Modeling Framework for Data Center Saurav Talukdar,1 Shilpa Marti,1 Kumaraguru Prabakar,2 and Deepthi Vaidhynathan2 1 Verrus LLC 2 National Laboratory of the Rockies The National Laboratory of the Rockies is a national Technical Report laboratory of the U.S. Department of Energy, NLR/TP-2C00-97716 Office of Critical Minerals and Energy Innovation, February 2026 operated under Contract No. DE-AC36-08GO28308. This report is available at no cost from the National Laboratory of the Rockies (NLR) at www.nlr.gov/publications. Modeling Framework for Data Center Saurav Talukdar,1 Shilpa Marti,1 Kumaraguru Prabakar,2 and Deepthi Vaidhynathan2 1 Verrus LLC 2 National Laboratory of the Rockies Suggested Citation Talukdar, Saurav, Shilpa Marti, Kumaraguru Prabakar, and Deepthi Vaidhynathan. 2026. Modeling Framework for Data Center. Golden, CO: National Laboratory of the Rockies. NLR/TP-2C00-97716. https://www.nlr.gov/docs/fy26osti/97716.pdf. The National Laboratory of the Rockies is a national Technical Report laboratory of the U.S. Department of Energy, NLR/TP-2C00-97716 Office of Critical Minerals and Energy Innovation, February 2026 operated under Contract No. DE-AC36-08GO28308. National Laboratory of the Rockies 15013 Denver West Parkway This report is available at no cost from the National Laboratory Golden, CO 80401 of the Rockies (NLR) at www.nlr.gov/publications. 303-275-3000 • www.nlr.gov This work was authored in part by the National Laboratory of the Rockies for the U.S. Department of Energy (DOE), operated under Contract No. DE-AC36-08GO28308. Support for the work was provided by Verrus LLC under TSA2431261. This work was also supported by the Laboratory Directed Research and Development (LDRD) Program at the National Laboratory of the Rockies. The views expressed herein do not necessarily represent the views of the DOE or the U.S. Government. This report is available at no cost from the National Laboratory of the Rockies (NLR) at www.nlr.gov/publications. U.S. Department of Energy (DOE) reports produced after 1991 and a growing number of pre-1991 documents are available free via www.OSTI.gov. Cover photos (clockwise from left): Josh Bauer, National Laboratory of the Rockies 61725; Visualization from National Laboratory of the Rockies Insight Center; Getty-181828180; Agata Bogucka, National Laboratory of the Rockies 91683; Dennis Schroeder, National Laboratory of the Rockies 51331; Werner Slocum, National Laboratory of the Rockies 67842. The National Laboratory of the Rockies prints on paper that contains recycled content. List of Acronyms AI artificial intelligence BESS battery energy storage system CBEMA Computer and Business Equipment Manufacturers Association CHIL controller-hardware-in-the-loop DER distributed energy resource EMT electromagnetic transient ERCOT Electric Reliability Council of Texas IEEE Institute for Electrical and Electronics Engineers IT information technology ITIC Information Technology Industry Council NERC North American Electric Reliability Corporation p.u. per unit POI point of interconnection SCADA supervisory control and data acquisition UPS uninterruptible power supply iii Executive Summary This report highlights the critical need for advanced modeling of data centers due to their rapidly increasing energy consumption and their impacts on power grid reliability. Driven by the demand for artificial intelligence (AI) applications, data centers are projected to consume a significant portion of U.S. energy by 2028, stressing an already challenged grid. This paper emphasizes the importance of “fast” timescale models to understand the dynamic interactions between data centers and the grid, especially given the rapid power fluctuations of AI workloads. It outlines a modeling framework that includes both offline and real-time electromagnetic transient domain simulations, detailing the necessary representations for various components, such as utility interfaces, transformers, information technology loads, uninterruptible power supply, cooling loads, battery energy storage systems, generators, protection systems, and higher-level control systems. Although standard simulation tools like PSCAD offer basic models, custom development is often required to accurately capture the unique and fast-changing behaviors of modern data centers. The paper also discusses key metrics and test cases for validating these models, focusing on transient load responses, protection relay coordination, and demand flexibility. Finally, it addresses the challenges of modeling large-scale data centers, such as computational complexity and the trade-off between model fidelity and practicality, suggesting hybrid modeling approaches as a solution. The overarching goal is to create a robust framework that helps assess data center impacts on grid stability, identify vulnerabilities, and inform the development of standards for these reliable integration of these large loads into the bulk power system. iv Table of Contents 1 Introduction ........................................................................................................................................... 1 2 Design/Operational Philosophy of Data Centers .............................................................................. 3 3 Modeling Framework Overview ........................................................................................................... 8 4 System Modeling Layers ..................................................................................................................... 9 4.1 Utility Interface and Medium-Voltage Bus ................................................................................... 9 4.1.1 Standard PSCAD Models Available ................................................................................ 9 4.1.2 Additional Model Requirements ...................................................................................... 9 4.2 Transformers ................................................................................................................................. 9 4.2.1 Standard PSCAD Models Available .............................................................................. 10 4.2.2 Additional Model Requirements .................................................................................... 10 4.3 Information Technology Loads ................................................................................................... 10 4.3.1 Standard PSCAD Models Available .............................................................................. 10 4.3.2 Additional Model Requirements .................................................................................... 10 4.4 Uninterruptible Power Supply ..................................................................................................... 11 4.4.1 Standard PSCAD Models Available .............................................................................. 11 4.4.2 Additional Model Requirements .................................................................................... 11 4.5 Cooling Loads ............................................................................................................................. 11 4.5.1 Standard PSCAD Models Available .............................................................................. 11 4.5.2 Additional Model Requirements .................................................................................... 11 4.6 Battery Energy Storage System................................................................................................... 12 4.6.1 Standard PSCAD Models Available .............................................................................. 12 4.6.2 Additional Model Requirements .................................................................................... 12 4.7 Generator Models ........................................................................................................................ 12 4.7.1 Standard PSCAD Models Available .............................................................................. 