Practical Guidance and Considerations for Large Load Interconnections

Introduction

The power grid is entering a period of transformation unlike anything seen in decades. After years of flat or even negative load growth, utilities across North America are suddenly facing large load interconnection requests measured not in megawatts but in hundreds and even thousands of megawatts per facility. Data centers — driven by the global expansion of cloud computing, artificial intelligence, and large language models — are the dominant driver of this shift, but they are far from the only one. This article lays out the key considerations our team applies when supporting utilities, developers, and regulators navigating this new landscape.

Why Large Loads Are Different

The grid has always accommodated industrial customers — arc furnaces, manufacturing plants, refineries, and mining operations. So why are today's large loads creating unprecedented engineering challenges?

Scale. Gigawatt-class campuses, once unheard of, are now appearing regularly in interconnection queues. In some regional analyses, data centers alone represent 80 percent or more of large load interconnection demand by capacity.

Speed. A modern data center can be designed and built in roughly two years, sometimes faster. Transmission infrastructure, generation, and substation expansion — by contrast — can take five to ten years once planning, permitting, and approval timelines are accounted for. This timing mismatch is now one of the central tensions in grid planning.

Load characteristics. Modern data centers are dominated by power electronic equipment — variable frequency drives, uninterruptible power supplies, rectifiers, and switching converters. Recent analysis suggests that 80 percent or more of a typical data center's load is power-electronic in nature. The dynamic behavior of these loads is fundamentally different from the motor-dominated industrial loads of the past. That distinction matters most when it comes to stability analysis and the studies required for safe integration.

Confidentiality. Unlike traditional industrial customers, hyperscale operators treat the internal architecture of their facilities as proprietary. This creates real friction in the interconnection process, where transmission planners need detailed equipment data to run meaningful power system studies.

Emerging Reliability Concerns

Two large load disconnection events in a major Mid-Atlantic region brought these issues into sharp focus. In one incident, approximately 1,500 MW of data center load tripped offline in response to a grid fault — even though the protection and control system performed exactly as designed. A subsequent event saw roughly 1,800 MW disconnect under similar conditions. In both cases, the trips occurred because of protection settings inside the customer facilities — voltage thresholds, UPS trip points, and timer settings — that the host utility had limited visibility into beforehand.

These events are leading indicators of a broader category of risks that large load interconnections can introduce:

• Ride-through performance failures, where loads trip offline in response to normal grid disturbances they should be able to withstand
• Voltage and frequency stability impacts from rapid, large-magnitude load swings
• Power quality concerns, including harmonics, flicker, and transient events caused by fast-switching power electronics
• Subsynchronous oscillations and torsional interactions with nearby synchronous generation
• Millisecond-scale load ramping — sub-second variations that can stress system controls and protection coordination
• Protection coordination challenges between utility-side and customer-side equipment

Public data curves from operating data centers show load variations of 5 to 10 MW occurring at sub-second timescales, with occasional drops from 30 MW to 5 MW and back within seconds. This stochastic behavior, multiplied across many facilities, introduces operational uncertainty that current planning and operating practices were never designed to handle.

Rethinking the Interconnection Process

The generator interconnection process has evolved significantly over the past two decades, particularly in response to the rapid growth of inverter-based resources. The large load side, by comparison, often lacks the same level of standardization, transparency, and rigor. Many utilities still administer load interconnections through serial, non-transparent processes with limited published data and inconsistent requirements across customers.

A robust large load interconnection framework should include the following elements.

1. Safeguarding Existing Customers

Before any new large load is brought online, the host utility and regulator must ensure that existing customers are not disadvantaged. This means careful attention to tariff design, cost allocation, and rate structure. Tools to consider include customer-paid interconnection and network upgrade costs, financial guarantees and application fees, exit penalties and minimum demand charges, demand ratchets tied to ramp-up schedules, and minimum load factor requirements.

2. Raising the Barrier to Entry

When the cost of submitting an application is low, speculative requests flood the queue and make it difficult for credible developers to get timely study results. A higher, more thoughtful barrier — financial, technical, and commercial — filters out projects that are unlikely to be built. The generation side has been moving in this direction for years. The load side can benefit from the same lesson.

3. A Transparent and Streamlined Process

A modern large load interconnection process should mirror the structure of generator interconnection: application, initial screens, feasibility study, system impact study (potentially clustered), facility study, interconnection agreement, construction, commissioning, and post-interconnection monitoring. Each phase needs clear timelines, documented deliverables, and consistent treatment across all customers.

