In short
- IST is commissioning Level 5: all systems together, at design load, with failures simulated.
- Pass criterion: the IT load never drops and temperatures stay inside the design range.
- Load banks replace the IT: resistive for kW, reactive for power factor, rack-mount heaters for airflow in the white space.
- Simulating 5 MW for 8 hours uses about 40 MWh, about 10,800 l of diesel on gensets, and rejects 5 MW of heat.
What an integrated systems test proves
Lower levels prove that each device and each system works. IST proves that they work together: that the UPS holds the load while the generators start, that the cooling restarts in the right order after a transfer, that controls and fire systems interact as the design intends. It is run with the building's own operators, so it also tests the procedures.
Typical IST scenarios and pass criteria
| Scenario | Expected response | Pass criterion |
|---|---|---|
| Utility failure | UPS on battery, gensets start and take load | No load drop; transfer inside the design window |
| Genset fails to start | Redundant sets share the load | Load carried; alarm at EPMS |
| UPS module fault | Redundant modules or the other path hold load | No transfer to unprotected bypass |
| Utility return | Retransfer to utility | Transition as designed; gensets cool down |
| Cooling unit failure | Standby chiller, CRAH or CDU starts | Supply air or water inside design range |
| Pump or valve failure | Standby pump takes over | Flow and pressure recover |
| Controls or network loss | Plant fails safe and keeps running | No unplanned stop; alarms raised |
| Fire alarm cause and effect | Dampers, shutdowns, releases | Actions match the matrix |
| Black building start | Sequenced restart | Restart without manual help |
The number of scenarios grows with topology: a 2N or distributed-redundant design has more failure paths to prove than an N+1 block. The IST library on the home page lists the same nine.
Load bank types
| Type | Use |
|---|---|
| Resistive (500 kW mobile, 1,000 kW container) | Full kW load on UPS, generators and switchgear; tests at unity power factor |
| Reactive (resistive-inductive or capacitive) | Leading or lagging power factor for generator synchronisation and voltage regulation |
| Rack-mount heater banks (~15 kW) | Simulate servers in the racks, so airflow, containment and liquid loops see real heat |
Large banks at the switchgear test the power train; rack-mount banks in the white space test the cooling. A full IST usually needs both.
Sizing the test, not just the plant
| Quantity | Formula |
|---|---|
| Load bank units | IT load to simulate ÷ unit rating, rounded up |
| Apparent power | kVA = kW ÷ PF |
| Energy during the test | kWh = kW × hours at full load |
| Diesel if gensets carry it | litres ≈ kWh × 0.27 l/kWh (check the genset datasheet) |
| Airflow through air-cooled banks | m³/h = kW ÷ (1.2 × 1.006 × ΔT) × 3,600 |
Simulating a 5 MW hall for 8 hours
| Quantity | Value |
|---|---|
| Units at 1,000 kW (or at 500 kW) | 5 (or 10) |
| Apparent power at PF 1.0 | 5,000 kVA |
| Energy = 5,000 kW × 8 h | 40,000 kWh |
| Diesel = 40,000 × 0.27 | 10,800 l |
| Heat to reject | 5,000 kW ≈ 1,422 RT |
| Airflow through banks at ΔT 15 K | ≈ 994,000 m³/h |
Banks outside the building need space, cabling, a fuel plan and somewhere for the hot air to go. Banks inside the white space put their heat into the cooling system, which is the point of the test, so the cooling must be ready first.
Running the IST day
- Prerequisites: L4 complete, issues log clear of critical items, controls and EPMS trending live.
- Scripts: every scenario has steps, expected responses, pass criteria and sign-off boxes, agreed before the day.
- Safety: a stop procedure, a single test director, radios, and clear isolation of live parts.
- Data: record temperatures, voltages and transfer times continuously; a pass needs evidence, not memory.
- Re-tests: a failed scenario is fixed and re-run, and anything it touched is re-tested.
Questions about IST and load banks
How big should the load banks be?
Large enough to load each hall to its design IT load, in units that can be stepped to test partial loads as well.
Why use reactive load banks?
Generators behave differently at leading or lagging power factor. Reactive banks check synchronisation and voltage regulation that resistive banks cannot show.
How long does an IST take?
Usually one to several days per hall, depending on the number of scenarios and re-tests. The planner estimates it with the other levels.
Is load bank testing repeated after handover?
Yes. Annual load bank tests on generators and UPS are recommended as part of maintenance.
Sources
- ASHRAE Guideline 0; common data center IST practice.
- Diesel 0.27 l/kWh is a typical full-load rate; air 1.2 kg/m³, cp 1.006 kJ/kg·K.
- Worked figures are screening values, not a test plan.