All five weekday entries (Monday September 14 through Friday September 18) are present in the research log. Four items are deduplicated across weeks rather than within this one and are reported as movement or as standing benchmarks rather than as new findings: the Octopus Energy–Uplight transaction was fully reported in last week’s digest from a different outlet and is carried here only for the closing mechanics the September 18 entry adds; Xcel Minnesota’s Capacity*Connect appears for the third consecutive digest and is carried as a standing benchmark in one paragraph; the New Jersey BPU virtual-power-plant straw proposal appears for the fourth consecutive digest and is carried in one line; and New York’s Dynamic Load Management order, reported in the August 28 digest, is carried only for the Bring-Your-Own-Battery rollout. One item — FERC’s six Section 206 show-cause orders of June 18 — has been referenced in this series repeatedly as context but never reported from the Commission’s own notice; it is treated here as a first primary-source account, not as new news. Several items carry pre-week publication dates — the Xcel Capacity*Connect piece from April 8, the New York DLM order from April, the PJM board proposal from July 28, the PJM load-trip account from August 12, the FERC show-cause suite from June 18, the Sunrun–Renew Home–Tesla framework from June 24, the DELTa tariff analysis from September 1, the PG&E Flex Connect account from September 2, the ComEd program piece and the Massachusetts executive order from earlier in 2026 — and they are reported on the day they entered the log with original dates shown. Three verification items were chased directly this week by targeted search and are noted where they land.

Four months of this series have tracked demand-side capacity as something being argued about. This week it was ordered.

Start with the instrument that pays nothing. On September 17 the Department of Energy issued Order No. 202-26-45 to PJM — effective through the end of September 18, authorizing the grid operator to activate backup generation and dispatch specified resources across 13 states and the District of Columbia, reaching more than 35 GW of backup generation during a late-season heat event. The order itself is unremarkable by now. The count is not: this is the seventh Section 202(c) emergency order issued to PJM in a single calendar year, after 202-26-06, -23, -32/33, -41 and the Duke Carolinas order 202-26-43 in an adjacent footprint. A tool designed for genuine emergencies has become a seasonal operating procedure, and legal analysts are reading the cadence the way a planner should: as evidence that capacity procurement and load forecasting are not keeping pace, not as seven separate weather events.

Then the instrument’s legal ceiling. On September 11 a unanimous D.C. Circuit panel vacated the first DOE §202(c) order forcing Consumers Energy’s 1,420-MW J.H. Campbell coal plant to delay its May 31 retirement, writing that it was “unpersuaded by DOE’s sweeping conception of its ’emergency’ authority” and that the order usurped state authority over generating resources. This closes a watch item this series has carried since June. The running cost of the broader regime — 10 units at six plants ordered to stay online — is past $547 million on the Sierra Club’s tracker, borne by ratepayers, for units several of which ran at sub-50% capacity factors or produced no power at all. Read the two items together and the conclusion is uncomfortable for anyone treating emergency authority as a reliability plan: it is invoked seven times a year and it can be vacated.

Then the size of the hole it is being used to fill. An independent resource-adequacy study commissioned by the Pennsylvania PUC — produced by Synapse Energy Economics, Mondre Energy and Aspen Technologies — models PJM’s 2030 loss-of-load expectation at 0.59 under the reference scenario built on PJM’s own 2026 load forecast, nearly six times worse than PJM’s 0.1 LOLE planning criterion, and at 13.20 in a high-load/low-supply case — more than a hundred times worse, implying more than thirteen loss-of-load days a year. Only the scenario with no new data centers meets the 0.1 target. That is the cleanest natural experiment this series has been handed: the gap is load-driven, it is quantified by an independent consultant under a state commission’s sponsorship, and every megawatt of firm demand-side reduction buys directly against it.

And then, in the same week, two jurisdictions issued the answer as an instrument with a date on it. Virginia’s utility-scale VPP mandate directs Dominion to file a program tariff by November 15, 2026 and a pilot to the State Corporation Commission by December 1, for up to 450 MW of DER aggregations across multiple regions, enrollable either directly with the utility or through a third-party aggregator — the first state to decline the make-versus-buy fork rather than pick a side. And California’s PUC approved the 2026 Avoided Cost Calculator update on September 3 in Decision D.26-09-007, refreshing the generation-energy, generation-capacity, ancillary-services, T&D-capacity and decarbonization value streams that convert a demand-side program into a defensible dollar figure. One state ordered the megawatts; another re-certified the price list.

The week’s argument, stated plainly: the adequacy gap is now measured, the avoided-cost method is now current, and the demand-side answer now has a filing deadline — while the emergency substitute has a seventh invocation and a judicial limit. Every element a DSM business case needs to be written this quarter arrived in the same five days.

⚡ Virtual Power Plants & Demand Flexibility

Virginia has ordered a 450-MW virtual power plant into existence with dual enrollment channels, and in doing so became the first state to refuse the utility-versus-aggregator choice that every other VPP proceeding this series has tracked was forced to make. Under the state mandate, Dominion must propose a VPP program tariff for residential, commercial and industrial customers by November 15, 2026, and file a pilot with the State Corporation Commission by December 1, consisting of DER aggregations totaling up to 450 MW sited across multiple geographic regions. Participants may enroll directly with the utility or through a third-party aggregator.

That last provision is the one worth copying. This series has spent two months documenting the same fight in different jurisdictions: Minnesota approved Xcel’s utility-owned, rate-based Capacity*Connect and declined to pair it with a third-party program, drawing objections from SEIA, MnSEIA and the community-solar coalition; New Jersey’s straw proposal is still working out compensation design; the Sunrun–Renew Home–Tesla framework is aggregator-led and has no utility counterparty at all. Virginia’s design simply admits both and lets the customer choose the channel. For a DERMS procurement this is a concrete requirements template rather than a policy preference: telemetry, dispatch and settlement must work identically whether the resource is enrolled by the utility or managed by an aggregator, which in practice means standards-based aggregator interfaces — IEEE 2030.5 — as a hard requirement rather than a scoring bonus, plus a settlement path that can attribute performance to two different counterparties against the same meter.

