Data Centres
Independent power-systems engineering for hyperscale and enterprise data centres. Tier III / IV, block load, N+1 / 2N, AI / HPC density, runtime-fuel matching.
Hyperscale to enterprise — Tier III / IV compliance, block-load capability, runtime / fuel matching. OE-independent second opinion.
Data-centre back-up power design has changed completely in the last five years. Hyperscale campuses are heading toward 1 GW; AI / HPC workloads are pushing rack density above 50 kW; block-load step demands (>60 %) are normal; hybrid topologies (diesel + gas + BESS) are mainstream. ES4PS provides OE-independent engineering for this new generation of data-centre load — from peer review through full design.
Real problems in data-centre power systems
1. Block-load capability
What does ISO 8528’s “rated load” mean under a single-step demand? Data centres ask for 60–80 % step load during UPS bypass switchover — can the generator’s frequency dip and recovery time tolerate that? Most sizing tools skip this question; they output “rated kVA” and move on.
Our approach: block-load capability calculated against ISO 8528-5 G3 / G4 class transient performance. Where required, we recommend oversizing or a parallel-running strategy.
2. Runtime / fuel matching
Uptime Tier III requires 12 hours of fuel; Tier IV requires 24–72 hours plus a 7-day refill plan. But against what load profile? 100 % rated or 50 % average? Tank sizing on the wrong assumption gives either unnecessary capex (oversized tank) or — in an actual crisis — fuel runs out.
Our approach: site’s actual load profile (PUE, IT load curve, seasonal variation) modelled. Fuel tank standard (EN 12285 / KIWA / BS / UL 142 / API 650) chosen per location — standards-flexible.
3. Hot / cold aisle and genset HVAC coordination
If genset-room HVAC is designed in isolation from IT-room climate, you get exhaust gas recirculation, intake hot spots, radiator fan jet noise consuming the IT-room HVAC budget. Four disciplines (enclosure, acoustics, HVAC, IT climate) must be co-designed.
Our approach: CFD analysis verifies exhaust dispersion and intake hot-spot risk. Custom Python computes recirculation ratio and IT-room load impact.
4. Redundancy topology (N+1 / 2N / 2N+1)
“Is N+1 enough or do we need 2N?” is not a budget question — it is an availability calculation. The answer comes from MTBF + MTTR + commissioning quality + operator training, computed together. Saying “Tier IV = 2N” is not sufficient.
Our approach: reliability block diagram (RBD), single-point-of-failure analysis, common-cause-failure assessment, operational availability calculation (including planned maintenance + human error).
5. Block-load + synchronisation
Multiple gensets in parallel + 5-second N+1 scenario (one unit fails, others redistribute load). Solved with paralleling switchgear + governor response + AVR cross-current compensation.
Our approach: ETAP + DigSilent PowerFactory transient stability analysis. Switchgear ATS make / break strategy. Sync-check criteria design.
6. Rating selection
Should the genset be Standby, Prime, or DCP (Data Centre Power)? Which rating applies under which conditions and how should you choose? Each class has a different usage profile, different annual hour limits, different maintenance interval — wrong selection can lead to uptime letter disputes and warranty rejection.
Our approach: correct rating selection from day one. Avoids uptime-letter friction with manufacturers down the road. Rating selected against load profile + annual run hours + redundancy topology.
Typical engagements
A. Peer review / second opinion (most common)
Customer has a design from their consultant or OE. Independent second pair of eyes: sizing right? Redundancy sufficient? Field surprises waiting? 2–3 week turnaround.
B. Full engineering package
Concept to commissioning: load analysis → sizing → equipment design (enclosure, acoustic, HVAC, fuel, switchgear) → standards compliance → FAT / SAT support. 3–6 months.
C. CFD analysis (standalone service)
Genset-room airflow, exhaust dispersion, intake hot spot. Often used to validate an existing design. 2–4 weeks.
D. Commissioning witness + punch list
Independent third-party witness — OE FAT / SAT verification. Independent sign-off for the customer’s operations team. Per project schedule.
E. Documentation audit
Existing data centre’s documentation health-check — can you answer “why was this calculated this way?” five years later? Traceability rebuild. 4–8 weeks.
Hyperscale + hybrid topology
Hybrid (diesel + gas + BESS)
Hyperscale projects are no longer single-source. BESS (lithium-ion or alternative) handles UPS ride-through and peak shaving. Gas genset (reciprocating) provides base load. Diesel is the last-resort backup. Three technologies need coordinated topology.
Engineering implications:
- 3-source automatic transfer scheme (different from classical ATS)
- Fuel-optimisation algorithm (gas vs diesel runtime cost)
- BESS thermal management (integrated with data-hall HVAC)
- Emission compliance (EU MCPD, US EPA Tier 4)
Note: BESS is currently outside our product segments but we provide system-integration assessment — BESS supplier evaluation, topology, control strategy, NFPA 855 fire-safety compliance.
Standards — data centre specific
| Subject | Standards |
|---|---|
| Sizing & block load | ISO 8528, NFPA 110, IEEE 446, Uptime Institute Tier Standard |
| HVAC (genset + IT room) | ASHRAE TC 9.9 Thermal Guidelines, ASHRAE 90.4 (DC energy) |
| Acoustic (urban DC) | ISO 3744, EU Outdoor Noise Directive 2000/14, local regulations |
| Fuel tank | EN 12285 / KIWA-BRL-K903 / BS 799 / UL 142 / API 650 (per location) |
| Switchgear / ATS | IEC 60947, IEEE C37, UL 1008, NEMA ICS |
| Arc flash | IEEE 1584, NFPA 70E |
| Commissioning | ASHRAE Guideline 0/1.1, NEBB, Uptime Tier Certification |
| Sustainability | EU CSRD, ISO 50001, LEED / BREEAM |
| HSE (DC-specific) | Arc-flash PPE (NFPA 70E), hot-aisle thermal exposure, lone worker, risk assessments, HAZOP analysis |