This week: the demand side stopped being an argument about value and became a price. Five grid operators filed their responses to FERC’s June 18 large-load show-cause orders — ISO New England on November 13, CAISO, MISO, NYISO and PJM on November 16, with SPP due around November 20 — and every one of them built the same thing: a 50-MW large-load threshold and a curtailable, non-firm transmission tier that can energize before firm upgrades are finished. The mechanics are where the week’s news actually lives, because the mechanics are a published price for interruptibility: PJM would charge $58,219/MW-year for non-firm service against $90,000/MW-year firm — a 35.3% discount — and cut its study cycle to about one year; MISO’s “Grid Responsive Integration Transmission Service” interrupts non-firm load whenever the nodal LMP tops $100/MWh, which MISO says still leaves service in more than 95% of hours; ISO-NE sets a $250/MWh redispatch trigger it says bound in about 1% of hours from 2021 through 2025; NYISO caps non-firm bids at $200/MWh. In the same week NERC’s 2026–27 Winter Reliability Assessment counted demand response at 43.6 GW, up from 41.7 GW, and batteries adding 9.2 GW of on-peak capacity — the largest single resource gain, ahead of natural gas at +8.7 GW — against peak demand up 21 GW (2.6%) it attributes explicitly to data centers, large industrial load and electrification. And Connecticut became the first state in this series to order a DERMS by implication: PURA’s November 4 decision in Docket 25-01-27 requires scheduled flexible interconnection across the distribution companies by May 1, 2027 and dynamic flexible interconnection from May 1, 2028, noting that dynamic operation requires “a grid-scale DERMS or an ADMS.”
🏗️ Data Centers & Large Load Growth
The filings landed, and the thing to read is not that they landed — it is that five independent operators converged on one structure and then priced it five different ways. (deduplicated — this item consolidates four distinct articles across three daily entries covering the same underlying filings; see Method)
Responding to FERC’s June 18 Section 206 show-cause orders (the rulemaking docket RM26-4 remains open), ISO New England filed on November 13 and CAISO, MISO, NYISO and PJM on November 16, with SPP due around November 20 and comments due roughly 30 days after each filing. The convergence is the headline: all five broadly adopt a 50-MW threshold to define a large load, and every operator stands up a curtailable, non-firm transmission tier that can energize before firm upgrades are complete.
The divergence is the usable content, because each operator has now published the parameters a DERMS or DR program already lives on — trigger, depth, frequency and term:
- MISO — Grid Responsive Integration Transmission Service (GRITS). Non-firm load is interrupted whenever the nodal LMP tops $100/MWh, a threshold MISO says still permits service in more than 95% of hours. Carries a 20-year revenue commitment and a $250,000 application fee, with implementation no earlier than January 1, 2029. Separately, MISO’s Large Load Addition Resource Study co-models loads at or above 200 MW with their associated generation, each represented in the other’s system model.
- PJM — Fast Energization Process. A single combined study of a large load and its generation in roughly one year, about twice as fast as today, with a non-firm rate of $58,219/MW-year against a $90,000/MW-year firm rate — a 35.3% discount at roughly 64.7% load factor. PJM serves about 10 GW of existing large computational load and forecasts up to 30 GW of additions by 2030.
- ISO-NE — three tiers, filed as Docket ER27-484-000. Firm contract demand service (same priority as firm load, requiring new capacity and network upgrades under a “bring your own new capacity” match), interim contract demand service, and a permanent non-firm “flexible load” service requiring no new-capacity designation. Non-firm and interim customers are re-dispatched whenever area energy prices reach $250/MWh — about 1% of hours — and are excluded from the Installed Capacity Requirement. Study process effective June 1, 2027; new services June 1, 2028.
- NYISO — Firm and Non-Firm Withdrawal Services plus a “State-Exempt Load Project” resource-adequacy exception, with non-firm bids capped at $200/MWh. New York transmission owners filed a complementary rate and cost-allocation proposal.
- CAISO — Flexible Interim Load Interconnection, approved by its Board on October 28, letting qualifying customers operate under project-specific operating restrictions while required upgrades are still being built.
Here is why this is the most consequential week for demand-side value in this series, and it is not a reliability argument. For three months this digest has argued that verified flexibility is worth something and struggled to find anyone who would put a number on it. Five grid operators just did, in tariff language, under a federal deadline. PJM’s 35.3% discount is the cleanest statement: a load willing to be curtailed pays $31,781/MW-year less than one that is not, which is a published market price for interruptibility set by the operator rather than inferred from an avoided-cost model. Put PJM’s two rates side by side with your own avoided-capacity benchmark — against the ~$66/kW-year anchor this series carries, PJM’s non-firm discount works out to roughly $31.78/kW-year of value assigned to curtailability alone, on transmission service, before a single capacity-market dollar.
The second implication is procurement, and it is immediate. Every one of these tiers requires the operator to know, in near-real time, that a load actually reduced when the trigger fired — which is the same telemetry, dispatch and settlement capability a utility buys for a DR portfolio, now underwriting large-load transmission service. A utility that has already built DR measurement infrastructure has already built most of what its large-load tariff will need; a utility that has not is now buying it twice. Put the large-load tariff team and the DSM team in the same room before either writes a specification.
The third is the one to be careful about. These are Section 205 filings FERC has not ruled on, with comments due about 30 days after each, and MISO’s own implementation date is no earlier than January 1, 2029. ISO-NE’s “bring your own new capacity” details remain in stakeholder work into 2027. So the prices are real as proposals and not yet real as rates — cite them as the operators’ own valuation of curtailability, which is what they are, and not as effective tariff terms. (Sources: POWER, November 19 — November 20 entry; Utility Dive, November 17 — November 18 entry; ISO Newswire, November 13 — November 17 entry; Construction Review Online, November 16 — November 17 entry. Four articles, one development, consolidated here)
PJM’s other filing is the reliability counterpart, and it converts controllability from an option into an interconnection condition. On November 16 PJM filed voltage- and frequency-ride-through requirements with FERC for a newly defined Large Computational Load (LCL) category — 50 MW or greater — effective January 16, 2027. The trigger was a July 22 transmission disturbance during which nearly 3.9 GW of data centers switched from utility power to onsite backup generation — the largest such load transfer in PJM’s history — affecting voltages, frequency, power flows and generator behavior across the Dominion zone and the Eastern Interconnection, though PJM’s control center managed it safely. New LCLs must demonstrate compliance as a condition of interconnection and operation; existing LCLs must comply within one year of the effective date or of commercial operation, whichever is later. PJM says the definition aligns with large-load standards under development at FERC and NERC, with ramp-rate requirements to follow.
Read the two PJM filings together, because together they say something neither says alone: the same capability is being priced as a discount in one docket and mandated as a condition in the other. A load that can ride through a disturbance and curtail on a price signal gets $58,219/MW-year instead of $90,000 and a one-year study; a load that cannot ride through will not be allowed to connect at all. That is controllability becoming table stakes rather than a premium product, and it is the clearest signal yet that the telemetry-and-verification layer this series keeps specifying is moving from “how you earn DR revenue” to “how you get served.” The article does not disclose the specific voltage and frequency thresholds, which would come from the filing itself, and this is distinct from PJM’s separate IRAS / “bring your own new capacity” market-design proposal, which FERC largely rejected on October 9 on jurisdictional grounds. (Source: PJM Inside Lines, November 16 — November 17 entry; distinct from the §206 response wave above)
Only about a fifth of the pipeline everyone is planning against is actually under contract, and the tracker that says so landed inside the window. Latitude Media’s data-center tracker, published November 13, finds utilities in 53 U.S. service territories have signed contracts for about 120 GW of data-center load — roughly 22% of the 537 GW they report in their pipelines, though still about 5.5 times the 22 GW operating in those territories today — while already soliciting about 37 GW of new generation whose cost ratepayers may partly bear. The reported-versus-signed gap is the planning risk, and the extremes are instructive: Georgia Power reports 80.1 GW against 12.8 GW signed; Dominion reports a 54.9 GW pipeline against roughly 12 GW under electric service agreements, with another 32.5 GW on non-firm engineering letters, only 41% of which Dominion includes in its forecast; and signed load reaches 235% of peak demand at AEP Ohio, 169% at NIPSCO and 145% at PPL.
The cost consequence is already booked. PJM’s market monitor attributes $29.4 billion — about 46% — of four auctions’ $63.6 billion in capacity cost to data-center load, even as PJM utilities just cut near-term large-load forecasts by a combined 6.5 GW, which could shrink PJM’s reliability-backstop procurement to as little as 630 MW from about 6.8 GW. And the speculative-load signals are contractual rather than rhetorical: Amazon can drop 1,200 MW from its NIPSCO deal by March 31, 2029, and the average announcement-to-power lag runs about 33 months.
Use the ratio, not the headline, and use it the same way every week: a capital plan or an IRP sized against an unscreened queue is sized against an artifact. This series has now found the same collapse at four different denominators — AEP Ohio’s 30 GW to 13 GW, Oncor’s 293 GW against ~45 GW conditionally classified under ERCOT’s Batch Zero, PPL’s 2030 adjustment falling from 13,343 to 12,764 MW between submissions, and now 537 GW reported against 120 GW signed nationally. The deferrable, cancellable demand-side alternative is the lower-regret purchase at every point on that curve, and the argument lands hardest in the utility’s own words where a filing supplies them. Keep the ERCOT figures reconciled rather than mixed: Oncor’s 293 GW transmission-queue share, ERCOT’s ~474 GW audited queue and the 191.8 GW of large-load studies are three different denominators, not one number — and this week adds a fourth, the national 537 GW reported / 120 GW signed pair. (Source: Tech Times, November 16, reporting Latitude Media data published November 13 — November 16 entry)
The quantitative case underneath all of it got its sharpest distribution-level number yet: Dominion’s distribution system averages a 41% load factor. At BNEF’s Houston Summit, Google’s Tyler Norris put U.S. grid utilization at 40–60% and called enabling non-firm transmission service the “real whitespace” — a framing the filings above have now operationalized. The figure for distribution planners is Dominion Energy Virginia’s 41% average distribution load factor, peaking just under 80% utilization, against a median transmission-facility utilization of roughly 33%. Systemwide load factor is put at about 60% today, rising to 66% by 2030 (Grid Strategies). The panel also supplied a replicable flexibility-backed-interconnection template: an Aligned Data Centers / Portland General Electric arrangement in which a 31 MW / 62 MWh battery tied to a 72 MW Hillsboro, Oregon data center let the project bypass a fully firmed network upgrade. NRG’s Jessica Hogle set a 25 GW demand-response, storage and VPP target by 2035.
A 41% distribution load factor is the business case, stated as a number, and it is the number to pull for your own system before the next distribution capital cycle. The latent headroom between a 41% average and a sub-80% peak is capacity a flexible-load or VPP program can monetize before a wires upgrade is triggered — which is the same “better use of the grid we already have” logic now carried by the CPUC’s directive, Connecticut PURA’s flexible-interconnection order below, and all five RTO filings above. Compute your own distribution and transmission load factors and put them in the IRP, because a utilization figure is the cheapest available argument for a non-wires alternative and it is one your own AMI and SCADA data already supports. The caveat is format: this is panel commentary, not a filed dataset, and the Aligned/PGE arrangement is reported without terms. (Source: Utility Dive, November 13 — November 18 entry; publication date precedes this window and is a first appearance in this series)
🌡️ Winter Reliability & Demand-Side Credit
NERC counted the demand side this winter, and the count is the best number the resource has: 43.6 GW of demand response and batteries as the single largest source of new on-peak capacity. NERC’s 2026–27 Winter Reliability Assessment, released November 18, reports aggregate peak demand across assessment areas rising 21 GW (2.6%) from last winter — exceeding the prior year’s 20 GW gain and NERC’s historic ~1% norm — attributed explicitly to “the expansion of data centers, large industrial loads, and broad electrification across end-use sectors.” The resource-mix detail is the demand-side data point: demand response grew to 43.6 GW this winter from 41.7 GW, and batteries added 9.2 GW of on-peak capacity — the single largest category of growth, ahead of natural gas at +8.7 GW — against total North American bulk-power resource growth of 26 GW (2.4%). NERC found no area short of resources under normal weather, and CEO Jim Robb said “the resource outlook has improved,” while flagging SERC at risk of shortfalls under more severe conditions as load growth erodes reserve margins, localized fuel and transmission constraints in ISO New England and parts of MISO and SPP, and — the line for long-range planning — a ten-year outlook in which many areas face a high risk of energy shortfalls and rolling blackouts by 2030. Freezing temperatures and an El Niño pattern are flagged as gas-supply risks.
