All five weekday entries (Monday September 7 through Friday September 11) are present in the research log. Four items are deduplicated across weeks rather than within this one and are reported as movement or as standing benchmarks rather than as new findings: Xcel Minnesota’s Capacity*Connect, reported in last week’s digest from a different outlet, is carried here only for the repayment mechanics the September 9 entry adds; the New Jersey BPU virtual-power-plant straw proposal now appears for the third consecutive digest and is carried as one line of context; the DOE generation-side 202(c) capacity-factor analysis (same June 23 source) was fully reported last week and is referenced, not re-reported; and EPRI’s DCFlex “flexibility for speed” piece (same June 26 source) was reported last week and anchors the EPRI Spotlight as a carry. Several items carry pre-week publication dates — the OATI/NCEMC deployment from February 20, GE Vernova’s GridOS launch from February 3, the Charles River Associates report from February 17, the Utility Dive VPP outlook from January 27, DOE Order 202-26-23 from May 18, the FERC ISO-NE order from May 29, the Columbia CGEP commentary from June 23, the PJM heat-wave account from July 20, and the FERC/PJM governance warning from July 24 — and they are reported on the day they entered the log with original dates shown. Two verification items were chased directly this week and are noted where they land.
For four months this series has tracked demand-side flexibility as a thing being designed — priced, tariffed, funded, litigated. This week the log produced something different and more useful: a performance record. Four separate items, arriving from a grid operator’s event log, a consultancy’s five-year retrospective, a capacity-auction breakdown and the IEA’s global accounting, all measure what demand-side resources actually delivered rather than what they were projected to deliver. The results do not flatter the newest constructs, and they flatter the oldest one considerably.
Start with the event. During the July 2–3 heat wave PJM issued the first “Emergency Use of Backup-Generator Warning” in its history — twice, to the Dominion zone, under the DOE Section 202(c) regime this series has tracked since June. It never escalated to an Action. No customer backup generator was dispatched. What did carry the peak was conventional demand response: roughly 6,113 MW shed on July 2 and 5,037 MW on July 3, against a served peak of 162,713 MW and an all-time unrestricted peak of 168,158 MW. The novel large-load reliability instrument stayed in its holster; the thirty-year-old one did the work, and did it with settlement-grade measurement behind it.
Then the retrospective. Charles River Associates read five years of FERC Order 2222 as a warning about the flexibility-centered large-load policies now being written, and the numbers are brutal: aggregated DERs account for under 0.2% of total system peak; PJM approved roughly 290 MW of aggregated backup generation against a system peak above 160 GW; FERC’s 270-day compliance clock became 2029 for MISO and Q2 2030 for SPP. Meanwhile the ~7 GW of legacy C&I demand response that predates Order 2222 keeps clearing, because it operates under established M&V. CRA’s formulation is the sentence of the week: “flexibility that cannot be verified, enforced, or translated into conservative planning assumptions is not a resource, it is a risk.”
Then the auction. In PJM’s 2026/2027 Base Residual Auction, capacity cleared at the $120,147/MW-year cap region-wide, and all roughly 8 GW of offered demand response cleared at 100%. DR’s ELCC accreditation slipped from 77% to about 69% — and a FERC-filed rule change lifts it to roughly 92% for 2027/2028 in exchange for DR moving to a 24/7 availability model on June 1, 2027. That is the price of admission stated plainly: a 23-point accreditation gain for an obligation that most residential aggregations are not currently built to meet.
And the global ledger. The IEA’s Electricity 2026 flexibility chapter finds only about 100 GW of demand response utilized worldwide against roughly 600 GW of residential air-conditioning peak and 160 GW of aluminum-smelting peak — attributing the gap to smart-meter penetration, weak price-signal exposure and aggregation rules that exclude small loads. Meanwhile utility-scale batteries added a record 63 GW in 2024 at costs down ~40% to ~$150/kWh.
Put them together and the week’s argument is uncomfortable but actionable. Demand response is the demand-side resource with a verified performance record, and virtual power plants are not yet — and the gap is measurement and obligation, not devices. Wood Mackenzie has North American VPP capacity growing only 13.7% to 37.5 GW in 2025, and EnergyHub’s peaker-equivalent “Turing Test” scores even the most advanced VPPs at roughly 2 out of 4 on plant-like performance. The programs that clear are the ones that can prove what they did.
The corollary runs the other way, and it is the constructive half. Every gate identified this week is a rule, not a physical limit. FERC spent the week removing one: its ISO-New England order in Docket ER26-1956 restored a qualification pathway for DERs that an Order 2023 compliance change had silently closed, leaving state-interconnected distributed resources with no method at all to establish capacity credentials in the Forward Capacity Market. Last week’s EL26-4 sampling order unlocked 4.9 GW the same way. Two consecutive weeks in which the binding constraint on distributed capacity turned out to be tariff plumbing is a pattern worth naming.
⚡ Virtual Power Plants & Demand Flexibility
Charles River Associates has produced the most rigorous skeptic’s case against the flexibility thesis this series has encountered, and every DR business case that assumes curtailable load substitutes for firm capacity should be tested against it. The February report, Grid under pressure: Flexibility-centered large load policies overlook recent lessons, reads the Order 2222 record as a direct roadmap for the large-load flexibility policies now being drafted, and its central distinction is the one planners most often elide: regulatory compliance is not market participation. Every RTO eventually filed an Order 2222 tariff. Aggregated DERs still account for under 0.2% of total system peak. PJM approved roughly 290 MW of aggregated backup generation in its 2023 auction against a peak above 160 GW. State-jurisdictional VPP and DER programs grew from fewer than 12 pilots to more than 100 — but overwhelmingly outside wholesale markets, which means outside the capacity revenue that makes them self-funding.
The timeline slippage is the part that should inform procurement schedules. FERC’s 270-day compliance target became five-to-ten-year implementations: MISO extended to 2029, SPP to Q2 2030, with only CAISO and ISO-NE near implementation. A utility writing a DERMS business case whose capacity revenue depends on wholesale aggregation in MISO or SPP is underwriting against a market that will not exist for the life of the first contract.
CRA’s prescription is a specification list, and it is nearly identical to the one this series has converged on from the opposite direction: hard telemetry, enforceable curtailment obligations backed by financial penalties, jurisdictional clarity, and cost certainty for the enabling infrastructure. The constructive reading is not that flexibility fails — it is that the ~7 GW of legacy PJM C&I demand response clears precisely because it has all four, while probabilistic aggregation without them does not. That is an argument for investing in the M&V and contracting layer before expanding device counts, and it is the strongest available rebuttal to a vendor pitch that leads with enrollment volume. (Source: Charles River Associates, February 17 — September 7 entry)
Utility Dive’s 2026 DER outlook supplies the counterpart measurement: VPP capacity is broadening faster than it is deepening, and by the industry’s own maturity test the best VPPs are roughly halfway to plant-like performance. Wood Mackenzie found North American VPP capacity grew only 13.7% in 2025, to 37.5 GW, at a moment when the AI data-center buildout is pulling utility capital toward gas, nuclear and coal. EnergyHub’s peaker-versus-VPP “Turing Test” — can a system operator tell the difference? — scores even the most advanced VPPs at about 2 on a 0-to-4 maturity scale, meaning they do not yet deliver reliable plant-like performance.