The siting is the other reason to care. This is the PJM/Virginia data-center load pocket — the same geography where the Pennsylvania PUC study models LOLE deterioration, where ~3,800 MW of data-center load tripped offline on July 22, and where six of this year’s seven §202(c) orders landed. If 450 MW of aggregated DER can be stood up and dispatched there on the mandated timeline, it is the strongest available evidence that demand-side procurement moves faster than large-load growth. If it cannot, that is worth knowing before another state writes the same statute.

The evaluation clock is explicit and unusually well designed: the pilot concludes July 1, 2028, after which the commission must assess its effectiveness at providing peak-demand grid services and begin developing a permanent program with procurement targets and performance metrics. A pilot with a statutory obligation to convert into a permanent program with targets is a different animal from a pilot that expires into a report. That M&V-to-procurement pathway is exactly what this series has argued the avoided-cost case depends on, and Virginia has now written it into law rather than into a settlement. (Source: Utility Dive — September 18 entry)

The Sunrun–Renew Home–Tesla coalition is the largest aggregated-DER commitment on record at more than 16 GW, and it is still — five months after announcement — a framework rather than accredited capacity, which is precisely the question this series has been carrying it as a watch item to answer. Convened by Renew Home (a Sidewalk Infrastructure Partners company), the coalition says it can orchestrate more than 16 GW of dispatchable capacity across most major U.S. markets by pooling hundreds of thousands of Sunrun and Tesla home batteries with flexible peak capacity from more than 8 million Renew Home-managed smart thermostats and devices.

The pitch is explicitly an avoided-cost pitch, and it is aimed at data centers as directly as at utilities: “capacity-as-a-solution” available with no additional hardware, software, interconnection, water or land, switched on “in months, not years” while transmission, substations and gas plants take far longer to permit and build. The coalition cites Brattle Group analysis that better utilization of the existing grid could cut U.S. electricity bills by $110–170 billion over the next decade while accelerating data-center interconnection by several years — a figure worth adding to the filing library alongside the ~$66/kW-year DR capacity benchmark this series uses, because it reframes demand flexibility as an affordability instrument rather than a reliability one.

The caution is the one this series applied to VPPs generally two weeks ago and it has not been answered here. Sixteen gigawatts of orchestrable capacity is a device count, not an accreditation. EnergyHub’s peaker-equivalent “Turing Test” scores even the most advanced VPPs at roughly 2 on a 0-to-4 maturity scale; PJM’s own record shows aggregated DER participation under 0.2% of system peak after five years of Order 2222. Until some portion of this 16 GW clears a capacity auction or is credited in an IRP at a stated ELCC, it belongs in a business case as capability with a timeline, not as capacity. That conversion remains the standing watch item. (Source: Canary Media, June 24 — September 17 entry)

Octopus Energy’s majority investment in Uplight closed on September 1 with Schneider Electric retaining a significant minority stake — carried here from last week’s digest for the closing mechanics only, because the ownership structure that emerged is more consequential for DSM buyers than the headline was. The transaction (agreement signed March 24, closed September 1) leaves Uplight an independent company able to draw on Schneider’s grid-operations, DER-orchestration and One Digital Grid DERMS capabilities while operating under Octopus’s majority control and its Kraken platform. The stated targets are unchanged: $1 billion in customer savings and more than doubling flexible capacity to over 20 GW within five years.

What the closing adds is the shape of the vendor risk. A utility running DSM through Uplight is now contracting with a platform jointly backed by a fast-moving retail-energy operator and an industrial grid-software incumbent that also sells a competing DERMS. That is not necessarily bad — it is arguably the best-capitalized customer-engagement layer in the market — but it is a specific contractual condition worth naming in a renewal: confirm data-portability and model-export terms, confirm that the Schneider relationship does not create a preferred-integration path that narrows your own DERMS options, and confirm which entity holds the enrollment relationship with your customers. The 20-GW ambition is enrollment-dependent and faces the early-adopter-plateau risk VPP analysts have been flagging all year; it should not be modeled as committed capacity. (Source: GlobeNewswire, September 1 — September 18 entry; the same transaction was reported in the September 11 digest from Utility Dive’s press-release feed)

Carried — Xcel Minnesota. Capacity*Connect appears for the third consecutive digest and is carried as a standing benchmark, not re-reported: up to $430 million for up to 200 MW of batteries in 1–3 MW increments over two years at commercial, industrial and nonprofit sites, sited at strategic grid locations to relieve local distribution constraints; MISO energy and capacity revenues expected to repay nearly the entire deployment cost, holding the typical residential bill impact to $0.67–$1.50 per year through 2030, with a $50 million Google contribution. The standing watch item is unchanged and now reads differently beside Virginia: the commission’s mandate for specific distribution-benefit estimates by the November 2027 Integrated Grid Plan, with quarterly reports in the interim, is the only place in the country where a regulator has ordered a utility to put a number on avoided distribution investment from a VPP. Virginia’s 2028 evaluation obligation is the second. Those two deliverables, two years apart, are what will eventually settle whether locational siting earns its premium. (Source: Canary Media, April 8 — September 16 entry; previously reported in the September 4 and September 11 digests)

Carried — New Jersey. Docket QO26030099, comments closed August 17, launch targeted 2027, spanning PSE&G, JCP&L, Atlantic City Electric and Rockland Electric under Governor Sherrill’s Executive Order No. 2. Fourth consecutive digest, carried in one line: it remains the open compensation-design case — bill credits, discounts or direct payments — and Virginia’s dual-enrollment structure is now the comparator to test its eventual answer against. (Source: NJ BPU, July 27 — September 16 entry; previously reported August 28, September 4 and September 11)