This is the single most citable paragraph the demand side has had all quarter, and it should be quoted from NERC rather than from an advocate. Two of the three fastest-growing on-peak resource categories in NERC’s own winter accounting are demand response (+1.9 GW) and batteries (+9.2 GW), and batteries beat gas. That is the avoided-cost argument stated by the reliability organization: flexible demand and storage are deferring the gas and transmission build that NERC’s own 2030 risk picture would otherwise demand. Put the 43.6 GW figure and the batteries-over-gas comparison into every proceeding where demand-side credit is being set this winter, with NERC’s attribution attached.
And then note the asymmetry this series has documented all month, because this week sharpens it to a contradiction inside one institution. NERC counts 43.6 GW of demand response in its seasonal assessment — and NERC Project 2024-02, the energy assurance standard that will define “dependable energy under stress” for long-term planning, names demand response exactly once, at R2.2, as an input a Planning Coordinator’s methodology must address, with DERs, VPPs and load flexibility appearing nowhere. The same organization that counts the resource operationally has not written a rule that credits it in planning. That gap is the argument, and it now has a number on the operational side: use it. (Source: America’s Electric Cooperatives / NRECA, November 18 — November 19 entry. Distinct from NYISO’s New-York-only assessment below and from PJM’s winter outlook carried in the November 13 digest)
NYISO’s own assessment shows demand response running as a sustained operational resource, not an emergency backstop. NYISO’s 2026–27 Winter Reliability Assessment, reported November 16, finds 29,464 MW of resources against a 24,620 MW forecast peak — roughly 4,844 MW, about 20%, of headroom — and concludes the system has sufficient supply, while VP of operations Aaron Markham flags fuel availability as “a major concern.” The operating record is the demand-side signal: NYISO activated demand-response programs for six consecutive days during Winter Storm Fern last winter. Last winter’s peak was 24,317 MW (February 7, 2026), below the all-time winter peak of 25,738 MW (January 7, 2014).
Six consecutive days is the figure to carry, and it belongs beside Sunrun’s eighteen consecutive days of Puerto Rico dispatch from the November 13 digest. Both are multi-day endurance records on a resource everyone specifies for single events, and in a gas-constrained winter system the value driver is deliverability during a cold snap, not gross capacity margin. So the winter DR product in a fuel-constrained region is a multi-day, energy-limited product — which is the Class C profile a four-hour battery fleet does not serve and a behavioral or curtailable-load program does. Ask your own portfolio the NYISO question: what does available capacity look like on day five? (Source: Daily Energy Insider, November 16 — November 18 entry)
DOE issued another Section 202(c) order — the first since October 12 — and it names data-center load growth as the reason. On November 13 Secretary Chris Wright issued Order No. 202-26-51, directing Consumers Energy to keep its 1,420-MW J.H. Campbell coal plant — slated to retire May 31, 2025 and kept running by a chain of 90-day orders — available to MISO from November 15, 2026 through February 12, 2027, a notably longer term than the prior renewals. The finding is a “risk of substantial energy supply shortfall in the MISO region, namely rising electricity demand coupled with a decrease in the availability of dispatchable generation,” and the order explicitly ties that shortfall to data-center and computational-load growth — citing NERC’s 2026 assessments, the OMS-MISO survey, and Consumers’ own plan for a 1-GW hyperscale data center — and to winter risk, pointing to Winter Storm Fern’s January 24, 2026 EEA-2/MaxGen event, during which Campbell produced 201,962 MWh. The order states the plant “shall not be considered a capacity resource.”
The arithmetic this series carries is unchanged and it got one order heavier: emergency coal retention is the expensive, last-resort version of what a planned, compensated demand-response or VPP program delivers in an orderly way with a settlement record attached. The order’s own language — not a capacity resource — is the point: these are stopgap megawatts that build no accreditation, exactly the gap firm, dispatchable flexible demand is valued to fill against the ~$66/kW-year avoided-capacity benchmark. Note also where it sits in the week: item one above has large loads buying faster interconnection with curtailability, and this order has the federal government holding a coal plant open partly because of a 1-GW hyperscale load on the same system. Data-center load is both the driver of the emergency and, if made flexible, the remedy — and only one of those two framings earns the customer a 35.3% rate discount. The caveat is legal: the order notes a September 11, 2026 D.C. Circuit opinion DOE disputes and has petitioned to rehear en banc, so this signals stress and the value of firm flexibility rather than establishing a resource. The 2026 §202(c) docket now exceeds fifty orders and has accredited nothing. (Source: U.S. DOE/CESER Order No. 202-26-51, November 13 — November 20 entry. First new order in the docket since No. 202-26-50 on October 12; J.H. Campbell is a carried subject in this series and this is a new order, not a re-report)
🔌 DERMS & Grid Integration Technology
A state has now ordered a DERMS in all but name, with dates — and that is the eighth-week drought’s most useful break. The sharpest regulatory content of the week arrives inside a vendor recap, and it is third-party substance: Connecticut PURA’s November 4 decision in Docket 25-01-27 sets a flexible-interconnection timeline — scheduled flexible interconnection across the distribution companies by May 1, 2027; dynamic flexible interconnection at the most constrained substations from May 1, 2028; expanding systemwide by November 1, 2030 — with PURA noting that dynamic operation requires more advanced monitoring and control, “such as a grid-scale DERMS or an ADMS.”
This is the first instrument in this series where a state interconnection mandate converts directly into a DERMS or ADMS procurement with a date attached, and the structure is worth copying into your own advocacy. PURA did not order a platform; it ordered a capability — dynamic management of DER output during constrained hours — and then observed what that capability requires. That is the correct regulatory posture and it is the one to ask for: mandate the function and the deadline, let the utility select the system. The underlying argument is the same one running through the five RTO filings above and the 41% Dominion load factor: flexible interconnection lets a resource connect sooner and have its output managed during constrained hours, instead of waiting for upgrades sized for the single worst hour.
The recap’s other content is sequencing and one deployment fact. The phased DERMS path offered — DER forecasting and visibility first, then dispatch pilots, then flexibility in daily operations — matches the co-op adoption pattern below and is a defensible RFP structure. The open data question is named explicitly: whether real-time telemetry is required or 5-to-15-minute intervals suffice, which is the single most consequential unresolved line in any DERMS specification and the one this series has flagged as the primary spec line all quarter. On utility AI, attendees held that an operational tool “has to give answers it can explain and defend,” built on the utility’s own data with humans making the final call; CAISO is running its AI Genie live in its production outage-management system with staff retaining final authority, more than 250 utilities run their DERMS on OATI, and one utility said it does not yet trust AI for forecasting. The ISO-NE CELT winter peak under normal weather is given as 20.5 GW in 2026/27 rising to 26.4 GW by 2035/36.
Two cautions. This is a vendor-authored, sponsored recap of DTECH Northeast (Boston, November 2–4), so the platform claims are promotional even though the PURA decision, the CAISO deployment and the ISO-NE forecast are independently verifiable — go to the PURA docket for the order rather than citing the recap. And the PURA decision itself is dated November 4, before this window; it enters the series here, through an in-window recap, as a first appearance. (Source: OATI, November 16 — November 19 entry; the entry’s one sponsored/vendor source)
Two cooperatives published the first-steps sequence that every DERMS business case needs and nobody writes down. In an OATI-moderated Renewable Energy World webinar, Big Rivers Electric Corporation (Owensboro, KY — three member owners serving 22 counties and about 140,000 accounts) and Ozarks Electric Cooperative (more than 85,000 members across northwest Arkansas and northeast Oklahoma) detailed their first steps toward DERMS: building the business case, defining initial uses (DER visibility, demand response, EV-charging coordination), aligning IT and operations stakeholders, and choosing a scalable architecture. Big Rivers ran an internal “DERMS 101” education series for senior staff, member owners and its board before launching, and its first program is a smart-thermostat program operated like a MISO Load Modifying Resource — a few seasonal events called only during grid-constrained conditions, framed as “an insurance policy versus monthly peak-clipping” — with a behavioral program and conservation voltage reduction (CVR) planned next across its three members. Ozarks deliberately waited for device integration (water heaters, thermostats) to mature, built a technology roadmap with the National Rural Telecommunications Cooperative in 2023, and launched in 2026, proceeding only “if the numbers pencil out.”
The sentence to take is the gating constraint, and it is not a platform feature: both co-ops flagged that program complaints and low participation can render a DERMS program worthless. That is the same finding New Jersey’s BPU quantified from the other end at 82% of target reachable from already-funded devices at ~75% enrollment, the same one the Uplight op-ed below argues from enrollment design, and the same one Carnegie Mellon measured as nine or ten internal departments per device. Four independent sources, four methods, one conclusion: enrollment and trust are the binding constraint on residential and small-utility flexible capacity, not orchestration capability. So weight your DERMS evaluation criteria accordingly — telemetry, event design and M&V against customer experience — and run the “DERMS 101” internal education step before the RFP, because Big Rivers did it before launching and named it as the reason the launch held. Note the program design too: a MISO LMR-style insurance product is a different product from monthly peak-clipping, with different notification, duration and frequency parameters — which is the FlexMosaic class distinction the EPRI Spotlight below applies to this week’s RTO tiers. Caveats: the webinar named no DERMS vendor or platform, no MW figures and no firm go-live dates beyond “this year.” (Source: Renewable Energy World, November 13 — November 17 entry; publication date precedes this window, first appearance in this series)
And the integration problem underneath every DERMS RFP got its clearest statement from a vendor who does not sell DERMS. In an interview, IOTech CPO and co-founder Andy Foster argues that “a VPP is only as good as its weakest device integration,” and that the binding constraint on scaling distributed fleets is device heterogeneity — differences in manufacturer, model and firmware, where even a firmware update can change what a register means — rather than the capability of the orchestration platform. The proposed architecture is a two-tier, federated “decide centrally, execute locally” design: low-power Arm edge nodes sit beside each asset, speak its native protocol, enforce local limits, and normalize device data into a common model (using OPC UA for supervisory control, with AI now automating the historically manual mapping of device data to SunSpec), while forecasting, trading and portfolio optimization run centrally. Control stays on-site for three reasons: speed (fast frequency services need a full response within about one second), resilience (sites keep respecting grid limits if the external link drops), and data volume (normalizing at the asset beats streaming raw telemetry). One grid-scale storage provider is said to embed the software across roughly 20 GW worldwide.
