All five weekday entries (Monday September 28 through Friday October 2) are present in the research log. Six items are deduplicated across weeks rather than within this one and are reported as standing benchmarks or as movement rather than as new findings: NERC’s 2026 Summer Reliability Assessment — the ERCOT net-internal-demand reduction and the regional demand-response availability map — appears for the fourth time in this series from the same May 27 source, having been reported in the May 29, June 26 and August 21 digests, and is carried in one paragraph; Xcel Minnesota’s Capacity*Connect April 2 approval order has been covered in this series since May and was last treated at length in the July 17 digest, and is carried as a standing benchmark; GE Vernova’s GridOS for Distribution was reported in the September 11 digest from the February launch and is carried only for the EnergyHub grid-edge pairing that this week’s entry adds; the New Jersey BPU virtual-power-plant proposal appears for the fifth time and is carried in one line; New York’s Dynamic Load Management order, reported August 28 and carried September 18, is carried again in one line; and Virginia’s 450-MW Dominion VPP pilot appears as context inside this week’s inclusive-investment item rather than as an item. Two items are movement on tracked watch items rather than repeats: PJM’s reliability backstop procurement, reported in the September 18 and September 25 digests at the proposal and scheduled-open stages, was acted on by FERC this week; and the Pennsylvania PUC’s emergency load-control docket, reported September 25 at the September 11 vote with an October 1 tentative order flagged as imminent, produced that order inside this week’s window. Several items carry pre-week publication dates — the Georgia Power DER Customer Program tariffs, the Brattle/Uplight report of June 3, the NERC assessment of May 27, the New York order of April, the Xcel order of April 2, the GridOS launch, the New Jersey proposal of July 27, the Rocky Mountain Power filing of June 26, the FlexEnergi announcement of September 8, the CLEAResult acquisition of August 25, the Texas PUC ruling of July 23, the microgrid feature of September 23, the ratemaking feature of September 24, the Solar for All ruling of September 25 and the inclusive-investment and data-center-obstacles pieces of September 29 — and they are reported on the day they entered the log with original dates shown. Four verification items were chased by targeted search this week and three of them landed, including one that materially corrects the log’s own characterization; they are noted where they appear. Mechanical check: zero source URLs overlap last week’s Sources section.
Last week the denominator got audited. This week the queue got written.
For five months this series has argued the demand side on price — a megawatt of verified flexibility against a megawatt of new gas, EE at roughly $20.70/MWh against $45–108/MWh for a new combined cycle, a 70-MW portfolio deferring $150M of CapEx. That argument still holds. But three separate venues this week did something different and more consequential: they assigned rank. Not what flexibility costs, but where a resource — or a load — stands in line when the system is short.
Start with the backstop, because the supply side just lost five months. On September 29, in Docket ER26-3380, FERC acted on PJM’s reliability backstop procurement — the mechanism meant to secure roughly 6,831 MW of new capacity for the 2028/29 delivery year at a weighted-average offer cap of $555/MW-day. The Commission accepted the filing and then suspended it for five months, to February 28, 2027, subject to refund and further proceedings, finding that three components may be unjust and unreasonable: the cost-allocation methodology, the rules governing transmission-owner exit from PJM, and the collateral requirements imposed on load-serving entities. It accepted the collateral rules for power suppliers while questioning those for LSEs, and it struck outright PJM’s provision letting qualifying cooperatives and municipal utilities opt out, on discrimination grounds. PJM pulled the September 30 auction open. Chairman Laura Swett said the Commission had no time “to rehabilitate the mess we received” and would not be “forced into accepting a deeply flawed, eleventh-hour procurement mechanism with billion-dollar implications for consumers.” The collateral problem had a face: Northern Virginia Electric Cooperative indicated it would have had to post roughly $2 billion.
A note on accuracy before going further. The research log records this as FERC having “partially rejected” the proposal. The verification search this week establishes the more precise and more useful characterization: the filing was accepted and suspended, with a new effective date of February 28, 2027 and three elements set for further proceedings, plus one provision struck. The distinction matters in a filing. “Rejected” implies PJM must start over; “accepted and suspended subject to refund” means the mechanism survives with a five-month hole in the calendar and a live record on cost allocation and collateral. The second is both true and a better argument.
Then the curtailment order, which descends from the same number. On October 1 the Pennsylvania PUC voted 4-0 to approve a tentative order in Docket M-2026-3064961 directing every electric distribution company in the state to establish an order of operations for emergency load control in which “certain large new loads that have not secured sufficient new electric capacity” are curtailed before existing residential and commercial customers — and in which data centers are barred from being classified as exempt “critical load.” EDCs must build and maintain large-load registries visible to the utility, PJM and the Commission, establish notification procedures for curtailment obligations, and update their emergency plans to shed these loads first. Comments run 30 days from Pennsylvania Bulletin publication with 15 days for replies; the final order is targeted for the January 28, 2027 public meeting; and a technical conference on large-computational-load cost allocation is set for November 17, 2026. The order cites the same 6,831-MW 2028/29 shortfall the suspended backstop was built to close, alongside the Commission-sponsored study warning that modeled reliability risk could reach roughly six times PJM’s planning standard by 2030.
And then the planning mandate, which is the one procurement teams will feel first. California’s regulatory week, captured in a September 30 Stoel Rives digest, produced two decisions that change planning inputs rather than headlines. In R.25-09-004 the CPUC authorized 2028–2029 demand-response bridge funding for the major IOUs, moved the application deadline to November 1, 2028, and adjusted Emergency Load Reduction Program pilot parameters to keep the pilot available through the bridge year. In the DER-orchestration proceeding R.21-06-017 it directed utilities to retain consultants to vet proposals for a common DER-orchestration framework and — the operative item — mandated a new “Load Flexibility Scenario” as a fourth required scenario in the Distribution Planning and Execution Process, beginning with the 2027–2028 cycle. A companion CEC interoperability workshop on November 4 advances OpenADR 3.1, IEEE 2030.5, OCPI and OCPP.
Read those three together. PJM decided who can bid and when. Pennsylvania decided who gets shed first. California decided which scenario the planners are required to model. None of the three is a price.
Meanwhile the supply side’s own schedule deteriorated independently of regulation. A September 29 Utility Dive deep dive put Goldman Sachs’ estimate at only 50–60% of planned data-center capacity coming online within two years, against forecasts of 17–20% of U.S. electricity by 2030–2035: standard power transformers at 2–3-year lead times, gas-turbine backlogs exceeding 100 GW pushing delivery into the early 2030s, and a shortage of roughly 500,000 construction workers. Public opposition hardened — a May 2026 poll found 71% of Americans would oppose a nearby data center, up 30 points year over year, with at least 20 projects killed in Q1 2026. And the counter-move inside that piece is the one to note: Portland General Electric’s GridCARE deployment freed roughly 80 MW of interconnection headroom — about one-fifth of its expected 2029 additions — by squeezing the existing system rather than building new.
The week’s argument, stated plainly: the demand side stopped being priced as an alternative to supply and started being positioned as a rank in a queue. A multi-billion-dollar supply-side backstop lost five months and its cost-allocation and collateral design simultaneously; a state commission put unsecured large load at the front of the shed order and took “critical load” off the table for data centers; a second commission made load flexibility a required planning scenario; and the physical build-out is landing at roughly half of what was announced. Firm, M&V-verified demand-side capacity is now the instrument that buys position — and position, unlike price, cannot be matched by a cheaper turbine.
⚡ Virtual Power Plants & Demand Flexibility
The enrollment constraint finally got a financing answer with published cost math, and Illinois supplied the numbers. Clean Energy Works’ Zuha Maryam Shaikh, writing September 29, frames the problem this series has treated as the ceiling on every VPP capacity figure: avoided-capacity value only becomes megawatts if enough customers own controllable devices. The baseline is stark — roughly 28% of U.S. homes now have a smart thermostat while utility DR participation runs far below that, and home energy upgrades are affordable for fewer than one in ten households under current market conditions. The inclusive utility investment model is the proposed fix: the utility pays for the DER at the customer’s premises and recovers the cost through a fixed monthly charge set below estimated bill savings, with no upfront cost, no credit check and no homeownership requirement — which the author argues could lift affordability above nine in ten households.
Illinois’s Equitable Energy Upgrade Program, whose final guidelines issued September 10, is the first version with a published cost stack: a 7-kW solar plus 13.5-kWh storage system costs $34,530 upfront, the financing gap after utility rebates is $11,913, cost recovery is capped at 80% of estimated savings, and the system could earn roughly $770 per year from PJM capacity, energy-arbitrage and ancillary-service markets once enrolled in a VPP.