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 program-design parameters that matter most (the 50 kW participant minimum and the three-year minimum program term that makes storage economic to enroll). The new detail this week is the delivery: Con Edison and Central Hudson roll out Bring-Your-Own-Battery options later in 2026, paying customers to discharge storage during utility call events. That extends DLM from load curtailment into distributed-storage dispatch and moves New York’s program suite from a DR portfolio into a DERMS orchestration problem. (Source: NY DPS, April 2026 — September 14 entry; program previously reported August 28)

Massachusetts began measuring against its 3.5-GW demand-management target this month, and the baseline-before-procurement sequencing is worth copying regardless of what the inventory finds. Governor Healey’s March 13 executive order directs the commonwealth to develop 3.5 GW of demand-management resources by 2035 — explicitly spanning energy efficiency, VPPs, microgrids, managed EV charging and demand response — and called for a September report inventorying which programs already exist before the state pursues new ones.

The scale framing is the usable part: 3.5 GW is roughly 13% of the entire six-state New England peak of 26.1 GW, and it dwarfs California’s largest VPP network at ~0.5 GW. A state codifying demand management at that share of peak positions coordinated demand-side resources as a genuine alternative to fossil peakers rather than a marginal program line, and it validates the avoided-cost premise this series works from (energy efficiency at ~$20.70/MWh against $45–108/MWh for new gas combined cycle) at policy scale rather than program scale.

The caveat is the honest one: a target and an inventory are planning artifacts, not procured megawatts. But the sequencing — establish the baseline, size the gap, then procure — is the discipline most demand-side targets skip, and it is the reason this one will produce a defensible number for how much of the 3.5 GW must be newly built. Utilities in states considering a percentage-of-peak demand-side target should watch what the September inventory actually counts, because the definitional choices made there (does an existing efficiency program count toward a demand-management target?) will be copied. (Source: Mass.gov — September 14 entry)

🔌 DERMS & Grid Integration Technology

Two vendor releases landed in the same log entry and together they draw the fault line this series has been tracking in every ADMS-versus-DERMS procurement debate: the ADMS incumbent is absorbing demand-response execution into the control room, while the standards-based independent is selling interoperability to the utilities the incumbent does not serve.

Oracle shipped a generative-AI upgrade to its Utilities Network Management System whose headline feature is storm-scenario outage forecasting and automated crew-dispatch planning — but the DER-relevant part is the enhanced DER-orchestration module that auto-builds demand-response and DER-event dispatch strategy templates and optimizes behind-the-meter, customer-owned resources directly from the NMS. Oracle states the platform supports six of the top 10 U.S. utilities and more than 61 million customers. When DR event templates, DER dispatch and crew management live in one platform at that installed base, it stops being a product feature and becomes a de facto reference architecture that every smaller utility’s RFP gets measured against — and it changes the integration diagram, and the total cost of ownership, that a DR business case has to assume.

Survalent moved the other way. It launched the Helix time-series historian and expanded SurvalentONE DERMS to extend DER orchestration and integration using the IEEE 2030.5 standard — the same secure utility-to-aggregator protocol adopted in California’s Rule 21 — explicitly targeting cooperative and municipal utilities managing rising behind-the-meter solar and storage. The pairing is the point: DERMS value depends on assessing real-time DER behavior, understanding its grid impact, and identifying constraint violations before they occur, all of which require a high-resolution operational-data store fused with the SCADA/ADMS layer rather than parked in a separate analytics silo. A historian is not a glamorous product announcement; it is the thing that makes both real-time constraint management and after-the-fact avoided-cost verification possible from the same data.

For a utility writing requirements, the two releases resolve into two separately testable scoring criteria rather than a vendor preference. First, native IEEE 2030.5 and IEEE 1547 support preserves aggregator interoperability — and Virginia’s dual-enrollment mandate just turned that from a preference into a functional necessity for anyone operating in a state that lets customers enroll through either channel. Second, a historian or data-platform integration path with stated resolution and retention, because DER telemetry has to feed operations and settlement from the same store. The suite-versus-best-of-breed question these two releases pose should be settled before requirements are written, not during evaluation, and the commercial terms that matter — data governance, open APIs, model-and-data exit terms — belong in the contract rather than in a schedule exhibit. (Sources: PR Newswire — September 17 entry, items 1 and 2)

A third market-sizing figure entered the log this week, and the series now has three circulating numbers that measure three different things — which is worth reconciling once rather than citing inconsistently. SNS Insider values the Advanced Distribution Management System market at $3.85B in 2025, projecting $19.80B by 2035 at a 20.2% CAGR, with DERMS contributing roughly 17% of ADMS spend and the control-room convergence of SCADA, OMS and DER management named as the structural driver.

Set that against the two figures this series already carries: T&D World’s DERMS-only sizing of $557M (2025) → $1.36B (2031) at ~16% CAGR, and Research and Markets’ digital power utility market at $142.68B (2026) → $213.88B (2031) at 8.42%. These are not in conflict; they are nested. The digital-power-utility number is the whole capital envelope, ADMS is the control-room platform inside it, and DERMS is the grid-edge slice inside that. Use the DERMS-only number for a DERMS line item, the ADMS number for a platform replacement, and the digital-utility number only as context for why the budget exists. The practical procurement figure this series has found most useful remains neither: $2 million to $5 million per 1,000 customers for full-scale deployment, with a recovery period that can exceed a regulatory cycle.