Three of those belong in the evaluable requirements section of an RFP, not the feature list: interoperability across manufacturer and firmware revisions, firmware-change resilience, and local fallback behavior when the link to the platform drops. The firmware point is the subtle one and it is a real operational risk — a resource whose register map changes under a vendor firmware push is a resource whose settlement quantity changes silently, which is a measurement-integrity problem before it is an integration problem. Add a contractual requirement that the aggregator or platform notify the utility of device firmware changes affecting telemetry, and test it. Pair this with the real-time-versus-5-to-15-minute question PURA’s timeline raises above: the answer differs by service, and the one-second frequency-response figure here is the evidence that it does — so specify the interval per service, not once for the platform. The caveats are that this is a software vendor’s own framing and the ~20 GW scale claim is not independently verified. (Source: ESS News, November 12 — November 16 entry; publication date precedes this window, first appearance in this series)
⚡ Virtual Power Plants & Demand Flexibility
Two Northeast utilities said the thing this series has been waiting for a utility to say: they are dispatching VPPs at named distribution constraints, and the gap stopping them is visibility, not capability. At a Latitude Media / EnergyHub webinar, executives from National Grid and Eversource described aggregated DERs enrolled in ConnectedSolutions and the higher-incentive ConnectedSolutions+ being dispatched increasingly for location-specific distribution constraints rather than system peak-shaving. The fleet sizes: National Grid has about 100 MW of DER load in New England and more than 350 MW upstate — roughly 2% and 5% of regional peak — mostly thermostats, with storage and managed EV charging growing; Eversource runs about 100,000 thermostats and 25,000 commercial and industrial customers across a territory that includes ISO New England’s largest load zone, with CS+ pilots targeting Greater Boston substations and a southeastern Massachusetts substation absorbing midday solar.
The operational tell is the one to carry into every DERMS procurement, because two large utilities just stated it on the record: Eversource cannot yet confidently say which customers have air conditioning, and only a few unmanaged EVs can overload the small pole-top transformers common in its territory. Targeted managed charging could defer those transformer replacements once AMI improves asset-level load data — “we will be in a much better position in a year.” That is the distribution-level observability gap, named by the utility that has it, and it is the precondition for the entire non-wires-alternative business case: you cannot target a constraint you cannot see. So sequence the investment honestly — AMI-derived asset-level load data and customer end-use attribution come before locational dispatch, not after — and if your own program is already claiming locational deferral value, ask what your evidence is that the enrolled devices sit on the constrained asset.
EnergyHub supplied the orchestration figures, and one of them answers a long-standing DR problem directly. EnergyHub reports managing about 4 GW of flexible capacity for load-serving clients that called roughly 3,500 load-shifting events this past summer, and that across 70 dynamic load-shaping events for 12 utilities it delivered up to 90% consistency for the full event duration. Full-duration consistency is the metric that matters and it is rarely published — the classic DR failure mode is an event that starts firm and decays as customers drop out, so a 90% full-duration figure is a more useful claim than any aggregate MWh or peak-MW number, because it speaks to the shape of delivery rather than its size.
But it is still a vendor figure from a sponsored webinar, so convert it before using it. Apply this series’ four tests: the measurement point (grid interface, service point or per asset), the baseline method, delivered-versus-committed MW per event, and the named utility whose filed evaluation will confirm it. None is disclosed here — no dollar figures, no dockets, no CS+ pilot MW — and the executives themselves remained “cautiously optimistic” that behind-the-meter flexibility can be defined precisely enough to solve named constraints. The right use of the 90% figure is as the acceptance criterion you write into your own contract, not as evidence anyone has met it. (Source: Utility Dive, November 18 — November 20 entry)
PSE&G published a clean operating result, and the detail that makes it citable is when the events were called. PSE&G reports its demand-response programs grew to nearly 45,000 enrolled customers by the end of the 2026 season, roughly double the ~23,000 at the end of 2025, with enrolled devices rising to about 60,000 from just over 30,000. Over the June 1 – October 30 season the utility called 16 demand-response events, one of which cut demand by 38.15 MW — the utility’s analogy is taking about 18,000 homes off the grid for an hour — and the programs were dispatched during each of PJM’s five highest grid-peak hours. Program design is the two-track model this series keeps flagging: FlexPower enrolls smart thermostats, EVs and chargers and certain electric water heaters ($50 per new device type plus $25 for each year enrolled), while SmartFlex is behavioral, paying $1.25 per kWh saved during an event with or without a smart device. PSE&G ties the experience explicitly to New Jersey’s emerging VPP program.
“Dispatched during each of PJM’s five highest grid-peak hours” is the sentence to steal, because it is a coincidence claim and coincidence is what capacity accreditation actually pays for. Most program reports publish enrollment and event counts; this one publishes alignment with the hours that set the regional capacity price and the system risk. Add peak-hour coincidence to your own program’s published metrics — it is derivable from data you already hold, it is the metric an accreditation proceeding cares about, and almost nobody reports it.
And read this beside last week’s “PowerUp New Jersey” approval, because it is the same state and it is the evidentiary half of that order. The BPU found 82% of each utility’s 3%-of-peak target reachable from already-funded devices at roughly 75% enrollment; PSE&G has now shown what the enrollment curve actually looks like in that state — a doubling in one year to 45,000 customers and 60,000 devices. Put the two numbers in one table and the sensitivity writes itself: a program that doubled to 60,000 devices in a year is a long way from the installed base the 82% assumes, which is exactly why this series recommends modeling the 82% at 40% and 55% enrollment and filing the sensitivity before an intervenor does. The caveats: this is a utility press release, so figures are self-reported, and the 38.15 MW is a single-event peak rather than a season average or a firm capacity value. (Source: PR Newswire / PSE&G, November 17 — November 19 entry. The November 12 “PowerUp New Jersey” BPU approval was reported in the November 13 digest and is carried here as context, not re-reported)
A renter VPP went live on Con Edison, and it attacks the one segment installed solar-and-storage structurally cannot reach. Jackery and Brooklyn-based aggregator Every Electric launched what they call North America’s first plug-in-battery VPP aimed at renters — roughly 2,000 portable Jackery Explorer 2000 v2 units across more than 1,000 Con Edison households, providing about 4 MWh of flexible storage with nearly 50 MWh cycled to date, enrolled in Con Edison’s “Smart Usage” demand-response program to shift air-conditioning load off peak, running since September. The design removes the install barrier: units are drop-shipped, need no electrician, permit or utility agreement, and run a window or portable AC through a passthrough outlet, with the aggregator’s cloud software charging off-peak and discharging at peak. The economics map onto a DR value stack: a $0-upfront tier (refundable $30 deposit) in which the aggregator keeps the battery and the Con Edison DR revenue and pays the resident $100–150 per AC per year, or a $650 purchase (off a $999 list) earning up to $250 per AC per year — with roughly $200,000 in seasonal payouts reported and a waitlist of more than 15,000 air conditioners.
Take the channel and the segment, not the megawatt-hours. 4 MWh is small and the figures are company-reported, but the structural point is large: aggregator-owned, no-install hardware plus a utility DR tariff can unlock dispatchable behind-the-meter capacity in the renter and multifamily population most DR portfolios miss entirely. This series has tracked the same barrier from four directions — Walmart’s investment in Base Power’s provider-owned model, SOLRITE’s no-upfront VPP agreements, ConnectDER’s 15-to-30-minute meter-socket adapter, and the Uplight op-ed’s behavioral track — and every one of them is an answer to the same question: who owns the hardware and who does the install. A 15,000-unit waitlist against 1,000 served households is the demand signal; the constraint is capital and logistics, not customer interest. If your service territory is heavily multifamily, price a no-install, aggregator-owned tier explicitly, because your installed-base enrollment math excludes those households today and your peak does not.
The caveats are substantial and should travel with any citation: every figure — 4 MWh, ~1,000 households, ~$200,000 in payouts, the 15,000-AC waitlist — is company-reported and printed without independent verification, the program is small and single-season, and this is a product-and-services launch rather than a filed program evaluation. One identity note for the record: “Every Electric” is a Brooklyn plug-in-battery aggregator and is a different company from “Everyday Electric,” the aggregator formerly known as Renew Home, which this series referenced in its October 14–23 entries. (Source: ESS News / pv magazine, November 19 — November 20 entry)
The breaker layer underneath all of this got a product, and it extends load control past the battery’s reach. SOLRITE — which finances, builds and owns residential solar-plus-storage under no-upfront Virtual Power Plant Agreements and operates in Texas, California, Connecticut, Massachusetts and Illinois — and battery maker sonnen added ADAPT Energy-managed Eaton AbleEdge smart breakers to their “VPP Node,” pairing the sonnenHub with breaker-level load control. During a VPP dispatch event the system sheds only pre-approved household loads (the example given is a pool pump) so the battery can export more of its capacity to the grid; during an outage it automatically sheds large loads to protect the battery from drain and inrush-current failure, restoring them when grid power returns. Controllable loads are reassignable by the homeowner without rewiring, with no separate loads subpanel required, and participation is opt-in and permission-based.
The capability worth noting is the direction of the value: shedding at the breaker frees battery capacity for export, which converts a backup asset into a larger grid asset without adding a kilowatt-hour of storage. For a utility program that is incremental dispatchable flexibility from loads that previously sat outside the battery’s reach, and the no-subpanel, no-rewiring design attacks install complexity — the same grid-edge friction the Arizona meter-socket approvals addressed one layer down. Caveats for sizing are severe: no pricing, home counts, MW capacity or named utility offtakers are disclosed, and the announcement does not use the term “DERMS” — it is a vendor product-and-partnership launch, not a quantified deployment. It is also distinct from SOLRITE’s earlier ComEd / Ameren Illinois expansion and the DTE–sonnen Michigan pilot already carried in this series. (Source: CleanTechnica, November 14 — November 16 entry; publication date precedes this window, first appearance in this series)
And an Uplight executive restated the enrollment thesis with the mechanism spelled out — reinforcement of a theme this series established last week, not a new finding. In a Utility Dive op-ed, Uplight VP of Virtual Power Plants Gisela Glandt argues the barrier to demand-response scale is opt-in enrollment design, not customer interest: roughly 80% of customers say they want help managing their energy while only ~10% enroll under opt-in programs, and shifting to default/opt-out and point-of-sale enrollment (with easy self-service opt-out) pushes participation into the 70–100% range — one Midwestern utility that moved a program from opt-in to opt-out scaled capacity 4–5× faster at 70–80% participation. The recommended design pairs behavioral demand response that needs no hardware and therefore reaches renters with OEM partnerships that enroll thermostats, water heaters, EVs and batteries by default at setup, using low-friction programs as an on-ramp to automated DR, time-varying rates and managed EV charging.
Report it as what it is: the fifth independent arrival at the same conclusion, from a vendor rather than a regulator, and the one with the sharpest number on the lever. The 70–100% opt-out range is the direct counterpoint to the ~75% enrollment assumption New Jersey’s 82% finding rests on — it is the only figure in this series that makes 75% look achievable, and it comes from the party with the most interest in it being true. That is precisely why it should go in the record as a contested range rather than a planning assumption: model 40%, 55% and 75%, cite the op-ed for the high case and Octopus’s 50–60% among smart-device owners for the realistic one, and let the commission see the spread. The usual caveat applies and matters here: this is a vendor executive’s op-ed whose figures are drawn from the author’s own industry experience rather than independently cited, and it is distinct from Uplight’s November 11 summer-portfolio results and the November 6 Octopus/Uplight participation piece, both reported in the November 13 digest. (Source: Utility Dive op-ed, November 11 — November 18 entry; publication date precedes this window, first appearance in this series. Reported as reinforcement of the enrollment theme established in the November 13 digest, not as a new development)
📋 Regulatory & Policy
North Carolina closed the large-load ratemaking thread this series has tracked for six weeks, and the data-center tariff is in the order. The North Carolina Utilities Commission issued written orders on November 16 approving the multi-year rate plans and performance-based-regulation framework for Duke Energy Carolinas (DEC) and Duke Energy Progress (DEP), finalizing settlements filed in July and August 2026. Rate-case terms: a 9.8% return on equity on a 53% equity layer, a historic base-case retail rate base of roughly $25.7B (DEC) and $17.8B (DEP) — about $43.5B combined — and approximately $3.8B (DEC) and $3.4B (DEP) of capital under the two-year plan, with revised rates effective January 1, 2027 and a conditional one-year rate-case stay-out (no new base-rate filings before November 1, 2028) tied to the commission granting deferrals for certain new generating-asset costs.