For IRP work the usable move is narrow and specific: model the financing mechanism as a lever on the enrollment curve rather than holding enrollment as a fixed assumption. A VPP forecast that assumes a participation rate and then prices incentives against it has the causality backwards in a market where fewer than one household in ten can finance the device in the first place. Two cautions. The $770/year figure is a market-revenue estimate for a specific system in a specific RTO and should not be generalized; and the 80%-of-savings recovery cap is the design parameter that makes or breaks the model — set it at 100% and the customer has no reason to participate, set it too low and the utility cannot recover. (Source: Utility Dive, September 29 — September 30 entry)
Virginia is the demand-side counterpart in the same piece, and it is carried here rather than re-reported. The commonwealth hosts nearly 600 data centers, the Dominion zone faces PJM’s largest projected absolute summer-peak increase through 2030, and legislative auditors warn of a possible $444-per-year residential bill increase by 2040. This series has reported the 450-MW Dominion VPP pilot, its dual-enrollment design, the November 15 tariff filing and the December 1 pilot filing repeatedly since September; the one element new to the series here is HB 1062’s inclusive-investment feasibility study, due November 1, 2026 — which lands two weeks before the VPP tariff and should be read beside it, because a commonwealth that studies utility-financed DER while standing up a 450-MW aggregation is deciding both halves of the same enrollment question in the same quarter. Separately, the piece notes 34 states now have programs requiring utilities to deploy DERs and at least a dozen have passed or are considering VPP legislation.
Carried — Minnesota. Xcel’s utility-owned Capacity*Connect program, approved by the Minnesota PUC on April 2, 2026 — up to 200 MW of distributed storage in 1–3 MW increments across the distribution grid by 2028, a $430M full-capacity budget, deployment services from Sparkfund, with a Commission-required interim assessment at 50 MW and a comprehensive independent evaluation — has been covered in this series since May and analyzed at length in the July 17 digest, including the sole-source procurement objection, the interconnection-queue asymmetry and the November 2027 evaluation date. It is carried here as the standing benchmark for the utility-ownership pole of the VPP ownership question: rate-based, utility-dispatched batteries are firm system capacity in the same tier as owned generation, and the model’s durability turns on whether the 50-MW checkpoint shows it beats a contracted-aggregation alternative on cost. Nothing in this week’s log advances it.
Carried — New Jersey. The BPU’s statewide Virtual Power Plant Program proposal of July 27 — home batteries, smart thermostats and EV chargers aggregated to shave peak and defer capacity, projected to save ratepayers millions annually — appears for the fifth time in this series and remains a proposal in stakeholder process. One line of standing relevance: New Jersey is still the undecided entry in the four-way enrollment-channel comparison this series tracks (Virginia dual-enrollment, Minnesota utility-only, Nevada utility-only with deferral, New Jersey undecided), and its eventual dispatch and enrollment terms are what will set its firm capacity contribution.
Carried — New York. The PSC’s Dynamic Load Management enhancements across CSRP, DLRP, DLC, Term-DLM and Auto-DLM were reported in the August 28 digest for the parameters that matter — the 50 kW participant minimum and the three-year minimum program term that makes storage economic to enroll — and carried September 18 for the Con Edison and Central Hudson Bring-Your-Own-Battery rollout. Nothing new this week; the item re-entered the log from the original April order.
🔌 DERMS & Grid Integration Technology
Georgia Power has put a DERMS into production and wired it directly to two dispatchable demand-response tariffs — and this is the cleanest integrated-architecture example the series has recorded. Under a Georgia PSC-approved stipulation stemming from its most recent IRP, Georgia Power is deploying a DERMS providing visibility, forecasting, control and optimization of behind-the-meter DER in coordination with real-time grid operations, with spend projected across 2025–2028. It did not procure that platform as a standalone IT purchase. It paired it with a DER Customer Program built on two tariffs:
- Resiliency Asset Service (RAS) — the utility designs, procures, installs, owns, operates and maintains a dispatchable resource behind a commercial or industrial customer’s meter to meet that customer’s resiliency need.
- Demand Response Credit (DRC) — the customer takes a bill credit in exchange for Georgia Power’s right to dispatch that resource during system reliability events.
The design solves the DR problem this series keeps returning to. Because the utility owns and operates the asset and dispatches it through its own DERMS, the megawatts are firm and dispatchable rather than notified-and-hoped-for — which is the profile that lets DR be scored against avoided generation and T&D CapEx at something approaching the Brattle ~$66/kW-year capacity value rather than at a discounted, enrollment-contingent figure.
Three things to take from it. First, the structural lesson: the DERMS is not an IT system, it is the control layer that makes a fleet of customer resiliency assets countable as system capacity — and a utility that procures the platform without the tariff that creates the dispatch right has bought the instrument without the asset. Second, RAS/DRC is a replicable tariff pair, and it is the one to copy in a vertically integrated jurisdiction with no wholesale market to monetize against: it converts a customer’s resilience spend into a stacked-benefit demand-side resource on the utility’s books. Third, the caveat is in the ordering: the resiliency framing puts customer-outage value first and grid dispatch second, so the realized system-capacity contribution depends on enrollment depth and on how often the utility actually calls the assets — which means the M&V and call-frequency record is what a commission will eventually test. (Source: Georgia Power DER Customer Program tariff documentation — September 28 entry; first appearance in this series)
The best DERMS number of the week is that a quarter of the available demand-side capacity comes from orchestration alone, with no new enrollment. Brattle, in work prepared for Uplight and reported June 3, modeled a representative Southwest Power Pool utility and found that stacking demand response, energy efficiency and time-of-use rates as a coordinated portfolio rather than program-by-program could unlock roughly 90 MW of incremental demand-side capacity by 2030 — lifting demand-side capability from about 3% of system load in a status-quo case to roughly 5%. The decomposition is the finding: about 66 MW comes from the stacked programs themselves, and another 23 MW comes purely from improved forecasting and staggered asset dispatch.
That 23 MW is the operational bridge between this series’ two threads. It is capacity obtained from the control-and-forecasting layer — exactly what Georgia Power is buying — rather than from signing another customer. In a business case, it is also the most defensible line item available, because it does not depend on an enrollment forecast. A 90-MW demand-side gain sits squarely in the range where this series’ illustrative economics apply: a ~70-MW DR portfolio deferring ~$150M in CapEx at roughly $20M in program cost for ~$130M net benefit. The report’s transferable recommendation is portfolio-level management with clear system-level performance targets and explicit modeling of interactive effects across programs — and the caveat is that these are modeled results for a representative utility, so the method transfers and the megawatt figure does not. (Source: Utility Dive, June 3 — September 28 entry; first appearance in this series)
A vendor announcement worth reading for its commercial model rather than its technology. FlexEnergi announced field deployment of Ora Edge, a DERMS that manages flexible load at the feeder and substation level — locational, continuous optimization rather than system-wide event calls — across residential, small-and-medium business, agricultural, large C&I and data-center customers, the last through Bring Your Own Capacity arrangements. It both sheds and adds load, which matters in a high-renewable territory. Dispatch triggers include wholesale price signals, weather events, and direct signals from utility SCADA/ADMS responding to grid constraints — that last one being precisely the ADMS-to-edge-DERMS interface this series’ RFP work treats as a core integration requirement.
The commercial structure is the real item: FlexEnergi sells capacity-as-a-service, with the utility procuring flexible capacity in defined MW blocks and paying on delivered performance. That shifts performance risk to the vendor and makes the purchase look more like a capacity PPA than a software license — a different contract, a different risk allocation, and a different line in the capital plan. For RFP evaluation the implication is direct: compare performance-based capacity contracts against platform-license DERMS on the same avoided-cost basis, because the two are usually scored on incompatible criteria and the comparison is where the money is. The announcement names no utility partners, no MW quantities and no device counts, so it is a design reference and a commercial-model data point, not a proven deployment. (Source: T&D World, September 8 — September 30 entry)
Vendor consolidation continued, and this one crosses a category line. CLEAResult — which describes itself as North America’s largest provider of energy-efficiency, energy-transition and sustainability services — acquired Connected Energy on August 25, terms undisclosed. Connected Energy, founded in 2004, brings a DERMS platform for utility demand-management programs (event orchestration, optimization, data visualization, performance reporting), hardware-based direct load control for C&I customers, and a specialized agricultural demand-response capability. CLEAResult’s stated rationale is that data-center load growth and faster DER adoption are making utility demand-management needs “more acute.”
