📋 Case Study
Data Center Electrical Design: 40 MW Hyperscale Facility in Singapore
Selecting optimal cable sizes for 2×20 MW primary feeders (20 kV) and critical 400 V bus duct/cable trunking systems while meeting IEC 60502-2 and CP 33:2021 requirements—balancing ampacity, voltage drop (<1%), short-circuit withstand (1 s, 50 kA), installation constraints (limited underground duct bank space), and lifecycle cost under tropical ambient conditions.
🏗️ Project Overview
A 40 MW hyperscale data center located in Jurong East, Singapore, designed to support high-density AI and cloud workloads. The facility comprises eight server halls, each fed by dual 20 kV utility feeds, with redundant 10 kV medium-voltage switchgear and 400 V low-voltage distribution. Ambient temperature averages 31°C year-round, with high humidity (75–85% RH), impacting thermal derating.
🎯 Challenge
Selecting optimal cable sizes for 2×20 MW primary feeders (20 kV) and critical 400 V bus duct/cable trunking systems while meeting IEC 60502-2 and CP 33:2021 requirements—balancing ampacity, voltage drop (<1%), short-circuit withstand (1 s, 50 kA), installation constraints (limited underground duct bank space), and lifecycle cost under tropical ambient conditions.
🔧 Design Approach
Iterative ampacity modeling using IEC 60287-1-1 with site-specific thermal resistivity (ρt = 1.2 K·m/W for compacted clay-sand mix), derating for grouping (3 circuits per duct bank), ambient temperature (40°C design basis), and harmonic content (THD ≤ 8% at LV). Employed Neher-McGrath thermal circuit modeling validated via ETAP v22.1 transient thermal simulation. Evaluated copper vs. aluminum conductors, single-core vs. multi-core, and alternative insulation (XLPE vs. EPR) under fire safety (IEC 60331) and smoke toxicity (IEC 61034) mandates.
📐 Design Diagram
AI-generated project design illustration
📐 Key Calculations
Adjusted Ampacity for 20 kV Feeder
I_z = I_t × k1 × k2 × k3 × k4 × k5
Result: 1,842 A
Accounts for ambient (k1=0.84), grouping (k2=0.82), soil thermal resistivity (k3=0.95), burial depth (k4=1.03), and harmonic derating (k5=0.98); ensures continuous 20 MW/20 kV feeder operation without thermal overload.
Voltage Drop at Full Load (400 V Bus Duct)
ΔU = √3 × I_b × L × (R' × cosφ + X' × sinφ) / 1000
Result: 0.78 %
Confirms compliance with CP 33:2021 <1% limit for critical IT loads; avoids undervoltage tripping of UPS inverters and server PSUs.
Short-Circuit Withstand (1 s, 50 kA)
S_min = I_sc × √t / k
Result: 226 mm² (copper, k=143)
Minimum cross-section required to avoid thermal failure during fault; drives selection of 300 mm² Cu XLPE cables (exceeding minimum by 33%) for margin and future capacity.
📊 Results
Metrics: Cable size reduced by 18% vs. conservative default tables, Projected CAPEX savings: SGD 2.1M across HV/LV systems, Thermal margin increased from 2.1°C to 7.4°C at peak load, Voltage drop compliance achieved at 92% loading without active compensation
Optimized 20 kV feeders use 300 mm² Cu/XLPE single-core cables (vs. baseline 400 mm²), and 400 V distribution employs 2×(1×630 mm² Cu/XLPE) parallel runs per phase—achieving full 40 MW capacity within footprint constraints, 12% lower conductor mass, and verified thermal stability under Singapore’s tropical conditions.
💡 Lessons Learned
- •Soil thermal resistivity field measurement is non-negotiable—assumed ρt values caused 11% ampacity overestimation in preliminary models
- •Harmonic current contribution (especially 5th/7th) must be included in IEC 60287 derating; omitting it risked 6.3°C excess conductor temperature
✅ Key Takeaways
- 1Ampacity optimization in tropical data centers requires integrated thermal modeling—not tabular derating—and must co-validate voltage drop, fault rating, and fire performance simultaneously.