4. Technical Requirements

Until recently, very few US jurisdictions had formal technical requirements for large loads — a stark contrast to European practice, where grid codes have imposed structured requirements on large load customers for over a decade. Requirements should address, at minimum: ride-through behavior under voltage and frequency excursions, reactive power and voltage response, power quality limits, protection coordination, monitoring and data sharing obligations, and load ramping characteristics. Without a coordinated regulatory framework, utilities that impose strong requirements risk losing development to neighboring jurisdictions with weaker standards — a dynamic that pushes everyone toward the lowest common denominator.

5. Quality Models and Studies

Sound decisions depend on sound models. For large load interconnections, that means a layered study approach scaled to the size and grid impact of the facility:

• Power flow studies to identify thermal overloads, congestion, and voltage problems driving network upgrade needs
• Positive-sequence transient stability studies for ride-through, voltage stability, and electromechanical oscillations
• Electromagnetic transient (EMT) studies for power electronic interactions, controller dynamics, phase-locked loop behavior, and subsynchronous phenomena
• Short circuit studies for breaker duty and protection coordination, especially where co-located or behind-the-meter generation is present

A 50 MW standalone load may not justify full EMT analysis. A gigawatt-scale AI campus sited near a large inverter-based resource almost certainly does.

6. Post-Interconnection Monitoring

Once a facility is energized, the engineering work is not finished. Operational monitoring is essential for catching performance issues before they escalate into system events. This requires defined data sharing agreements, continuous telemetry, and a feedback loop between the customer's operations team and the utility's planning and protection functions.

The Modeling Challenge

A modern data center is not a simple lumped load. It contains power distribution units, cooling systems (increasingly served by variable frequency drives), lighting, UPS systems, backup generation, filters, and increasingly advanced devices such as behind-the-meter battery energy storage and e-STATCOMs used to smooth load variability for AI training workloads.

Representing this internal complexity in grid models is non-trivial. The level of detail required scales with the size and grid impact of the facility. Critically, much of the data needed to build accurate models — equipment specifications, control settings, protection setpoints — resides with the customer or the original equipment manufacturer. Establishing data sharing protocols with appropriate confidentiality protections is therefore a prerequisite to good modeling, not an afterthought.

The same modeling discipline that the industry has applied to inverter-based generation over the past decade now needs to extend to large loads. Protection settings, ramp characteristics, and ride-through behavior all need to be documented, verified, and kept current as equipment and firmware evolve — exactly the approach that Keentel Engineering's power system studies practice applies across generation, transmission, and now large load interconnections.

Mitigation Technologies

Several technologies are emerging or maturing to help integrate large loads more gracefully into the bulk power system.

E-STATCOMs — energy-augmented static synchronous compensators — combine traditional reactive power support with a small energy buffer behind the power electronics. This allows real power injection on millisecond timescales, enough to smooth out the fastest load variations even though they cannot sustain large injections for long. For AI training workloads that cycle rapidly, this capability addresses a grid reliability concern that conventional reactive compensation cannot.

Behind-the-meter battery energy storage can shave peak demand, provide ride-through support during grid disturbances, and enable flexible operation of the underlying compute or industrial load. As battery costs continue to fall, co-located storage is becoming a standard design element for large load interconnections rather than an optional add-on.

Grid-forming inverters and integrated energy parks — combining generation, storage, and load behind a coordinated control system — are increasingly being considered for the largest facilities, particularly where transmission capacity is constrained or reliability requirements are high.

Load flexibility itself is a powerful tool. Most transmission capacity is unused most of the time; the binding constraint is peak conditions. A load that can shift, curtail, or back off during those hours can often be accommodated using existing capacity that a firm, inflexible load could not. The technology to enable this is generally mature. The regulatory and commercial structures to support it are still evolving.

The Regulatory Question

A recurring question in this space is: who has the authority to impose technical requirements on large load customers?

Under current frameworks, transmission providers can establish interconnection requirements for any entity connecting to their system. Historically, those requirements have focused on generation. There is nothing preventing utilities and ISO/RTOs from extending similar requirements to large loads today — but doing so on a utility-by-utility basis produces inconsistent rules across regions and creates the race-to-the-bottom dynamic mentioned earlier.

A more uniform approach would benefit from action at the federal level. Reliability organizations face an additional barrier: their standards apply to registered entities, and there is currently no registration classification for large load customers. Until that gap is closed, even well-crafted reliability standards cannot be enforced against the facilities they are intended to govern.

The direction of travel across multiple industry forums and regulatory bodies is clear. The pace remains uneven. In the meantime, utilities acting under existing authority — by publishing clear technical requirements, requiring meaningful data from applicants, and enforcing ride-through and power quality standards at the interconnection agreement stage — can move faster than the federal process without waiting for it.

Frequently Asked Questions