The agricultural piece is the underrated asset: irrigation load control is a large, predictable summer-peak resource in western and plains territories and has appeared in recently approved program portfolios. The structural signal is that EE program implementation and DR/DERMS dispatch are increasingly sold as one contract — which genuinely simplifies procurement and M&V, and genuinely raises a lock-in question. The RFP discipline that follows is the same one this series has specified against suite consolidation all year: require data portability, platform independence and model-export terms as commercial conditions, so that DR capacity is not tied to a single implementer relationship that a later competitive solicitation cannot unwind. No MW-under-management or customer counts were disclosed. (Source: CLEAResult, August 25 — September 30 entry)
Carried — GE Vernova. GridOS for Distribution, unifying ADMS, DERMS and grid-planning analytics on a governed, federated grid data fabric, was reported at length in the September 11 digest from the February launch, including the vendor roster, the subscription-licensing shift and the single-vendor lock-in judgment it forces. The one element this week’s entry adds is the EnergyHub partnership, which connects GridOS DERMS to grid-edge control over batteries, solar, EVs and connected devices for single-pane-of-glass DR dispatch alongside every other operator resource, with GridOS DERMS citing 90-plus customers and 127 million-plus managed service points as installed scale. Read beside the CLEAResult transaction above, the two describe the same market motion from opposite ends: the platform vendor acquiring the customer-engagement layer, and the program implementer acquiring the platform. The procurement consequence is identical in both directions, and it is the reason data governance, open APIs and model and data exit terms belong in the contract rather than in the requirements matrix.
🔋 Energy Storage
Section 48E is quietly doing more for the customer-sited DER base than most state incentive programs, and the window is finite. A September 30 Utility Dive piece documents building operators using Section 48E investment tax credits — which can return 30–50% of project cost — to make battery storage and solar-plus-storage pencil out, with Alliant’s Matthew Noll reporting “a lot of battery energy storage systems” being deployed to store cheaper off-peak power for use during expensive peak periods. The economics are concrete rather than illustrative: a Colorado rural hospital’s $65M expansion, in which heating and cooling account for roughly 20% — about $13M — of cost, could capture $4–8M in credits. Private firms monetize credits by selling them at 85–93 cents on the dollar; nonprofits and public facilities take direct Treasury payments.
The demand-side detail is the one to carry: small manufacturers are adopting storage specifically to shift high-energy operations — lathes, furnaces, kilns — to off-peak hours. That is dispatchable, price-responsive industrial flexibility arriving on the system as a byproduct of a federal tax credit, in a customer class that is far easier to enroll and meter than residential aggregations. For the avoided-cost business case, 48E lowers the effective cost of the customer-sited assets that populate the value stack — which means a DR or DERMS program designed today is aggregating a device base that is cheaper than the program’s own incentive analysis probably assumes.
The caveat is a deadline. The One Big Beautiful Bill Act eliminated standalone solar credits while preserving solar-plus-storage and geothermal, with a phase-out beginning in 2032. A storage-dependent enrollment forecast that extends past the phase-out is assuming economics that are legislated to disappear, and a program whose participation ramp depends on 48E-financed installs should state that dependency explicitly rather than carry it as a trend line. (Source: Utility Dive, September 30 — October 2 entry)
And the cost-effectiveness anchor this week came from a real filing rather than a representative model. Rocky Mountain Power’s Semi-Annual DSM Forecast Report to the Utah PSC, filed June 26, projects a total DSM budget rising to roughly $110.1M in 2026 from about $98.9M in 2025, with the Wattsmart Homes program forecast at ~69,696 MWh of savings at a ~$25M budget — revised down from ~84.9 million kWh after commercial midstream savings were found to be double-counted. The headline figure is a forecast Utility Cost Test of 1.30–1.45 for the whole Wattsmart Homes program: every dollar spent avoids $1.30–$1.45 of future supply-side cost.
Two reasons this is more useful than a consultant’s benchmark. First, it is a jurisdiction-specific, utility-filed benefit-cost ratio in the range that supports this series’ framework — energy efficiency at roughly $20.70/MWh against $45–108/MWh for new gas combined cycle — and it is citable as a regulated utility’s own forecast rather than as advocacy. Second, RMP explicitly positions Wattsmart Batteries as a planning-relevant DSM resource delivering peak shaving, load shifting for renewable integration and fast-response capacity through coordinated battery dispatch. That is distributed storage counted as a DSM resource inside a DSM filing, which is the categorization fight this series has tracked in half a dozen other jurisdictions, settled here by practice rather than by order.
The discipline the filing imposes on itself is the lesson: the double-count correction is the reason the defensible IRP input is the verified ex-post benefit-cost ratio, not the filed forecast. A UCT above 1.0 is a projection subject to realized participation, and this filing demonstrates that these projections get revised downward by material amounts for mundane accounting reasons. Cite the 1.30–1.45 range with the forecast caveat attached, and ask for the ex-post number. (Source: Rocky Mountain Power / Utah PSC, June 26 — September 29 entry)
Two financing developments restored and expanded the low-income DER base this week, and both are legally contingent. On September 25, U.S. District Judge Tanya Chutkan ruled that the EPA’s August 2025 termination of the $7 billion Solar for All program was “arbitrary and capricious, contrary to law, and in excess of statutory authority” — and applied the ruling program-wide rather than only to the plaintiff. She is the second federal judge to overturn the cancellation within a single week. The local amounts are concrete: the Texas Solar for All Coalition was awarded nearly $250M, with Harris County alone receiving over $54M in 2024 funding, directed at low-income distributed solar, storage and solar-powered community-resilience hubs.
Read beside the Illinois EEUP math above and the 48E economics, this completes a three-part financing stack for the enrollment problem: a federal grant channel for households that cannot self-finance, a federal tax-credit channel for commercial and institutional sites, and a utility on-bill channel for everyone in between. A VPP enrollment forecast that treats device adoption as a function of incentive level alone is ignoring all three. The caveat on Solar for All specifically is that this is trial-court litigation subject to appeal, so the funding — and the DER it would deploy — remains legally contingent, and should enter a planning scenario as a sensitivity rather than as a baseline. (Source: Utility Dive, September 25 — October 1 entry)
🏗️ Data Centers & Large Load Growth
The build-out is landing at roughly half of what was announced, and the reasons are physical rather than regulatory. The September 29 Utility Dive deep dive is the most useful single assembly of deployment-risk evidence this series has acquired, because every constraint it names is independent of the policy fights the series otherwise tracks. Goldman Sachs expects only 50–60% of planned data-center capacity to come online within two years, against forecasts that data centers reach 17–20% of U.S. electricity by 2030–2035. The binding constraints:
- Standard power transformers at 2–3-year lead times.
- Gas-turbine backlogs exceeding 100 GW, pushing new delivery into the early 2030s.
- A shortage of roughly 500,000 construction workers.
- 71% of Americans would oppose a data center built near them — a May 2026 poll result, up 30 percentage points year over year — with at least 20 projects killed by local opposition in Q1 2026, and marquee projects from Microsoft, Google and Oracle hitting permitting, legal and pipeline roadblocks.
The developer response splits two ways, and only one of them is good for a utility. The first is bring-your-own-capacity: OpenAI’s 8 GW in Ohio, Amazon’s 7.7-GW West Texas gas proposal. The second is flexibility and headroom, and the example is the number to carry: Portland General Electric’s GridCARE deployment freed roughly 80 MW of interconnection headroom — about one-fifth of PGE’s expected 2029 additions — by extracting more capacity from existing infrastructure rather than building new.
For DSM and DERMS planning, this is the empirical floor under the speed-to-power argument. Every delayed transformer and every stalled interconnection extends the period during which flexible load and DER headroom are the only resources available on the relevant timeline — not the cheapest, the only. But the caveat is sharp and it runs against the utility: bring-your-own-capacity routes around the utility entirely. The demand-side value accrues to a utility only where flexibility is captured inside the regulated planning and interconnection process, which is the same conclusion the California AB 2383 analysis reached two weeks ago and the same arithmetic as MISO’s ZGIA fast track. A utility that does not offer a flexible-interconnection or headroom product is not holding the line against DER — it is ceding the customer to self-supply. (Source: Utility Dive, September 29 — October 2 entry)
Texas has now ruled on whether mandated large-load curtailment is a compensable grid service, and the answer is no. This is the first test case under Senate Bill 6, it entered the log this week from a July ruling, and it closes a question this series’ large-load thread has carried open since June. In Docket 59220, decided July 23, the Texas PUC approved net metering between Ensign Infrastructure’s Goodnight campus in Armstrong County — Crusoe Energy Systems as data-center operator, Google as technology partner — and the 265.5-MW Goodnight 1 wind farm. The condition: the entire 525.5-MW campus (including the 260-MW Crusoe Load 2) must curtail within 30 minutes during ERCOT emergencies.
