Land-Based Backup Generator Cooling for Arctic Data Center

Scenario

Project Type: Critical infrastructure cooling design for a Tier IV data center in Utqiaġvik, Alaska (71.3°N). Location Context: Extreme cold (−45°C winter lows), permafrost foundation limiting buried piping depth, and reliance on glycol-water mix (30% propylene glycol) due to freeze protection requirements. Constraints: No onsite water source — all coolant must be closed-loop recirculated; pump must operate continuously at −30°C ambient without preheat; maximum allowable pressure drop ≤ 80 kPa to preserve low-energy chiller integration.

Given Data

• Heat Load: 132 kW (peak exhaust heat rejection from 2 MW diesel backup generator)
• Specific Heat Capacity of Water: 3.92 kJ/(kg·K) (measured value for 30% propylene glycol solution at 10°C mean loop temp)
• Temperature Difference: 6.2 K (increased from standard 5 K to reduce flow velocity and mitigate erosion-corrosion in low-temp glycol)
• Density of Water: 1032.1 kg/m³ (glycol-water mixture density at 10°C)
• Gravitational Acceleration: 9.81 m/s²
• Pump Efficiency: 68% (reduced due to viscous losses in glycol blend and oversized motor for cold-start torque)

Calculation

Using the same thermal balance formula:

Flow Rate (m³/h) = (Heat Load × 3600) / (Specific Heat Capacity × Density × Temperature Difference × Pump Efficiency)

• Numerator = 132 × 3600 = 475,200 kJ/h
• Denominator = 3.92 × 1032.1 × 6.2 × 0.68 ≈ 16,942.1
• Flow Rate = 475,200 / 16,942.1 ≈ 28.05 m³/h

Rounded per tool precision: 28.05 → 28.05 m³/h

Result and Decision

A magnetically coupled, sealless canned-motor pump (Grundfos MAGNA3 65-160/C) was specified — rated 28.5 m³/h @ 12.5 m head, 68% efficiency at design point, with -40°C-rated elastomers and integrated variable-speed drive synchronized to generator load. Piping used insulated, double-jacketed stainless steel with trace heating only at valve manifolds. Commissioning verified stable flow ±1.2% over 72-hr continuous test at −32°C ambient.

Lesson

Glycol concentration drastically alters both specific heat capacity and density — using pure-water defaults would underestimate required flow by ~18% and risk thermal overload. Always validate fluid properties at actual operating temperature, not reference tables at 20°C.