The operative provision for this series is that the orders create a “DC Schedule” for certain data-center customers — the large-load tariff tracked from Duke’s October 6 settlement filing (NCUC E-7 Sub 1329 / E-2 Sub 1380, repurposing Schedule HLF with a ≥50 MW / 80%-load-factor-or-≥150 MW threshold, 75% minimum billing demand, and 10–15-year minimum terms) through Walmart’s October 12 motion to reopen comments — now approved, though Duke says it is still assessing that portion and will engage stakeholders on implementation.
This is movement on a tracked item, reported as the decision rather than as a new tariff, and the significance is that a state has locked in the principle that large new loads pay upfront for the grid they trigger rather than socializing those costs. That protects the ratepayer base funding DSM programs and sharpens the avoided-cost comparison between a firmed-and-upfront-paid data center and a demand-side alternative — which is the comparison the Cornell study below quantifies nationally. But note the threshold design against this week’s federal convergence: North Carolina’s DC Schedule triggers at ≥50 MW at an 80% load factor or any load ≥150 MW, while all five RTO filings use a flat 50 MW. The load-factor condition is the lever, and it cuts both ways — it screens for the firm, high-utilization loads a tariff is meant to capture, and it writes out exactly the flexible, lower-load-factor large load the RTO non-firm tiers are designed to attract. If you are drafting or intervening in a large-load tariff, reconcile the state threshold against the RTO’s: a customer can be inside one regime and outside the other, and a flexible load that escapes the state tariff by running a low load factor is the customer the non-firm transmission tier now rewards.
Keep the reconciled threshold list this series maintains, now with the week’s additions: 50 MW (FERC-approved for PJM; all five RTO §206 filings; NERC-mandatory at ≥50 MW AND ≥100 kV; Florida’s first SB 484 tariff at ≥50 MW firm; PJM’s new LCL ride-through category), 75 MW (ERCOT Batch Zero), 200 MW (MISO’s Large Load Addition Resource Study co-modeling), 50-at-80%-or-any-150 (North Carolina), and 5 MW (TVA). Screen one reconciled customer list against all of them and size by the lowest. The caveat on the source: it frames the order primarily as a rate case and reports the DC Schedule only in outline, with the detailed tariff mechanics carried over from the previously logged settlement record. DEC and DEP are scheduled to combine into a single utility on January 1, 2027. (Source: The Energy Mag, November 17 — November 19 entry. The October 6 settlement filing and Walmart’s October 12 motion were reported in the October 26 digest; this is the order resolving that thread)
And a peer-reviewed study put the cost-allocation argument in Nature Energy with a 22-hour remedy attached. A Cornell study published November 9 in Nature Energy finds that across four major U.S. grids data centers paid only 38% ($3.12 billion) of the $8.25 billion in 2025 energy-infrastructure costs tied to their growth, leaving 62% ($5.13 billion) to fall largely on other ratepayers — with the lowest-income households seeing added energy costs of up to 28% of income and rural high-energy-burden families’ costs rising 76% more than urban residents’. The build-out shows in transmission investment climbing from $2 billion in 2020 to $34 billion in 2025, wholesale prices rising roughly 10% or more in the most-affected hubs — about 15% in Virginia and 20% in the southern San Francisco Bay Area — and cross-state subsidy, with residents of New Jersey, Pennsylvania and Ohio helping pay for grid upgrades driven by Virginia data centers, amplified because costs are allocated on past consumption, so households underwrite infrastructure built for future demand.
The modeled remedy is the finding to carry: if data centers curtailed power use during just 22 of each region’s peak demand hours per year — or shifted operations to another region at peak — grid strain and the need for new infrastructure would ease materially. Twenty-two hours. That is the whole ask, and it is now in a peer-reviewed journal rather than a trade panel. It is the same “curtail a few dozen hours to unlock dozens of gigawatts of headroom” logic this series has carried from Norris’s 40–60% utilization figure and the 41% Dominion distribution load factor — now quantified as a ratepayer-protection instrument as well as a reliability one, which is a materially stronger framing in a rate case than reliability alone. The authors frame implementation and coordination of that flexibility, not its technical feasibility, as the binding challenge — which is precisely the orchestration-and-verification layer a DERMS supplies — and pair it with policy options (charging data centers directly for the demand they add, directing funds to low-income households).
Put the 22-hour figure beside this week’s filings, because the filings are the implementation mechanism the study says is missing. MISO’s GRITS interrupts when nodal LMP exceeds $100/MWh and still serves more than 95% of hours; ISO-NE’s $250/MWh trigger bound in about 1% of hours over five years. One percent of 8,760 hours is 88 hours; the study needs 22. The curtailment regimes being filed at FERC are, on the operators’ own numbers, more than sufficient to deliver the remedy a Nature Energy paper says would materially reduce new-build need — and that sentence is the single most useful thing to put in front of a commission this quarter. Caveats: this is a modeling study of 2025 data across four grids rather than an operational demonstration, and it is a distinct paper from the Cornell/Applied Energy “trading power reductions” flexibility study carried in this series in October — different team, journal and framing. (Source: Cornell Chronicle, November 2026, reporting Nature Energy, November 9 — November 16 entry; publication date precedes this window, first appearance in this series)
🏢 Benchmark Hygiene & Carried Context
Four reconciliations this week, and the first one is new and load-bearing.
1. The non-firm price stack is now a thing this series can quote, and it is a different stack from the residential incentive stack. Keep them apart. Residential flexibility incentives (six utility-set points, all above the ~$66/kW-year avoided-cost anchor by roughly 1.5× to 3×): $100 per average kW discharged (Alliant IA/WI) · $105/kW-year (DTE pilot) · up to $110/kW on average contribution across events (APS Storage Rewards) · $125 per average kW/year (Entergy Louisiana) · $200/kW new and $100/kW existing, one-time, trued up to delivered output (New Jersey Phase 2 Block 1) · plus $350,000/MW-year participant-facing in Google–Voltus, which is not a utility price. Large-load non-firm transmission pricing — new this week: PJM $58,219/MW-year non-firm against $90,000/MW-year firm (35.3% discount at ~64.7% load factor) · MISO GRITS interruption at nodal LMP > $100/MWh with service in 95%+ of hours · ISO-NE redispatch at $250/MWh, ~1% of hours 2021–25 · NYISO non-firm bid cap $200/MWh · CAISO project-specific operating restrictions, no published price. These are not comparable to each other — one stack is a payment to a residential customer per kW, the other a discount or trigger price on transmission service to a large load — and mixing them is the easiest error available this week.
2. The 50-MW threshold convergence is now five-for-five, and the outliers are the story. All five RTO §206 filings use 50 MW, as do PJM’s LCL ride-through category, NERC’s registry criteria (at ≥50 MW and ≥100 kV) and Florida’s first SB 484 tariff (≥50 MW firm). The outliers: ERCOT at 75 MW, MISO’s own Large Load Addition Resource Study at 200 MW (a study trigger, not a service threshold — do not conflate it with MISO’s 50-MW large-load definition), North Carolina at 50-at-80%-load-factor-or-any-150 MW, and TVA at 5 MW. One reconciled customer list, screened against all of them, sized by the lowest.
3. Winter DR capacity figures come from three different operators and three different scopes. NERC’s continent-wide 2026–27 Winter Reliability Assessment: demand response 43.6 GW, up from 41.7 GW; batteries +9.2 GW on-peak; gas +8.7 GW; peak demand +21 GW (2.6%); total resources +26 GW (2.4%). NYISO, New York only: 29,464 MW of resources against a 24,620 MW forecast peak, ~20% headroom; DR activated six consecutive days during Winter Storm Fern; last winter’s peak 24,317 MW (February 7, 2026) against an all-time winter peak of 25,738 MW (January 7, 2014). PJM, carried from the November 13 digest: a high-stress cold peak studied up to 152,200 MW against a record winter peak of 143,700 MW (January 22, 2025), with reliability holding at the stress case only with pre-emergency and emergency demand response in the stack. Three operators, three footprints, three assessments — cite the one whose footprint matches your argument and never add the DR figures together.
4. Carried benchmarks cited once this week rather than re-reported: the ~$66/kW-year and ~$20.70/MWh avoided-cost anchors · Brattle’s 4% benchmark (≈40 MW per GW of system peak) and its 10 GW → 35 GW OEM-enrollment finding · New Jersey’s 82%-at-~75%-enrollment finding and the 547.1 MW PowerUp target (November 13 digest) · Octopus Energy US’s 50–60% participation among smart-device owners against a 5–10% historical benchmark · Oracle’s ~1 GW across 73 events and Uplight’s 735 MW peak and 15,629 MWh, all vendor-reported without third-party M&V (November 13 digest) · Sunrun’s Puerto Rico fleet — ~70,000 batteries, 45 MW over four-hour windows, 18 consecutive days of dispatch from October 13 · PJM’s nine DR days in 2026 against eight across all of 2015–2024, and PJM’s still-unpublished deployed DR megawatts · PJM’s recovered qualifying-offset criteria and the $555/MW-day backstop cap · the MISO LMR non-performance penalty effective June 1, 2027 · SB 913’s June 2028 device-level-telemetry precondition · NERC Project 2024-02’s R2.2 and R3, and the RR27-1-000 registry petition at ~68,288 MW / ~86% coverage · EPRI FlexMosaic’s five classes · the 2026 §202(c) docket exceeding fifty orders with nothing accredited · Carnegie Mellon’s nine-or-ten-departments-per-device finding and the ~800 co-ops / ~2,000 munis without VPP programs · ConnectDER’s 15-to-30-minute meter-socket install · FERC’s twice-held ruling that emergency-order costs fall on the load inside the emergency footprint (~$26M across two orders for a never-dispatched 729.9 MW unit).
🔬 EPRI Research Spotlight
No EPRI publication dated inside the window — the nineteenth consecutive week — and this week the carried FlexMosaic framework has its most concrete application yet, because five grid operators just published the class parameters without naming the classes. A direct check returned no EPRI publication dated November 16–20. The drought remains cadence rather than omission: EPRI’s large-load and flexibility work is the substrate beneath the computational-load standards NERC’s board adopted on October 26 and beneath the curtailable tiers filed this week, and the framework is again the instrument the week’s biggest story is missing.
Recall what FlexMosaic is and why it fits this week exactly. Announced at CERAWeek in March 2026 and built with 65-plus utilities, system operators, regulators and hyperscalers, FlexMosaic defines five flexibility classes for large loads — Class A, support during infrequent extreme system stress; Class B, management of daily or weekly demand peaks; Class C, response to long-lasting energy shortages; Class D, response to sudden supply or demand swings; Class E, frequency stabilization. Class membership is determined by four parameters: notification time, response duration and frequency, and the depth and speed of the response. Demonstrations have delivered up to 40% flexibility.
Now put this week’s filings against those four parameters, because every one of the five operators published them and none of them named a class. This is the exercise to run before commenting on any of the filings:
- MISO GRITS — trigger is nodal LMP above $100/MWh, frequency is the residual of “service in more than 95% of hours” (so up to ~438 hours a year), duration is undefined, depth is full interruption of non-firm load. An up-to-438-hour, price-triggered, indefinite-duration obligation is not Class A insurance — it is Class B daily-and-weekly peak management with a Class C tail, and it is priced and procured as a single undifferentiated product.
- ISO-NE non-firm / interim — trigger is area energy prices at $250/MWh, frequency ~1% of hours (about 88 hours a year on the 2021–25 record), depth is consumption reduction or eligible generation increase, and the service is excluded from the Installed Capacity Requirement. Eighty-eight hours at a high price trigger is closer to Class A, which is why the ICR exclusion is coherent — ISO-NE is declining to count a resource it is also declining to rely on.