Two findings matter. First, the curtailment obligation is not capped at the paired generator’s output — ERCOT can compel curtailment of roughly twice the wind farm’s nameplate capacity, a 2:1 load-to-generation ratio. Second, and this is the one to write down: the Commission ruled that mandatory emergency curtailment is a reliability obligation, not a compensated service, so the load cannot collect demand-response payments for it.
For project finance the reported consequence is that models assuming curtailment proportional to generator capacity are wrong, which pushes developers toward full on-site backup generation or storage to ride through extended emergencies — more behind-the-meter assets, procured for a reason that has nothing to do with any utility program. For IRP accounting the consequence is a clean rule: count mandated large-load curtailment as emergency risk reduction, not as DR capacity with an avoided-cost value. The two are not interchangeable and a filing that books the former as the latter is inflating its demand-side portfolio with megawatts it does not control and cannot settle. The next test is Docket 59399 — Amazon/Vistra at Comanche Peak — which will show whether firm nuclear co-location draws different curtailment treatment than intermittent wind. (Source: Tech Times, July 30, on the July 23 ruling — September 30 entry; first appearance in this series)
Carried — NERC’s 2026 Summer Reliability Assessment, fourth appearance. The regional demand-response availability map (SERC Central +172.3%, ERCOT +54.9%, SPP +25.8%, New England −13.3% versus 2025) and the 3.7-GW / 4.6% reduction in NERC’s ERCOT net internal demand forecast “because more data centers can be curtailed by grid operators when needed to prevent grid emergencies” were reported in the May 29, June 26 and August 21 digests, all from this same May 27 source. It is carried here in one paragraph and flagged because it illustrates the dedupe failure mode this series guards against: a four-month-old reliability assessment re-entering a daily log as a fresh item, and the ERCOT figure is attractive enough — it is still the single most useful sentence available for a large-load flexibility filing, because it comes from the entity whose institutional job is conservatism — that it will keep re-entering. Cite it as the May assessment, not as a current finding, and keep the standing qualifier: the counted capacity is only as firm as the curtailment terms behind it, which is exactly what the Texas Docket 59220 ruling above has now begun to define.
The PJM governance question this series has carried as its first open verification item landed, and it landed unresolved. The tracker’s priority chase for this run was whether PJM adopted governance reforms by FERC’s end-of-September deadline or whether FERC would impose them. The answer, from two sources: PJM members voted on a governance term sheet on September 30, and the result has not yet been filed with FERC and awaits Commission approval. What the members agreed to: state executive branches would participate in the Organization of PJM States for the first time (OPSI today comprises only utility regulators), and one additional state seat goes on the board nominating committee — against a competing staff proposal that would have added two state seats plus an independent-appointee option. Critically, PJM’s members committee retains final decision-making authority over Section 205 filings, and OPSI proposals require two-thirds member approval to advance.
That last provision is why the states are not satisfied. Maryland Governor Wes Moore said FERC “gave clear direction to PJM and its members to give states a real voice, but PJM’s members responded with a term sheet that stifles the states by handing those same members a veto over any proposal the states bring forward,” and State Senator Katie Fry Hester called the plan one that leaves states “spectators.” For demand-side work the relevance is procedural but real: the stakeholder body that would have to approve any PJM-wide demand-response or flexibility market reform just confirmed it keeps the 205 pen, and a two-thirds threshold for state-originated proposals sets the realistic bar for anything a state commission wants to advance through PJM rather than through its own jurisdiction. Read beside the backstop suspension, the same week produced a Commission unwilling to accept PJM’s rushed filing and a membership unwilling to cede filing authority. The verification item moves from “unresolved” to “resolved and contested” — track the FERC filing, which is where it actually gets decided. (Verification searches this week: Maryland Matters, October 1; RTO Insider, PJM stakeholder term sheet)
📋 Regulatory & Policy
Pennsylvania wrote the curtailment hierarchy, and it is the most directly actionable state order of the quarter. The full content of the October 1 tentative order in Docket M-2026-3064961, approved 4-0, is covered in this digest’s opening. The four directives to every EDC in the commonwealth, restated as a compliance list:
- Identify and maintain registries of large loads — with visibility to the utility, PJM and the Commission.
- Establish procedures for notifying large-load customers of potential curtailment obligations.
- Prioritize reducing eligible new large loads that lack sufficient secured capacity before shedding other customers.
- Update emergency procedures to prevent data centers from being classified as critical loads exempt from curtailment.
The procedural calendar: 30 days for comments from Pennsylvania Bulletin publication, 15 days for replies, final order targeted January 28, 2027, and a separate technical conference on large-computational-load cost allocation on November 17, 2026. The order’s factual basis is PJM’s 6,831-MW 2028/29 shortfall and the Commission-sponsored finding that modeled reliability risk could reach roughly six times PJM’s planning standard by 2030 — and, from the August PUC report this series reported in September, 18.56% projected industrial load growth over five years led by PPL, while total state electricity usage declined from 2024 to 2025.
Two readings. The constructive one: curtailment rank is now a siting and investment variable, and a large customer that brings firm, M&V-verified demand-side capacity or on-site dispatchable DER moves itself out of the first-to-be-shed tier. That is a value proposition a utility can sell to a large-load customer directly, priced against the cost of being first in the queue — which is a materially easier sale than a bill-savings pitch. The limiting one: a data center able to island on on-site generation or storage blunts the rule’s reach on exactly the loads it targets. An order that puts unsecured large load first in the shed order gives every such load a reason to procure behind-the-meter capacity, which removes it from the shared forecast and from the utility’s addressable flexibility market in the same motion. That is the same self-supply dynamic the GridCARE and bring-your-own-capacity items above describe, arriving here as a second-order effect of a consumer-protection order. (Sources: RTO Insider, October 1; PA PUC press release, October 1 — October 2 entry, with verification search)
California made load flexibility a required planning scenario, which is the most consequential procurement sentence in this week’s log. The two CPUC actions captured in the September 30 Stoel Rives digest, restated for what each one changes:
R.25-09-004 — demand-response bridge funding. Continued funding authorized for the major IOUs’ 2028–2029 demand-response programs, application deadline moved to November 1, 2028 to align with pending policy updates, and Emergency Load Reduction Program pilot parameters adjusted to keep the pilot available through the bridge year. The value is unglamorous and real: program continuity de-risks the multi-year payback assumptions in a DR avoided-cost model. A DR business case whose payback runs past the program’s authorized horizon is carrying a termination risk it usually does not disclose, and a bridge-year authorization removes it for a defined period.
R.21-06-017 — DER orchestration. Utilities directed to retain consultants to vet proposals for a common DER-orchestration framework, and — the operative item — a new “Load Flexibility Scenario” mandated as a fourth modeling scenario in the Distribution Planning and Execution Process beginning with the 2027–2028 cycle.
That fourth scenario is the structural win this series has been arguing for since May. Flexible load stops being an alternative proposal submitted against the distribution plan and becomes a scenario the distribution plan is required to contain — which puts it inside the capacity-planning math that drives distribution CapEx deferral, rather than outside it as an intervenor’s counterfactual. And a standardized orchestration framework is precisely the normalization layer this series has specified as the central DERMS RFP deliverable; the CEC interoperability workshop on November 4 advancing OpenADR 3.1, IEEE 2030.5, OCPI and OCPP indicates the standards such a platform will be expected to speak.
The caveat is honest: these are directional decisions — bridge funding, consultant vetting, a new modeling scenario — not a procurement award. The value shows up in planning assumptions now and in dispatchable megawatts later. But planning assumptions are what this series has repeatedly identified as the gate, and California has just moved the gate. (Source: Stoel Rives, September 30 — October 2 entry)
Michigan is moving toward performance-based ratemaking, and that changes the one incentive that most determines whether a utility wants demand-side solutions at all. Michigan residential rates rose 10.42% from June 2025 to June 2026 — 20.82¢ to 22.99¢/kWh — more than double the 4.98% national increase over the same period (17.47¢ to 18.34¢/kWh). Michigan PSC Chair Dan Scripps’ first affordability recommendation is multi-year rate plans paired with performance-based ratemaking. Pending Senate Bill 768 would extend the mandatory minimum interval between rate cases from one year to three.
The mechanism is the point. Under traditional cost-of-service ratemaking a utility earns its return by building rate-based generation and wires, which means a DR program that defers CapEx reads on the income statement as foregone profit — the structural reason demand-side programs lose internal arguments they win on levelized cost. Under PBR that rewards peak-load reduction and reliability, the same DR portfolio becomes a way to earn. Commonwealth Edison has already adopted metrics including peak-load reduction and grid-reliability incentives, which makes this a design that exists rather than a theory. And the three-year rate-case interval lengthens the planning horizon over which avoided-cost demand-side investments pay back — a second-order effect that matters as much as the metrics, for the same reason New York’s three-year minimum program term matters to storage.