The standard caution applies — these are analyst estimates, not utility spend disclosures, so they anchor scale and direction and have no business in an avoided-cost calculation. But the direction is consistent across all three sources and it says the same thing: DER orchestration is still a minority of control-room spend and the fastest-growing slice of it, which is the condition under which vendor competition drives standards adoption and integration costs down. (Source: GlobeNewswire / SNS Insider, September 9 — September 14 entry)

Flexible interconnection now has an operating performance record at two utilities, and the numbers are good enough to change how curtailable capacity is treated in a business case.

PG&E’s Flex Connect offers larger distribution-connected loads expedited hookups — roughly four to eight months against a one-to-three-year firm-interconnection wait — in exchange for letting PG&E limit their capacity when the local grid is constrained. A participant links its energy-management system to PG&E’s grid-management system, which issues scheduled or real-time capacity limits based on grid availability. The operating record after two years is the persuasive part: participants have drawn their needed capacity in 90% of hours, with operational loads actually impacted less than 1% of the time. Demand is customer-pulled rather than commission-mandated — five active and roughly 85 prospective sites, typically 2–5 MW with some data-center and manufacturing prospects near 10 MW, mostly EV charging but including small data centers, advanced manufacturing and market-facing storage — and PG&E is now moving to formalize the pilot under commission oversight. Two customers have “graduated” to firm service after about 18 months.

Note that this is a different program from PG&E’s SHARE proof-of-concept reported in the September 4 digest, which is the Google-funded residential-device orchestration pilot. Flex Connect is the large-load interconnection instrument. A utility benchmarking against PG&E should not conflate them.

ComEd is graduating the same approach from pilot to full program, onboarding roughly 50 MW of flexible-interconnection customers per year starting in 2026 and building to about 240 MW by 2027 at the end of its grid plan, by connecting DER under terms allowing occasional, targeted curtailment during peak generation instead of triggering multi-year network upgrades. Its affordability framing is unusually explicit and unusually quotable for an IRP or rate narrative: ComEd evaluates outcomes against a goal that customers’ total home energy costs not exceed 3% of average household income — 6% for space-heat customers — tying interconnection design to an energy-burden target rather than a generic reliability rationale.

The number that should move a business case is PG&E’s <1% curtailment against 90% full-capacity availability. That is the empirical basis for treating curtailable interconnection as firm enough for most commercial load profiles, which is the assumption a deferral case rests on and which has until now been asserted rather than measured. And both programs reinforce the procurement point this series keeps arriving at from different directions: curtailable capacity is only bankable if the utility owns the real-time control layer to call it. PG&E’s program is run out of its “execution operations and distributed energy resource management systems” function. Flexible interconnection and DERMS are complementary investments, not substitutes — the utility can only offer the deal because it can enforce the limit. (Sources: Utility Dive, September 2 — September 15 entry; Renewable Energy World — September 14 entry)

🏗️ Data Centers & Large Load Growth

The Pennsylvania PUC now has an independent, state-sanctioned number for how far PJM’s 2030 reliability misses its own planning criterion, it attributes the miss to data-center additions, and it is the single most citable exhibit this series has acquired for a demand-side filing. The study — Synapse Energy Economics, Mondre Energy and Aspen Technologies, modeling 2027–2030 — finds that under the reference scenario, built on PJM’s own 2026 load forecast, the 2030 loss-of-load expectation reaches 0.59, described as “nearly six times worse” than PJM’s 0.1 LOLE planning criterion, the standard designed to hold shortages to roughly one event every ten years. In a high-load, low-supply “worst-case future,” modeled LOLE hits 13.20 — “over 100 times worse,” implying more than 13 days of loss-of-load events per year. Shortfalls are attributed “largely” to surging data-center additions, and only the no-new-data-center scenario meets the 0.1 target.

Three things make this more useful than the market-analyst load forecasts this series has been citing. First, it is commissioned by a state commission and produced by independent consultants, which means it is admissible in the proceedings where a vendor deck is not — the same reason the CRS report acquired in September was worth flagging. Second, it uses PJM’s own load forecast as the reference input, which forecloses the standard rebuttal that the modeler inflated demand. Third, the scenario structure isolates the variable cleanly: when the only scenario that meets the reliability criterion is the one without new data centers, the policy question stops being whether to build supply and becomes how much of the new load can be made flexible, which is the demand-side question.

The energy-balance finding travels further than the LOLE numbers. Pennsylvania falls from a large net exporter of about 91 TWh in 2025 to roughly 69 TWh by 2035 and 38 TWh by 2040, becoming a ~5-TWh net importer by 2040 in the high-load/low-supply case. A state watching its export surplus erode into an import position on a twenty-year horizon is a state whose IRP filings will start treating demand-side hedges as a resource rather than a program, and that transition is where DSM business cases get their most favorable hearing.

It also arrives in a live regulatory context rather than as a standalone report: the PUC’s September 11 unanimous motions directed staff to propose updates to emergency-curtailment rules and to convene a technical conference on data-center cost allocation. Analysis moving into rulemaking within a month is fast by commission standards, and it means the curtailment terms and cost-allocation principles that emerge in Pennsylvania will be written against these numbers. (Source: Utility Dive, September 15 — September 16 entry)

PJM’s board has proposed making curtailment a condition of service for new large loads, and — the part demand-side planners should read twice — has structured a 6.8-GW backstop auction so that demand-side commitments reduce the amount of supply PJM must buy. The package has three components and each one matters differently.

The backstop auction would procure 6.8 GW to cure the reserve-margin shortfall from the last base auction for the delivery year starting mid-2028 — minus any bilateral contracts, new IRP supply, or large-load demand-response commitments that emerge first. If FERC approves, it runs September 30 to October 21 with 15-year commitments and an offer cap raised to $555/MW-day from $325/MW-day. That subtraction clause is a structural statement about how PJM now values demand-side commitments: they are fungible with procured supply at the portfolio level, which is exactly the substitution demand-side advocates have been arguing for and rarely see written into a procurement mechanism.