- NYISO non-firm — bid cap $200/MWh; frequency and duration undefined in the reporting, with a “State-Exempt Load Project” resource-adequacy exception alongside it.
- PJM — the non-firm rate ($58,219/MW-year vs $90,000 firm) is a price with no published trigger, frequency or duration in the coverage, while the separate ride-through filing specifies a Class-D-shaped obligation (response to sudden voltage and frequency swings) as a condition of interconnection, with ramp-rate requirements to follow.
- CAISO — project-specific operating restrictions under Flexible Interim Load Interconnection: by construction, a different class per project, with no published scheme for classifying them.
Here is the deliverable, and it is cheap, specific and available during the comment windows that are open right now. File FlexMosaic’s class definitions into the §206 comment dockets as the classification scheme for each operator’s non-firm service tier, and ask each operator to state which class its tier is. The reason is not tidiness; it is double counting and mispricing. A tier that spans Class A and Class B should be written as two products, because an operator that counts a Class A rare-stress resource in its Class B peak-management plan has counted the same megawatts twice — the identical error this series flagged when Georgia Power said it may curtail OpenAI’s 1,000 MW “during peak grid demand or local emergencies,” which is Class B and Class A, with no party saying which governs. MISO’s 95%-of-hours language is the same ambiguity in tariff form: it defines the service by what it is not (the 5% of hours) rather than by notification, duration and frequency. The cheapest question at any large-load negotiating table remains “which FlexMosaic class is this?”, because the class derives the rest of the terms for free.
Two supporting moves, both carried forward and both sharper this week. One: the R2 methodology filing remains the highest-value action on the board. NERC Project 2024-02’s R2.2 requires a Planning Coordinator’s methodology to address “Demand Response programs, capabilities, and operational limitations” and defines none of those terms, while R2.3.5 covers “energy storage resource constraints” and R3 scores risk on the frequency and duration of unserved-load events, maximum unserved load and total unserved energy. FlexMosaic’s four parameters are those terms, already agreed with 65-plus institutions and already public. File them as the quantification scheme before the methodology is documented, because an R2 methodology is written once and inherited for years, and one treating demand response as a single undifferentiated input will derate a Class C long-duration-shortage resource to a Class A rare-stress assumption and never show the difference. This week supplies the evidence R3 actually scores: NYISO activated DR for six consecutive days during Winter Storm Fern, and NERC counts 43.6 GW of DR this winter — duration and total energy, not peak coincidence. Two: argue R6 explicitly, because a Corrective Action Plan may consist of a planner’s own mitigating actions or recommendations to third parties including measures needing regulatory approval — which makes a CAP a potential demand-side procurement, and nobody in the demand-side industry appears to be making that argument. The caveat that must travel: Project 2024-02 is initial-draft stakeholder language, not an adopted standard, and the posted draft still reads “[Standard Name and Number TBD].” (Targeted gap-fill — EPRI direct check returned no publication dated November 16–20, the nineteenth consecutive week. FlexMosaic is a carried framework from this series’ spring and summer coverage, applied here to the five §206 filings)
🚩 Utility-Sector Relevance Flags
⚑ Five Grid Operators Just Published A Price For Interruptibility — PJM’s Is A 35.3% Discount, And That Is The Number This Series Has Been Asking For All Quarter
Topic: Large-Load Interconnection / Valuation of Curtailability
Relevance: Responding to FERC’s June 18 §206 show-cause orders (rulemaking RM26-4 still open), ISO-NE filed November 13 and CAISO, MISO, NYISO and PJM on November 16, with SPP due around November 20 and comments due ~30 days after each. All five broadly adopt a 50-MW large-load threshold and every one creates a curtailable, non-firm transmission tier that can energize before firm upgrades are complete. PJM: a one-year “Fast Energization Process” (~2× faster) and a non-firm rate of $58,219/MW-year against $90,000/MW-year firm — a 35.3% discount at roughly 64.7% load factor; PJM serves ~10 GW of large computational load and forecasts up to 30 GW of additions by 2030. MISO GRITS: interruption when nodal LMP > $100/MWh (service in 95%+ of hours), 20-year revenue commitment, $250,000 application fee, implementation no earlier than January 1, 2029; plus a Large Load Addition Resource Study co-modeling loads ≥200 MW with their generation. ISO-NE (ER27-484-000): three tiers — firm with a “bring your own new capacity” match, interim, and a permanent non-firm “flexible load” service — with re-dispatch at $250/MWh (~1% of hours 2021–25) and exclusion from the Installed Capacity Requirement; study process June 1, 2027, services June 1, 2028. NYISO: Firm and Non-Firm Withdrawal Services, non-firm bid cap $200/MWh, plus a “State-Exempt Load Project” RA exception. CAISO: Flexible Interim Load Interconnection with project-specific operating restrictions, board-approved October 28. These are §205 filings FERC has not ruled on.
Action Signal: Implement — Put PJM’s two rates side by side with your own avoided-capacity benchmark today. $31,781/MW-year of rate difference is roughly $31.78/kW-year assigned to curtailability alone, on transmission service, before a single capacity-market dollar — set against this series’ ~$66/kW-year anchor, that is an operator-published, tariff-stated valuation of flexibility rather than one inferred from a model. Three things follow. One: put the large-load tariff team and the DSM team in the same room before either writes a specification, because every tier above requires the operator to verify that a load actually reduced when the trigger fired — the same telemetry, dispatch and settlement capability a DR portfolio already has. A utility with DR measurement infrastructure has already built most of what its large-load tariff needs; one without is now buying it twice. Two: classify each tier before commenting — see the EPRI Spotlight: MISO’s up-to-438-hour price-triggered obligation and ISO-NE’s ~88-hour $250/MWh trigger are different products sold as one category each, and a tier spanning FlexMosaic Class A and Class B should be written as two products or the same megawatts get counted twice. Three: cite them as proposals, not rates — comments are open ~30 days after each filing and MISO’s own implementation is no earlier than January 1, 2029.
⚑ The Same Capability Is A Discount In One PJM Docket And A Condition Of Service In The Other — Controllability Just Became Table Stakes
Topic: Grid Reliability / Ride-Through Standards
Relevance: On November 16 PJM filed voltage- and frequency-ride-through requirements with FERC for a newly defined Large Computational Load (LCL) category at 50 MW or greater, effective January 16, 2027. The trigger was a July 22 transmission disturbance in which nearly 3.9 GW of data centers switched from utility power to onsite backup generation — the largest such load transfer in PJM’s history — affecting voltages, frequency, power flows and generator behavior across the Dominion zone and the Eastern Interconnection, managed safely by PJM’s control center. New LCLs must demonstrate compliance as a condition of interconnection and operation; existing LCLs within one year of the effective date or commercial operation, whichever is later. PJM says the definition aligns with large-load standards under development at FERC and NERC, with ramp-rate requirements to follow. Specific voltage and frequency thresholds are not disclosed in the coverage. Distinct from PJM’s separate IRAS / “bring your own new capacity” market-design proposal, largely rejected by FERC on October 9 on jurisdictional grounds.
Action Signal: Engage — Read the two PJM filings as one policy and say so in your comments: a load that can ride through and curtail on a price signal pays $58,219/MW-year instead of $90,000 and gets a one-year study, while a load that cannot ride through will not be allowed to connect. That is the telemetry-and-verification layer this series keeps specifying moving from “how you earn DR revenue” to “how you get served” — the strongest available argument that demand-side measurement infrastructure is infrastructure, not program overhead. Get the filing itself for the thresholds, because the ride-through envelope is the engineering term your large-load customers will have to design to, and watch for the ramp-rate filing, which is the one that will convert a Class D obligation into a measurable performance standard. And screen your own interconnection queue now: every existing ≥50 MW computational load in PJM has a one-year compliance clock, and the customer conversations are easier before January 2027 than after.
⚑ Connecticut Ordered A DERMS By Implication, With Dates — Mandate The Function And The Deadline, Not The Platform
Topic: Flexible Interconnection / State DERMS Mandate
Relevance: Connecticut PURA’s November 4 decision in Docket 25-01-27 sets a flexible-interconnection timeline: scheduled flexible interconnection across the distribution companies by May 1, 2027; dynamic flexible interconnection at the most constrained substations from May 1, 2028; expanding systemwide by November 1, 2030 — with PURA noting that dynamic operation requires more advanced monitoring and control, “such as a grid-scale DERMS or an ADMS.” Reported through an OATI recap of DTECH Northeast (Boston, November 2–4), which also offers a phased DERMS path — DER forecasting and visibility, then dispatch pilots, then flexibility in daily operations — names the open data question as real-time telemetry versus 5-to-15-minute intervals, and reports CAISO running its “AI Genie” live in its production outage-management system with staff retaining final authority, 250+ utilities on OATI DERMS, one utility that “does not yet trust AI for forecasting,” and an ISO-NE CELT normal-weather winter peak of 20.5 GW in 2026/27 rising to 26.4 GW by 2035/36. Vendor-authored, sponsored recap; the PURA decision, CAISO deployment and ISO-NE forecast are independently verifiable.
Action Signal: Implement — Copy PURA’s structure into your own advocacy: it mandated a capability and a date, not a product, and then observed what the capability requires. That is the posture to ask a commission for, because it survives a procurement cycle and a vendor change. Then do three things. Get the order from the PURA docket rather than citing the recap — this is third-party regulatory substance inside a sponsored document, and the citation should point at the source. Use the May 1, 2028 dynamic date as a planning anchor if you operate in New England, since a constrained-substation deployment in 18 months means the forecasting-and-visibility phase has to be underway now. And settle the telemetry-interval question per service rather than once for the platform — the one-second frequency-response figure from the IOTech item below is the proof that the answer differs by service, so write the interval requirement into each use case (operations, dispatch, settlement) separately. This is the first instrument in this series where a state interconnection mandate converts directly into a DERMS or ADMS procurement with a deadline attached.
⚑ NERC Counted 43.6 GW Of Winter Demand Response And Batteries Beat Gas — Quote The Reliability Organization, Not The Advocate
Topic: Demand Response as Capacity / Winter Reliability
Relevance: NERC’s 2026–27 Winter Reliability Assessment, released November 18, reports aggregate peak demand up 21 GW (2.6%) year-over-year (against +20 GW last winter and a ~1% historic norm), attributed explicitly to “the expansion of data centers, large industrial loads, and broad electrification across end-use sectors.” Demand response grew to 43.6 GW from 41.7 GW; batteries added 9.2 GW of on-peak capacity — the single largest category of growth, ahead of natural gas at +8.7 GW — against total North American bulk-power resource growth of 26 GW (2.4%). No area short under normal weather (“the resource outlook has improved,” CEO Jim Robb); SERC at risk under more severe conditions; localized fuel and transmission constraints in ISO-NE and parts of MISO and SPP; and a ten-year outlook in which many areas face a high risk of energy shortfalls and rolling blackouts by 2030. Freezing temperatures and an El Niño pattern flagged as gas-supply risks.
Action Signal: Implement — Put the 43.6 GW figure and the batteries-over-gas comparison, with NERC’s own attribution, into every proceeding where demand-side credit is being set this winter. Two of the three fastest-growing on-peak resource categories in NERC’s own accounting are demand response (+1.9 GW) and batteries (+9.2 GW), and batteries beat gas — which is the avoided-cost argument stated by the reliability organization rather than by the industry that benefits from it. Then name the contradiction, because it is inside one institution and it is this quarter’s throughline: NERC counts 43.6 GW of demand response in its seasonal assessment, and NERC Project 2024-02 — the energy assurance standard that will define “dependable energy under stress” for long-term planning — names demand response exactly once, at R2.2, as a methodology input, with DERs, VPPs and load flexibility appearing nowhere. The same body that counts the resource operationally has written no rule that credits it in planning. That gap is now assertable with a number on the operational side, which it was not a month ago. Use it in the Project 2024-02 comment period and in your Planning Coordinator’s R2 methodology.