The caveat is the whole game: PBR’s effect depends entirely on which metrics a commission adopts and how it weights them. Poorly designed metrics reward the same spending they were meant to curb, and a peak-reduction metric without a cost cap can be satisfied by expensive peak reduction. A utility or advocate entering a PBR docket should arrive with the metric definitions drafted rather than react to someone else’s. (Source: Utility Dive, September 24 — October 1 entry)
And the federal litigation picture on distributed-solar funding improved twice in one week. The Solar for All restoration is reported in the Storage section above for its enrollment effect. Its regulatory significance belongs here: two federal judges overturned the same cancellation within one week, the second ruling applying program-wide, which means the $7 billion is currently restored nationwide rather than for a single plaintiff. The enforcement question is appellate, and the planning treatment is a sensitivity rather than a baseline.
🔬 EPRI Research Spotlight
A direct check again confirms no new EPRI publication dated inside this week’s window — the twelfth consecutive dry week — but the drought is beside the point this week, because EPRI showed up in the log through the channel this series predicted it would. The targeted search of EPRI’s media resources surfaced nothing dated September 28 through October 2; the most recent substantive items remain the DCFlex nine-site expansion, the European DCFlex extension, and the Aurora materials carried since July. As established three months ago, this is reported as a finding about publication cadence, not as a coverage omission — and this week supplies the confirming instance: the week’s most useful EPRI contribution arrived inside a trade-press feature, attributed to an EPRI staff lead, not through a research release.
EPRI’s argument this week is that microgrids and the DERs inside them should be dispatched as everyday grid resources rather than held idle for emergencies — and the barrier it identifies is a DERMS integration problem with a specific name. In a September 23 Utility Dive feature, Jackie Baum, EPRI’s senior team lead for DER integration, makes the case directly: “the majority of the time, the power is on,” so utilities should be using these assets for routine peak management and voltage support rather than treating them as emergency-only backup.
The scale tension in the same piece is the clearest statement of the problem this series has seen from a single utility. PNM has roughly 40 MW of balancing-authority flexibility today. Proposed data centers in its territory could reach 300 MW. That is not a gap that can be closed by holding distributed assets in reserve; it is a gap that requires everyday use of distributed flexibility as an operational necessity rather than an option.
And PNM’s Jon Hawkins names the obstacle precisely: standardization. Every microgrid differs in its loads, its generation mix and its controls, which complicates utility integration in a way that standardized solar or standardized storage does not. The resource base is real — Sandia National Laboratories on power-electronics testing, and Kit Carson Electric Cooperative running five microgrid projects including batteries and hydrogen — but heterogeneous controls are exactly what keeps a real resource base from becoming a dispatchable one.
Two requirements follow for a DERMS specification, and both are testable. First, the platform must normalize heterogeneous microgrid controls into a single dispatchable layer — which is the same requirement the CPUC’s new common-orchestration framework is asking consultants to define, arriving from the operational side rather than the regulatory one, and it is worth noting that the two landed in the same week. Second, and this is the sentence to carry into a business case: distributed flexibility earns avoided-cost value only when it is dispatched continuously, not when it is reserved for outages. An asset called twice a year has a resilience value and essentially no capacity value; the same asset in routine peak and voltage service has both. Most utility microgrid portfolios are currently specified for the first case and then evaluated against the second.
The FlexMosaic gap is unchanged at six products, and this week added no new flexibility instrument to price. The DCFlex Initiative’s FlexMosaic framework — flexibility classes keyed to notification time, duration, frequency, depth and speed of response — remains the only published vocabulary precise enough to compare competing flexibility products. The six products this series tracks (Virginia’s 450-MW dual-enrollment VPP; PJM’s proposed IRAS curtail-or-backup; interruptible riders in roughly a quarter of large-load tariffs; the §202(c) 15-minute compulsion; the PJM DR 24/7 availability obligation effective June 1, 2027; and Texas SB 6 mandatory uncapped curtailment) stand without addition. But two of them moved this week in ways that affect the mapping. Pennsylvania’s order of operations is not a seventh product — it is a priority rule applied across the others, which is a different and harder thing to represent in a flexibility taxonomy organized around resource characteristics rather than dispatch sequence. And the Texas Docket 59220 ruling supplies the missing attribute on product six: mandatory curtailment is now established as uncompensated and uncapped by paired generation, which answers FlexMosaic’s depth key in the worst possible way — depth is whatever the operator requires.
The Aurora sequencing problem is unchanged and remains closed as an argument. The 96-MW Manassas, Virginia reference design still targets late 2026; this week’s search surfaced no acceleration and no further slip. NERC’s computational-load standards closed their comment period and ballot September 18, board adoption is scheduled for December, and the FERC filing deadline is December 31 — so the standards get filed before the demonstration that was meant to supply their data is operating, and the only large-scale measured event on the record remains the 3,800-MW July 22 northern Virginia load trip in which the data centers disconnected too early. The coalition’s estimate that the reference design adopted nationwide could unlock roughly 100 GW on the existing system is why the sequencing matters rather than merely annoys.
A note on the NERC Rules of Procedure track, which was this run’s fifth verification priority. The tracker flagged an October 20-day comment period on the registry changes that create the Computational Load Owner and Computational Load Operator entity types — the track that determines whether the December standards bind anyone at all. This week’s search located the August posting of the proposed Rules of Procedure changes and the summary of changes document, but surfaced no confirmation that the October comment period has opened on schedule. Treat the schedule as unconfirmed rather than as slipped, and chase it first next run: if the registry track slips past December, the December 31 FERC filing is a paper milestone and the 50 MW / 100 kV threshold applies to no registered entity.
Standing watch: Aurora commissioning data and any further slip; the CLO-001-1/-002-1/-003-1 ballot results and whether the Rules of Procedure registry track holds its October comment / November ballot / December adoption schedule — unconfirmed as of this week; whether FlexMosaic or any response-class vocabulary appears in the ER26-3515 and RM26-4 compliance records; whether the EPRI/Open Compute Project standardized utility–data-center agreement produces publishable contract language; the EPRI/Dominion V2G demonstration’s 12-volt state-of-charge and feeder-hosting findings; whether EPRI’s DER integration team publishes the microgrid-standardization work that Baum’s comments imply exists; and an EPRI.com direct check again next run — thirteenth week.
🚩 Utility-Sector Relevance Flags
⚑ FERC Suspended PJM’s 6.8-GW Backstop Five Months and Opened Its Cost Allocation and Collateral Design — Not “Rejected,” and the Difference Is the Argument
Topic: Resource-Adequacy Gap / Backstop Procurement / Demand-Side Opening
Relevance: On September 29, in Docket ER26-3380, FERC accepted and then suspended for five months — to February 28, 2027, subject to refund and further proceedings — PJM’s reliability backstop procurement for roughly 6,831 MW of new 2028/29 capacity at a $555/MW-day weighted-average offer cap. Three components were found potentially unjust and unreasonable: cost allocation, transmission-owner exit rules, and collateral requirements on load-serving entities (supplier collateral was accepted). FERC struck the co-op and municipal opt-out on discrimination grounds. PJM pulled the September 30 auction. Northern Virginia Electric Cooperative would have posted roughly $2 billion. Chairman Swett: no time “to rehabilitate the mess we received,” and the Commission will not accept “a deeply flawed, eleventh-hour procurement mechanism with billion-dollar implications for consumers.” Results, previously expected December 2, are now untethered from that date.
Action Signal: Implement — Two actions. First, correct the characterization in any internal or filed material: this series’ own research log recorded a “partial rejection,” and the accurate framing — accepted, suspended to February 28, 2027, three elements in further proceedings, one struck — is both defensible and stronger, because it means the shortfall persists with no procurement mechanism running for five months rather than a proposal headed back to the drawing board. Second, price the gap on the new calendar. A resource that can be enrolled, metered and settled before February 2027 is competing against nothing; after it, against a redesigned mechanism whose collateral terms FERC has already said may be unlawful. The LSE-collateral finding is the one to read closely — a $2-billion posting requirement is what makes a backstop structurally hostile to anyone other than a large IOU, and it is the clearest available contrast with a demand-side portfolio that requires no posting at all.