13 4.7.2 Additional Model Requirements .................................................................................... 13 4.8 Protection Systems ...................................................................................................................... 13 4.8.1 Standard PSCAD Models Available .............................................................................. 14 4.8.2 Additional Model Requirements .................................................................................... 14 4.9 Higher-Level Control Systems .................................................................................................... 14 4.9.1 Standard PSCAD Models Available .............................................................................. 14 4.9.2 Additional Model Requirements .................................................................................... 15 4.10 Automatic Transfer Switches/Static Transfer Switches .............................................................. 15 4.10.1 Standard PSCAD Models Available .............................................................................. 15 4.10.2 Additional Model Requirements .................................................................................... 15 5 Key Metrics/Test Cases for Model Evaluation ................................................................................. 17 6 Challenges and Considerations ........................................................................................................ 21 6.1 Computational Complexity ......................................................................................................... 21 6.2 Model Fidelity vs. Practicality .................................................................................................... 21 6.3 Trade-Off Decisions .................................................................................................................... 21 7 Conclusion .......................................................................................................................................... 23 References ................................................................................................................................................. 24 Appendix .................................................................................................................................................... 28 v List of Figures Figure 1. Overview of data center power system architectures across various entities. ............................... 5 Figure 2. ITIC curve ..................................................................................................................................... 6 List of Tables Table 1. Key Metrics and Test Cases for Model Evaluation ...................................................................... 17 Table A-1. Test Case Description and Technical Relevance for User Reference....................................... 28 vi Introduction There is tremendous activity in data center interconnection requests throughout the United States, with some utilities having queues exceeding their total current operational capacity by a few multiples. This is fueled by a significant demand for data center computing capacity for artificial intelligence (AI) applications. A recent report from Lawrence Berkeley National Laboratory (Shehabi et al. 2024) projects data center energy consumption to be 6.7−12% of the total U.S. energy consumption by 2028, compared to 4.4% of the total U.S. energy consumption in 2025. The Long-Term Reliability Assessment report from the North American Electric Reliability Corporation (NERC) (2024) identifies that most of the North American bulk power system faces significant resource adequacy challenges in the next 10 years due to surging energy demand, the retirement of thermal generators, and the time to completion of new power plant projects. In most locations, power grids have limited available capacity to support data center campuses ranging from several hundred megawatts to the gigawatt scale (or, as NVIDIA refers to them, AI factories). To address these capacity constraints in the near term, there are significant conversations around data center load flexibility. A 2025 report by Duke University’s Nicholas Institute estimates that modest curtailment of only 0.25–1% of annual consumption during critical hours could enable the grid to absorb up to 100 GW of new load without the need for new generation or other major upgrades (Norris et al. 2025). To mitigate the shortage of grid power, certain data center campuses are colocating energy generation capabilities, with the possibility of selling excess generation back to the grid (Bloomberg 2025). Some hyperscalers are demonstrating their load flexibility capabilities through workload scheduling (Terrell 2025) or through virtual power plant contracts (Voltus 2025). A recent PJM stakeholder review called out the need for load flexibility, the use of backup power generation in data center campuses, and load curtailment as options for assisting the grid (PJM 2025). In summary, the rapid demand for data centers is already stressing the grid and poses questions about how to continue to operate the grid reliably. The interaction between grid-side events and data center power systems at various timescales needs to be better understood, as data centers are on track to be a significant load on the grid. This understanding would help inform interconnection studies, policies, and regulations to assist with grid reliability. There are already new regulations and interconnection policies; a notable one is SB6 in the Electric Reliability Council of Texas (ERCOT) territory, which can cause data centers to be shut down during grid emergency periods. Data center load has significantly evolved from being central processing unit-heavy uncorrelated workloads to being graphics processing unit-dominated and coordinated workloads. These workloads can go from an idle power state (10–20% of rated power) to a peak power state (80– 100% of rated power) within 10–100 milliseconds, exposing an oscillatory point of interconnection (POI) demand signal to the grid. AI workload frequencies fall into the subsynchronous regime. These can excite resonant modes in transmission networks, potentially leading to subsynchronous resonance or interarea oscillations. Further, rapid, cyclical power swings can lead to voltage flicker that is visible on lighting systems and frequency modulations that impact grid frequency regulation, particularly in isolated or constrained regions. A recent 1 NERC incident review (NERC 2025) reports approximately 1,500 MW of data center loads disconnecting from the grid due to short-duration grid faults (solely due to data center protection schemes) causing overvoltages (to 1.07 per unit [p.u.]) and overfrequencies (60.047 Hz), thus exposing risks of wide-area outages. With significant expansion on the demand and supply side, including behind-the-meter generation or energy storage capabilities to come, along with dynamic workloads, it is important to view this challenge from the lens of grid reliability. There is a clear need to put together a framework to specify ramp rates, acceptable oscillations, and allowable protection schemes, etc., that would assist in sustaining reliable grid operation and not exacerbate grid outage risks. This necessitates the need for power system studies that can model grid and data center dynamics/interactions in a unified framework. To enable the study of data center and grid interactions, a high-fidelity power system model of the data center needs to be developed, particularly one that can cater to the fast-timescale needs of dynamic AI loads and fast-timescale grid instability events (such as those reported in the NERC incident review [NERC 2025]). Two types of modeling frameworks are necessary to evaluate the impacts of data centers on grid stability: 1. Offline electromagnetic transient (EMT) domain simulations (Badrzadeh n.d.) (using tools such as PSCAD and EMTP) of the data center power system with appropriate representations of the grid 2. Real-time EMT domain simulations (in a controller-hardware-in-the-loop [CHIL] framework [IEEE 2025]) with the inclusion of the data center and grid-side controls. In this regard, a technical report from the National Laboratory of the Rockies performed EMT domain CHIL simulations for the 70-MW data center design of Verrus (Vaidhynathan et al. 2025). 