The cost argument runs the other way and remains strong: former DOE advisor Jen Downing notes VPPs are a proven, cost-effective option for peak reductions of up to 20%, and they compete directly against the diesel gensets hyperscalers install for a daily two-hour peak. That framing matters more this week than last, given that PJM’s July event showed backup generation being warned but never called — the gensets are an insurance product with no operating record, and a VPP with a settlement history is a different kind of asset.
Two signals are directly actionable. The procurement one: utilities are positioned to make foundational DERMS and data-management investments in 2026 because pressure to serve new load drives system-capability adoption, and “visibility will be key to VPP proliferation.” The valuation trap is the more important one — PJM raised the ELCC value for demand-side resources in the 2027-28 auction while simultaneously narrowing eligibility in ways that exclude many residential and small-C&I participants, and PJM’s own market monitor argues the rules still fail to link compensation to actual availability. A higher accreditation value on a narrower eligibility base is not unambiguously good news for a residential program, and a business case that models the ELCC improvement without modeling the eligibility screen will overstate revenue. (Source: Utility Dive, January 27 — September 7 entry)
The IEA devoted the flexibility chapter of Electricity 2026 to the demand side and called demand response a largely untapped breakthrough — roughly 100 GW utilized worldwide against a residential air-conditioning pool six times that size. This is the authoritative global anchor for a DR filing that has to defend demand-side capacity against firm alternatives, and this year’s edition pivots deliberately toward demand flexibility. Current utilization by sector: industry ~75 GW, buildings ~30 GW, transport under 5 GW, against roughly 600 GW of residential-AC peak and 160 GW of aluminum-smelting peak available in principle.
The IEA’s explanation for the shortfall is the same stack this series keeps arriving at: thin penetration of smart meters, controllable appliances and home-energy-management systems; limited customer exposure to short-term price signals; and aggregation rules that restrict small-load participation. That third item is the same barrier FERC removed in PJM last week and in ISO-NE this week, which is worth noting — the IEA is describing as a global structural condition something that two Commission orders have now shown to be a tariff choice.
The storage comparison sharpens the value stack rather than undermining it. Utility-scale battery additions hit a record 63 GW in 2024 (total 124 GW) with project costs down roughly 40% to about $150/kWh, and California’s batteries now equal nearly 25% of peak load. Cheap four-hour storage competes with DR for the same peak-shaving role — which is precisely why the program-design parameters this series keeps flagging (term length, availability obligation, minimum size) determine whether a DR portfolio remains the cheaper option or becomes a legacy program defended on sunk cost. The companion report, The Value of Demand Flexibility (December 2025), supplies the international best-practice framing regulators increasingly expect alongside Brattle-style avoided-cost methods. (Source: IEA, Electricity 2026 — September 8 entry)
Octopus Energy closed its majority investment in Uplight on September 1 and set a target of doubling flexible capacity from 8.5 GW to 20 GW in five years — a demand-side platform consolidation that reshapes the vendor list utilities buy DSM services from. Octopus, which operates the world’s largest virtual power plant across 11 million households in 27 countries, takes majority control with Schneider Electric remaining an investor, and Nick Chaset — CEO of Octopus Energy US — becomes Uplight CEO. The stated targets are $1 billion in customer savings and more than doubling flexible capacity to 20 GW over five years.
The installed base is the reason this is a market-structure event rather than a financing item. Uplight already manages 8.5 GW of flexible load across more than 85 utilities — including eight of the ten largest U.S. utilities — with 75-plus ecosystem partners, crossed one million enrolled customer devices in 2026, tripled program flexible capacity in the 2025 season, and has returned $586 million to customers since 2020. A utility running a DSM program through Uplight now has a vendor with different ownership, different capital, and a five-year growth mandate; that belongs in the next contract review regardless of how well the program is performing.
Two imported products sharpen the enrollment-and-dispatch problem a DERMS has to solve. Octopus Shift is a single U.S. enrollment and engagement application — the customer-facing layer that determines whether a nominally eligible device is actually enrolled and actually responds. Octopus PowerStore is a turnkey residential battery with no upfront customer cost, which is the third-party-funded route to dispatchable storage this series flagged in the PG&E/Google SHARE structure, now available as a vendor product rather than a bespoke corporate partnership. Read against the week’s performance theme, the “Demand Stack” framing — efficiency, rates, electrification and flexibility packaged into near-term capacity with “planning-grade reliability” — is a claim to exactly the credential CRA says aggregation lacks. It is worth asking a vendor making that claim what evidence supports the adjective. (Source: Uplight / Octopus Energy Group, September 1 — September 8 entry)
Carried items — Xcel Minnesota and New Jersey. Xcel’s Capacity*Connect appeared in last week’s digest as the rate-based point in the three-funding-structure comparison; the September 9 entry adds the repayment mechanics, which change how the program should be cited. The Minnesota PUC approved up to $430 million for up to 200 MW of batteries in 1–3 MW increments over two years at commercial, industrial and nonprofit sites, sited “at strategic locations on the grid” to test relief of local distribution constraints rather than territory-wide peak shaving. The economics are the new detail: MISO energy and capacity revenues are expected to repay almost the entire deployment cost, holding the typical residential bill impact to $0.67–$1.50 per year through 2030, with a $50 million Google contribution tied to its in-state data center. That materially softens the ~$2,150/kW headline this series has been using as the rate-based comparable — the gross capital number is not the ratepayer number, and a fair comparison against the $624/kW Colorado third-party VPP has to net the market revenue. The ownership objection from SEIA, MnSEIA and CCSA stands, as does the standing watch item: specific distribution-benefit estimates by the November 2027 Integrated Grid Plan, with quarterly reports before then, are what will eventually test whether the siting premium bought locational value. (Source: Canary Media, April 8 — September 9 entry; previously reported in the September 4 digest)
New Jersey (Docket QO26030099, comments closed August 17, launch targeted 2027) appears for the third consecutive digest and is carried in one line: it remains the one jurisdiction in this series’ compensation-design comparison where the performance-versus-enrollment question is still genuinely open, and this week’s performance evidence is exactly what should settle it. (Source: NJ BPU, July 27 — September 9 entry; previously reported August 28 and September 4)
🔌 DERMS & Grid Integration Technology
North Carolina Electric Membership Corporation is moving an existing OATI DERMS from forecast-based scheduling to real-time, sub-minute orchestration across 26 distribution cooperatives — the cleanest deployment-stage DERMS datapoint this series has logged, and it comes from the co-op segment rather than an IOU. NCEMC and OATI are extending a multi-year partnership to enable real-time communication and dynamic dispatch to batteries, solar, demand-response assets and five islandable microgrids, coordinated with OATI’s GridMind microgrid controls, so resources can be called proactively under program schedules and reactively under emergency conditions.