⚑ Eversource Cannot Say Which Of Its Customers Have Air Conditioning — And It Is Dispatching VPPs At Named Distribution Constraints Anyway
Topic: Distribution-Level Observability / Non-Wires Alternatives
Relevance: At a Latitude Media / EnergyHub webinar, National Grid (about 100 MW of DER load in New England and more than 350 MW upstate — roughly 2% and 5% of regional peak, mostly thermostats with storage and managed EV charging growing) and Eversource (about 100,000 thermostats and 25,000 C&I customers, in a territory including ISO-NE’s largest load zone) described shifting ConnectedSolutions and ConnectedSolutions+ from peak-shaving to location-specific distribution-constraint relief, with CS+ pilots targeting Greater Boston substations and a southeastern Massachusetts substation absorbing midday solar. The binding gap, stated by the utilities: Eversource cannot yet confidently say which customers have air conditioning, and only a few unmanaged EVs can overload the small pole-top transformers common in its territory — targeted managed charging could defer those replacements once AMI improves asset-level load data (“we will be in a much better position in a year”). No dollar figures, dockets or CS+ pilot MW disclosed; executives remained “cautiously optimistic”; sponsored webinar.
Action Signal: Implement — Sequence the investment honestly: AMI-derived asset-level load data and customer end-use attribution come before locational dispatch, not after. You cannot target a constraint you cannot see, and two large Northeast utilities have now said on the record that they cannot see it yet. Then audit your own claims: if your program already books locational deferral value, state your evidence that the enrolled devices sit on the constrained asset — that is the question an intervenor will ask and the one this admission makes fair game everywhere. Two things to build. A customer end-use attribution layer (which premises have central AC, heat pumps, EVs, water heaters) is a prerequisite deliverable, not an analytics nice-to-have — put it in the DERMS or AMI business case explicitly with the transformer-deferral benefit attached, because the pole-top-transformer case is the cleanest small-dollar non-wires alternative in this series. And carry the maturity curve as the headline finding: VPPs are moving from a capacity-only peak resource to a non-wires alternative that defers specific distribution upgrades, which is a higher-value, locational use the avoided-cost framework can credit against T&D capital, not just generation — so run your avoided-cost calculation locationally or you are underpricing your own program.
⚑ “Up To 90% Consistency For The Full Event Duration” Is The Right DR Metric And Almost Nobody Publishes It — So Contract On It
Topic: M&V / Program Performance Metrics
Relevance: EnergyHub reports managing about 4 GW of flexible capacity for load-serving clients that called roughly 3,500 load-shifting events this past summer, and that across 70 dynamic load-shaping events for 12 utilities it delivered up to 90% consistency for the full event duration. Separately this week, PSE&G reported 16 events over a June 1 – October 30 season with one event cutting 38.15 MW and the portfolio dispatched during each of PJM’s five highest grid-peak hours. All vendor- and utility-self-reported; EnergyHub’s figures come from a sponsored webinar with no measurement point, baseline method, named utility or docket disclosed, and PSE&G’s 38.15 MW is a single-event peak rather than a season average or firm capacity value.
Action Signal: Implement — Full-duration consistency is the metric that matters, because the classic DR failure mode is an event that starts firm and decays as customers drop out — and a 90% full-duration figure says more about deliverability than any aggregate MWh or peak-MW claim. But convert it before using it, with this series’ four tests: the measurement point (grid interface, service point or per asset), the baseline method, delivered-versus-committed MW per event rather than aggregate MWh or a percentage of commitment, and the named utility whose filed evaluation will confirm it. None is disclosed. So the right use of the 90% figure is as the acceptance criterion you write into your own aggregator contract, not as evidence that anyone has met it — and pair it with a measured attrition clause, because the open question this series carries from Octopus’s Nick Chaset is whether participation holds as call frequency rises. Second action, and it is nearly free: add peak-hour coincidence to your own published program metrics. PSE&G’s “dispatched during each of PJM’s five highest grid-peak hours” is a coincidence claim, and coincidence is what capacity accreditation actually pays for — it is derivable from data you already hold, it is what an accreditation proceeding cares about, and almost no program reports it.
⚑ A Nature Energy Paper Says 22 Peak Hours Of Data-Center Curtailment Would Materially Cut New-Build Need — And This Week’s Filings Already Exceed It
Topic: Large-Load Cost Allocation / Flexibility Remedy
Relevance: A Cornell study published November 9 in Nature Energy finds that across four major U.S. grids data centers paid only 38% ($3.12B) of the $8.25B in 2025 energy-infrastructure costs tied to their growth, leaving 62% ($5.13B) largely on other ratepayers — lowest-income households seeing added costs up to 28% of income, rural high-energy-burden families’ costs rising 76% more than urban residents’, transmission investment climbing from $2B (2020) to $34B (2025), wholesale prices up ~15% in Virginia and ~20% in the southern SF Bay Area, and cross-state subsidy (NJ/PA/OH ratepayers helping fund Virginia upgrades), amplified because costs are allocated on past consumption. The modeled remedy: curtailing data-center use during just 22 of each region’s peak demand hours per year — or shifting operations across regions at peak — would ease grid strain and reduce new-infrastructure need. The authors name implementation and coordination of that flexibility, not technical feasibility, as the binding challenge. A modeling study of 2025 data, not an operational demonstration; distinct from the Cornell/Applied Energy “trading power reductions” paper carried in October.
Action Signal: Engage — Put the 22-hour figure next to the operators’ own trigger frequencies, because that single comparison is the strongest thing available to file this quarter. MISO’s GRITS interrupts above $100/MWh nodal LMP and still serves more than 95% of hours; ISO-NE’s $250/MWh trigger bound in about 1% of hours over 2021–25 — and one percent of 8,760 hours is 88 hours against the study’s 22. The curtailment regimes being filed at FERC are, on the operators’ own numbers, more than sufficient to deliver the remedy a peer-reviewed journal says would materially reduce new-build need. File that into the §206 comment dockets and into any state large-load proceeding. Then take the framing shift, which is the real value: this reframes large-load flexibility as a ratepayer-protection and affordability instrument rather than only a reliability one, and in a rate case the affordability frame is materially stronger — especially with the equity numbers attached (28% of income at the low end, rural costs rising 76% more than urban). And note what the authors say the barrier is — implementation and coordination — because that is the orchestration-and-verification layer a DERMS supplies, which makes this paper a citation in a DERMS business case, not just a policy brief.
⚑ North Carolina’s Data-Center Tariff Triggers On Load Factor — Which Writes Out Exactly The Flexible Load The RTO Non-Firm Tiers Reward
Topic: Large-Load Cost Allocation / Threshold Reconciliation
Relevance: The NCUC issued written orders November 16 approving multi-year rate plans and a performance-based-regulation framework for Duke Energy Carolinas and Duke Energy Progress: 9.8% ROE on 53% equity; combined rate base ~$43.5B ($25.7B DEC + $17.8B DEP); two-year capital ~$7.2B ($3.8B + $3.4B); new rates January 1, 2027; a conditional one-year rate-case stay-out (no new base-rate filings before November 1, 2028) tied to deferrals for certain new generating-asset costs. The orders create a “DC Schedule” for certain data-center customers — the tariff tracked from Duke’s October 6 settlement filing (E-7 Sub 1329 / E-2 Sub 1380, repurposing Schedule HLF at ≥50 MW with an 80% load factor or any load ≥150 MW, 75% minimum billing demand, 10–15-year minimum terms) through Walmart’s October 12 reopen-comments motion — now approved, with Duke still assessing implementation and engaging stakeholders. DEC and DEP combine into one utility January 1, 2027. Source frames it primarily as a rate case and reports the DC Schedule in outline.
Action Signal: Engage — Reconcile the state threshold against the RTO’s, because this week they diverged in a way that creates an arbitrage: North Carolina triggers at ≥50 MW at an 80% load factor or any load ≥150 MW, while all five RTO §206 filings use a flat 50 MW. The load-factor condition screens for firm, high-utilization loads — and in doing so it writes out precisely the flexible, lower-load-factor large load the RTO non-firm tiers are built to attract. A customer can sit inside one regime and outside the other, and a flexible load that escapes the state tariff by running a low load factor is the same customer the non-firm transmission tier now rewards with a 35.3% discount. If you are drafting or intervening, say that out loud in the record and ask which outcome the commission intends. Keep the reconciled threshold list and screen one customer list against all of it, sized by the lowest: 50 MW (PJM/FERC, all five §206 filings, NERC at ≥50 MW and ≥100 kV, Florida SB 484 firm, PJM’s LCL ride-through) · 75 MW (ERCOT Batch Zero) · 200 MW (MISO’s Large Load Addition Resource Study — a study trigger, not a service threshold) · 50-at-80%-or-any-150 (North Carolina) · 5 MW (TVA). Report this as the decision on a six-week tracked thread, not as a new tariff — and take the principle it locks in: large new loads pay upfront for the grid they trigger rather than socializing those costs, which protects the ratepayer base that funds DSM programs.
⚑ 537 GW Reported, 120 GW Signed — The Fourth Independent Demonstration That A Financial Filter Collapses A Large-Load Pipeline
Topic: Large-Load Forecast Risk / IRP Overbuild
Relevance: Latitude Media’s data-center tracker, published November 13, finds utilities in 53 U.S. service territories have signed contracts for about 120 GW — roughly 22% of the 537 GW they report in their pipelines, still about 5.5× the 22 GW operating today — while already soliciting ~37 GW of new generation. Reported-versus-signed extremes: Georgia Power 80.1 vs 12.8 GW; Dominion 54.9 GW pipeline vs ~12 GW under service agreements (plus 32.5 GW on non-firm engineering letters, only 41% in its forecast); signed load as a share of peak — AEP Ohio 235%, NIPSCO 169%, PPL 145%. Cost consequence: PJM’s market monitor attributes $29.4B (~46%) of four auctions’ $63.6B capacity cost to data-center load, even as PJM utilities cut near-term large-load forecasts by a combined 6.5 GW, potentially shrinking PJM’s reliability-backstop procurement to as little as 630 MW from ~6.8 GW. Speculative-load signals: Amazon can cut 1,200 MW from NIPSCO by March 31, 2029; average ~33-month announcement-to-power lag.
Action Signal: Watch — Use the 537-to-120 ratio, not the headline, and use it the way this series has used the other three: a capital plan or an IRP sized against an unscreened queue is sized against an artifact. The collapse now has four independent demonstrations at four denominators — AEP Ohio 30 → 13 GW, Oncor’s 293 GW against ~45 GW conditionally classified under ERCOT Batch Zero, PPL’s 2030 adjustment falling 13,343 → 12,764 MW between submissions, and now 537 GW reported against 120 GW signed nationally. The deferrable, cancellable demand-side alternative is the lower-regret purchase at every point on that curve, and the argument lands hardest in the utility’s own words where a filing supplies them — make “risk of unrealized loads” the frame. Keep the denominators reconciled rather than mixed: Oncor’s 293 GW transmission-queue share, ERCOT’s ~474 GW audited queue, the 191.8 GW of large-load studies, and now the national 537 GW reported / 120 GW signed pair are four different measures, not one number. And note the two-sided risk the PJM figures expose: forecasts that come down by 6.5 GW can shrink a backstop procurement by an order of magnitude, so a demand-side resource counting on that backstop should not plan around the 6.8 GW figure.