2 1 Design/Operational Philosophy of Data Centers Data center designs are entirely focused on providing a high uptime for computing/information technology (IT) (Uptime Institute n.d.) and void any consideration for demand flexibility and grid resilience. A typical data center uptime requirement can range from 99% to 99.999% (leading to appropriate tier categories) (Uptime Institute n.d.), presenting a very short window of acceptable downtime, outside of which means significant losses for businesses. Figure 1 shows the power system architecture of various entities. (a) (b) 3 (c) (d) (e) 4 (f) Figure 1. Overview of data center power system architectures across various entities. (a) Simple data center power distribution from Google with a central uninterruptible power supply (UPS) in 2007 (Fan 2007). (b) N+1 radial architecture (Sakalkar 2020). (c) Meta’s power distribution architecture in 2016 (Wu 2016). (d) Medium-voltage architecture from Google with a 2N utility and generators and pooling of IT loads on a medium-voltage power distribution network (Sakalkar 2020). (e) Distributed redundant 4N/3 power distribution from Microsoft (Zhang et al. 2021). (f) Medium-voltage architecture from Verrus with a 2N utility, battery energy storage system (BESS), and generators and pooling of loads in a power distribution network (Vaidhynathan et al. 2025). Based on the representation shown in Figure 1, next, we discuss the common attributes across these designs: • Utility/grid: These data centers are powered by the utility/grid as the primary source of energy. The data centers are typically fed by a medium-voltage (13.8-kv–34.5-kV) supply from a utility substation, which is then stepped down to low-voltage (~400-V) supply for the IT and cooling loads in the data center. • Transfer switches: Depending on the uptime requirements, appropriate redundancies are implemented in the power distribution system feeding the IT and cooling loads. These redundancies facilitate, e.g., concurrent maintainability and automatic failover to an alternate power feed to deliver to the uptime requirements. The presence of the redundancy is managed through appropriate automatic or static transfer switches, as shown in Figure 1 (a), (e). • IT load: Data center IT hardware (e.g., computing servers, network servers, storage servers) are voltage-sensitive equipment, and their power quality requirements are governed by the Information Technology Industry Council (ITIC) curve (Elphick and 5 Smith 2010). The ITIC curve is published by ITIC (formerly known as the Computer and Business Equipment Manufacturers Association [CBEMA]), and it provides an AC voltage boundary that most IT equipment can tolerate or ride through without experiencing unexpected shutdowns or malfunctions. Figure 2 shows the ITIC curve: It shows that the input AC voltage below 0.7 p.u. is considered a dropout event, and a 1.2- p.u. overvoltage is acceptable for up to 3 milliseconds. The typical IT load is approximately 400 V, so the appropriate transformation of the medium-voltage supply from the substation needs to be incorporated into the design. • UPS: To mitigate AC supply voltage fluctuations outside the allowable region within the ITIC curve, most data center power systems have UPS units. The UPS has batteries and can support the IT load for a few minutes. Figure 2. ITIC curve Source: Voltage Disturbance. n.d. • Backup power (generators): To meet the (tier-specific) uptime requirements, data centers need backup energy sources to mitigate risks of losses due to grid outages causing downtime. Most data centers rely on generators (diesel, natural gas, hydrotreated vegetable oil) as a backup energy source (Uptime Institute. n.d) (Figure 1 (a) to (e)), and some newer designs are relying on BESS (Figure 1 (f)). In particular, the generators are to be used only when the utility is not available, and the transition is managed through automatic transfer switches (see Figure 1 (a), (b), (c), (d), (e)). The backup energy source (generators and/or BESS) are connected at a low voltage in some designs (such as in Figure 1 (a), (b), (c), (e)) and at the medium-voltage level in others (Figure 1 (d) and (f)). The UPS helps keep the IT hardware powered when the data center transitions from utility power to a backup source (generator power) during a utility outage. A recent PJM event called on dispatching on-site generators in hyperscale data centers to serve the grid. Most data centers are designed to operate on generators only when the utility is not available (Figure 1 (a) to (e)) and not with the intent to parallel generators with the grid 6 for demand response. Such changes in operational requirements require modifications at the physical hardware level (possibly higher ratings of switchgear) as well as modifications of the controls software. In short, this is riskier to implement in brownfield sites. • BESS: Recently, owing to grid capacity constraints, data centers are increasingly being considered for grid services such as demand response. A recent report from EPRI (2025) compares the different approaches for data center load flexibility; in the report, BESS are particularly suitable compared to generators because of both reliability and emission constraints. In Figure 1 (f), the system design from Verrus accounts for grid paralleling of the BESS to respond to demand response requests (Vaidhynathan et al. 2025) through appropriate control systems. • Cooling load: The cooling system in data centers comprises pumps, chillers, air handling units, and coolant distribution units, which are low-voltage loads (~400–480 V). Many of these have variable-frequency drives. In most data centers, the IT and cooling loads are separate power distribution systems (see Figure 1 (a) to (e); cooling systems are not included), unlike the design from Verrus shown in Figure 1 (f), where cooling loads and IT loads are fed from the common medium-voltage bus. Appropriate backup power systems are put in place for cooling loads to meet the uptime requirements of the data center. In recent times, cooling systems in data centers are being considered for load flexibility (EPRI 2025) and power usage effectiveness optimization (Lazic et al. 2018). Most traditional data center design and operational philosophies cater only to the uptime, inherently leading to data centers being a passive participant on the grid. The operations must keep the IT hardware operating 24-7, which can sometimes exacerbate grid instability during voltage events (as described in the NERC incident review [NERC 2025]) based on the protection schemes implemented on the load side. The NERC incident review found almost 1,500 MW of data center load disconnecting from the grid during approximate 10-millisecond-scale grid-side faults, causing overvoltages and overfrequencies on the grid, with the possibility of causing a wide-area outage. Given the current growth of data centers and the grid constraints, it is very important to understand how data center design and operations interact with the grid and what types of grid disturbances/events a data center can “ride-through” and assist the grid during instability events. In recent times, there is a lot of attention on developing data centers as a flexible load on the grid using a variety of approaches, such as on-site generation, on-site energy storage, and compute job scheduling, and by dispatching cooling systems. As noted, every system architecture shown inFigure 1 has generators except that from Verrus, which also has BESS (Figure 1 (f)). A comparison of the effectiveness of these different approaches for data center load flexibility is discussed in a recent report from EPRI (2025). From the grid operator’s perspective, it is important to be able to better understand how such components of the data center power system could be leveraged to improve grid reliability. The next section describes a model of individual components that need to be integrated to develop an EMT domain model of any of the data center power system architectures shown in Figure 1. 