load. Engage us at the design stage for best results -- generator specification changes made at late design stage are expensive and time-consuming.

Leading Power is a CE-certified diesel generator manufacturer based in Fu'an, Fujian, China. Established in 2008. 200kW-2,500kW data center generator configurations. Cummins QSK and Volvo Penta engines. Stamford low-reactance alternators. ComAp InteliGen NT parallel control. Factory non-linear load test. 24-hour technical response.