The Interim Resource Adequacy Service (previously “connect and manage”) is the demand-side core. New large loads that do not bring their own supply would be required to reduce load or switch to on-site backup when the system nears emergency conditions, starting June 1, backed by a new large-load registry of location, ramp schedule and capacity supply used to set load-reduction priorities. This is the same curtailment obligation the §202(c) orders impose by emergency directive, converted into a standing condition of interconnection — which is the difference between an instrument that builds a telemetry and settlement record and one that does not.

The forecast exclusion is the quietest and most consequential: the board would direct staff to exclude incremental new large loads without new supply from the demand forecasts used in future auctions. Jefferies expects this to lower capacity prices over time. For a utility, it also changes the denominator in any avoided-cost calculation anchored to a PJM load forecast — worth checking before the next filing cycle rather than after.

The counterweight belongs in the same paragraph. PJM’s independent market monitor calculates that keeping data-center load in the capacity auction added $29.4 billion over the last four auctions, and warns that interconnecting loads without adding generation could raise energy costs and reduce reliability. Large loads could grow roughly 70 GW by 2038 across PJM. Curtailment-based adequacy is a bridge, not a substitute for supply — but a bridge with a registry, a priority order and a settlement path is a considerably better bridge than a seventh emergency order. (Source: Utility Dive, July 28 — September 16 entry)

On July 22 roughly 3,800 MW of northern Virginia data-center load tripped offline on a normally cleared fault — the largest such event in PJM history — and the reliability-standards response it triggered will land before anyone has measured how flexible data centers actually are. A normally cleared fault on a 230-kV line in Dominion’s northern Virginia zone caused approximately 3,800 MW of data-center load to disconnect; PJM’s overall load fell about 3.8%, from 99,984 MW to 96,205 MW. Comparable events of roughly 1,500 MW occurred in the same zone in 2024 and 2025. PJM’s operating committee concluded the data centers “should not disconnect from the grid … they are disconnecting too early,” and the grid operator is evaluating expanded interconnection reliability requirements including ride-through standards.

The regulatory clock is the planning input. Under a FERC directive, NERC must finalize by December 31 the registry criteria and initial reliability standards for adding computational loads to the bulk power system, with a March plan for additional standards — but voltage and frequency ride-through requirements are deferred to next year, which PJM warns is too late: “you really need these requirements in place before the load comes on the system.”

Two implications for this series’ threads. First, the same ride-through and voltage/frequency discipline being imposed on inverter-based generation under the NERC IBR standards (full implementation January 1, 2030) is now extending to inverter-heavy computational load — utilities designing DER hosting and flexibility programs in high-growth zones should assume large-load ride-through obligations are coming and write them into interconnection agreements before the standard forces a retrofit. Second, note the sequencing problem this creates alongside the EPRI item below: the standards will be filed on December 31 and the Aurora AI Factory’s demonstration data arrives after that, so computational-load standards will be written on modeling rather than on measurement — and a 3,800-MW trip is the only large-scale empirical datapoint currently on the record. (Source: Utility Dive, August 12 — September 15 entry)

The large-load tariff population is now counted three different ways by three different trackers, and the divergence is not an error — it is three different definitions, which matters when you cite one in a filing. The NC Clean Energy Technology Center and SEPA report through their DELTa database that tracked large-load tariffs and service rules grew from 41 in July 2025 to 104 by July 2026 across more than 70 utilities and 37 states, of which 69 are approved and 35 pending, and 45% specify a large-load threshold of at least 50 MW — aligning with FERC’s June 2026 large-load definition.

Reconcile that against the two counts this series already carries. SEPA’s earlier count was 77 tariffs across 60 utilities in 36 states. Halcyon’s tracker, via the LBNL/Brattle brief reported September 11, catalogued 264 tariff filings as of August 17. These are counting different objects: DELTa tracks tariffs and service rules (the instrument), Halcyon tracks filings (including multiple filings per tariff and pending amendments), and the earlier SEPA count is an earlier snapshot of the DELTa methodology. Cite the 104 figure for how many large-load tariffs exist, the 264 figure for how active the filing docket is, and do not present them as a trend. Getting this wrong in a rate case is the kind of error opposing counsel is paid to find.

The operative datapoint for the demand-side case is unchanged and now better sourced: as of Q2 2026, one-quarter of tracked tariffs include a concrete option for dispatchable large-load flexibility or codify curtailment pathways — most commonly through optional interruptible-service riders (Pennsylvania’s model 50 MW+ tariff tied to PJM’s Emergency Load Response Program; Xcel Minnesota’s interruptible rate for 100 MW+ customers with 3 MW+ controllable load), and also through utility dispatch of behind-the-meter generation or demand-response programs.

The analysis usefully situates tariffs as one of several flexibility pathways rather than the whole field — alongside wholesale-market rules (SPP’s non-firm “CHILL” service, FERC’s §206 show-cause orders), state legislation (Texas S.B. 6; Virginia H.B. 284/S.B. 371 requiring load-flexibility programs for 25 MW+, 75%+-load-factor customers), cross-agency working groups, and bespoke agreements (Georgia Power’s contract flexibility terms; Google’s load-shifting partnerships with OPPD, TVA and I&M). A utility positioning its own DR or interruptible design should know which of those five pathways its state has already used, because the one it has used is the one its commission will expect to see extended. (Source: DSIRE Insight / NCCETC & SEPA, September 1 — September 15 entry)

📋 Regulatory & Policy

A unanimous D.C. Circuit panel vacated the first DOE Section 202(c) order compelling a coal plant to delay retirement, closing a watch item this series has carried since June and establishing that the federal reliability-emergency tool has a legal ceiling. The panel vacated DOE’s order forcing Consumers Energy’s 1,420-MW J.H. Campbell plant to delay its May 31 retirement, writing that it was “unpersuaded by DOE’s sweeping conception of its ’emergency’ authority” and that the order usurped state authority over generating resources.