The number that makes this a procurement reference rather than a press release is the penetration figure: roughly 15% of NCEMC’s peak need is already served by DERs, utility- and customer-owned. That is approximately the threshold at which fixed scheduling stops working — forecast error at 15% of peak is no longer a rounding item in the day-ahead position — and it gives a planning team a concrete trigger point to cite when arguing that real-time orchestration has become load-bearing rather than aspirational. The capability progression is itself the specification: an RFP should distinguish scheduling from dispatch explicitly, because a platform can satisfy the first for years before the second is tested.
Two further procurement signals. First, microgrid integration as a grid-supporting resource rather than islanded backup — the five islandable microgrids are orchestrated within the same platform, not managed as separate assets, which is a requirement worth writing rather than assuming. Second, resilience as the funding driver: the post-Hurricane Helene state commitment of $5 million to develop 26 microgrids is what moved this from roadmap to deployment, a reminder that storm-recovery capital is an underused funding channel for orchestration infrastructure that a reliability business case alone may not clear. For vendor evaluation, the co-op reference base broadens the comparable set meaningfully — a G&T cooperative coordinating 26 member distribution utilities has a multi-party governance problem that closely resembles a large IOU’s multi-jurisdictional one. (Source: Microgrid Knowledge / OATI, February 20 — September 11 entry)
GE Vernova’s GridOS for Distribution bundles ADMS and DERMS with GIS network modeling, field execution and visual intelligence on a single data fabric — collapsing the integration decision this series has been tracking into a single-vendor buy, with Alabama Power among early adopters. Announced February 3 and showcased at DTECH 2026, the platform combines Smallworld Geo Network Management, GridOS ADMS, GridOS DERMS, GridOS Field and GridOS Visual Intelligence on a governed, federated grid data fabric, with the pitch that a utility can operate distribution “as one intelligent, orchestrated system.”
For RFP scoping this reframes the central evaluation axis. The question is no longer only what functions does the DERMS perform but best-of-breed integrated by the utility versus a pre-integrated suite from one supplier — a build-versus-buy and lock-in judgment that OT, IT and IRP teams weigh differently and that should be settled before requirements are written rather than discovered during scoring. It also confirms the reading this series recorded from the Itron essay: DERMS has moved from the innovation budget into the capital plan alongside ADMS, and a vendor packaging both as a single capital purchase — paired with a shift toward subscription licensing — is the commercial evidence for that claim.
The buyer-side caution is structural and belongs in the contract rather than the evaluation matrix. A unified data fabric concentrates capability and dependency in the same place, which makes data governance, cybersecurity, open-standard APIs and exit terms first-order commercial terms rather than schedule exhibits. A utility that cannot extract its own network model, DER registry and settlement history in a usable form has made a longer commitment than the contract term states. (Source: GE Vernova, February 3 — September 8 entry)
A September 7 market forecast sizes the capital wave DERMS procurement sits inside, and supplies the one number most useful to a small utility or cooperative: $2 million to $5 million per thousand customers for full-scale deployment. Research and Markets puts the global digital power utility market at $142.68 billion in 2026 rising to $213.88 billion by 2031, an 8.42% CAGR, with North America at 37.45% of 2025 revenue and Asia-Pacific growing fastest at 12.25%. Notably for this series, the explicit demand drivers named include “data-centre flexibility procurement surge” and FERC Order 2222‘s opening of wholesale markets to aggregated DERs — the two threads this digest runs, appearing as line items in a market-sizing model.
The barrier figure is the part with operational value. Full-scale smart-grid deployments running $2M–$5M per 1,000 customers, with cost-recovery periods that can exceed a standard regulatory cycle, quantifies why third-party funding structures and avoided-cost stacking matter disproportionately for smaller utilities and cooperatives — and it is the financial context behind NCEMC’s Helene-funded microgrid build above. A recovery period longer than the rate cycle is a regulatory-risk problem, not just a financing one.
The vendor roster — ABB, GE, Hitachi Energy, Itron, Landis+Gyr, Oracle, OSI, Schneider Electric, Siemens and AutoGrid — is a serviceable RFP long-list, and the report flags open-standard APIs and component-level cybersecurity as rising procurement requirements, converging with the GridOS caution above. As a subscription market-research promotion carrying no primary deployment data, it is weighted accordingly and used for sizing and drivers rather than for any single claim. (Source: Research and Markets via GlobeNewswire, September 7 — September 7 entry)
🏗️ Data Centers & Large Load Growth
PJM issued the first “Emergency Use of Backup-Generator Warning” in its history during the July 2–3 heat wave, never escalated it to an Action, and dispatched no customer generation — while conventional demand response shed roughly 6,113 MW and 5,037 MW on the two days. The new large-load reliability tool got its first operational test and the old one carried the peak. This is the operational sequel to the Section 202(c) orders this series has tracked since June, and it is the most valuable single item in the week’s log because it converts an asserted capability into a measured one — in both directions.
The reliability context frames the stakes. Served peak reached 162,713 MW and estimated unrestricted demand hit 168,158 MW, an all-time unrestricted record. PJM’s Market Monitor characterized the warnings as “precautionary” while noting the system “was getting very close to a shortage of primary reserves.” The companion DOE emissions order spanned 51 plants / 594 units / ~20.4 GW, of which 17 plants (~3.25 GW) ran beyond normal emissions limits for 1,107 combined hours.
For DR valuation, the 6,113 MW and 5,037 MW figures are the kind of in-event performance data that does more for a capacity value in an IRP filing than any amount of modeled availability — dispatched, verified, and delivered at the system’s annual maximum. Set them against the CRA finding above: this is the legacy, metered, M&V-established C&I fleet performing exactly as accredited, at the moment of highest system value, in the same footprint where aggregated DERs sit under 0.2% of peak.
The gap the series should flag is on the other side of the ledger. PJM has not formally adopted the 50 MW large-load threshold into its governing documents, and it declined to disclose how many megawatts of backup generation were even available. So the “flexible large load as reliability resource” proposition remains asserted rather than measured: the instrument was invoked, held in reserve, and never tested, and no one outside PJM knows the size of the fleet it would have called. That is the strongest available argument for pre-contracted, metered demand-side capacity over untested emergency backstops — not because the backstop is wrong, but because after its first real exercise it still has no performance record, and a resource with no performance record cannot be planned against. (Source: Data Center Knowledge, July 20 — September 11 entry)
MISO’s “zero injection” interconnection framework drew support filings from Google, Xcel, AES Indiana, Ameren, Dairyland, SEIA and Advanced Energy United on September 8 — a 90-day fast track for generation colocated behind the substation with the large load it serves, and the first product of MISO’s new Large Load Working Group. The Zero Injection Generator Interconnection Agreement (ZGIA) proposal, filed August 18, creates an expedited review path for generation serving a colocated large load at the same substation and voltage level with “little or no effect on the transmission system.” It is a partial response to FERC’s mid-June show-cause orders finding RTO large-load interconnection rules inadequate, with full responses due mid-November.