⚑ Pennsylvania Put Unsecured Large Load First in the Shed Order and Barred Data Centers From “Critical Load” — File in This Docket, and the Window Is Roughly Six Weeks
Topic: Curtailment Priority / Emergency Load Control / DR Dispatchability
Relevance: The PA PUC approved a tentative order 4-0 on October 1 in Docket M-2026-3064961 directing every EDC to establish an order of operations in which new large loads that have not secured sufficient capacity are curtailed before existing residential and commercial customers, and in which data centers cannot be classified as exempt “critical load.” EDCs must build large-load registries visible to the utility, PJM and the Commission, establish curtailment notification procedures, and update emergency plans accordingly. Comments are due 30 days after Pennsylvania Bulletin publication with 15 days for replies; the final order is targeted for January 28, 2027; a technical conference on large-computational-load cost allocation is set for November 17, 2026. Basis: PJM’s 6,831-MW 2028/29 shortfall and a Commission-sponsored finding of modeled reliability risk at ~6× the planning standard by 2030, against 18.56% projected five-year industrial load growth led by PPL while total state usage declined 2024 to 2025.
Action Signal: Engage — This series has carried “check the emergency-action sequence so enrolled DR is not called before unsupported large load” as a standing action item since August. Pennsylvania has now written that sequence the right way round, and the comment window is the opportunity to make it durable and specific. File on two points. First, that “secured sufficient new electric capacity” must be defined by a verified instrument — a posted security amount with a verification date — rather than by a contractual assertion, because the ComEd $1 posting is the precedent for what an unverified commitment is worth. Second, that enrolled, M&V-verified demand-side capacity should be credited as “secured capacity” for purposes of the shed order, which turns the registry into an enrollment incentive rather than merely a compliance list. Then use the order commercially: curtailment rank is now a priceable product for a large-load customer, and the cost of being first in the queue is an easier sale than bill savings. Utilities outside PJM should read the tentative order regardless — a state that rewrites curtailment priority against a quantified RTO shortfall has produced the template, and the November 17 cost-allocation conference is where the companion half gets decided.
⚑ California Made Load Flexibility a Required Distribution-Planning Scenario From 2027–28 — This Is the Gate This Series Has Been Arguing to Move
Topic: Distribution Planning / DER Orchestration Framework / DERMS RFP Requirements
Relevance: The CPUC in R.21-06-017 directed utilities to retain consultants to vet proposals for a common DER-orchestration framework and mandated a new “Load Flexibility Scenario” as a fourth required scenario in the Distribution Planning and Execution Process beginning with the 2027–2028 cycle. Separately, in R.25-09-004, it authorized 2028–2029 DR bridge funding for the major IOUs, moved the application deadline to November 1, 2028, and adjusted ELRP pilot parameters to keep the pilot available through the bridge year. A CEC interoperability workshop on November 4 advances OpenADR 3.1, IEEE 2030.5, OCPI and OCPP.
Action Signal: Implement — The scenario mandate is the structural change: flexible load moves from an intervenor’s counterfactual submitted against the distribution plan to a scenario the plan is required to contain, which puts it inside the math that drives distribution CapEx deferral. Three things follow. One, in any jurisdiction, propose the same mechanism — a required load-flexibility scenario in distribution planning is a smaller, more winnable ask than a procurement target and it produces the deferral numbers that justify procurement later. Two, treat the common orchestration framework as the published statement of what a DERMS RFP must deliver, and put OpenADR 3.1 and IEEE 2030.5 in the requirements document as pass/fail rather than scored items — the November 4 workshop names the standards the framework will be built on. Three, use the bridge-year authorization as the model for a continuity ask: a DR business case whose payback extends past its program’s authorized horizon carries an undisclosed termination risk, and an explicit bridge authorization removes it.
⚑ A Quarter of the Available Demand-Side Capacity Comes From Orchestration With Zero New Enrollment — 23 of 90 MW
Topic: DSM Portfolio Value Stack / DERMS Business Case / Avoided-Cost Method
Relevance: Brattle, for Uplight, modeled a representative Southwest Power Pool utility and found that managing DR + EE + TOU rates as a coordinated portfolio rather than program-by-program unlocks roughly 90 MW of incremental demand-side capacity by 2030, lifting demand-side capability from about 3% to roughly 5% of system load. Decomposition: ~66 MW from the stacked programs, and ~23 MW from improved forecasting and staggered dispatch alone. The recommendation is portfolio-level management with clear system-level performance targets and explicit modeling of interactive effects across programs.
Action Signal: Implement — The 23-MW line is the most defensible item in a DERMS capital request, because it is the only capacity in the stack that does not depend on an enrollment forecast. Build the request around it: state the orchestration-derived megawatts separately from the enrollment-derived megawatts, and price the platform against the former. The 90-MW figure also sits squarely in this series’ illustrative range — a ~70-MW portfolio deferring ~$150M of CapEx at ~$20M program cost — so the comparison is directly usable, with the standing caveat that the method transfers and the specific megawatt number does not. One discipline this report forces that most DSM filings skip: model the interactive effects, because a portfolio that counts EE savings and DR reduction against the same customer hour is double-counting, and an intervenor who finds it discredits the whole stack.
⚑ Georgia Power Wired a DERMS Directly to Two Dispatchable DR Tariffs — Copy the Tariff Pair, Not Just the Platform
Topic: DERMS Deployment / C&I Demand Response / Utility-Owned BTM Resource
Relevance: Under a GA PSC-approved stipulation from its most recent IRP, Georgia Power is deploying a DERMS for visibility, forecasting, control and optimization of behind-the-meter DER coordinated with real-time grid operations, spend across 2025–2028, paired with a DER Customer Program built on two tariffs: Resiliency Asset Service (RAS), under which the utility designs, procures, installs, owns, operates and maintains a dispatchable resource behind a C&I customer’s meter to meet that customer’s resiliency need; and Demand Response Credit (DRC), under which the customer takes a bill credit for the utility’s right to dispatch that resource during system reliability events. Because the utility owns and dispatches the asset, the megawatts are firm rather than notified, which is the profile supporting scoring against avoided generation and T&D at something near the Brattle ~$66/kW-year value.
Action Signal: Implement — This is the integrated architecture this series has specified all year, in production, in a vertically integrated jurisdiction with no wholesale market to monetize against — which makes it the most transferable DERMS precedent available to a utility that cannot point to PJM revenues. The replicable asset is the tariff pair, not the software: RAS creates the asset and DRC creates the dispatch right, and a DERMS procured without the tariff that creates the dispatch right has bought the instrument without the resource. Draft both together. Then anticipate the test a commission will eventually apply: the resiliency framing puts customer-outage value first and grid dispatch second, so the realized system-capacity contribution depends on enrollment depth and actual call frequency — instrument the call record from day one, because that is the number that will be challenged.
⚑ Texas Ruled Mandated Large-Load Curtailment Is Not Compensable — Stop Booking It as DR Capacity
Topic: Curtailment Accounting / Emergency vs. Compensated DR / Co-Location
Relevance: In Docket 59220, decided July 23, the Texas PUC approved the first co-located data-center net-metering arrangement under SB 6 — Ensign Infrastructure’s 525.5-MW Goodnight campus (including 260-MW Crusoe Load 2), Crusoe as operator and Google as technology partner, paired with the 265.5-MW Goodnight 1 wind farm. Conditions: the entire campus must curtail within 30 minutes in ERCOT emergencies — roughly twice the paired generator’s nameplate, a 2:1 load-to-generation ratio, not capped by the generator’s size — and the Commission ruled that this mandatory emergency curtailment is a reliability obligation, not a compensated service, so the load cannot collect demand-response payments for it. Reported consequence: project-finance models assuming proportional curtailment are wrong, pushing developers toward full on-site backup generation or storage. Next test: Docket 59399 (Amazon/Vistra, Comanche Peak) on whether firm nuclear co-location draws different treatment than intermittent wind.
Action Signal: Implement — Adopt the accounting rule now, before it is imposed: count mandated large-load curtailment as emergency risk reduction, never as DR capacity with an avoided-cost value. The two are not interchangeable, and a filing that books compelled curtailment into a demand-side portfolio is claiming megawatts it does not control, cannot schedule and cannot settle. This also supplies the attribute that was missing from the SB 6 analysis two weeks ago: the obligation’s depth is uncapped, which is the answer to one of FlexMosaic’s five keys and the reason this product is the hardest of the six to map. Use it competitively as well — where curtailment is compelled and uncompensated, the procured alternative is the only instrument that produces defined depth, defined notification, settlement-grade M&V and a reusable telemetry record, and that is now a documented regulatory distinction rather than a marketing claim.
