What we make of it

Problem

As the share of synchronous generators in the power system declines, the rotational inertia they provide also decreases. The same power imbalance may therefore produce a higher initial rate of change of frequency, motivating research into control algorithms that enable inverter-based generation to provide virtual inertia and frequency support. More recently, researchers have also investigated the use of power-electronic resources on the load side to support the grid [1], [2].

This approach is becoming increasingly important as large loads, such as data centers, expand. Their power consumption and response to disturbances can substantially affect the surrounding grid. In particular, a load that maintains constant power consumption tends to increase its input current when the voltage falls. Depending on grid conditions, this response can deepen the voltage drop or reduce the stability margin. This behavior is known as the constant-power load (CPL) characteristic and exhibits negative incremental resistance in the relationship between voltage and current changes, as illustrated in Fig. 1. The characteristic applies within the operating range in which constant-power regulation is maintained; the response can change when current limits or protection functions intervene [3].

To address these load characteristics, control strategies are being developed to use the power-electronic resources embedded in large loads to mitigate their impact on the grid and contribute to grid support.

Approach

Research on grid support through load-side power-electronic interfaces has taken several forms [4]. These approaches can be divided into two cases: one in which changes in power drawn from the grid are reflected in load power consumption, and one in which the power drawn from the grid changes while the load's power demand continues to be met. In the first case, the load adjusts its consumption to accommodate changes in grid input power. In the second, an energy storage system compensates for the difference between grid input power and load power. Figure 2 shows the two configurations.

The first case, in which changes in grid input power are reflected in load power consumption, is shown in Fig. 3. The grid-side converter adjusts the power drawn from the grid through grid-support control, such as droop control, and inner voltage and current control loops. The load-side converter adjusts load power consumption to suppress DC-link voltage deviations caused by the mismatch between incoming power and consumed power. Thus, load consumption must decrease when grid input power decreases and increase when it increases, maintaining the power balance at the DC link [5].

During this process, the DC-link capacitor buffers the instantaneous power mismatch, whereas sustained changes in grid input power must be accommodated by continuously adjusting load consumption. The approach is therefore limited by the load's allowable power range and response capability. The detailed inner-loop arrangement in Fig. 3 depends on the control implementation. The focus here is on the load-side converter accommodating the power variation to maintain the DC-link voltage.

In the second case, in which changes in grid input power are not instantaneously reflected in load power consumption, a battery energy storage system (BESS) mainly compensates for the power difference. Adjusting grid input power while still meeting the load's power requirement calls for another component to supply the deficit or absorb the surplus. The configuration in Fig. 4 connects a BESS to the DC link through a DC/DC converter for this purpose [6], [7].

As in the first case, the grid-side converter adjusts the power drawn from the grid to provide grid support. In this configuration, however, the BESS converter is primarily responsible for maintaining the DC-link voltage, rather than the load-side converter. To satisfy load demand consistently, a reduction in grid input power is accommodated by modulating the battery's charging and discharging power. An increase in grid input power requires the opposite adjustment. This allows a mismatch between grid input power and load power, while the load's original operating requirements continue to be met. Maintaining load power here means avoiding an additional change in consumption for grid support; it does not exclude changes in the load's own workload.

Grid support using load-side resources can therefore be understood as an arrangement in which the grid-side converter adjusts grid input power, while either the load or the BESS accommodates the resulting power mismatch to maintain the DC-link voltage. This approach can also be considered for data centers equipped with uninterruptible power supplies (UPSs) and batteries [8], [9]. The Pacific Northwest National Laboratory (PNNL) report Electromagnetic Transient Modeling of Large Data Centers for Grid-Level Studies (PNNL-38817, Alpha Release, December 2025) presents data center power architectures with AC and DC distribution. A double-conversion UPS in an AC distribution system comprises an input rectifier, a DC link, and an output inverter, with the battery connected on the DC side. In a DC distribution system, the output DC/AC inverter can be omitted, while energy storage can still be connected on the DC side. The architectures in Fig. 5 provide a basis for considering the second approach, in which storage compensates for the difference between supply and demand while load operation is not disturbed [10].

Nevertheless, structural similarity between a published configuration and a data center power system does not mean that the same control scheme can be applied without modification. Data center power consumption can change with workload or demand response, while the supply voltage and frequency for AC loads and the supply voltage for DC loads must remain within their respective allowable ranges. The UPS must therefore provide reliable power to the IT load while enabling grid input power to be adjusted for grid support [11].

