More Than 3 GW of Load Dropped Off the PJM Grid: What Does the Incident Teach Us About Data Center Transfer Logic?

Introduction

• According to reports published on July 22, 2026, data centers switched to backup power following the loss of a transmission line in Northern Virginia.

• PJM data indicated that more than 3 GW of demand suddenly disappeared from the grid.

• The event caused voltage and frequency disturbances that were detectable across a wide area, but system reliability was not compromised.

• The incident highlights the importance of coordinating UPS systems, generators, switchgear and protection settings.

The event raises an important engineering question: when is it justified to disconnect a data center from the public grid, and under what conditions is it safe to reconnect later? A critical facility's primary objective is to maintain continuous power to its own loads. However, where substantial data center capacity is concentrated, the simultaneous response of many similarly configured control systems may also become visible at grid level.

What happened on the PJM grid?

According to the available reports, data center control systems transferred to backup power after a transmission line in Northern Virginia went out of service. Based on data published by PJM, more than 3 GW of load suddenly dropped off the public grid during the event.

This significant change in demand caused voltage and frequency disturbances that could be detected across a wider area. According to the research summary, system reliability was not compromised. Nevertheless, the event clearly demonstrates that the operation of data center protection and automation systems cannot be assessed solely within the facility boundary.

It is important to distinguish between reported facts and engineering conclusions. The public summary does not provide a complete account of the specific settings or switching sequences at every affected facility. The incident should therefore be treated not as the failure of a single equipment type, but as a case study for reviewing transfer philosophies.

Who is this for, and why is it relevant?

The topic is primarily relevant to data center operators, facility managers, electrical designers, project owners, colocation providers and technical procurement teams. Its engineering focus is the interaction between medium- and low-voltage distribution, UPS systems, generators, ATS or STS equipment, protection systems and supervisory controls.

The industry's exposure results from the growing concentration of loads. A protective response that appears appropriate for one facility may produce a different outcome at a larger scale if multiple sites operate with nearly identical thresholds and delays. Availability therefore requires not only redundant equipment but also coordinated dynamic behavior.

What does this mean for operations?

The first lesson is that a short voltage disturbance, a sustained grid outage and a breach of power-quality limits may require different responses. If every deviation triggers immediate disconnection, the facility may enter generator operation unnecessarily. If the logic is too permissive, loads outside the UPS supply path, cooling systems or other auxiliaries may be placed at risk.

The review should therefore cover the entire power supply chain:

• which voltage and frequency thresholds initiate a transfer;

• which time delays and reset conditions apply;

• which loads the UPS supports during generator start-up;

• how the generators start and synchronize;

• which interlocks prevent incorrect or unintended parallel switching;

• how operators receive information about the actual sequence of events.

Individual settings may be acceptable in isolation while the complete sequence still introduces risk. This can only be assessed using time-stamped event logs, switching matrices and integrated testing.

Disconnection is only the first half of the process

Transfer to backup power often receives more attention than the return to the normal grid supply. During retransfer, however, grid stability, generator operating status, UPS behavior and the staged restoration of loads must all be considered at the same time.

A single return of grid voltage is not necessarily a sufficient condition for reconnection. It is advisable to define a stability period, an acceptable power-quality range and unambiguous permissive logic. Hysteresis, time delays and appropriate interlocking may be required to prevent rapid switching back and forth.

Load restoration may not be suitable as a single-step process either. The starting behavior of cooling systems, battery chargers, transformers and other auxiliaries can create a combined transient load. The sequence should be aligned with the actual system architecture and operating states.

Common mistake

A common mistake is to treat the settings of the ATS, UPS, generator controls and medium-voltage protection as separate project packages. Each subsystem may pass its own functional test while the event sequence of the complete facility remains insufficiently validated.

A further risk arises when settings documentation is not updated after modifications. A firmware update, switchgear replacement, generator expansion or new load block may alter the original coordination. Nominal redundancy may remain in place even though automatic operation no longer follows the approved design intent.

It is also incorrect to assume that a successful disconnection test automatically confirms the adequacy of the return process. The two directions involve different states, timing requirements and load transients.

What should be reviewed from a technical perspective?

The first requirement is an approved and current single-line diagram together with an operating description. Without these documents, it is difficult to demonstrate that protection selectivity, transfer logic and operating procedures all describe the same system.

The following areas are particularly important:

Protection coordination and voltage sag ride-through

Thresholds and delays should be checked to determine whether they distinguish between a grid transient that the facility can ride through and an actual loss of supply. The objective is not to prevent disconnection at all costs, but to establish the right balance between an unnecessary response and a delayed response.

UPS and generator capacity under dynamic conditions

In addition to rated power, the assessment should consider start-up time, stepped load acceptance, actual battery condition and the availability of auxiliary systems. A backup power source provides effective redundancy only if it can support the entire transition process.

Event logging and time synchronization

Logs from the UPS, ATS, generator controls, switchgear and building management system must be comparable. Without reliable time synchronization, it is difficult to determine after an incident which event was the initiating cause and which was merely a consequence.

Recommended next step

Operators should consider initiating a targeted transfer-system audit. The first phase can be a document-based review comparing single-line diagrams, protection settings, the cause-and-effect matrix, manufacturer parameters and operating procedures.

In the second phase, disconnection, islanded operation and reconnection can be verified through simulation or risk-controlled on-site testing. In a live data center, the test should be prepared in stages, with an approved recovery plan, clearly defined abort criteria and the involvement of all affected operations teams.

Under the Digital Technologies engineering approach, this type of assessment is not limited to checking a single ATS or UPS. It involves a vendor-neutral, end-to-end evaluation of medium- and low-voltage distribution, UPS systems, generators, switchgear, supervisory systems and operating procedures. In an existing environment, the objective is to prioritize risks and implement modifications in phases with minimal downtime.

Conclusion

The PJM load loss of more than 3 GW is a reminder that the effects of data center transfer logic can extend beyond the facility boundary. Redundant UPS systems and generators do not by themselves guarantee correct operation: protection functions, delays, interlocks and reconnection sequences must operate as a coordinated system.

The practical response is not simply to make protection settings less sensitive, but to review the complete event chain. Documented design intent, accurate event logging, integrated testing and controlled return to the grid can collectively reduce the risk of an external disturbance initiating an unnecessary or difficult-to-control internal transfer process.

Sources

• https://www.reuters.com/business/energy/massive-disconnect-power-roiled-largest-us-electric-grid-2026-07-22/

• https://www.investing.com/news/stock-market-news/massive-disconnect-of-power-roiled-largest-us-electric-grid-4807202

• https://www.pjm.com/-/media/DotCom/planning/planning-criteria/dominion-planning-criteria.ashx