The eligibility guardrails are what planners should read closely, because they define the boundary of the behind-the-meter carve-out: a qualifying project may have no more nameplate capacity than the load it serves and cannot trigger network upgrades beyond the substation (except adjacent-substation protection). Google is already pushing MISO to extend the concept to two-substation configurations to unlock more generation on an expedited basis — which is the tell that this boundary will be renegotiated upward, not held.
For DERMS, IRP and hosting-capacity work the consequence is direct and underappreciated: ZGIA changes how much of the load surge ever lands on the shared grid. Load served by colocated zero-injection generation does not appear in the same place in a distribution or transmission forecast as load served from the system, which changes hosting-capacity headroom, peak-forecast assumptions and the avoided-cost denominator against which every demand-side program is evaluated. It is also, bluntly, a competitor: a 90-day path to power is a faster answer to interconnection backlog than any orchestration program, and a utility whose flexibility pitch rests on speed-to-power should expect this comparison. The Sierra Club and NRDC objection — that fast-tracked colocated generation could “operationalize ongoing cost shifts” onto existing customers and let ZGIA projects jump ahead of Definitive Planning Phase resources — is the same ratepayer-equity tension attached to every large-load accommodation in this series, and MISO is separately building a fast-track Load Addition and Resource Study process alongside it. (Source: Utility Dive, September 9; MISO filing August 18 — September 10 entry)
LBNL and the Brattle Group found large-load tariff filings have multiplied to 264 catalogued items, with minimum contract terms jumping from an average of five years to twelve — and a new layer of “emerging” risk protections that read as a ratepayer-protection playbook. Drawing on Halcyon’s large-load tariff tracker, which had catalogued 264 tariff filings as of August 17, the researchers analyzed 55 tariffs and isolated 18 rate-design elements. Eight “established” practices — minimum demand thresholds, minimum contract durations, collateral — are now joined by five “emerging” ones: nonrefundable upfront payments for system-impact studies, mandatory ramp-to-full-load schedules, “hold harmless” cost-recovery makeup provisions, resizing conditions, and substantial early-exit fees.
The duration finding is the headline and it deserves to be quoted precisely: tariffs proposed before 2025 averaged five-year minimum terms; those after 2025 average twelve. Demand thresholds range from under 1 MW to 150 MW, site-by-site or aggregated — which is the threshold-divergence problem this series has flagged for three weeks, now measured across a real population rather than inferred from a handful of RTO filings. Benchmarks at the extremes: Entergy Louisiana at five years, El Paso Electric proposing twenty.
This cuts two ways for the demand-side business case, and both are worth having ready. It validates the stranded-asset argument that underpins deferring generation and T&D capital with demand response — utilities are demanding firm, long-dated commitments precisely because speculative load is a ratepayer risk, which is the same logic in a different direction. And it raises the bar a data center must clear, making flexible or interruptible service and behind-the-meter resources structurally more attractive than a firm tariff loaded with exit fees and a twelve-year term. Read alongside ZGIA above, the picture is coherent and slightly alarming for anyone selling flexibility: large loads facing twelve-year firm commitments have two escape routes — become flexible, or go behind the substation — and only one of them helps the utility. (Source: Utility Dive, September 8 — September 10 entry)
Columbia’s Center on Global Energy Policy argues data-center-driven bill increases are not inevitable, naming grid-enhancing technologies and demand response as the near-term levers — and identifies the structural reason they lag: the ~9–10% allowed return rewards steel in the ground over system optimization. The diagnosis is that prices rise “when new demand triggers high-cost infrastructure buildout, reflects inefficient planning, or shifts costs unevenly across customers,” and that data-center load shaping manages peak “without relying on high-cost, low-utilization generation” — the same avoided-cost logic behind this series’ ~$66/kW-year DR benchmark.
The affordability backdrop is the material worth carrying into a rate proceeding: residential retail prices rose roughly 6% nominally last year, more than twice inflation; investor-owned utilities sought $18 billion in rate increases, the most since the mid-1980s; and regulators approved 66% of the requested dollar value. The load-growth sizing from the companion LBNL work: data centers reaching 9.5–15.3% of U.S. electricity by 2030, up from 4.7% in 2024.
The structural finding is the one this series has not previously stated so directly. Under traditional cost-of-service regulation utilities earn roughly 9–10% on capital, so “this regulatory structure rewards capital deployment more than system optimization” — and planning frameworks oriented toward proving need for new capital, rather than evaluating lower-cost alternatives, will keep GETs and DR deployment lagging regardless of their cost-effectiveness. That is an argument for performance-based regulation and for weighting demand-side alternatives explicitly in IRP screening, and it is worth pairing with Virginia’s HB434 non-wires screening mandate from last week’s digest: a screening obligation is the procedural fix for exactly the incentive problem CGEP describes. On the GETs side, advanced-conductor reconductoring is credited with up to $180 billion in savings by 2050 in the cited PNAS work — a reminder that the demand-side portfolio’s nearest competitor for the “cheaper than a new build” slot is not always another demand-side program. (Source: Utility Dive, June 23 — September 11 entry)
📋 Regulatory & Policy
FERC restored a capacity-market qualification pathway for distributed energy resources in ISO-New England that an Order 2023 compliance change had quietly closed — the second consecutive week in which the binding constraint on distributed capacity turned out to be tariff plumbing. Approved May 29, 2026 in Docket ER26-1956-000, the revisions create equivalent Network Resource Capability (NRC) and Capacity Network Resource Capability (CNRC) qualification rules for DERs. The gap arose in April 2024, when ISO-NE moved its Capacity Network Resource Interconnection Service study into the cluster process to comply with Order No. 2023: because DERs interconnect through state procedures rather than ISO interconnection, they were left with no method whatsoever to establish the credentials required to sell capacity. FERC found the fix just and reasonable, holding that it provides comparable treatment to transmission-connected resources and could “potentially allow for broader participation.”
The reason this belongs alongside last week’s EL26-4 sampling order rather than in a footnote is the pattern. Two consecutive weeks, two Commission orders, and in both cases a substantial block of distributed capacity was excluded from a wholesale market not by physics, economics or performance but by a procedural artifact — one a metering requirement aggregators could not satisfy, the other a qualification pathway that compliance drafting inadvertently deleted. For a business case, the lesson is a diligence step: before modeling wholesale capacity revenue for a DER portfolio, confirm that the specific qualification pathway for state-interconnected resources exists in the current tariff and has been exercised. It is a question most financial models assume away.