7 2 Modeling Framework Overview The modeling framework builds on high-fidelity EMT representations developed through offline simulations and validated in real-time CHIL environments. To ensure the reliable and scalable integration of data centers into the grid, a flexible and detailed modeling approach is required. The framework should support: • Offline EMT simulations—comprehensive time domain models using tools like PSCAD for detailed disturbance studies • Real-time EMT simulations—CHIL environments enabling closed-loop control and dynamic studies, including actual devices and controllers. 8 3 System Modeling Layers 3.1 Utility Interface and Medium-Voltage Bus An accurate model of the utility interface must: • Model the data center POI as a dynamic, voltage-/frequency-sensitive node, parameterized with a realistic grid impedance (resistance and reactance) to simulate both strong and weak utility grid (e.g., N-1 contingency) scenarios. • Incorporate medium-voltage protection relays and utility coordination schemes to ensure reliable fault detection and grid compliance. • Simulate voltage sag/swells, frequency excursions, and POI oscillations, including those driven by rapid changes in AI loads. 3.1.1 Standard PSCAD Models Available The PSCAD standard library includes: • Three-phase medium-voltage bus and basic node models • Standard relays (overcurrent/undervoltage) • Circuit breakers and line elements • Modules for simulating common voltage/frequency disturbances. 3.1.2 Additional Model Requirements Custom or advanced model enhancements might be required to: • Parameterize the grid impedance at the POI to reflect the dynamic strength of the grid, enabling accurate representation of strong (high short-circuit ratio) and weak (low short- circuit ratio) grid behaviors and their effects on system stability and control interactions (Shah 2024). • Implement high-fidelity POI models capturing fast transient effects, relay coordination, and voltage/frequency dynamics unique to large data center and inverter-dominated systems. • Integrate programmable protection schemes aligned with utility standards and coordination routines. • Design and execute test routines for rapid load changes, harmonic distortion analysis, and oscillatory impacts that are not present in default PSCAD models. • Use impedance scan and system strength tools for frequency domain validation, and support advanced simulations for black-start, ride-through, and grid support controller performance. 3.2 Transformers To accurately represent transformers in data center interconnections, models must capture: • Fast transient phenomena, such as inrush current and saturation during energization or reclosing • Detailed tap changer control behavior under varying load and voltage conditions 9 • Protective relaying schemes for transformer overcurrent, differential protection, and under-/overvoltage trips • Transformer interaction with dynamic, fast-varying data center loads and grid disturbances, including fault scenarios and overvoltage risk. 3.2.1 Standard PSCAD Models Available The PSCAD library provides comprehensive transformer models, including three-phase power transformers with magnetization characteristics: • Saturation modeling • Tap changer functionality • Protection elements. These components support dynamic simulations of steady-state and transient conditions, such as load changes and switching events. 3.2.2 Additional Model Requirements Given the diverse data center power system architectures, additional custom model development might be required for: • Transformers in medium-voltage power plane architectures, where the integration of BESS and pooled IT and cooling loads creates complex interactions • Specialized transformer configurations found in distributed redundant 4N/3 power distribution or designs involving medium-voltage-level backup power connections • Transformer protections tailored for rapid fault response and load transfer schemes unique to data center power redundancy approaches • Integrating transformer models with complex transfer switch logic and protection schemes, as seen in modern hyperscale data centers • Capturing the impact of combined IT and cooling load feeding from a common medium- voltage bus, which requires coordinated transformer behavior modeling. Custom models should ensure compatibility with dynamic load changes from AI workloads, on- site generation, and demand flexibility strategies prevalent in evolving data center designs. 3.3 Information Technology Loads IT load models need to represent typical power consumption dynamics of the server and network hardware, including: • Time-varying load transitions from idle to peak demand with rapid ramp capabilities • Load sensitivity to voltage and frequency variations. 3.3.1 Standard PSCAD Models Available PSCAD offers generic composite and resistive load components that can approximate aggregated IT loads. These models capture basic steady-state and dynamic load characteristics. 3.3.2 Additional Model Requirements For data center applications, custom models should address: 10 • Fast transient behavior associated with AI-driven workload fluctuations • Power demand diversity reflecting heterogeneous server types (central processing units, graphics processing units) • Interaction effects with cooling and environmental systems to simulate coupled demand changes • Support for real-time adaptive load control mechanisms for grid support functions. 3.4 Uninterruptible Power Supply UPS models must accurately capture: • Dual-mode operation—grid-following during normal operation and grid-forming during islanded or backup modes • Transition logic for seamless switching between utility power and battery backup in response to faults or power quality events • Compliance with ride-through requirements during voltage disturbances. 3.4.1 Standard PSCAD Models Available PSCAD does not include specific UPS library components. Common practice is to represent UPS units either as ideal voltage sources or by building custom inverter models with integrated battery systems. 3.4.2 Additional Model Requirements Custom UPS inverter and battery management system models are required to emulate real UPS behavior during dynamic load changes, battery charge/discharge cycles, and power quality events. Integration with data center energy management systems and coordination with BESS are necessary for the accurate simulation of backup power architectures in data centers. 3.5 Cooling Loads Cooling load models must encompass: • Voltage and frequency sensitivity of motor-driven components, such as pumps and fans • Control logic for staged startup, load shedding, and curtailment during grid events or demand response • Thermal inertia effects influencing cooling system recovery times after disturbances • Coordination with facility management systems to reflect realistic operational sequences. 3.5.1 Standard PSCAD Models Available PSCAD includes generic motor load components that can be used to model variable-frequency drive-driven pumps, chillers, and fans commonly found in cooling systems. 3.5.2 Additional Model Requirements Customized models should be developed to: • Capture detailed variable-frequency drive control schemes and their dynamic response to grid disturbances. 11 • Model interactions between cooling loads and IT loads, including common medium- voltage bus influences. • Represent emergency backup operation modes powered by UPS or generators. • Support integration with data center energy management and supervisory control systems for demand flexibility and resilience strategies. 