Alarm Bells Are Sounding for the Grid: Answering Them Takes More Than Megawatts

Alarm Bells Are Sounding for the Grid: Answering Them Takes More Than Megawatts

Mon, August 10, 2026 at 6:45 PM GMT+3 7 min read

The PJM Interconnection set an all-time peak demand record on July 2 of this year, preliminarily estimated at more than 168,000 MW. That broke a mark that had stood since 2006. There were no rolling blackouts, and the commentary that followed carried a note of relief, even congratulation: the system held up. That is true. It is also dangerously incomplete.Consider the ledger of what holding up required:

• Two federal emergency orders under Section 202(c) of the Federal Power Act, one waiving environmental limits so plants could run past permit restrictions and one authorizing forced curtailment of data centers onto backup generation.

• A recall of generators from scheduled maintenance.

• Activation of emergency demand response.

• Operating reserves roughly halved in a single day to about 5,100 MW.

• And day-ahead prices topping $2,000/MWh in parts of the footprint, roughly triple a comparable peak a year earlier.

COMMENTARY

It was the third federal emergency intervention for PJM in 2026 alone, among dozens of such orders issued across North American grid operators this year. As I write this, another hot weather alert and emergency order request are in effect for the same footprint.When federal emergency powers become routine operating procedure, the system is not holding up. It is operating at its ceiling. And the customers who kept their lights on paid triple for the privilege, which means the event that operations survived is one that affordability did not.

EP+DPX:

[evtx_block slug="ep-dpx-26-textblock"]

None of this should surprise us. NERC's latest Long-Term Reliability Assessment projects summer peak demand could surge by 224 GW and winter peak by 245 GW over the next decade, upward revisions of 69% and 65% from projections made one year earlier. And as Power Magazine's coverage of the report captured, NERC's John Moura has already named the required response: the industry "must move beyond margin-only thinking" toward probabilistic and energy risk analysis, treating large loads as part of system planning rather than mere demand forecasts.Moura is right, and in my experience most utility leaders already agree with him. The harder question is no longer whether to adopt probabilistic planning. It is how to actually execute it. Because July 2 was not an anomaly; it was a preview. Customers are deploying flexible resources, from rooftop solar and batteries to managed EV charging, faster than utility planning and operations can absorb them. Capacity constraints are colliding with a new era of risk defined by extreme weather and cyber threats. The planning processes, operating practices and regulatory constructs that served the industry for a century were built for a world of stable, predictable demand, and that world is gone. Utilities are being pushed to the limit of what the current operating system can deliver, and regulators need to see that cliff before the industry reaches it, not through another emergency order, but through the plan.

Why the Old Approach Breaks

For decades, the demand curve for power in North America was barely a curve at all. Growth was stable and could be projected years in advance, so the industry planned deterministically: pick a forecast, plan to it, add a margin. Deterministic planning works when the future is predictable, volatility is low and the rate of change is small. But when change accelerates and variables multiply, its rigid assumptions fall out of date faster than plans can be revised.Load projections that get revised upward 65% in a single year are kryptonite to deterministic planning. So are climate-driven disasters, accelerating adoption of variable resources and fast-changing load shapes. The result is exactly what NERC describes: resource development falling behind demand and vulnerabilities rising across infrastructure the entire economy depends on.

What Probabilistic Planning Actually Means in Practice

Probabilistic, systems-based planning is built for volatility, but the phrase gets used loosely, so it is worth being concrete. In practice it means four things:1. Inputs become distributions rather than points: load, weather, outage rates, DER adoption and market prices are modeled as ranges with probabilities, so planners can speak in P50 and P90 terms rather than defending a single number.2. The system is evaluated at operational fidelity, across all 8,760 hours of the year, against a realistic model of the actual grid, so investments are tested against how the system truly behaves rather than a single peak snapshot.3. Candidate solutions are evaluated across a library of futures, not one forecast. The investments that hold up across many scenarios form a defensible foundation of no-regrets commitments, while single-scenario bets are exposed for what they are.4. Lastly and least appreciated, the plan stops being an annual document and becomes a continuously updated model, one that can be re-run as conditions change and opened to regulators and stakeholders so they can interrogate assumptions directly.A static report is stale the day it prints. A living model builds the credibility that rate cases increasingly demand, and demand it they do. For example, utilities requested $9.2 billion in rate increases in the second quarter alone, up 26% from a year earlier according to the nonprofit PowerLines. And commissions are interrogating the financial logic behind those requests with a rigor the industry has rarely faced.

Where First Attempts Fail

In our work helping utilities stand up these capabilities, the same failure modes appear so consistently that they are worth naming.Boiling the ocean: Teams wait for a perfect, all-encompassing system model before producing anything. The utilities that succeed start with the data they already have, typically years of AMI and SCADA history, and build fidelity iteratively.Models that do not govern: A sophisticated probabilistic model gets built, and capital decisions continue to be made the old way. Unless model outputs are wired into the capital approval process, the exercise produces insight and changes nothing.Scores creeping back in: Qualitative factors like safety and customer satisfaction get converted into weighted scores that make plans look precise while breaking the line from investment to value. Probabilistic planning replaces that scoring with explicit, interrogable assumptions. Guard against the old habits returning under new branding.Technology without operating change: The hardest lesson: modeled savings materialize only when the utility also changes how it plans and operates. A probabilistic model bolted onto deterministic practices delivers deterministic results.

A Practical Sequence

For utilities beginning the transition, the work sequences naturally into three horizons.

• First 90 days: Build the baseline from existing data. Stand up an integrated inventory across transmission, distribution and DER assets, and run an initial hourly analysis from AMI and SCADA history to establish how the system actually performs today.

• First year: Build the scenario library and connect it to decisions. Develop probabilistic forecasts for demand, weather and generation variability; stress-test the portfolio against extreme weather, cyber events and demand surges to locate cost-risk thresholds; and require the next capital cycle's major business cases to demonstrate performance across multiple futures.

• Ongoing: Operate the living model. Compare scenario projections against actual performance and refine iteratively, publish assumptions transparently to regulators and internal stakeholders, and invest in the data management, system modeling and version control infrastructure that keeps the model current.

Making the Cliff Visible

NERC has done the industry a service by quantifying the demand shock. The industry's job now is to be equally honest about the other side of the equation: the ceiling of the current operating system. That means bringing regulators inside the models, exposing the real trade-offs among reliability, affordability and risk, and building the shared case for the deeper transformation of people, processes and technology that is now required by one of the most complex systems humanity has ever built.Probabilistic planning is not the end of that transformation. It is the instrument that makes the need for it visible, measurable and impossible to ignore. The alarm bells have been rung, and on July 2 they were nearly drowned out by applause for a system that survived on emergency powers. The next July 2 is already on the calendar. The question is whether the industry answers it with more megawatts and more emergency orders, or with a fundamentally better way of deciding.—Dr. Matthew Green is Senior Vice President of Integrated Planning & Advisory at TRC Companies, a global professional services firm providing integrated strategy, consulting, engineering and applied technologies in support of the energy transition.