The timing matters locally. This lands as ISO-NE works toward its November 1, 2026 Order 2222 energy and ancillary-services implementation, with capacity-market DERA participation following February 1, 2027 for the 2028/29 capacity year — both standing watch items in this series. A companion provision grants all resources a one-time two-and-a-half-year extension of the three-year window to resume commercial operation after a forced outage where long-lead equipment is on binding order, which is a supply-chain accommodation directly relevant to storage project timelines and to the equipment-availability risk flagged in last week’s bulk-power executive-order item. (Source: Troutman Pepper, Washington Energy Report, May 29 order — September 10 entry)
PJM’s 2026/2027 capacity auction gives demand response its clearest valuation datapoint of the year — 100% of roughly 8 GW offered cleared at the $120,147/MW-year cap — and a FERC-filed change trades a 23-point accreditation increase for a 24/7 availability obligation starting June 1, 2027. Capacity cleared at the FERC-approved cap region-wide (against a $64,693/MW-year floor), and an uncapped simulation would have cleared at $141,828/MW-year — meaning the cap, not the market, set the price, and the underlying scarcity is worse than the clearing price shows. Reserve margin landed at 18.9% against a 19.1% target, short by 0.2 points (~309 MW), on a target that itself rose from 17.8%; peak load forecast moved from roughly 154 GW to 159 GW, with data-center growth the primary driver.
The DR mechanics are the operative content. Roughly 8 GW of demand response was offered and 100% of it cleared. DR’s ELCC accreditation fell from 77% to about 69% for 2026/2027 — and a FERC-filed rule change lifts it to approximately 92% for 2027/2028 in exchange for DR moving to a 24/7 availability model as of June 1, 2027.
That trade is the single most consequential forward-looking item in this week’s digest for anyone running a DR portfolio, and it has to be modeled as a package rather than as a favorable accreditation revision. Capacity prices at the cap raise the revenue attributable to each accredited megawatt sharply; the ELCC derate and the coming availability obligation determine how many nameplate megawatts become accredited ones. A summer-afternoon-only residential program is not a 24/7 resource. Programs whose enrollment terms, notification windows or opt-out provisions assume a seasonal peak-hour obligation will need contract amendments before June 2027, and the customer-facing terms are the long pole — not the platform. Read against this week’s performance theme: PJM is raising the accreditation value of demand response toward parity while simultaneously raising the standard of proof, which is consistent with everything else the week produced. (Source: Enel North America, July 22 — September 11 entry)
FERC’s deadline for PJM governance reform is this month, and a direct check confirms the dispute-resolution forum convened on September 1. FERC Chair Laura Swett told the July 23 technical conference that PJM must agree to governance and stakeholder reforms by the end of September 2026 or the Commission will impose them, describing a “grave legitimacy crisis” rooted in two capacity auctions that failed to attract significant new generation as data-center demand outran supply — auctions that hit the price cap and drove 20%-plus rate increases for some utilities, with AEP and other transmission owners openly weighing departure from the RTO. Reforms under discussion: board independence, an advisory MISO-style stakeholder process replacing sector-weighted voting that lets two sectors block measures, and Section 205 filing rights for states.
A targeted verification search this week (not from the daily log) confirms the process is live and on schedule: FERC’s Director of Dispute Resolution Services was directed to convene an interest-based forum commencing September 1, 2026, with PJM invited to circulate a preliminary set of proposed reforms to registered participants on or before August 26. The Commission also took post-conference comment under a Federal Register notice issued August 4. The end-of-September deadline is therefore roughly three weeks out at publication with a structured negotiation already underway.
Why this sits in a DER digest: PJM’s demand-side rules are set through exactly this machinery. ELCC values, demand-side eligibility screens, availability-linked compensation and the sampling methodology FERC ordered under EL26-4 are all products of the stakeholder process now under Commission pressure — and the market monitor’s standing criticism is that the rules fail to link compensation to actual availability, which is the same question the 24/7 ELCC trade above is trying to answer. For any PJM-footprint DR or VPP business case, near-term capacity value is a moving target attached to a governance fight with a September deadline, which argues for value stacks that do not over-rely on a single auction’s demand-side rules. A FERC-imposed outcome would also set precedent for how aggressively federal regulators reshape RTO governance under AI-driven load growth. (Source: Utility Dive, July 24 — September 8 entry; process status verified via FERC notices)
The Section 202(c) record expanded again on both sides of the meter — a new Duke Energy Carolinas order in the Southeast, and the first detailed account of the May PJM curtailment authorization. Order No. 202-26-43, issued to Duke Energy Carolinas on September 3 in response to a same-day application and effective September 3–8, authorizes Duke to dispatch specified units and, in coordination with its Transmission Owners, to direct backup generation at large loads to operate as a last resort before or during an EEA 3. The filing package includes a Southern Power consent letter and a VACAR South reliability-coordinator endorsement — the multi-party coordination a real curtailment event requires across an integrated Southeastern footprint, and a useful template for what an actual event would demand operationally.
The May order supplies the detail this series lacked. DOE Emergency Order 202-26-23, issued May 18 at PJM’s Sunday request, allowed PJM to direct transmission owners and utilities across its Mid-Atlantic and Midwest footprint to curtail data centers and other large loads that have backup generation, as a last resort before rolling blackouts, for three days of unseasonable May heat. The triggering arithmetic is a textbook shoulder-season squeeze: under 5,800 MW of projected reserves at the May 18 peak with more than 40 GW offline for planned spring maintenance, and three-day peak forecasts of 134,027 / 135,961 / 119,103 MW, with Maryland and Virginia most stressed. PJM concurrently activated pre-emergency demand response in the BGE, Dominion and Pepco zones to relieve local constraints — the same tiering that would recur in July.
DOE’s own framing is the part to carry into a large-customer conversation: backup generation at hyperscaler and industrial sites has “remained largely untapped during grid emergencies” and can “prevent avoidable blackouts.” That is the federal government describing behind-the-meter large-load backup as an emergency demand-response fleet — the same orchestration logic a DERMS applies to distributed storage and interruptible load, invoked without compensation, telemetry or a settlement record. Three observations follow. First, the recurrence is now plannable: May, July and September orders across PJM and the Carolinas in a single year is a pattern, not a series of surprises. Second, the generation-side use of the same statute continues to perform poorly — the 65%-down capacity-factor record reported in last week’s digest is unchanged and remains the best available rebuttal to “just keep the plant running.” Third, and most usefully after this week’s PJM heat-wave account: in the one event where the load-side authority was actually exercised to the warning stage, it was not used, and metered demand response was. (Sources: U.S. Department of Energy, September 3; Utility Dive, May 19 — September 7 and September 10 entries; generation-side analysis previously reported September 4)
🔬 EPRI Research Spotlight
A direct check confirms no new major EPRI publication entered the research log this week — the ninth consecutive dry week. A targeted search of EPRI’s media resources surfaced nothing newer than the DCFlex expansion and Aurora materials this series has been carrying since July. As established three weeks ago, the gap is now reported as a finding about publication cadence rather than as a coverage omission: EPRI’s substantive data-center-flexibility output is landing through the DCFlex program and partner announcements rather than through standalone research releases, and a utility relying on EPRI publications as its flexibility-research feed should adjust where it is looking.