The scale of what the ruling puts in question: 10 generating units across six plants have been ordered to stay online, at a running cost the Sierra Club’s tracker puts past $547 million, borne by ratepayers — for units that are largely uneconomic, several of which ran at capacity factors well below 50% or produced no power at all, with deferred maintenance raising the cost of compliance.

The ruling does not end the regime. DOE can seek rehearing, continues reissuing 90-day orders elsewhere, and — as this week’s Order 202-26-45 demonstrates — continues issuing load-side orders to PJM on a different theory. But the reasoning cuts at the core of the generation-retention application, and the strategic read for IRP and capacity planning is direct: if federal authority to compel uneconomic supply-side retention is legally constrained, the avoided-cost case for demand-side and distributed alternatives strengthens as the durable, litigation-proof path to resource adequacy. A commission weighing a demand-side program against “just keep the plant running” now has an appellate decision saying the second option is not reliably available. (Source: Utility Dive, September 11 — September 15 entry)

DOE issued its seventh Section 202(c) emergency order to PJM of 2026 on September 17, and the count is now a planning input rather than a news item. Order No. 202-26-45, effective September 17 through the end of September 18, authorizes PJM to activate backup generation and dispatch specified resources across 13 states and Washington, D.C., reaching more than 35 GW of backup generation during a late-season heat event. The order is consistent with the earlier 2026 PJM directives allowing curtailment of large loads of at least 50 MW at a single delivery point as a last resort before an Energy Emergency Alert Level 3.

Put the year’s record in one place, because it is the strongest single argument this series can make for planned demand-side capacity: 202-26-06, -23 (May 18, three days, reserves projected under 5,800 MW), -32/33 (June 30, 15-minute backup-gen compulsion), -41 (September 1), -43 (Duke Carolinas, September 3–8) and now -45 (September 17–18) — seven orders to PJM in one calendar year plus one in an adjacent footprint, spanning three seasons. Against that: the load-side instrument has been warned twice, never escalated to an Action, and has dispatched no customer generation, while PJM has declined to disclose how many megawatts of backup generation were even available and has not adopted the 50 MW threshold into its governing documents.

So the honest characterization for any proceeding where emergency authority is offered as a reliability answer is unchanged and now better supported: it is invoked routinely, it has produced no performance record on the load side, and it was just narrowed by an appellate court on the generation side — while metered demand response shed 6,113 MW in PJM’s July heat event with settlement-grade measurement behind it. (Source: Discovery Alert, September 2026 — September 18 entry; order confirmed by direct search)

California re-certified the machinery that converts a demand-side program into a dollar figure, and any utility benchmarking avoided costs in this IRP cycle should be reading the 2026 update rather than the prior vintage. The CPUC approved the 2026 Avoided Cost Calculator update on September 3, 2026 in Decision D.26-09-007, followed by a Draft Calculator Workshop on September 9. The update refreshes the value streams that determine demand-side cost-effectiveness: avoided generation energy, generation capacity, ancillary services, transmission and distribution capacity, and decarbonization-policy compliance.

Why this matters outside California. The ACC remains the reference implementation of a rigorous, regulator-blessed avoided-cost framework — the transparent, documented version of the calculation that every other jurisdiction performs less formally. The values are jurisdiction-specific and reflect California’s decarbonization mandates, so they do not travel; the method does, and it is the method that out-of-state IRP advocates benchmark against. A 2026 refresh means the benchmark is current for this filing cycle.

The structural point for planners is the linkage: the ACC is directly tied to IRP procurement strategy, and the 2024–2026 IRP cycle treats the IOUs’ existing shed (conventional, peak-event) demand-response programs as baseline resources. That means a change in the calculator’s T&D-capacity or generation-capacity components moves the entire net-benefit calculation for a DR portfolio — including the kind of illustration this series works from, where a 70-MW portfolio defers roughly $150M in CapEx at about $20M program cost. If your business case cites an avoided-cost figure derived from ACC methodology, the derivation should be re-run against the 2026 update before the next filing. (Source: CPUC — September 18 entry)

FERC’s six simultaneous Section 206 show-cause orders have anchored this series’ large-load thread since June as background; here is the primary-source account, because the fourth reform category is the federal analogue of everything this digest tracks at the utility level. On June 18, 2026 the Commission issued six contemporaneous orders to show cause under Section 206 of the Federal Power Act — one each to PJM, CAISO, MISO, NYISO, ISO-NE and SPP — each preliminarily finding the RTO’s tariff unjust and unreasonable with respect to large-load integration, and each giving the RTO/ISO and its transmission owners 60 days to justify why current tariffs remain just and reasonable or to file tariff changes, plus a 30-day report on how it intends to ensure resource adequacy for existing and new large loads.

The five reform categories tee up exactly the levers a DSM business case depends on: efficient study processes; cost-shifting and transparency safeguards; accommodation of co-location and behind-the-meter generation; new transmission services for flexible large loads; and a process to study generation serving electrically proximate large loads. The fourth category is the one to watch — it is the direct federal analogue of the utility-level flexible-interconnection pilots in this digest (PG&E Flex Connect, ComEd), and if it produces standardized curtailable-interconnection terms at the wholesale level, DR and VPP resources acquire a clearer market home than the state-by-state patchwork they currently occupy.