3.6 Battery Energy Storage System BESS models must comprehensively capture: • Fast transient response and dynamic control during grid events, including seamless operation in both grid-connected and island modes • Performance criteria outlined in the Institute for Electrical and Electronics Engineers (IEEE) Std 1547 (IEEE 2018), such as autonomous frequency response, voltage support, and disturbance ride-through • Accurate battery state-of-charge management and interactions with site controllers, especially under rapid load changes. 3.6.1 Standard PSCAD Models Available PSCAD includes generic battery storage models and inverter-based templates that simulate charge/discharge cycles, grid-following and grid-forming controls, and interactions with AC/DC systems. Increasingly, most original equipment manufacturers now provide factory-tested, IEEE Std 1547-compliant BESS models that are ready for integration with simulation platforms. These models reflect current grid support functions—including ride-through, voltage and frequency regulation, and advanced controls—ensuring alignment with utility interconnection requirements. 3.6.2 Additional Model Requirements Custom development might still be required to: • Implement advanced control strategies tailored for data center microgrids and integrate with other backup assets, such as UPS and generators. • Support parallel operation and coordinated dispatch, unique to large-scale deployments. • Enable hardware-in-the-loop or real-time simulation for the comprehensive validation of BESS performance within complex, fast-varying data center loads. 3.7 Generator Models Generator models should incorporate the following key elements: • Startup delays and ramp rates—reflecting the response times of diesel, natural gas, or other backup generation asset • Load acceptance dynamics—capturing realistic backup power connection times during transitions • Protective relays—including overcurrent, undervoltage, overvoltage, and frequency deviation protections 12 • Synchronization and reconnection mechanisms—for smooth transitions back to grid operation during utility restoration • Automatic generation control—for continuous regulation of generator output to maintain system frequency and overall power balance across multiple generators • Automatic voltage regulation—to control excitation and regulate voltage stability and reactive power output at the POI • Control modes: o Isochronous control for maintaining constant frequency/speed, critical during islanded or isolated operation o Droop control for proportional load sharing among multiple parallel generators in response to frequency changes o Load-sharing schemes coordinating power dispatch among generators based on capacity and system demands. 3.7.1 Standard PSCAD Models Available PSCAD offers detailed synchronous and asynchronous generator models, including excitation systems, governors, and protection schemes; however, these models are often challenging to integrate because the detailed dynamic characteristics and control parameters are typically not provided by manufacturers, making it difficult to accurately represent actual generator behavior. 3.7.2 Additional Model Requirements Custom development is needed to: • Capture real-world generator dynamics reflecting site-specific configurations and unique backup power operations. • Integrate tightly with UPS and BESS models for coordinated backup power responses under fault and transient conditions. • Include advanced control schemes allowing smooth switching between control modes (e.g., from droop to isochronous) depending on the operational scenarios. • Model fault ride-through capabilities and grid code compliance for reliable operation during disturbances. • Enable the simulation of coordinated load sharing and fault response among multiple generators to avoid overload and ensure stable power delivery. 3.8 Protection Systems Every breaker in the modeled system is assumed to have its own programmable thresholds. These thresholds define whether a specific subsystem remains connected or is isolated during an abnormal condition. Accurate protection system modeling requires: • Implementation of protective relay logic with customizable thresholds for voltage, current, and frequency • Circuit breaker models supporting single-pole and three-pole operations, with realistic opening/closing time delays • Simulation of relay coordination and cascading trip sequences under various fault scenarios 13 • Integration of current transformer and voltage transformer saturation effects and harmonic filtering for refined fault discrimination. 3.8.1 Standard PSCAD Models Available PSCAD offers a range of protection system modeling elements, including circuit breakers, relays, and protective devices that can replicate fundamental protection functions, such as overcurrent, undervoltage, and distance protection. But it does so primarily through a combination of predefined components and user-defined or custom-built logic. 3.8.2 Additional Model Requirements Custom development is needed to incorporate: • Advanced relay functions, such as relay coordination—including timing, multizone protection, and cascading trip sequences—which are often implemented through user- developed logic or scripting • Communication-assisted protection features, such as breaker failure detection and interlocking • Breaker control and reclose sequences tailored to the operational requirements of data centers. 3.9 Higher-Level Control Systems Higher-level control systems in data centers include supervisory control and data acquisition (SCADA) systems, building management systems, energy management systems, paralleling switchgear controllers, and microgrid controllers, among others. These systems coordinate breaker operations and control distributed energy resources (DERs), loads, and backup generation to ensure reliable, efficient, and flexible operation. Incorporating supervisory control logic allows the assessment of how coordinated dispatch and automated restoration can mitigate widespread disconnections observed in past incidents. These control systems implement a complex sequence of operations to facilitate automated operation of the data center. Modeling of higher-level controls requires: • Implementation of custom control sequences for the operation of breakers, load/DER dispatch, and demand response • Real-time communication and monitoring interfaces between control layers and power equipment • Integration of fault detection, system isolation, and black-start control strategies. 3.9.1 Standard PSCAD Models Available PSCAD does not provide ready-made library models for higher-level data center control systems, such as microgrid controllers, SCADA systems, building automation systems, and energy management systems. Instead, models must be developed using custom logic and control blocks tailored to specific project requirements. For reference frameworks, it is worthwhile to consider established standards: 14 • IEEE Std 2030.7 (IEEE 2017): defines functional specifications and interoperability requirements for microgrid controllers, covering dispatch, transitions between grid- connected and islanded states, and cyber-physical system integration • ASHRAE Guideline 13 (ASHRAE 2024): standardizes the design, documentation, specification, and integration of building automation systems, ensuring support for system architecture, commissioning, performance monitoring, and communication between heating, ventilating, and air-conditioning/mechanical and electrical systems. 3.9.2 Additional Model Requirements Custom development is generally needed to: • Build site-specific supervisory controllers with interfaces conforming to IEEE Std 2030.7 and ASHRAE Guideline 13, ensuring robust communications and fail-safe operation. • Integrate with third-party building automation systems or energy management system solutions to support holistic facility management. • Validate controller performance using advanced simulation environments, such as CHIL, which enables real-time interaction and robust predeployment testing of the control hardware under realistic operational conditions. 