The DCFlex thesis returns to the log this week, and against the week’s performance evidence it reads differently than it did in September. The DCFlex Initiative and its FlexMosaic framework — organized around flexibility classes keyed to notification time, duration, frequency, depth and speed of response — conclude that data centers willing to be flexible can maximize their “speed to power,” directly addressing the five-to-seven-year interconnection queues throttling load growth. Pilots with NVIDIA and Emerald AI software have delivered up to 40% load flexibility by pausing or slowing training workloads and redirecting inference queries to less-stressed systems, with the 96-MW Aurora AI Factory and a Silicon Valley Power pilot cited as templates. The linchpin, per the program’s own framing, is a standardized binding agreement between utilities and data centers, now forming through EPRI and the Open Compute Project Foundation — with NARUC’s Ann Rendahl arguing that such pilots let state regulators require data centers to take flexibility seriously in interconnection.
Here is the week’s uncomfortable pairing. DCFlex demonstrates 40% workload flexibility under test conditions. CRA demonstrates that five years of Order 2222 produced aggregated DER participation under 0.2% of system peak, and PJM’s July event demonstrates that the large-load backup instrument was warned but never called. The distance between demonstrated technical capability and dispatched, settled megawatts is the entire subject of this digest, and DCFlex’s own answer — the binding standardized agreement — is precisely the right one. Technical flexibility without an enforceable contract, verified telemetry and a settlement record is the same category of asset CRA calls a risk rather than a resource. That is not a criticism of the program; it is the argument for prioritizing the contracting workstream over the demonstration workstream when both compete for the same staff.
Aurora’s timing problem is unchanged and now fully consequential. The coalition — NVIDIA, Emerald AI, EPRI, Digital Realty and PJM — still targets late 2026 for the Manassas, Virginia facility, having originally slated it for the first half of the year, and this week’s verification search surfaced nothing indicating either acceleration or further slip. NERC’s computational-load standards are due at FERC by December 31, which means the reference design’s demonstration data — precise, real-time responses to simulated grid-stress events including mimicked heat-wave demand spikes and sudden renewable drops — will very likely arrive after the standards are filed. The standards will be written on modeling rather than measurement. The coalition estimates the reference design, adopted nationwide, could unlock roughly 100 GW on the existing system, a figure Duke’s Nicholas Institute reaches by an entirely independent method.
FlexMosaic remains the only published vocabulary precise enough to price this week’s three separate flexibility products against one another: PJM’s warned-but-unexercised backup-generation transition, MISO’s ramp-limit and ride-through interconnection preconditions, and the dispatchable-flexibility options now appearing in roughly a quarter of large-load tariffs. Those are three different response characteristics being procured in three venues with no shared grammar, and the DR ELCC 24/7 availability change adds a fourth axis — continuous obligation versus event-based response — that FlexMosaic’s notification-and-duration keys handle and that tariff language generally does not. The DCFlex coalition spans Compass Datacenters, Constellation, Emerald AI, Google, National Grid, Nebius, NVIDIA, Oracle and PADO AI across nine demonstration sites including Europe.
Standing watch: Aurora commissioning data and any further slip; the CLO-001-1 comment record after the September 18 close and the fall ballot results; FlexMosaic uptake in the ER26-3515 and RM26-4 compliance records; and an EPRI.com direct check again next run — tenth week.
🚩 Utility-Sector Relevance Flags
⚑ Demand Response Now Has a Performance Record and VPPs Do Not — Use the Right One in the Right Proceeding
Topic: DR Valuation / VPP Business Case / Evidentiary Strategy
Relevance: In a single week the log produced four independent measurements pointing the same direction. PJM’s July 2–3 heat wave: conventional DR shed 6,113 MW and 5,037 MW at an all-time unrestricted peak of 168,158 MW, while the new large-load backup-generator instrument was warned twice and never dispatched. CRA: five years of Order 2222 produced aggregated DER participation under 0.2% of system peak and ~290 MW of approved aggregated backup in PJM, while ~7 GW of legacy C&I DR with established M&V keeps clearing. PJM’s 2026/2027 auction: 100% of ~8 GW of offered DR cleared at the $120,147/MW-year cap. EnergyHub’s Turing Test: the most advanced VPPs score about 2 of 4 on peaker-equivalent maturity. The difference between the two categories is not device capability; it is verified performance under an enforceable obligation.
Action Signal: Implement — Separate the two resources in filings rather than presenting a blended “demand-side portfolio.” Cite dispatched DR performance where a proceeding demands proven capacity, and present VPP programs as what they currently are: capability being built toward an accreditation standard, with a stated timeline for reaching it. Blending them invites an intervenor to apply the weaker record to the whole portfolio. Internally, put the July event numbers into the standing DR evidence file — an in-event shed at the annual peak is worth more than any availability model.
⚑ PJM’s DR Accreditation Goes to ~92% on June 1, 2027 — and the Price Is a 24/7 Availability Obligation
Topic: Capacity Accreditation / Program Terms / Contract Amendment
Relevance: DR’s ELCC fell from 77% to ~69% for 2026/2027 and a FERC-filed change lifts it to roughly 92% for 2027/2028 in exchange for a 24/7 availability model effective June 1, 2027. That is a 23-point accreditation gain — meaningful at a capacity price sitting at the $120,147/MW-year cap — conditioned on an obligation most residential and small-C&I programs are not structured to meet. PJM separately raised the demand-side ELCC in the 2027-28 auction while narrowing eligibility, and the market monitor still argues compensation is not tied to actual availability. A model that takes the accreditation improvement without the eligibility screen and the availability obligation will overstate revenue materially.
Action Signal: Implement — Model the accreditation change as a package, not a revision, and start the contract work now. The binding constraint is customer-facing terms — notification windows, seasonal restrictions, event caps, opt-out provisions — not platform capability, and amending enrollment terms across a residential population takes longer than a capacity year. Identify which programs in the portfolio can credibly carry a continuous obligation and which cannot, and decide deliberately which ones you will stop accrediting rather than discovering it at qualification.
⚑ Two Consecutive FERC Orders Found the Binding Constraint on Distributed Capacity Was Tariff Plumbing
Topic: Wholesale Market Access / Qualification Diligence / Revenue Modeling
Relevance: Last week FERC ruled PJM’s interval-meter requirement unjust and unreasonable in EL26-4, unlocking at least 4.9 GW. This week’s item is structurally identical: in Docket ER26-1956 (order May 29) FERC restored NRC/CNRC qualification pathways for DERs in ISO-New England after an April 2024 Order 2023 compliance change left state-interconnected resources with no method at all to establish capacity credentials. Neither exclusion was about performance, economics or physics. Both were drafting artifacts that silently removed gigawatts from a capacity market, and both took a Commission proceeding to reverse.