All five commissioners concurred. That bipartisan unanimity makes the resulting reforms considerably more durable than a §202(c) emergency order — which, as this week’s D.C. Circuit ruling demonstrates, can be vacated. FERC acknowledged regional differences and did not order PJM and SPP to respond on co-location issues, which they had already addressed. Response deadlines were extended to mid-November for several RTOs; MISO’s ZGIA proposal (filed August 18, supported September 8 by Google, Xcel, AES Indiana, Ameren, Dairyland, SEIA and Advanced Energy United) is the first substantive product of the process. A targeted search this week surfaced no consolidated public account of what was filed in the ~September 16 tranche; the mid-November record remains the place these terms get set. (Source: FERC, June 18 — September 17 entry; response status confirmed by direct search)

🔬 EPRI Research Spotlight

A direct check confirms no new major EPRI publication entered the research log this week — the tenth consecutive dry week. A targeted search of EPRI’s media resources this week surfaced nothing newer than the DCFlex nine-site expansion and the Aurora materials this series has carried since July. As established a month ago, this is now reported as a finding about publication cadence rather than as a coverage omission: EPRI’s substantive data-center-flexibility output is landing through the DCFlex program and partner announcements rather than through standalone research releases. A utility relying on EPRI publications as its flexibility-research feed is looking in the wrong place — the material is in the program’s demonstration coalition and its partners’ press channels.

The DCFlex thesis is unchanged and this week’s items make its central gap sharper. The DCFlex Initiative and its FlexMosaic framework — organized around flexibility classes keyed to notification time, duration, frequency, depth and speed of response — conclude that data centers willing to be flexible can maximize their “speed to power,” directly addressing the five-to-seven-year interconnection queues throttling load growth. Pilots with NVIDIA and Emerald AI software have delivered up to 40% load flexibility by pausing or slowing training workloads and redirecting inference queries to less-stressed systems. The coalition spans Compass Datacenters, Constellation, Emerald AI, Google, National Grid, Nebius, NVIDIA, Oracle and PADO AI across nine demonstration sites including Europe. The linchpin, per the program’s own framing, is a standardized binding agreement between utilities and data centers, now forming through EPRI and the Open Compute Project Foundation.

FlexMosaic remains the only published vocabulary precise enough to price this week’s four separate flexibility products against one another, and this week the list got longer rather than shorter:

Virginia’s 450-MW dual-enrollment VPP — an aggregated, dispatchable resource with two possible counterparties against the same meter.
PJM’s proposed IRAS — a curtail-or-switch-to-backup obligation triggered near emergency conditions, prioritized by a registry.
The interruptible riders in a quarter of large-load tariffs — bilateral, tariffed curtailment with utility-specific notification and duration terms.
The §202(c) backup-generation transition — a 15-minute compulsion with no compensation, no telemetry requirement and no settlement record.

Those are four different response characteristics, procured in four venues, with no shared grammar. A utility that has committed capacity in more than one of them cannot currently answer the question “how much flexibility do we actually have, and under what notification?” without a mapping exercise, and FlexMosaic’s notification-and-duration keys are the only published structure that performs it. The PJM DR 24/7 availability change effective June 1, 2027 adds a fifth axis — continuous obligation versus event-based response — that tariff language generally does not handle.

Aurora’s timing problem is unchanged and now fully consequential. The coalition — NVIDIA, Emerald AI, EPRI, Digital Realty and PJM — still targets late 2026 for the 96-MW Manassas, Virginia facility, having originally slated it for the first half of the year; this week’s verification search surfaced nothing indicating either acceleration or further slip. NERC’s computational-load standards are due at FERC by December 31, which means the reference design’s demonstration data will very likely arrive after the standards are filed. Set that beside the 3,800-MW July 22 trip reported in this digest and the position is stark: the only large-scale empirical evidence on the record about how data centers behave under grid stress is an event in which they disconnected too early, and the standards will be written before the evidence that they can also be made to respond usefully. The coalition estimates the reference design, adopted nationwide, could unlock roughly 100 GW on the existing system.

Standing watch: Aurora commissioning data and any further slip; the CLO-001-1 comment record after the September 18 close (today) and the fall ballot results; FlexMosaic uptake in the ER26-3515 and RM26-4 compliance records; whether the EPRI/OCP standardized utility–data-center agreement produces publishable contract language; and an EPRI.com direct check again next run — eleventh week.

🚩 Utility-Sector Relevance Flags

⚑ Virginia Refused the Make-vs-Buy Fork — Dual Enrollment Is Now a DERMS Requirement, Not a Policy Preference
Topic: VPP Program Design / DERMS Procurement / Aggregator Interfaces
Relevance: Virginia’s mandate directs Dominion to file a VPP tariff by November 15 and a pilot by December 1 for up to 450 MW across multiple regions, with customers enrolling either directly with the utility or through a third-party aggregator. Every other jurisdiction this series tracks picked a side: Minnesota approved a utility-owned VPP and declined a parallel third-party program; the Sunrun–Renew Home–Tesla 16-GW framework is aggregator-only with no utility counterparty; New Jersey has not yet decided. Virginia’s hybrid means telemetry, dispatch and settlement must work identically across two enrollment channels against the same meter — and the pilot concludes July 1, 2028, after which the SCC must develop a permanent program with procurement targets and M&V metrics.
Action Signal: Implement — Write dual-channel enrollment into the DERMS requirements document as a demonstrable capability with acceptance criteria: standards-based aggregator interfaces (IEEE 2030.5, IEEE 1547) as a pass/fail requirement rather than a scoring bonus, and a settlement path that can attribute performance to either counterparty. A platform built for utility-direct enrollment can be retrofitted for aggregators only at the data model, which is the most expensive place to discover the gap. If your state has not yet chosen a VPP structure, Virginia is now the design that avoids having to choose.