⚑ Half the Announced Data-Center Capacity Will Not Exist in Two Years, and One Utility Freed 80 MW Without Building Anything
Topic: Speed-to-Power / Interconnection Headroom / Supply-Side Lead Times
Relevance: Goldman Sachs expects only 50–60% of planned data-center capacity online within two years against forecasts of 17–20% of U.S. electricity by 2030–2035. Constraints: power transformers at 2–3-year lead times, gas-turbine backlogs >100 GW pushing delivery into the early 2030s, a ~500,000-worker construction shortage, 71% public opposition to a nearby data center (May 2026, +30 points YoY) and ≥20 projects killed by local opposition in Q1 2026, with Microsoft, Google and Oracle projects hitting permitting, legal and pipeline roadblocks. Developer response splits into bring-your-own-capacity (OpenAI 8 GW Ohio, Amazon 7.7 GW West Texas) and flexibility/headroom — Portland General Electric’s GridCARE freeing ~80 MW of interconnection headroom, about one-fifth of PGE’s expected 2029 additions.
Action Signal: Implement — The 80-MW GridCARE figure is the number to carry, because it is a measured headroom recovery expressed as a share of a real interconnection pipeline, which is the form a commission can act on. Two disciplines. One, price every demand-side and headroom option as cost per kilowatt available in 2028 rather than per kilowatt-year — on a 2–3-year transformer and early-2030s turbine calendar, availability date is the differentiating variable and levelized cost is not. Two, read the bring-your-own-capacity column as the competitive threat it is: self-supply routes around the utility entirely, so demand-side value accrues only where flexibility is captured inside the regulated planning and interconnection process. A utility without a flexible-interconnection or headroom product is not resisting DER; it is ceding the customer to self-supply — the same arithmetic as MISO’s ZGIA fast track and California’s AB 2383.
⚑ Michigan’s PBR Move Attacks the Reason Demand-Side Programs Lose Internal Arguments They Win on Cost
Topic: Ratemaking Structure / DSM Incentive Alignment / Business-Case Framing
Relevance: Michigan residential rates rose 10.42% from June 2025 to June 2026 (20.82¢ → 22.99¢/kWh) against 4.98% nationally (17.47¢ → 18.34¢/kWh). Michigan PSC Chair Dan Scripps’ first affordability recommendation is multi-year rate plans with performance-based ratemaking. ComEd already uses metrics including peak-load reduction and reliability incentives. Pending Senate Bill 768 would extend the minimum interval between rate cases from one year to three.
Action Signal: Engage — Under cost-of-service ratemaking a CapEx-deferring DR program reads on the income statement as foregone return, which is the structural reason demand-side proposals lose internally after winning on levelized cost; under PBR tied to peak reduction and reliability, the same portfolio becomes a way to earn. Where a PBR or multi-year-plan docket is open or opening, engage with metric definitions drafted rather than reacting to someone else’s — a peak-reduction metric without a cost cap can be satisfied by expensive peak reduction, and poorly designed metrics reward the spending they were meant to curb. Note the second-order effect too: a three-year rate-case interval lengthens the payback horizon demand-side investments are evaluated over, which is the same design parameter as New York’s three-year minimum program term, and this series’ standing finding holds — term length does more work than incentive level.
⚑ A Real Utility Filed a 1.30–1.45 Utility Cost Test and Also Filed a Downward Revision — Use the Second Fact as Carefully as the First
Topic: DSM Cost-Effectiveness / Benchmark Hygiene / Storage as DSM
Relevance: Rocky Mountain Power’s Semi-Annual DSM Forecast to the Utah PSC (June 26) projects total DSM budget at ~$110.1M in 2026 (from ~$98.9M in 2025), with Wattsmart Homes forecast at ~69,696 MWh at a ~$25M budget — revised down from ~84.9 million kWh after commercial midstream savings were found to be double-counted — and a forecast Utility Cost Test of 1.30–1.45 program-wide. RMP explicitly positions Wattsmart Batteries as a planning-relevant DSM resource delivering peak shaving, load shifting for renewable integration and fast-response capacity through coordinated dispatch.
Action Signal: Implement — Add the 1.30–1.45 range to the benchmark library as a regulated utility’s own filed forecast, which is harder to attack than a consultant’s modeled case and supports the framework’s EE cost of ~$20.70/MWh against $45–108/MWh new gas CC. Then attach the discipline the same filing demonstrates: the defensible IRP input is the verified ex-post benefit-cost ratio, not the filed forecast, because this filing shows these projections get revised by material amounts for mundane accounting reasons — a ~15 GWh downward correction for double-counted midstream savings. When citing a UCT, cite the vintage and whether it is forecast or verified, and ask any vendor or consultant quoting one which it is. Separately, note the categorization win: distributed storage counted as a DSM resource inside a DSM filing, settled by practice rather than by order — useful precedent in any jurisdiction where that classification is still contested.
⚑ The Enrollment Constraint Now Has a Three-Part Financing Stack — Model It as a Lever, Not an Assumption
Topic: DER Enrollment Economics / VPP Capacity Forecasting / Program Design
Relevance: Three channels converged in this week’s log. Utility on-bill: Illinois’s EEUP (final guidelines September 10) — a 7-kW solar + 13.5-kWh system at $34,530, an $11,913 financing gap after rebates, recovery capped at 80% of estimated savings, no upfront cost, credit check or homeownership requirement, and ~$770/yr potential PJM VPP revenue; baseline context, ~28% of U.S. homes have a smart thermostat while DR participation runs far lower, and upgrades are affordable for <1 in 10 households versus a claimed >9 in 10 under coordinated inclusive investment. Federal tax credit: Section 48E returning 30–50% of project cost, credits monetized at 85–93¢ on the dollar or taken as direct Treasury payments by nonprofits and public facilities, with small manufacturers shifting lathes, furnaces and kilns off-peak — and a phase-out beginning 2032 after OBBBA cut standalone solar. Federal grant: $7B Solar for All restored program-wide by a second federal judge in one week (September 25, Judge Chutkan), ~$250M to the Texas Solar for All Coalition, ~$54M to Harris County. Also: 34 states require utilities to deploy DERs; 12-plus have passed or are considering VPP legislation; Virginia HB 1062’s inclusive-investment feasibility study is due November 1, 2026, two weeks before Dominion’s VPP tariff.
Action Signal: Engage — Model DER adoption as a function of financing access rather than incentive level. Every VPP capacity forecast in this series’ library assumes a participation rate and then prices incentives against it, which is backwards in a market where fewer than one household in ten can finance the device. Three concrete moves: evaluate an on-bill, below-savings recovery mechanism for your own territory using the EEUP parameters as the starting template, and treat the 80%-of-savings cap as the design variable that decides whether the model works; state the 48E dependency explicitly in any storage-dependent enrollment ramp, because the 2032 phase-out is legislated rather than forecast; and enter Solar for All restoration as a sensitivity, not a baseline, since it is trial-court litigation subject to appeal. For Virginia specifically, read the November 1 HB 1062 study beside the November 15 VPP tariff — the commonwealth is deciding the financing half and the aggregation half of the same enrollment question within a fortnight.
⚑ PNM Has 40 MW of Flexibility and 300 MW of Proposed Data Centers — and EPRI Says the Blocker Is Standardization
Topic: Microgrid Orchestration / DERMS Normalization Layer / Dispatch Frequency
Relevance: In a September 23 feature, EPRI’s Jackie Baum (senior team lead, DER integration) argues microgrids and their DERs should be dispatched for routine peak management and voltage support rather than held as emergency-only backup — “the majority of the time, the power is on.” The scale tension: PNM has ~40 MW of balancing-authority flexibility today while proposed data centers could reach ~300 MW. PNM’s Jon Hawkins names the barrier as standardization — every microgrid differs in loads, generation and controls, unlike standardized solar or storage. Participants include Sandia National Laboratories (power-electronics testing) and Kit Carson Electric Cooperative (five microgrid projects, batteries and hydrogen). This is also the week’s only substantive EPRI contribution, and it arrived through trade press rather than a research release — the twelfth consecutive week with no EPRI publication in the window.
Action Signal: Engage — Two requirements, both testable in an RFP. One, require the platform to normalize heterogeneous microgrid and DER controls into a single dispatchable layer, with the vendor demonstrating it against at least two dissimilar control architectures rather than asserting it — this is the same deliverable the CPUC’s common-orchestration framework is asking consultants to define, arriving the same week from the operational side. Two, carry this sentence into every microgrid and resilience business case: distributed flexibility earns avoided-cost value only when dispatched continuously, not when reserved for outages. An asset called twice a year has resilience value and essentially no capacity value. Most utility microgrid portfolios are specified for the emergency case and then evaluated against the capacity case, and the gap between those two is where the business case fails. Fix it at specification, where it costs nothing.