3.10 Automatic Transfer Switches/Static Transfer Switches Automatic transfer switch and static transfer switch models should capture: • Fast transfers between independent utility and backup sources (utility, generator, UPS) • Sensing and logic-driven transfer initiation (utility loss, voltage/frequency threshold crossing) • Transfer time, transient response, and interruption characteristics unique to automatic transfer switch (mechanical) and static transfer switch (solid-state, subcycle response) • Voltage synchronization, inrush control, and load acceptance routines 3.10.1 Standard PSCAD Models Available PSCAD provides standard switch components and logic blocks that can emulate basic automatic transfer switch/static transfer switch operation. Users can deploy programmable breakers and switching elements to represent both manual and automatic transfer events. The standard library includes: • Basic three-phase switch models for utility-to-backup source transitions • Logic blocks for emulating status sensing and transfer decision processes • Time delay and sequential switching features. 3.10.2 Additional Model Requirements Custom development might be required to: • Accurately represent the transient switching dynamics, including the difference between a mechanical automatic transfer switch (typical transfer time in seconds) and solid-state static transfer switch (typical transfer time in milliseconds). 15 • Integrate programmable protection coordination to prevent back-feed, ensure source synchronization, and comply with utility interlock requirements. • Implement fault ride-through and seamless load transition during critical events, particularly important for Tier IV and hyperscale data centers with stringent uptime demands. • Design and run test cases for switching under abnormal conditions, such as voltage sags, swells, and faulted source transfer. 16 4 Key Metrics/Test Cases for Model Evaluation This section outlines critical metrics and essential test scenarios for validating data center power system models, particularly emphasizing the dynamic behaviors associated with fast-changing load scales, such as AI-driven workload spikes. The metrics align with observations from NERC incident reports (NERC 2025) and recommendations for data center integration to ensure grid stability and reliable operation. The following test metrics ensure that data center models can capture fast transient behaviors, validate protection settings, and assess coordinated DER and load cycling responses under abnormal grid events. Note that these are to be viewed as recommendations, not as a comprehensive list. Table 1. Key Metrics and Test Cases for Model Evaluation Test Category Test Specification Reference Notes/Description Standard Grid overvoltage/ Simulate grid transients leading to NERC 2025 Includes simulation of undervoltage 1.07-p.u. overvoltage. grid-forming inverter disturbance response Validate disconnection thresholds behavior under IEEE 2018 (NERC incident driven) and ride-through per IEEE Std overvoltage/ 1547 and IEEE Std 2800. undervoltage events, IEEE 2022a including autonomous Test relay coordination logic and voltage support and protection responses. fault ride-through. Integrates with NERC fault scenarios using IEEE Std 2800, IEEE Std 1547, and grid- forming control modes Grid underfrequency/ Simulate frequency excursions IEEE 2018 Tests DER and large overfrequency across a 57- to 63-Hz range. load frequency disturbance response Test DER and load ride-through tolerance under fast IEEE 2022a (NERC incident driven) against defined trip curves (e.g., grid events, distinct 59.5 Hz for six cycles, 62 Hz for from voltage ride- 300 seconds). NERC 2014 through. Aligns with NERC findings from Validate controller logic for underfrequency system frequency exit/reentry, recovery separations. Focused ramp, and hysteresis. on trip logic Include momentary cessation, compliance, frequency reconnection delay, and droop response coordination, response tests. and post-event Model NERC frequency recovery behavior disturbance scenarios (e.g., under IEEE/NERC- ERCOT 2021, WECC 2020 defined curves frequency dips). Fast-frequency Model BESS/UPS fast-frequency IEEE 2018 Captures fast-frequency response and ancillary support (<0.5-Hz deviation). support from both grid- services capability following and grid- IEEE 2022a forming DERs. 17 Test Category Test Specification Reference Notes/Description Standard Implement droop and automatic Incorporates inertial generation control-based control NERC 2015 response, synthetic schemes. inertia, and Evaluate frequency-watt support autonomous droop functions. control. Reflects capabilities defined in the grid-forming white paper (NERC 2015) and IEEE Std 2800 Grid impedance change Simulate short-circuit ratio drop IEEE 2022a Emulates generation and contingency from 15 to 2 to emulate plant loss. loss or grid topology simulation Assess inverter stability, changes by altering Koralewicz harmonics, and control behavior. system impedance. 2020 Important for modeling Perform impedance scans and weak grid interactions short-circuit ratio sensitivity and resonance issues analysis. that might arise with inverter-based resources Subharmonic and Simulate 10- to 100-ms ramping NERC 2017 Captures potential flicker behavior under cycles (0.1–10 Hz). subharmonics due to AI AI load Evaluate subsynchronous workloads, which can resonance and voltage flicker be linked to impacts. subsynchronous resonance and flicker Measure harmonics and excitation events in transmission of network resonance. grids. Essential for inverter-based generation and long transmission lines Grid support functions Validate fault ride-through with IEEE 2022a Tests standard- (IEEE/NERC aligned) voltage/frequency events. conforming ride-through Implement volt-volt ampere and grid support IEEE 2018 reactive (VAR), volt-watt, and functionalities, such as frequency-watt functions. volt-VAR, volt-watt, and frequency-watt Confirm compliance with IEEE Std response. Ensures 1547 and IEEE Std 2800 profiles. compliance with IEEE grid codes and readiness for interconnection approval Demand flexibility and Reduce X% load for Y hours on IEEE 2017 Tests distinct demand curtailment (demand the grid signal. flexibility strategies: response tests) PJM 2025 18 Test Category Test Specification Reference Notes/Description Standard Include day-ahead demand 1. Scheduled day- response for capacity planning. Norris et al. ahead curtailments Include N-1 dispatch-based fast 2025 2. Fast curtailment demand response for contingency based on N-1 transmission relief (response transmission within 5–10 minutes). contingency dispatch. Test dispatch under independent This aligns with real- system operator/regional time grid relief, transmission organization replacing generalized curtailment scenarios. load dispatch testing with time-specific curtailment windows. Ensures model compliance with demand response use cases DER and load cycling Simulate SCADA coordinated IEEE 2017 Validates end-to-end coordination control of DERs (BESS, gensets, DER control logic under UPS) and data center loads under coordinated and normal and contingency fallback modes. It conditions. includes SCADA- Validate black-start sequence as triggered sequencing of IEEE Std 2030.7, including delay DER units for black- timers, source priority, and loading start, islanding, and stages. real-time load cycling. Also tests controller Implement grid-to-island and integration with building island-to-grid transitions. management Evaluate DER inverter interactions systems/SCADA, with heating, ventilating, and air- inverter ramp rate conditioning, and IT loads during enforcement, and startup and ride-through. critical load-shedding logic. Ensures compliance with IEEE Std 2030.7 controller functional requirements and NERC operational coordination guidance Inrush current Simulate energization of IEEE 2021 Verifies the data center characterization transformers. power