⚑ Dominion’s Distribution System Averages A 41% Load Factor — That Is The Non-Wires Business Case, Stated As A Number You Already Have
Topic: Distribution Planning / Grid Utilization
Relevance: At BNEF’s Houston Summit, Google’s Tyler Norris put U.S. grid utilization at 40–60% and called enabling non-firm transmission service the “real whitespace.” The distribution figure: Dominion Energy Virginia’s distribution averages a 41% load factor, peaking just under 80% utilization, against a median transmission-facility utilization of roughly 33%. Systemwide load factor is put at about 60% today rising to 66% by 2030 (Grid Strategies). A flexibility-backed-interconnection template: an Aligned Data Centers / Portland General Electric arrangement in which a 31 MW / 62 MWh battery tied to a 72 MW Hillsboro, Oregon data center let the project bypass a fully firmed network upgrade. NRG’s Jessica Hogle set a 25 GW demand-response, storage and VPP target by 2035. Panel commentary, not a filed dataset; the Aligned/PGE arrangement is reported without terms.
Action Signal: Implement — Compute your own distribution and transmission load factors and put them in the IRP and the next distribution capital filing. A utilization figure is the cheapest available argument for a non-wires alternative and your own AMI and SCADA data already supports it. The headroom between a 41% average and a sub-80% peak is capacity a flexible-load or VPP program can monetize before a wires upgrade is triggered — the same “better use of the grid we already have” logic now carried by the CPUC’s directive, Connecticut PURA’s flexible-interconnection order, and all five RTO filings. Then treat battery- or flexibility-backed interconnection as a replicable deferral mechanism rather than a one-off: the 31 MW / 62 MWh for 72 MW arrangement is a sizing ratio worth testing against your own queue — roughly 43% of nameplate load in power and about two hours of duration, bought instead of a firmed network upgrade. Ask the obvious follow-up in any proceeding where it appears: who holds the dispatch right, and is the battery credited as the utility’s deferral or the customer’s? That question is the difference between a non-wires alternative and a private workaround, and this series has found the answer is the whole value of the arrangement.
⚑ DOE’s Latest §202(c) Order Names A 1-GW Hyperscale Load As Part Of Its Reason — And Says The Plant It Saves Is Not A Capacity Resource
Topic: Section 202(c) / Emergency Retention vs Planned Flexibility
Relevance: Order No. 202-26-51, issued November 13 by Secretary Chris Wright, directs Consumers Energy to keep the 1,420-MW J.H. Campbell coal plant available to MISO from November 15, 2026 through February 12, 2027 — a notably longer term than the prior 90-day renewals (202-25-3/-7/-9, 202-26-16/-22/-39) and the first new order in the docket since No. 202-26-50 on October 12. The finding is a “risk of substantial energy supply shortfall in the MISO region, namely rising electricity demand coupled with a decrease in the availability of dispatchable generation,” citing NERC’s 2026 assessments, the OMS-MISO survey, and Consumers’ own planned 1-GW hyperscale data center, plus Winter Storm Fern’s January 24, 2026 EEA-2/MaxGen event, during which Campbell produced 201,962 MWh (2,441,255 MWh net January–July 2026). The order states the plant “shall not be considered a capacity resource.” Contested legal footing: a September 11, 2026 D.C. Circuit opinion DOE disputes and has petitioned to rehear en banc.
Action Signal: Watch — Carry the arithmetic unchanged, because it got one order heavier and the order’s own words make the argument: emergency coal retention is the expensive, last-resort version of what a planned, compensated DR or VPP program delivers in an orderly way with a settlement record attached — and DOE itself says these megawatts are “not a capacity resource,” so they build no accreditation. The 2026 §202(c) docket now exceeds fifty orders and has accredited nothing, which remains the cleanest statement of why directive flexibility builds no asset while compensated flexibility does. The sharper point this week is the pairing: large loads in item one are buying faster interconnection with curtailability at a 35.3% discount, while the federal government holds a coal plant open partly because of a 1-GW hyperscale load on the same system. Data-center load is both the driver of the emergency and, made flexible, the remedy — and only one of those two framings earns the customer a rate discount. Put the two documents on one page in any MISO-region proceeding. Treat the order as a signal of stress and of the value of firm flexibility, not as evidence of a new resource, and note the legal contest travels with any citation.
⚑ A Renter VPP Is Running On Con Edison With No Electrician, No Permit And A 15,000-Unit Waitlist — Price The No-Install Tier
Topic: VPP Deployment / Hard-to-Reach Segments
Relevance: Jackery and Brooklyn aggregator Every Electric launched what they call North America’s first plug-in-battery VPP aimed at renters — roughly 2,000 portable Jackery Explorer 2000 v2 units across 1,000+ Con Edison households, about 4 MWh of flexible storage with nearly 50 MWh cycled, enrolled in Con Edison’s “Smart Usage” DR program to shift air-conditioning peaks, running since September. Drop-shipped, no electrician, permit or utility agreement, running a window or portable AC through a passthrough outlet. Economics: a $0-upfront tier ($30 refundable deposit) where the aggregator keeps the battery and the Con Ed DR revenue and pays the resident $100–150 per AC per year, or a $650 purchase (off a $999 list) earning up to $250 per AC per year; roughly $200,000 in seasonal payouts and a waitlist of more than 15,000 air conditioners. Every figure is company-reported and printed without independent verification; small, single-season, a product launch rather than a filed evaluation. “Every Electric” is not “Everyday Electric,” the aggregator formerly Renew Home.
Action Signal: Watch — Take the channel and the segment, not the megawatt-hours. 4 MWh is immaterial; the structural point is that aggregator-owned, no-install hardware plus a utility DR tariff reaches the renter and multifamily population installed solar-and-storage structurally excludes — and that is the population your installed-base enrollment math leaves out while your peak does not. This series has now tracked the same barrier from five directions, all answering one question — who owns the hardware and who does the install: Walmart’s investment in Base Power’s provider-owned model, SOLRITE’s no-upfront VPP agreements, ConnectDER’s 15-to-30-minute meter-socket adapter, the behavioral track in the Uplight op-ed, and now drop-shipped portable storage. A 15,000-AC waitlist against ~1,000 served households is a demand signal; the constraint is capital and logistics, not customer interest. So if your territory is heavily multifamily, price a no-install, aggregator-owned tier explicitly and carry it as a separate enrollment channel in the business case. And watch the one thing a launch cannot tell you: whether portable, customer-sited units survive a second season with their dispatch performance intact — ask for the degradation and attrition record next summer before crediting the channel with capacity.
⚑ A Vendor Firmware Push Can Silently Change What A Register Means — Which Means It Can Silently Change A Settlement Quantity
Topic: DERMS/ADMS Specification / Device Integration
Relevance: IOTech CPO and co-founder Andy Foster argues “a VPP is only as good as its weakest device integration,” and that the binding constraint on scaling distributed fleets is device heterogeneity — manufacturer, model and firmware differences, where even a firmware update can change what a register means — rather than orchestration-platform capability. The proposed design is two-tier, federated, “decide centrally, execute locally”: low-power Arm edge nodes beside each asset speaking native protocols and enforcing local limits, normalized to a common model with OPC UA for supervisory control and AI automating the historically manual mapping of device data to SunSpec, while forecasting, trading and portfolio optimization run centrally. Control stays on-site for speed (fast frequency services need a full response within about one second), resilience (sites keep respecting grid limits if the link drops) and data volume. One grid-scale storage provider is said to embed the software across roughly 20 GW worldwide. Vendor framing; the ~20 GW claim is not independently verified.
Action Signal: Implement — Put three things in the evaluable requirements section of the RFP rather than the feature list: interoperability across manufacturer and firmware revisions, firmware-change resilience, and defined local fallback behavior when the link to the platform drops. The firmware point is the one with money attached and it is rarely specified: a resource whose register map changes under a vendor firmware push is a resource whose settlement quantity changes silently — that is a measurement-integrity failure before it is an integration failure, and it is invisible until an audit. Add a contractual obligation that the aggregator or platform notify the utility of any device firmware change affecting telemetry, and test the notification path, not just the device. Then settle the interval question per service: the one-second frequency-response figure here is the evidence that a single platform-wide telemetry interval is wrong, and it pairs directly with the real-time-versus-5-to-15-minute question Connecticut PURA’s timeline raises — specify the interval separately for operations, for dispatch and for settlement, because this series’ standing rule holds that inferred feeder-level state serves operations while measured per-asset telemetry serves accreditation and settlement, and a vendor offering the first as a substitute for the second is offering an operating tool and a disallowance.
