| Data centres
to the operation of the data centre and sensitive to voltage disturbances. To ride through voltage disturbances on the electric grid, data centres employ uninterruptible power supply (UPS) systems that will instantaneously take over providing power to the data centre equipment when a grid disturbance occurs.
The differing types and designs of these UPS systems cause differences in the characteristics of the data centre responses to a voltage disturbance.
A centralised design typically uses UPS systems at the load centre level that are in the range of 2–5 MW. The UPS uses power electronics to switch the load to a battery bank connected to the UPS. These battery banks are not designed to supply the load for long periods but rather to power the load for the relatively short duration of a disturbance, or — in the case of a complete electric grid outage — long enough to start a backup generator that will then provide power to the UPS.
A decentralised UPS design uses many smaller UPS systems at the rack level. These rack- mounted UPS systems are typically in the range of 3–4 kW.
Decentralised UPS systems operate in a similar way to the centralised UPS, just on a smaller scale.
Another type of UPS is a dynamic/diesel rotary uninterruptible power supply (DRUPS). See Figure 4. These systems use a flywheel to provide uninterruptible power and a clutch system to quickly start and connect a diesel engine upon encountering a disturbance in the electric grid. The load characteristics of these types of UPS systems differ in response to a transient disturbance on the electric grid.
For the static centralised and decentralised UPS systems that utilise batteries, the load will be taken over by the battery when a transient voltage disturbance occurs. Since it is a transient disturbance, such as a temporary fault on the electric grid, the grid voltage will typically return to normal in milliseconds.
Once the grid voltage returns to normal, the load will then be transferred back to the grid. Upon detecting a transient voltage disturbance, a DRUPS system will immediately transfer the load to the flywheel/ac generator and start the engine, which will act as the prime mover for the generator before the flywheel exhausts its kinetic energy.
This system will not quickly transfer the load 246 244 242 240 238 236 234 232 7/10/2024 7:02:00 PM
19:00
19:05
19:10 Figure 3. 230 kV line voltages (source: NERC)
back to the grid after the transient disturbance has cleared and the grid voltage returns to normal. Typically, transferring the load back to the grid from the DRUPS system must be done manually.
Summarising, the typical static UPS system load characteristic, as seen by the grid, is a short- duration loss of load that returns quickly after the transient disturbance clears. In contrast, the typical DRUPS system load characteristic, as seen by the grid, is a loss of load that does not return quickly after the transient disturbance clears. NERC’s discussions with the data centre owners/operators also identified another protection/control scheme that impacts the response of data centre load to voltage disturbances on the grid. This scheme detects and counts voltage disturbances on the grid. If a certain number of voltage disturbances are seen within a given time, the data centre will transfer its load to the backup system, and it will remain there until it is manually reconnected to the grid. The typical number of voltage disturbances that trigger this scheme is three, and a typical time is one minute. As such, three voltage disturbances within one minute will result in data centres using this protection/control scheme, transferring their load off the grid and staying off until they manually transfer back. This scheme can be deployed with both centralised and decentralised UPS designs.
While the three load characteristics described above were predominant in the July 2024 event,
Figure 4. Rolls-Royce mtu DRUPS (Kinetic PowerPack) (source: Rolls-Royce mtu)
many variations in load characteristics exist. These characteristics are determined by the numerous vendor-supplied controls within a data centre, including vendor-specific UPS controls. Additionally, controls under the purview of data centre owner/operators, such as responding to a particular number of disturbances within a certain period, also determine the characteristic response to system disturbances. In the case of the July 2024 incident, most of the sustained load reduction occurred simultaneously with the third voltage depression, which coincided with the third automatic reclosing attempt. At that time, approximately 1260 MW of load dropped off the electric grid and did not return for hours. Most of the load loss in this event can be attributed to the interaction between the automatic reclosing sequence on the faulted transmission line and a data centre protection/control scheme that counts the number of voltage disturbances within a specified period.
While the July 2024 incident did not present any significant issues in terms of the reconnection of the loads, the potential exists for issues in future incidents if the load is not reconnected in a controlled manner. Significant amounts of load being reconnected to the system presents challenges to balancing authorities and transmission system operators. Ramp rates for load connection are just as critical to system operations as generation ramping, says NERC.
Understanding load
The July 2024 incident has also highlighted potential reliability risks to the Bulk Electric System with respect to the voltage ride-through characteristics of large data centre loads. Similar incidents have occurred in other US Interconnections with cryptocurrency mining loads as well as oil/gas loads. While these loads are different from the data centre loads of the July 2024 incident, they present the same challenges to operators and planners of the Bulk Electric System. Understanding the changing dynamic nature of load is critical to its future operation, concludes NERC.
www.modernpowersystems.com | May 2025 | 27
19:15
19:20
19:25
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49