system and Measure inrush current magnitude protection system and duration up to 2 s. stability during transformer startup Evaluate voltage sag at the POI events. Ensures power and its effect on the downstream quality during transient equipment. load 19 Test Category Test Specification Reference Notes/Description Standard IT load behavior Measure power factor, harmonics, IEEE 2022b Models the electrical and step loading effects of IT fingerprint of AI/IT loads equipment under real-world beyond just kilowatts. Quirk, workloads. Supports UPS design, Davidson, Characterize switched-mode filter sizing, and DER and Schmidt power supply behavior at peak ride-through logic under 2022 central processing unit/graphics nonlinear load stress. processing unit use using Helps validate power harmonic spectrum (IEEE Std quality metrics and 519). inverter interactions during real compute Simulate fast-compute load ramp- workload fluctuations ups to observe power surges, flicker, and subharmonic behavior. Evaluate inverter + UPS interaction under clustered rack activity and server restarts. Use waveform capture to validate against total harmonic distortion <5% and PF> 0.9 targets. Protection system Test undervoltage/overvoltage, IEEE 2022c Validates coordination coordination overfrequency/underfrequency of protection zones relay thresholds and breaker across breakers and NERC 2014 timing per IEEE Std C37.2 and relays in grid-connected C37.013. systems, particularly Validate cascading trip logic, NERC 2022 under DER-dominated protection zone overlap, and disturbance conditions. breaker reclosing sequences. Simulates cascading trips, rate-of-change-of- Include miscoordination scenarios frequency-based reported in the NERC Odessa disconnects, and disturbance report (NERC 2021). breaker reclose timing Assess rate-of-change-of- errors. Rooted in NERC frequency settings and Odessa findings (NERC misoperations under fast inverter 2021) and conforms to disconnect conditions. NERC protection and Use time-current characteristic control and IEEE C37 analysis for protection staging and series standards. islanding response. These scenarios collectively validate whether modeled data centers exhibit the same vulnerability patterns noted in prior NERC large load loss reviews. 20 5 Challenges and Considerations Modeling large-scale data centers poses significant challenges due to the sheer number of electrical components involved. 5.1 Computational Complexity Simulating this scale with detailed component-level models significantly increases computational time and resource requirements. Each individual device, such as an inverter or a breaker, introduces additional states and equations into the EMT simulation, potentially leading to long run times and memory constraints. 5.2 Model Fidelity vs. Practicality • Detailed modeling involves representing each device with high fidelity. This offers precise insight into fast transients (e.g., inrush currents, dynamic load ramping) and control system interactions, which is critical for the robust evaluation of reliability and protection coordination. • Aggregate modeling simplifies groups of similar devices (e.g., IT loads, BESS, or generators) into lumped equivalent models, reducing the computational burden but potentially losing individual device behaviors and interacting dynamics, which are crucial to assessing certain fast-timescale phenomena. 5.3 Trade-Off Decisions • The need for model accuracy must be balanced with simulation efficiency and the purpose of the study. For system-level studies assessing the overall load impacts or grid interaction, aggregate models might suffice. • Detailed models should be reserved for critical components or subsystems where individual device behavior strongly influences outcomes (e.g., specific IT load or UPS system dynamics affecting grid event studies). • Hybrid approaches combining detailed and aggregated models can effectively target computational resources while retaining sufficient fidelity for key metrics, such as inrush current, load ramp-up profiles, and protection system coordination. Hybrid modeling approaches that combine detailed subsystem representations with aggregated load equivalents provide a practical balance between fidelity and scalability. For example, Figure 1 illustrates the power system architecture of a 70-MW data center featuring medium-voltage power planes, distributed redundant power sources, coordinated IT and cooling loads, BESS, multiple UPS units, and utility feeds. • Detailed modeling for this 70-MW system provides high-fidelity insight into fast transient events, such as inrush currents during startup, load ramping linked to AI workload cycles, detailed UPS transition behaviors, and precise relay coordination. This is crucial for accurate studies on grid stability, fault response, and protection setting validation; however, the computational burden is substantial, as simulating a 70-MW system (with 22 IT load domains and 8 cooling domains), each with unique electrical devices (with controls and measurements), requires significant central processing unit time, memory, and data management efforts. 21 • Aggregate modeling offers a practical alternative by reducing the model complexity through lumping similar devices, enabling more rapid simulations that are suitable for high-level scenario planning or grid impact studies—for example, lumping the IT racks with UPS, which are spread over several rows, into a single IT load with a large inline UPS. • A hybrid approach is often effective, employing detailed modeling for critical components (e.g., BESS, UPS, critical transformers) while aggregating less critical or more uniform loads (e.g., bulk IT and cooling loads) to optimize fidelity and efficiency trade-offs. 22 6 Conclusion This framework provides a structured, modular approach for representing data center power systems within EMT platforms such as PSCAD or a digital real-time simulator. By integrating component-level dynamics, protective coordination, and high-level control logic, it captures both transient and steady-state behavior under abnormal grid events. The inclusion of grid-following, grid-forming, and black-start capabilities ensures applicability across evolving data center architectures. Aligning with recent reliability observations, the framework enables NERC and utilities to assess aggregate load behavior at the POI, identify vulnerability thresholds, and support the development of standards for reliable large load integration. 23 References ASHRAE. 2024. ASHRAE Guideline 13-2024: Specifying Building Automation Systems. www.ashrae.org/technical-resources/bookstore/ashrae-guideline-13-specifying-building- automation-systems. Badrzadeh, Babak. n.d. “Electromagnetic Transient Simulation Models for Large-Scale System Impact Studies in Power Systems Having a High-Penetration of Inverter Connected Generation.” Australian Energy Market Operator. www.nerc.com/globalassets/programs/documents/emt_simulation_models_for-large- scale_system_impacts.pdf. Bloomberg. 2025. “Meta Enters Power Trading To Support its AI Growth Energy Needs.” EnergyWire. November 24, 2025. subscriber.politicopro.com/article/eenews/2025/11/24/meta- enters-power-trading-to-support-its-ai-growth-energy-needs-00664523?ref=wheresyoured.at. Electric Power Research Institute (EPRI). 2025. Grid Flexibility Needs and Data Center Characteristics. www.epri.com/research/programs/063638/results/3002031504. Elphick, Sean, and Vic Smith. 2010. “The 230 V CBEMA Curve—Preliminary Studies.” Presented at the 2010 20th Australasian Universities Power Engineering Conference, December 5–08, 2010. https://ieeexplore.ieee.org/abstract/document/5710795. Fan, Xiaobo, Wolf-Dietrich Weber, and Luiz Andre Barroso. 2007. “Power Provisioning for a Warehouse-Sized Computer.” ACM SIGARCH Computer Architecture News 35.2: 13–23. dl.acm.org/doi/10.1145/1250662.1250665. Institute of Electrical and Electronics Engineers (IEEE). 2017. IEEE 2030.7-2017 – IEEE Standard for the Specification of Microgrid Controllers. Piscataway, New Jersey. https://ieeexplore.ieee.org/document/8340204. ———. 