📌 Sources
November 16, 2026 Entry
– Cornell Chronicle — Data Centers Need Power: Lower-Income, Rural Residents Pay the Price (November 2026), reporting a study published in Nature Energy (November 9, 2026) (first appearance; publication date precedes this window and entered the log through the November 16 entry’s weekend reach. Distinct from the Cornell/Applied Energy “trading power reductions” flexibility study (Mahuze et al., September 22) carried in this series’ October 15–16 coverage — different team, different journal, different framing: cost allocation and equity rather than inter-data-center flexibility trading. Checked by name as well as URL)
– CleanTechnica — Sonnen, SOLRITE & ADAPT Energy, Integrating VPP Breakers to Turn Home Loads into Flexible Grid Assets (November 14, 2026) (first appearance; publication date precedes this window. Distinct from SOLRITE’s earlier ComEd / Ameren Illinois expansion and from the DTE–sonnen Michigan pilot already carried in this series — same vendors, different arrangements)
– ESS News — INTERVIEW: “A VPP is only as good as its weakest device integration” (November 12, 2026), interview with IOTech CPO and co-founder Andy Foster (first appearance; publication date falls inside the prior week’s calendar range but the item entered the log only in the November 16 entry and does NOT appear in the November 13 digest’s Sources — verified by name and by URL)
– Tech Times — Latitude Finds Data Center Power Contracts Cover 22% of Reported Load as Utilities Seek Power Plants (November 16, 2026), reporting Latitude Media data published November 13 (first appearance; the Oncor 293 GW / ERCOT ~474 GW / 191.8 GW figures carried in prior digests are different denominators and are reconciled in Benchmark Hygiene rather than mixed)
November 17, 2026 Entry
– PJM Inside Lines — PJM Proposes New Rules to Protect Grid Reliability as Data Center Demand Grows (November 16, 2026) (first appearance; the reliability/ride-through track, distinct from PJM’s §206 large-load response consolidated above and from PJM’s separate IRAS / “bring your own new capacity” market-design proposal that FERC largely rejected on October 9, carried in prior digests)
– ISO Newswire — ISO-NE Files Proposal With FERC to Address “Large Loads” (November 13, 2026) (first appearance; publication date precedes this window. Consolidated into the single large-load filings item (deduplicated) — Docket ER27-484-000, the one clearly-confirmed filing at the November 16 deadline)
– Construction Review Online — U.S. Grid Operators Rework Power Connection Rules as Data Center Demand Surges (November 16, 2026) (first appearance; the pre-deadline cross-RTO snapshot, consolidated into the single large-load filings item (deduplicated). Distinct from FERC’s June 18 six-RTO show-cause orders themselves, carried in the October 7 coverage)
– Renewable Energy World — Electric Cooperatives Detail First Steps of DERMS Implementation (November 13, 2026) (first appearance; publication date precedes this window. OATI-moderated webinar; no DERMS vendor, platform, MW figure or firm go-live date disclosed)
November 18, 2026 Entry
– Utility Dive — Grid Operators File Closely Watched Large Load Interconnection Reforms at FERC (November 17, 2026) (first appearance; the post-deadline cross-RTO synthesis, consolidated into the single large-load filings item (deduplicated) with the POWER, ISO Newswire and Construction Review Online pieces)
– Utility Dive — Utility, Tech Officials See Opportunities to Increase Grid Utilization (November 13, 2026) (first appearance; publication date falls inside the prior week’s calendar range but the item entered the log only in the November 18 entry and does NOT appear in the November 13 digest’s Sources, which contains no utilitydive.com URL — verified by name and by URL)
– Utility Dive — No Home Left Behind: Scaling VPP Participation While Maintaining Customer Control (op-ed by Gisela Glandt, VP of Virtual Power Plants, Uplight, November 11, 2026) (first appearance; publication date precedes this window. Reported as REINFORCEMENT of the enrollment theme established in the November 13 digest, not as a new finding. Distinct from Uplight’s November 11 summer-portfolio results release and from the November 6 Octopus/Uplight participation piece, both carried in the November 13 digest — same company, three different documents)
– Daily Energy Insider — NYISO Assessment Says New York Has Adequate Capacity to Meet Demand This Winter (November 16, 2026) (first appearance; New-York-only assessment, distinct from NERC’s continent-wide Winter Reliability Assessment below and from PJM’s winter outlook carried in the November 13 digest — three operators, three footprints, reconciled in Benchmark Hygiene)
November 19, 2026 Entry
– America’s Electric Cooperatives (NRECA) — NERC: Winter Reliability Outlook Improves, but Longer-Term Risks Persist (November 18, 2026) (first appearance; NERC’s continent-wide 2026–27 Winter Reliability Assessment. Distinct from the NYISO assessment above and from the November 17 entry’s NERC computational-load items)
– PR Newswire (PSE&G) — PSE&G Demand Response Participation Nearly Doubles as Second Season Concludes (November 17, 2026) (first appearance; utility-self-reported. The November 12 “PowerUp New Jersey” BPU approval reported in the November 13 digest is carried here as context and is NOT re-reported)
– OATI — DTECH Northeast Takeaways: DERMS, Flexibility and AI, by Amy Bartak (November 16, 2026) (first appearance; sponsored/vendor source, used for the Connecticut PURA November 4 decision in Docket 25-01-27, the CAISO AI Genie production deployment and the ISO-NE CELT forecast — all independently verifiable — rather than for platform claims. The PURA decision itself is dated November 4, before this window, and enters the series here through an in-window recap)
– The Energy Mag — Duke Energy Secures North Carolina Approval for Rate Base and Data Center Tariff (November 17, 2026) (movement — the NCUC’s November 16 written orders resolving a six-week tracked thread. Distinct from Duke’s October 6 settlement filing (E-7 Sub 1329 / E-2 Sub 1380) and Walmart’s October 12 reopen-comments motion, both reported in the October 26 digest; this is the order, not the filing)
November 20, 2026 Entry
– POWER Magazine — Five Grid Operators Respond to FERC With Distinct Plans for Large-Load Integration, by Sonal C. Patel (November 19, 2026) (first appearance; the anchor of the consolidated large-load filings item (deduplicated) — the operator-by-operator analysis carrying the rate, trigger and timeline mechanics the Utility Dive, ISO Newswire and Construction Review Online pieces did not. Same underlying filings, new reporting)
– Utility Dive — How National Grid, Eversource Are Using VPPs to Solve Local-Level Constraints, by Brian Martucci (November 18, 2026) (first appearance; Latitude Media / EnergyHub sponsored webinar — utility-executive and vendor claims, no dollar figures, dockets or CS+ pilot MW disclosed)
– U.S. Department of Energy (CESER) — Federal Power Act Section 202(c): Midcontinent Independent System Operator (MISO) Order No. 202-26-51, issued November 13, 2026, effective November 15, 2026 – February 12, 2027 (a new order, the first in the docket since No. 202-26-50 on October 12; publication date precedes this window. J.H. Campbell and the 90-day order chain (202-25-3/-7/-9, 202-26-16/-22/-39) are carried subjects in this series, cited as such — this order is reported as new, the chain is not re-reported)
– ESS News (pv magazine) — Jackery and Every Electric Launch 4 MWh Plug-In Battery VPP for New York Renters, by Tristan Rayner (November 19, 2026) (first appearance; all figures company-reported without independent verification. “Every Electric,” a Brooklyn plug-in-battery aggregator, is a different company from “Everyday Electric,” the aggregator formerly Renew Home, referenced in this series’ October 14–23 coverage)
Method: All five weekday entries — Monday November 16 through Friday November 20 — are present in the research log; no days are missing. The window is strict Monday–Friday and no item from a prior week is reported as new. Items are reported on the day they entered the log with their original publication dates shown; nine of this week’s twenty sources carry pre-window publication dates from inside the daily entries’ trailing-window reach, and each is marked in Sources — the Uplight/Glandt op-ed of November 11, the ESS News IOTech interview of November 12, Utility Dive’s BNEF piece and ISO Newswire’s ISO-NE filing and Renewable Energy World’s co-op webinar and DOE’s Order 202-26-51 all of November 13, CleanTechnica’s sonnen/SOLRITE piece of November 14, the Cornell/Nature Energy study of November 9, and Connecticut PURA’s decision of November 4 reached through an in-window recap. Every one was checked against the November 13 digest’s Sources section BY NAME as well as by URL, because the November 16 entry’s own italic note says the weekend proper yielded little fresh trade news and the entry drew from the Nov. 9–16 window — which is exactly the quiet-window reach-back shape this series has flagged for three runs and which caught Cornell on November 6 and Toronto Hydro on November 13. It fired again this week and the check cleared: ZERO source URLs overlap the November 13 digest’s Sources, and no item is a renamed repeat. Two near-misses worth recording because a URL diff would have passed both: the ESS News IOTech interview of November 12 and Utility Dive’s BNEF grid-utilization piece of November 13 both carry publication dates inside the PRIOR week’s calendar range, yet neither appears in the November 13 digest — they entered the log only in the November 16 and November 18 entries respectively, so they are first appearances here rather than reach-back repeats, and the November 13 digest’s Sources contains no utilitydive.com or ess-news.com URL at all. The third, the Cornell item, is the one that needed a name check rather than a URL check: it is a DIFFERENT Cornell study (Nature Energy, November 9, cost allocation and ratepayer equity) from the Cornell/Applied Energy “trading power reductions” paper (Mahuze et al., September 22) carried in this series’ October 15–16 coverage — different team, journal and framing. Within-log deduplication, which is where this week’s work was: THE FERC LARGE-LOAD SHOW-CAUSE FILINGS APPEAR AS FOUR DISTINCT ARTICLES ACROSS THREE DAILY ENTRIES and are consolidated into ONE item (deduplicated) — POWER’s November 19 operator-by-operator analysis as the anchor because it alone carries the rate, trigger and timeline mechanics, with Utility Dive’s November 17 post-deadline synthesis, ISO Newswire’s November 13 ISO-NE filing notice and Construction Review Online’s November 16 pre-deadline snapshot folded in; all four report the same underlying November 13 and November 16 filings and the daily log itself flagged the distinction between them, which is the within-log note the consolidation acts on. PJM’s November 16 ride-through filing is kept SEPARATE from that item on purpose: it is a reliability-track filing in a different proceeding, and the digest’s analytic point is precisely that the two PJM filings price and mandate the same capability in parallel. Cross-week handling: the NCUC Duke orders are reported as the DECISION on a thread tracked since the October 6 settlement filing and Walmart’s October 12 motion (October 26 digest), never as a new tariff; DOE Order 202-26-51 is reported as a NEW order while J.H. Campbell and the 90-day order chain are carried subjects cited in one line; the Uplight op-ed is reported as REINFORCEMENT of the enrollment theme the November 13 digest established from the New Jersey BPU’s 82% finding, Brattle and the CEC, not as a new finding — a vendor arriving fifth at a conclusion three institutions reached first; PSE&G’s season results are reported as the evidentiary half of the PowerUp New Jersey order carried last week, with the order itself cited as context; and the NYISO and NERC winter assessments are kept distinct from each other and from PJM’s winter outlook, with all three reconciled in Benchmark Hygiene rather than mixed. Sunrun’s eighteen consecutive days of Puerto Rico dispatch, Octopus’s 50–60% participation rate, Oracle’s ~1 GW, Uplight’s 735 MW peak and New Jersey’s 547.1 MW and 82% finding are all cited as carried benchmarks from the November 13 digest, in one line each, and are not re-reported. The DERMS and ADMS searches in all five daily entries again returned mostly evergreen explainers (Uplight “What Is DERMS,” Oracle, GE Vernova “ADMS vs. DERMS”), DistribuTECH 2026/2027 session listings and 2025–2036 market-sizing reports — the November 5 GlobeNewswire “DERMS Market to 2036” release and the Krishnan & Associates go-to-market guide excluded as marketing — making this the EIGHTH consecutive week with no standalone DERMS vendor or deployment item. The DERMS section is therefore built, as this series has learned to build it, on a state regulatory mandate (Connecticut PURA Docket 25-01-27, the first instrument here that converts an interconnection mandate directly into a DERMS or ADMS procurement with dates), a cooperative adoption sequence, and a device-integration architecture — and the FlexEnergi “Ora Edge” edge-DERMS product and the NCEMC–OATI real-time DERMS rollout were both checked and rejected as already-logged or months old (September 8 and February 18, 2026). Several other fresh-looking items were rejected inside the daily entries and are recorded here so they are not re-chased: a POWER “Winter Reliability Concerns Spur Fresh DOE Intervention” piece that proved to be a November 2025 article on long-expired orders; an MGRID “long-term data-center contracts” item that is an aggregator rewrite of the same FERC large-load filings consolidated above; FranklinWH’s RE+ “VPP Services” release as thin product marketing; the New Jersey “PowerUp” approval as already logged in the November 13 entry; and the Austin Energy Power Partner Battery pilot, the Google–Voltus >1 GW PJM VPP and Dominion/Generac “Concerto” as already logged or years old. Targeted search was used only to verify or fill clear gaps, EPRI prioritized, and it resolved less than last week. The EPRI direct check found NO publication dated inside the window — the NINETEENTH consecutive week — so the EPRI Spotlight again applies the carried FlexMosaic five-class framework, this time as a classification scheme for the five RTO non-firm tiers, which is the most concrete application the framework has had in this series: every operator published the four class-determining parameters (notification time, response duration and frequency, depth and speed of response) and none named a class, and MISO’s up-to-438-hour price-triggered obligation and ISO-NE’s ~88-hour $250/MWh trigger are plainly different products sold as one undifferentiated category each. What targeted search did NOT resolve this run, carried forward: the CPUC’s November 19 voting-meeting outcomes in R.22-11-013 (Avoided Cost Calculator guiding principles plus the Distributional Equity Analysis) and Resolution E-5455 — the meeting falls one day before this window closes and no decision document or secondary report was locatable at run time, so the item is held rather than reported, and the primary documents should be pulled next run; SPP’s large-load filing, due around November 20, was not confirmed as filed; the Pennsylvania PUC’s November 17 technical conference on large-computational-load cost allocation is confirmed as scheduled on the PUC’s own page but no output or transcript was found; the Pennsylvania Bulletin publication date that starts the comment clock in Docket M-2026-3064961 remains unverified after six runs and has been escalated — stop spending searches on it; Dominion’s November 15 VPP tariff filing and CAISO’s November 16 EL26-71 compliance filing were not locatable through search and should be pulled from the SCC and FERC dockets directly; the ER26-3380 amended tariff sheets were again NOT read directly, so whether PJM’s prior-delivery-year DR exclusion and three-year registration-history minimum survive into the amended tariff remains open and should be read from pjm.com rather than through coverage; NERC Project 2024-02’s further comment period and final ballot remain TBD and Board adoption is unconfirmed; and PJM’s deployed demand-response megawatts for the 2026 shoulder-season events remain unpublished, which is still the inversion worth naming — vendors and now an orchestration platform publish depth figures while the operator does not.