Because both grid disturbances and load variations can cause a power imbalance within the UPS, the grid-support function and the function that maintains the load supply must be coordinated to avoid conflicting actions. This requires clear allocation of responsibilities for power sharing and DC-link voltage control among the grid-side converter, load-side converter, and BESS, together with consideration of the battery's available energy and its reserve for backup power. Grid-support performance and continuity of load operation must be assessed together under these conditions to determine whether the approach is suitable for practical data center applications [12].

Takeaways

Changes in grid input power for grid support can be accommodated either by adjusting load consumption directly or by using energy storage to absorb the difference while load operation is not disturbed. Data centers equipped with UPSs and batteries provide a structural basis for considering the second approach. In practical applications, the roles of the converters and the BESS must be defined with both grid input power adjustment and reliable load supply in mind, so that both functions can operate stably within the available power and energy limits.

References

1. P. Denholm, T. Mai, R. W. Kenyon, B. Kroposki, and M. O’Malley, “Inertia and the power grid: A guide without the spin,” Nat. Renewable Energy Lab., Golden, CO, USA, Tech. Rep. NREL/TP-6A20-73856, May 2020. [Online]. Available: https://www.nrel.gov/docs/fy20osti/73856.pdf
2. H. Jain, B. Mather, A. K. Jain, and S. F. Baldwin, “Grid-supportive loads—A new approach to increasing renewable energy in power systems,” IEEE Trans. Smart Grid, vol. 13, no. 4, pp. 2959–2972, Jul. 2022, doi: 10.1109/TSG.2022.3153230.
3. A. Emadi, A. Khaligh, C. H. Rivetta, and G. A. Williamson, “Constant power loads and negative impedance instability in automotive systems: Definition, modeling, stability, and control of power electronic converters and motor drives,” IEEE Trans. Veh. Technol., vol. 55, no. 4, pp. 1112–1125, Jul. 2006, doi: 10.1109/TVT.2006.877483.
4. Q. Sun, J. Hu, R. Wang, and H. Zhang, “Source-load cooperative frequency regulation based on virtual synchronous/asynchronous machine concept,” Energy Internet, vol. 1, no. 2, pp. 188–199, 2024, doi: 10.1049/ein2.12012.
5. O. Gomis-Bellmunt, S. Dadjo Tavakoli, V. Albernaz Lacerda, and E. Prieto-Araujo, “Grid-forming loads: Can the loads be in charge of forming the grid in modern power systems?,” IEEE Trans. Smart Grid, vol. 14, no. 2, pp. 1042–1055, Mar. 2023, doi: 10.1109/TSG.2022.3202646.
6. B. Wang, R. Burgos, and B. Wen, “A grid-supporting rectifier with frequency and voltage regulation capabilities,” in Proc. IEEE Energy Convers. Congr. Expo. (ECCE), 2023, pp. 590–597, doi: 10.1109/ECCE53617.2023.10362109.
7. Q.-C. Zhong and Z. Lyu, “Droop-controlled rectifiers that continuously take part in grid regulation,” IEEE Trans. Ind. Electron., vol. 66, no. 8, pp. 6516–6526, Aug. 2019, doi: 10.1109/TIE.2018.2863193.
8. A. Azizi et al., “Strengthening data center operations using grid-forming battery energy storage as a line-interactive uninterruptible power supply,” Int. J. Electr. Power Energy Syst., vol. 175, 2026, Art. no. 111638, doi: 10.1016/j.ijepes.2026.111638.
9. S. Kim, S. Jung, J. Lee, S. Oh, and J. Shin, “Development of fast response technology using data center on-line UPS,” Trans. Korean Inst. Electr. Eng., vol. 74, no. 11, pp. 1817–1823, Nov. 2025 (in Korean), doi: 10.5370/KIEE.2025.74.11.1817.
10. B. A. Ross and J. Follum, “Electromagnetic transient modeling of large data centers for grid-level studies,” Pacific Northwest Nat. Lab., Richland, WA, USA, Rep. PNNL-38817, Alpha Release, Dec. 2025. [Online]. Available: https://www.energy.gov/sites/default/files/2026-01/Data_Center_EMT_Models.pdf
11. K. Kaur, S. Garg, N. Kumar, G. S. Aujla, K.-K. R. Choo, and M. S. Obaidat, “An adaptive grid frequency support mechanism for energy management in cloud data centers,” IEEE Syst. J., vol. 14, no. 1, pp. 1195–1205, Mar. 2020, doi: 10.1109/JSYST.2019.2921592.
12. Y. Lin et al., “Research roadmap on grid-forming inverters,” Nat. Renewable Energy Lab., Golden, CO, USA, Tech. Rep. NREL/TP-5D00-73476, Nov. 2020. [Online]. Available: https://www.nrel.gov/docs/fy21osti/73476.pdf