2018. IEEE 1547-2018 – IEEE Standard for Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces. Piscataway, New Jersey. https://ieeexplore.ieee.org/document/8332112. ———. 2021. IEEE C57.12.90-2021 – IEEE Standard Test Code for Liquid-Immersed Distribution, Power, and Regulating Transformers. Piscataway, New Jersey.https://standards.ieee.org/ieee/2004/11300/ https://ieeexplore.ieee.org/document/9707753. ———. 2022a. IEEE 2800-2022 – IEEE Standard for Interconnection and Interoperability of Inverter-Based Resources (IBRs) Interconnecting with Associated Transmission Electric Power Systems. Piscataway, New Jersey. https://ieeexplore.ieee.org/document/9762253. ———. 2022b. IEEE 519-2022 – IEEE Standard for Harmonic Control in Electric Power Systems. Piscataway, New Jersey. https://ieeexplore.ieee.org/document/9848440. 24 ———. 2022c. IEEE C37.2-2022 - IEEE Standard for Electrical Power System Device Function Numbers, Acronyms, and Contact Designations. Piscataway, New Jersey. https://ieeexplore.ieee.org/document/9984140. ———. 2025. IEEE 2004-2025 – IEEE Recommended Practice for Hardware-in-the-Loop (HIL) Simulation-Based Testing of Electric Power Apparatus and Controls. Piscataway, New Jersey. https://standards.ieee.org/ieee/2004/11300/. Koralewicz, Przemyslaw, Shahil Shah1, Vahan Gevorgian, Robb Wallen, Kapil Jha, Dale Mashtare, et al. 2020. “Impedance Analysis and PHIL Demonstration of Reactive Power Oscillations in a Wind Power Plant Using a 4-MW Wind Turbine.” Frontiers in Energy Research. https://docs.nrel.gov/docs/fy20osti/76053.pdf#page=1.00&gsr=0. Lazic, Nevena, Tyler Lu, Craig Boutilier, Moonkyung Ryu, Eehern Wong, Binz Roy, and Greg Imwalle. 2018. “Data Center Cooling Using Model-Predictive Control.” Presented at the 32nd Conference on Neural Information Processing Systems (NeurIPS 2018), Montréal, Canada. https://proceedings.neurips.cc/paper_files/paper/2018/file/059fdcd96baeb75112f09fa1dcc740cc- Paper.pdf. Norris, Tyler H., Tim Profeta, Dalia Patino-Echeverri, and Adam Cowie-Haskell. 2025. Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in US Power Systems. Nicholas School of the Environment, Duke University. https://nicholasinstitute.duke.edu/publications/rethinking-load-growth. North American Electric Reliability Corporation (NERC). 2014. Generator Frequency and Voltage Protective Relay Settings. Number: PRC-024-2. www.nerc.com/globalassets/standards/projects/2014-01/prc-024-2_clean_2015_jan_12.pdf. ———. 2015. Essential Reliability Services Task Force Measures Framework. Reporthttps://www.nerc.com/globalassets/programs/rapa/ra/erstf-framework-report---final.pdf. ———. 2017. “Chapter 3: Recommended Practices & Mitigation Measures.” In Reliability Guideline Forced Oscillation Monitoring & Mitigation. https://www.nerc.com/globalassets/who- we-are/standing-committees/rstc/smwg/reliability_guideline_-_forced_oscillations_-_2017-07- 31_-_final.pdf#page=17.06. ———. 2022. 2022 Odessa Disturbance Texas Event: June 4, 2022—Joint NERC and Texas RE Staff Report. https://www.nerc.com/globalassets/our-work/reports/white- papers/nerc_2022_odessa_disturbance_report-1.pdf. ———. 2024. “2024 Long-Term Reliability Assessment.” Presented December 2024, corrected July 11, 2025. https://www.nerc.com/globalassets/our-work/assessments/2024- ltra_corrected_july_2025.pdf. ———. 2025. “Incident Review: Considering Simultaneous Voltage-Sensitive Load Reductions.” https://www.nerc.com/globalassets/our-work/reports/event- reports/incident_review_large_load_loss.pdf. 25 PJM. 2025. “Joint Stakeholder Options PJM CIFP – Large Load Additions.” October 1, 2025. https://www.pjm.com/-/media/DotCom/committees-groups/cifp-lla/2025/20251001/20251001- item-05d---joint-stakeholder-options---amazon-calpine-constellation-google-microsoft-talen.pdf. Quirk, David, Tom Davidson, and Roger Schmidt. 2022. “Data Centers: ASHRAE’s Data Center Thermal Guidelines: Fifth Edition Highlights.” ASHRAE. store.accuristech.com/standards/data- centers-ashrae-s-data-center-thermal-guidelines-fifth-edition-highlights?product_id=2258176. Sakalkar, Varun, Vasileios Kontorinis, David Landhuis, Shaohong Li, Darren De Ronde, Thomas Blooming, et al. 2020. “Data Center Power Oversubscription with a Medium Voltage Power Plane and Priority-Aware Capping.” Presented at ASPLOS 2020, March 16–20, 2020, Lausanne, Switzerland. https://dl.acm.org/doi/pdf/10.1145/3373376.3378533. Shah, Shahil. 2024. “Stability Analysis of IBR-Dominated Power Systems Using Grid Impedance Scan Tool.” Presented at the PSERC webinar, October 23, 2024. https://documents.pserc.wisc.edu/documents/general_information/presentations/pserc_seminars/ webinars_2024/fall_2024/Shah___Webinar_Slides_10_23_24v2.pdf. Shehabi, Arman, Alex Newkirk, Sarah Smith, Alex Hubbard, Nuoa Lei, Md Abu Bakar Siddik, et al. 2024. 2024 United States Data Center Energy Usage Report. Berkeley, CA: Lawrence Berkeley National Laboratory. https://escholarship.org/uc/item/32d6m0d1. Terrell, Michael. 2025. “How We’re Making Data Centers More Flexible to Benefit Power Grids.” Google, The Keyword, Innovation and AI. August 4, 2025. https://blog.google/innovation-and-ai/infrastructure-and-cloud/global-network/how-were- making-data-centers-more-flexible-to-benefit-power-grids/. Uptime Institute. n.d. “Tier Classification System.” https://uptimeinstitute.com/tiers. Vaidhynathan, Deepthi, Kumaraguru Prabakar, Gregory Martin, Anand Ramesh, Ben Wheeler, Christopher Coco, et al. 2025. Vulcan Test Platform: Demonstrating the Data Center as a Flexible Grid Asset. Golden, CO: National Renewable Energy Laboratory. https://docs.nrel.gov/docs/fy25osti/94844.pdf. Voltage Disturbance. n.d. “ITIC Curve.” https://voltage-disturbance.com/voltage-quality/itic- curve/. Voltus. 2025. “Voltus Launches “Bring Your Own Capacity” Product to Support Data Center Growth and Grid Resiliency.” September 30, 2025. https://www.voltus.co/press/bring-your-own- capacity-data-centers. Wu, Qiang, et al. 2016. “Dynamo: Facebook’s Data Center-Wide Power Management System.” ACM SIGARCH Computer Architecture News 44.3: 469–480. https://dl.acm.org/doi/10.1145/3007787.3001187. 26 Zhang, Chaojie, Alok Gautam Kumbhare, Ioannis Manousakis, Deli Zhang, Pulkit A. Misra, Rod Assis, et al. 2021. “Flex: High-Availability Datacenters With Zero Reserved Power.” https://www.microsoft.com/en-us/research/wp-content/uploads/2021/04/FlexMA-DCs- ISCA21.pdf. 27 Appendix Table A-1. Test Case Description and Technical Relevance for User Reference Metric/Test Case Description Relevance Fast transient load response Simulate rapid load ramp-up Assess voltage and frequency from 10–20% to 80–100% within oscillations at the POI. milliseconds. Inrush current characterization Measure peak and duration of Identify potential stress on transformer and motor inrush equipment and grid during currents during transient events. energization/startup. Load ramp-up profile Verify model behavior under Ensure model realistically steep load increase and represents operational data decrease profiles. center power demand curves. Protection relay coordination Test relay threshold settings, Avoid unintended tripping time, and coordination to disconnections as highlighted by prevent cascading load NERC incidents involving large shedding. loads. IT equipment Evaluate for voltage sags, Confirm robustness of critical IT swells, and frequency equipment to maintain deviations. continuous operation. Dynamic stability under fault Simulate grid fault conditions Ensure system stability and and assess sustained operation minimize operational disruptions or controlled disconnection during faults. response. Demand flexibility and Test fast load shedding, Validate capability to provide curtailment compute job scheduling, and demand response and support cooling adjustments in response grid emergency operations. to grid signals. Coordinated DER and load Simulate distributed resources Optimize grid support and cycling (BESS, UPS) cycling in resilience during disturbances coordination with load and load fluctuations. management. 28