Action Signal: Implement — Add a qualification-pathway diligence step to every DER business case that books wholesale capacity revenue: confirm, in the current effective tariff, that a pathway exists specifically for state-interconnected resources, that it has been exercised by a real resource, and what the evidentiary requirement is. Financial models routinely assume market access as a given. Two orders in two weeks say it is a variable — and one worth checking in your own RTO before the next filing rather than after.
⚑ The PJM Governance Deadline Lands This Month and PJM Demand-Side Rules Are Downstream of It
Topic: Capacity-Market Governance / PJM Value Stack / Regulatory Risk
Relevance: FERC Chair Swett’s end-of-September deadline for PJM governance and stakeholder reform is roughly three weeks out, with a Dispute Resolution Services forum convened September 1 and PJM’s preliminary reform proposal circulated August 26. Reforms on the table — board independence, a MISO-style advisory stakeholder process replacing sector-weighted voting, Section 205 filing rights for states — would change who can block a demand-side rule change. ELCC values, eligibility screens, availability-linked compensation and the EL26-4 sampling methodology are all set through this machinery, and two capacity auctions clearing at the cap with 20%-plus downstream rate increases are what put it under Commission pressure.
Action Signal: Watch — Do not rebuild a value stack on an unresolved governance proceeding, but do stress-test any PJM-footprint DR or VPP business case against a scenario in which demand-side eligibility and accreditation rules change on a FERC-imposed timeline rather than a stakeholder one. If your utility participates in the PJM stakeholder process, the sampling-methodology proceeding under EL26-4 and the governance forum are now the same fight over who sets demand-side rules, and they should be staffed together.
⚑ Colocated “Zero Injection” Generation Is a Faster Answer to Interconnection Than Any Flexibility Program
Topic: Large-Load Interconnection / Load Forecasting / Hosting Capacity
Relevance: MISO’s ZGIA proposal (filed August 18, supported September 8 by Google, Xcel, AES Indiana, Ameren, Dairyland, SEIA and Advanced Energy United) creates a 90-day review path for generation colocated with large load at the same substation and voltage, capped at no more nameplate than the load served and no network upgrades beyond the substation. Google is already pressing to extend it to two-substation configurations. Set beside the LBNL/Brattle finding that large-load tariff minimum terms jumped from five to twelve years with new exit fees and upfront study deposits, large loads now have two ways out of a firm-service commitment: become flexible, or go behind the substation. Only one of those keeps the load in your forecast.
Action Signal: Engage — Quantify the colocation exposure in your own load forecast explicitly: which prospective large loads in the queue could be served by behind-the-substation generation, and what happens to hosting-capacity headroom, peak forecasts and the avoided-cost denominator if they are. A demand-side program justified against a load forecast that assumes system service is exposed to a forecasting error, not just a competitive one. If you are in MISO, the mid-November show-cause response tranche is where these terms get set.
⚑ The Large-Load Tariff Population Is Now Measurable — 264 Filings, Twelve-Year Terms, Five New Risk Provisions
Topic: Large-Load Rate Design / Comparables / Avoided-Cost Framing
Relevance: LBNL and Brattle catalogued 264 large-load tariff filings as of August 17, analyzed 55, and isolated 18 rate-design elements. The structural shift: minimum contract terms averaging five years pre-2025 versus twelve years post-2025, demand thresholds spanning under 1 MW to 150 MW, and five emerging protections — nonrefundable study deposits, mandatory ramp schedules, hold-harmless cost makeup, resizing conditions, early-exit fees — layered atop eight established ones. Entergy Louisiana at five years and El Paso Electric proposing twenty bracket the range. This supersedes the 77-tariff count this series has been citing as its comparables set.
Action Signal: Engage — Use the 264-filing population, not a handful of neighboring utilities, as the comparables set when defending your own large-load tariff terms, and reuse the same ratepayer-protection reasoning in the demand-side case: the argument that existing customers should not carry stranded-asset risk for speculative load is the argument for deferring capital with demand response, stated in the utility’s own filing. Check your tariff against the five emerging provisions specifically — they are the ones a commission will ask why you omitted.
⚑ Fifteen Percent DER Penetration Is Where Scheduling Stops Working — and There Is Now a Deployment to Cite
Topic: DERMS Procurement / Capability Staging / RFP Specification
Relevance: NCEMC and OATI are moving an existing DERMS from planning and forecast-based scheduling to real-time, sub-minute orchestration across 26 distribution cooperatives, including five islandable microgrids under GridMind controls, with roughly 15% of NCEMC’s peak already served by DERs. The trigger was resilience capital — a post-Helene state commitment of $5 million for 26 microgrids — not a reliability business case. Separately, GE Vernova’s GridOS for Distribution bundles ADMS, DERMS, GIS, field execution and visual intelligence on one data fabric with Alabama Power among early adopters, and market sizing puts full deployment at $2M–$5M per 1,000 customers with recovery periods that can exceed a regulatory cycle.
Action Signal: Implement — Write the scheduling-versus-real-time-dispatch distinction into the DERMS requirements document as separately demonstrable capabilities with acceptance criteria, because a platform can satisfy the first for years before the second is tested, and 15% penetration is a usable threshold for arguing when the second becomes load-bearing. On the suite-versus-best-of-breed question GridOS poses, settle it before requirements are written, and make data governance, open APIs and model-and-data exit terms commercial terms rather than schedule exhibits — a unified data fabric concentrates capability and dependency in the same place. For smaller utilities and co-ops, treat storm-recovery and resilience capital as a live funding channel for orchestration infrastructure.
⚑ The Allowed Return Is the Structural Obstacle to Demand-Side Deployment, and It Can Be Named
Topic: Regulatory Strategy / IRP Screening / Performance-Based Regulation
Relevance: Columbia CGEP identifies the mechanism plainly: utilities earn roughly 9–10% on capital, so “this regulatory structure rewards capital deployment more than system optimization,” and planning frameworks built to prove need for new capital rather than evaluate lower-cost alternatives will keep GETs and DR lagging regardless of cost-effectiveness. The affordability context is severe — residential prices up ~6% nominally, IOUs seeking $18 billion in increases (most since the mid-1980s), regulators approving 66% of the requested dollar value — and data centers reach 9.5–15.3% of U.S. electricity by 2030 from 4.7% in 2024.
Action Signal: Engage — When a demand-side alternative loses an IRP screen, examine whether it lost on economics or on a screening framework that only tests whether new capital is needed. That is a procedural finding a commission can act on, and it pairs directly with the non-wires screening mandates now in statute in Virginia and signalled in Rhode Island. The honest internal version of this flag is that a utility arguing for demand-side investment is arguing against its own earnings mechanism, which is why performance-based regulation and shared-savings structures belong in the same filing rather than a separate docket.
