Fast Frequency Response Potential of Data Centers Through Workload Modulation and UPS Coordination

Overview

The study asks whether data centres can help stabilise frequency in low-inertia systems thanks to controllable IT workloads and on-site UPS systems. A dynamic model combining data-centre power consumption and grid frequency dynamics was developed; frequency-deviation control strategies adjust server power and discharge UPS batteries during frequency events. The 20 MW data centre is integrated at Bus 18 in Area 1 of the IEEE 39-bus system.

Model parameters: server droop gain 10 MW/Hz, UPS droop gain 15 MW/Hz (aggregate 25 MW/Hz), UPS energy 5 MWh, response time constant 0.1 s, discharge/charge efficiency 0.95/0.9, PUE 1.2, deadband ±0.03 Hz. The authors state UPS response time is typically below 100 ms. After a 200 MW generation trip (H = 2 s), the nadir was 59.66 Hz with no data-centre response, 59.78 Hz with UPS only and 59.85 Hz with coordinated UPS + IT response; recovery was about 1.6 s faster.

In the coordinated case the UPS discharged up to +8 MW within 0.2 s while IT throttling cut 4 MW; measured total response latency was 0.15 s. The coordinated strategy cut nadir deviation by 57% and recovery time by 34%, with task delay below 2%. The paper says UPS batteries can sustain multiple FFR activations per hour without compromising backup if SOC is kept within 20-80%.

Limits: the model is idealised (a first-order data-centre block) and some reported energy figures are inconsistent: the 1.65 MWh (33.3% of 5 MWh) discharged in the first 10 s does not match the stated ~8 MW peak discharge. Results rest on one disturbance scenario and one test system, with no validation on a real data centre.

Key points

1. Model parameters: 20 MW data centre, server droop gain 10 MW/Hz, UPS droop gain 15 MW/Hz (aggregate 25 MW/Hz), UPS energy 5 MWh, response time constant 0.1 s, discharge/charge efficiency 0.95/0.9, PUE 1.2, deadband ±0.03 Hz.
2. The authors state UPS response time is typically below 100 ms, suitable for the first stage of fast frequency response.
3. After a 200 MW generation trip (H = 2 s), the nadir was 59.66 Hz with no response, 59.78 Hz with UPS only and 59.85 Hz with coordinated UPS + IT response; recovery was about 1.6 s faster.
4. In the coordinated case the UPS discharged up to +8 MW within 0.2 s while IT throttling cut 4 MW; total response latency was 0.15 s.
5. The coordinated strategy cut nadir deviation by 57% and recovery time by 34%, with task delay below 2%.
6. The paper says UPS batteries can sustain multiple FFR activations per hour without compromising backup if SOC stays within 20-80%, but it also reports 1.65 MWh (33.3% of 5 MWh) discharged in the first 10 s, which does not match the stated ~8 MW peak.

Using it in design

• For a UPS that will provide grid services, define the SOC window (for example 20-80%) and show by calculation that backup autonomy is preserved inside it.
• Keep UPS droop and server throttling gains as separate control inputs, and verify response latency (0.15 s) and the task-delay limit (2%) together.
• Check that MW peak discharge and discharged MWh are consistent in your own energy budget, and do not repeat the inconsistency found in the paper.

Who should read it

Power-system engineers and data-centre energy managers evaluating UPS and workload flexibility for grid frequency services.

Document contents

1. 1. Introduction
2. 2. Fast Frequency Response (2.1 Concept and Motivation, 2.2 Power System Frequency Dynamics, 2.3 Primary and Fast Frequency Response Control, 2.4 Integration of Data-Center-Based FFR, 2.5 Analytical Insights)
3. 3. Data Centers (3.1 Power Composition and Load Characteristics, 3.2 IT Server Load Modulation, 3.3 UPS Energy Storage Coordination, 3.4 Coordinated Control Strategy, 3.5 Integration into System Model)
4. 4. Case Study (4.1 Test System Description, 4.2 Simulation Scenarios, 4.3 Parameter Configuration, 4.4 Performance Metrics, 4.5 Implementation Notes)
5. 5. Simulation Results and Discussion (5.1 Frequency Response Performance, 5.2 Data Center Power Dynamics, 5.3 UPS State-of-Charge Evolution, 5.4 Sensitivity to Control Parameters, 5.5 Quantitative Performance Comparison, 5.6 Discussion)
6. 6. Conclusion and Future Work

Related standards

• IEC 62040-3
• IEEE 1547

Read the document

Publication details

Authors

Xiaojie Tao; Rajit Gadh

Publisher

IEEE Access (Institute of Electrical and Electronics Engineers (IEEE))

Year

2025

Document number

arXiv:2512.14128

Volume

13

Pages

214511-214519

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.1109/access.2025.3646120

Licence

CC BY 4.0 — share with attribution

Licence evidence

arxiv.org

Citation