⚑ PJM’s Members Kept the Section 205 Pen and Set a Two-Thirds Bar on State Proposals — Track the FERC Filing, Not the Term Sheet
Topic: RTO Governance / Market-Reform Pathway / Stakeholder Process
Relevance: This was the tracker’s first open verification item and it resolved as contested rather than settled. PJM members voted on a governance term sheet September 30; it has not yet been filed with FERC and awaits Commission approval. Agreed: state executive branches join the Organization of PJM States for the first time (OPSI today is utility regulators only), and one additional state seat goes on the board nominating committee — against a competing staff proposal for two state seats plus an independent-appointee option. Retained: the members committee keeps final decision-making on Section 205 filings, and OPSI proposals require two-thirds member approval to advance. Maryland Governor Wes Moore said members “responded with a term sheet that stifles the states by handing those same members a veto over any proposal the states bring forward”; Senator Katie Fry Hester said it leaves states “spectators.”
Action Signal: Watch — The practical consequence for demand-side work is a routing decision. Any PJM-wide DR or flexibility market reform still runs through a members committee that has just confirmed it holds the 205 pen, and a state-originated proposal now faces an explicit two-thirds threshold — which means a state commission’s faster path to a flexibility outcome is its own jurisdiction, exactly as Pennsylvania demonstrated this week by writing curtailment priority itself rather than seeking it through PJM. Track the FERC filing of the term sheet, because that is where it is actually decided and where the states’ objections become a record. And note the week’s shape: the same seven days produced a Commission unwilling to accept PJM’s rushed backstop filing and a membership unwilling to cede filing authority — a governance stalemate layered on top of a five-month procurement gap, in the RTO carrying the 6,831-MW shortfall that two state dockets are now responding to directly.
📌 Sources
September 28, 2026 Entry
– Georgia Power — DER Customer Program: Resiliency Asset Service and Demand Response Credit Tariffs (GA PSC-approved stipulation; DERMS spend 2025–2028) (first appearance in this series)
– Utility Dive — Customer Experience, Better Modeling Can Boost Demand-Side Portfolio: Report (Brattle Group for Uplight; June 3, 2026) (first appearance in this series)
– Utility Dive — Demand Management, Data Center Flexibility Boost Regional Reliability: NERC (May 27, 2026) (fourth appearance from the same source — fully reported in the May 29, June 26 and August 21 digests; carried here in one paragraph and cited as the May assessment, not as a current finding)
– NY Department of Public Service — Commission Improves Customer-Centered Electric Demand Response Programs (CSRP, DLRP, DLC, Term-DLM, Auto-DLM; order April 2026) (reported in the August 28 digest for the 50 kW minimum and three-year minimum term; carried September 18 for the Bring-Your-Own-Battery rollout; one line only)
September 29, 2026 Entry
– Utility Dive — Minnesota Approves Xcel’s Utility-Owned Virtual Power Plant (order April 2, 2026) (covered in this series since May and analyzed at length in the July 17 digest; carried here as a standing benchmark — 200 MW, 1–3 MW units, $430M, Sparkfund, 50-MW interim assessment, November 2027 independent evaluation)
– GE Vernova — GridOS for Distribution: Unified ADMS, DERMS and Grid-Planning Analytics on a Federated Grid Data Fabric (fully reported in the September 11 digest from the February launch; carried here only for the EnergyHub grid-edge pairing, 90+ customers and 127M+ managed service points)
– New Jersey Board of Public Utilities — NJ Proposes “Virtual Power Plant” Plan to Save Ratepayers Millions Annually (July 27, 2026) (fifth appearance in the series; carried in one line — still a proposal in stakeholder process, and still the “undecided” entry in the four-way enrollment-channel comparison)
– Rocky Mountain Power — Semi-Annual DSM Forecast Report to the Public Service Commission of Utah (June 26, 2026) (first appearance in this series)
September 30, 2026 Entry
– Utility Dive — The Virtuous Cycle of Inclusive Utility Investment and Virtual Power Plants (Zuha Maryam Shaikh, Clean Energy Works; September 29, 2026) (Illinois EEUP cost math is new; the Virginia 450-MW Dominion VPP pilot within it has been reported repeatedly since September and is carried as context, with only the HB 1062 November 1 feasibility study treated as new)
– T&D World — Edge DERMS Targets Locational Grid Management and Flexible Load (FlexEnergi “Ora Edge”; September 8, 2026) (first appearance; no named utility partners, MW quantities or device counts disclosed — design and commercial-model reference only)
– CLEAResult — CLEAResult Strengthens Utility Demand Management Portfolio With Acquisition of Connected Energy (August 25, 2026) (first appearance; terms, MW under management and customer counts undisclosed)
– Tech Times — Texas Sets Curtailment Precedent: Co-Located AI Campuses Must Run Off-Grid (July 30, 2026, on PUC Docket 59220 decided July 23, 2026) (first appearance in this series; next test is Docket 59399, Amazon/Vistra at Comanche Peak)
October 1, 2026 Entry
– Utility Dive — FERC Chair Slams PJM ‘Mess’ as Country’s Largest Grid Operator Delays Power Auction (September 30, 2026) (movement on a tracked watch item — the backstop proposal was reported in the September 18 and September 25 digests at the proposal and scheduled-open stages; see the verification note below, which corrects this entry’s “partially rejected” characterization)
– Utility Dive — Microgrids Must Push Beyond Pilots Into Everyday Grid Operations (EPRI, PNM, Sandia, Kit Carson; September 23, 2026) (first appearance; anchors this week’s EPRI Spotlight)
– Utility Dive — Affordability Concerns Drive Momentum for Performance-Based Ratemaking (September 24, 2026) (first substantive treatment in this series)
– Utility Dive — Second Federal Judge Overturns $7B Solar for All Cancellation by Trump Administration (September 25, 2026) (first appearance)
October 2, 2026 Entry
– RTO Insider — Pennsylvania PUC Approves Tentative Order on Emergency Load Control Procedures (Docket M-2026-3064961; October 1, 2026) (movement on a tracked watch item — the September 11 5-0 vote opening this docket was reported in the September 25 digest with the October 1 tentative order flagged as imminent)
– Stoel Rives — Energy Regulatory Updates, September 30, 2026 (covering September 23–29; CPUC R.25-09-004 DR bridge funding and R.21-06-017 DER orchestration) (first appearance)
– Utility Dive — The Data Center Boom Continues Apace, but Projects Face Mounting Obstacles (September 29, 2026) (first appearance; distinct from the September 25 digest’s data-center-speculation roundup at 830495)
– Utility Dive — Facilities Are Using 48E Investment Tax Credits to Make Energy Projects Pencil Out (September 30, 2026) (first appearance)
Verification Searches Conducted This Week
– mgrid.org — FERC Suspends PJM’s Backstop Procurement Five Months and PJM Pulls Its September 30 Auction (Docket ER26-3380; September 29, 2026) — landed, and it corrects the research log. FERC accepted PJM’s filing and suspended it five months to February 28, 2027, subject to refund and further proceedings, finding cost allocation, transmission-owner exit and LSE collateral potentially unjust and unreasonable (supplier collateral accepted), and struck the co-op/municipal opt-out on discrimination grounds. The log’s “partially rejected” framing is superseded. (Corroborated by RTO Insider — FERC Conditionally Approves PJM Backstop Procurement, Delays Implementation.)
– Pennsylvania PUC — PUC Advances Consumer Safeguards as Data Center Growth Adds New Demands to Electric Grid (October 1, 2026) — landed. Primary source adds the 4-0 vote tally, the requirement that large-load registries be visible to the utility, PJM and the Commission, and the November 17, 2026 technical conference on large-computational-load cost allocation.
– Maryland Matters — Maryland Officials Say Changes to Electric Grid Governance Aren’t Enough (October 1, 2026) and RTO Insider — PJM Stakeholders Present Terms of Governance Changes — landed. The series’ first open verification item resolves as contested: members voted September 30, the term sheet is not yet filed with FERC, state executive branches join OPSI for the first time, one added state seat on the nominating committee, the members committee retains Section 205 authority, and OPSI proposals need two-thirds member approval.
– EPRI media-resources direct check — no result. No EPRI publication dated September 28 through October 2 in either the research log or EPRI’s own press-release index; twelfth consecutive dry week. Most recent substantive items remain the DCFlex initiative materials and the nine-site and European expansions carried since July.
– NERC Rules of Procedure registry track — not confirmed. The August posting of proposed Computational Load Owner/Operator Rules of Procedure changes was located, but no confirmation surfaced that the expected October 20-day comment period has opened. Treat the October comment / November ballot / December adoption schedule as unconfirmed rather than slipped, and chase it first next run — if the registry track slips past December, the December 31 FERC filing is a paper milestone and the 50 MW / 100 kV threshold binds no registered entity.