⚑ The Load-Side 202(c) Instrument Still Has No Performance Record After Its First Exercise
Topic: Emergency Authority / Large-Load Contracting / Planning Assumptions
Relevance: PJM issued its first-ever backup-generator Warning on July 2 and 3, never escalated to an Action, dispatched no customer generation, declined to disclose how many megawatts of backup generation were available, and has not adopted the 50 MW large-load threshold into its governing documents. Meanwhile Order 202-26-43 (Duke Energy Carolinas, September 3–8) and Order 202-26-23 (PJM, May 18, three days, reserves projected under 5,800 MW with 40-plus GW on planned outage) establish the recurrence pattern across two footprints and three seasons in one year. The instrument is now routine to invoke and entirely unmeasured in performance.
Action Signal: Engage — Plan against recurrence rather than treating each order as an exception: build curtailment obligations, behind-the-meter dispatch visibility and backup-generation telemetry into large-load interconnection agreements now, so that when the authority is next invoked the utility has a metered fleet rather than an unknown one. In any proceeding where emergency authority is offered as a reliability answer, the accurate characterization is that it was warned, held, and never tested — while metered demand response shed 6,113 MW in the same event. Note also what an actual event requires operationally: the Duke package needed transmission-owner coordination, a Southern Power consent letter and a VACAR South endorsement.
⚑ Global Demand-Response Utilization Is ~100 GW Against a ~600 GW Residential Pool — and the Gap Is Rules
Topic: DR Potential / International Benchmarking / Filing Support
Relevance: The IEA’s Electricity 2026 flexibility chapter finds only about 100 GW of DR utilized worldwide as of 2024 — industry ~75 GW, buildings ~30 GW, transport under 5 GW — against roughly 600 GW of residential-AC peak and 160 GW of aluminum-smelting peak, attributing the shortfall to smart-meter and controllable-appliance penetration, weak short-term price-signal exposure, and aggregation rules that restrict small-load participation. Two FERC orders in two weeks have now demonstrated that the third barrier is a tariff choice. Meanwhile utility-scale storage added 63 GW in 2024 at costs down ~40% to ~$150/kWh, with California batteries near 25% of peak load.
Action Signal: Watch — Use the IEA figures as the international best-practice framing regulators increasingly expect alongside Brattle-style avoided-cost methods, particularly the sectoral breakdown, which shows where untapped potential actually sits (buildings and transport, not industry). Track the storage cost curve as the competitive benchmark rather than the complement: at $150/kWh and falling, four-hour storage is what a DR program’s cost-effectiveness will increasingly be tested against, and program design — term, availability, minimum size — is what determines which one wins.
📌 Sources
September 7, 2026 Entry
– U.S. Department of Energy — Federal Power Act Section 202(c): Duke Energy Carolinas, LLC, Order No. 202-26-43 (issued September 3, 2026)
– Charles River Associates — Grid Under Pressure: Flexibility-Centered Large Load Policies Overlook Recent Lessons (Dmitry Balashov, Margarita Patria; February 17, 2026)
– Utility Dive — In 2026, Virtual Power Plants Must Scale or Risk Being Left Behind (Herman K. Trabish; January 27, 2026)
– Research and Markets via GlobeNewswire — Global Digital Power Utility Market Analysis Report 2026-2031 (September 7, 2026)
September 8, 2026 Entry
– Uplight / Octopus Energy Group via Utility Dive — Octopus Energy Completes Investment in Uplight, Accelerating Plans to Save US Customers $1 Billion and Double Grid Flexibility (September 1, 2026)
– GE Vernova — GE Vernova Launches GridOS for Distribution, the Industry’s First Unified Solution for Grid Orchestration (February 3, 2026)
– Utility Dive — FERC Will Impose Reforms if PJM Fails to Adopt Changes by September, Chairman Warns (Ethan Howland; July 24, 2026)
– International Energy Agency — Electricity 2026, “Flexibility” chapter (2026; CC BY 4.0)
September 9, 2026 Entry
– Canary Media — Xcel Minnesota Is Building a First-of-its-Kind Virtual Power Plant (Jeff St. John; April 8, 2026) (program previously reported in the September 4 digest; carried here for the repayment mechanics)
– New Jersey Board of Public Utilities — NJ Proposes “Virtual Power Plant” Plan to Save Ratepayers Millions Annually, Docket QO26030099 (July 27, 2026) (previously reported August 28 and September 4; carried as one line of context)
– Utility Dive — Power Plants Under DOE Emergency Orders Are Producing Way Less Energy Than Before (June 23, 2026) (fully reported in the September 4 digest; referenced here, not re-reported)
– Utility Dive — Data Centers Are Ready to Negotiate Flexibility for Speed (Herman K. Trabish; June 26, 2026) (previously reported in the September 4 digest; anchors the EPRI Spotlight as a carry)
September 10, 2026 Entry
– Utility Dive — Google, Xcel, Others Back MISO’s ‘Zero Injection’ Large-Load Proposal (Ethan Howland; September 9, 2026; MISO filing August 18, 2026)
– Utility Dive — Large-Load Tariffs Increasingly Rely on Upfront Payments, Exit Fees, Ramp Schedules (Brian Martucci; September 8, 2026; LBNL/Brattle brief)
– Utility Dive — PJM Gets Emergency Approval to Curtail Data Centers, Large Loads During Hot Weather (Ethan Howland; May 19, 2026; DOE Order 202-26-23 issued May 18, 2026)
– Troutman Pepper, Washington Energy Report — FERC Approves ISO-NE Updated Market Rules for Distributed Energy Resources (order issued May 29, 2026; Docket ER26-1956-000)
September 11, 2026 Entry
– Microgrid Knowledge / OATI — North Carolina Electric Coop and OATI Connect Microgrids and DERs Under Real-Time Control Platform (February 20, 2026)
– Utility Dive — GETs, Demand Response Can Ease Near-Term Data Center Electricity Price Pressure: Report (Ethan Howland; June 23, 2026; Columbia Center on Global Energy Policy)
– Data Center Knowledge — PJM Issued First Backup-Generator Warnings During Heat Wave (Shane Snider; July 20, 2026)
– Enel North America — PJM 2026/2027 Capacity Auction Results (Erin Donohue; July 22, 2025)
Verification Sources (targeted search, not from the daily research log)
– FERC — PJM Governance and Stakeholder Processes (proceeding page)
– FERC — Notice for Dispute Resolution Services Proceedings for PJM Governance and Stakeholder Processes (forum commencing September 1, 2026)
– Federal Register — PJM Governance and Stakeholder Reforms; Notice Requesting Post-Conference Comment (August 4, 2026)
– EPRI — DCFlex Initiative Expands to Nine Demonstration Sites Across U.S. and Europe (most recent substantive EPRI release; ninth dry week confirmed)
– Emerald AI — Launching the First Power-Flexible AI Factory with NVIDIA (96-MW Aurora AI Factory, Manassas VA; ~100 GW reference-design estimate)