⚑ The Pennsylvania LOLE Numbers Are the Cleanest Adequacy-Gap Citation This Series Has Acquired
Topic: Resource Adequacy / IRP Evidence / DR Business Case
Relevance: A Pennsylvania PUC-commissioned study by Synapse, Mondre and Aspen models PJM’s 2030 LOLE at 0.59 under a reference case built on PJM’s own 2026 load forecast — about six times the 0.1 planning criterion — and at 13.20 in the high-load/low-supply case, over 100 times worse, implying >13 loss-of-load days per year. Only the no-new-data-center scenario meets 0.1. Pennsylvania’s net exports fall from ~91 TWh (2025) to ~38 TWh (2040) and to a ~5-TWh net import position in the worst case. The PUC’s September 11 motions moved emergency-curtailment rules and data-center cost allocation into active proceedings within a month of the study.
Action Signal: Implement — Put this in the standing evidence file and cite it in preference to vendor or trade-association load forecasts: it is state-commissioned, independently produced, and built on the RTO’s own forecast, which forecloses the “you inflated demand” rebuttal. Use the scenario structure rather than just the headline number — the argument that lands is the only modeled future that meets the reliability criterion is the one without the new load, which makes flexibility of that load the entire question. If you file in PJM, quantify your own demand-side portfolio as a fraction of the modeled gap.

⚑ California’s Avoided Cost Calculator Was Just Refreshed — Re-Run Any Derivation Built on the Prior Vintage
Topic: Avoided-Cost Methodology / DR Cost-Effectiveness / IRP Filing Hygiene
Relevance: The CPUC approved the 2026 ACC update on September 3 in Decision D.26-09-007, with a draft calculator workshop September 9, refreshing avoided generation energy, generation capacity, ancillary services, T&D capacity and decarbonization-policy compliance — the five streams that determine demand-side cost-effectiveness. The ACC is directly linked to IRP procurement strategy, and the 2024–2026 cycle treats existing shed (peak-event) DR as a baseline resource, so a change in the T&D-capacity or generation-capacity component moves DR net-benefit math directly. Values are California-specific; the method is the transferable part and remains the reference implementation regulators outside California benchmark against.
Action Signal: Implement — Audit every avoided-cost figure in the current business-case library for which vintage of ACC methodology it derives from, and re-run anything anchored to a pre-2026 version before the next filing. This is unglamorous filing hygiene, but an intervenor who notices that your avoided-cost derivation uses a superseded calculator gets to argue your whole cost-effectiveness case is stale, and the fix is a spreadsheet exercise rather than an argument.

⚑ Seven Emergency Orders in One Year, and an Appellate Court Just Narrowed the Authority
Topic: Emergency Authority / Planning Assumptions / Filing Strategy
Relevance: Order 202-26-45 (September 17–18) is the seventh §202(c) order to PJM in 2026 — after -06, -23, -32/33, -41 and the Duke Carolinas order -43 — authorizing backup-generation activation across 13 states and D.C., reaching >35 GW of backup generation. In the same month, a unanimous D.C. Circuit panel vacated the first generation-side order (the 1,420-MW J.H. Campbell retention), finding DOE’s “sweeping conception” of emergency authority usurped state generation authority; the broader regime — 10 units at six plants — has run past $547M in ratepayer cost. Meanwhile the load-side instrument has been warned twice, never escalated to an Action, dispatched no customer generation, and PJM has neither disclosed available backup megawatts nor adopted the 50 MW threshold into its governing documents.
Action Signal: Engage — Plan against recurrence, not exception: seven orders across three seasons in one footprint is a plannable pattern, so build curtailment obligations, behind-the-meter dispatch visibility and backup-generation telemetry into large-load interconnection agreements now, while the terms are still negotiable. In any proceeding where emergency authority is offered as the reliability answer, state the accurate characterization: invoked seven times, never tested on the load side, and narrowed by the D.C. Circuit on the generation side — against metered DR that shed 6,113 MW in PJM’s July event with settlement-grade measurement.

⚑ PJM’s IRAS Would Make Curtailment a Condition of Service — and Demand-Side Commitments Shrink the Backstop Buy
Topic: Capacity Procurement / Large-Load Interconnection / Demand-Side Substitution
Relevance: PJM’s board proposes a one-time 6.8-GW backstop auction (September 30 – October 21, 15-year commitments, offer cap raised to $555/MW-day from $325) explicitly reduced by bilateral contracts, new IRP supply, or large-load demand-response commitments that emerge first — a procurement mechanism that treats demand-side commitments as fungible with supply. The Interim Resource Adequacy Service would require new large loads without their own supply to curtail or switch to backup near emergencies from June 1, backed by a registry of location, ramp schedule and capacity supply setting reduction priorities. The board would also exclude incremental new large loads without new supply from the demand forecasts used in future auctions — which Jefferies expects to lower capacity prices, and which changes the denominator in any avoided-cost calculation anchored to a PJM forecast.
Action Signal: Engage — Two concrete actions. First, if you serve large loads in PJM, get into the IRAS registry design: location, ramp schedule and capacity-supply data determine reduction priority, which is the difference between your enrolled DR being called before or after an unsupported data center. Second, stress-test any avoided-cost figure anchored to a PJM load forecast against the forecast-exclusion change — a lower forecast lowers the deferral value your demand-side case is claiming, and it is better to model that yourself than to have it modeled for you.

⚑ Flexible Interconnection Now Has an Operating Performance Record — 90% of Hours, Under 1% Curtailed
Topic: Flexible Interconnection / Hosting Capacity / Deferral Case
Relevance: PG&E Flex Connect cuts interconnection for larger distribution loads from a 1–3-year firm wait to roughly 4–8 months in exchange for utility-imposed capacity limits when the grid is constrained — and after two years, participants have drawn needed capacity in 90% of hours with operational loads impacted less than 1% of the time. Pipeline: 5 active, ~85 prospective sites at 2–5 MW (some near 10 MW), mostly EV charging plus small data centers, manufacturing and storage; two customers have graduated to firm service. ComEd is moving the same model from pilot to program — ~50 MW/year to about 240 MW by 2027 — against an explicit affordability standard that total home energy costs not exceed 3% of household income (6% for space heat). Note Flex Connect is not the Google-funded PG&E SHARE pilot reported September 4; do not conflate them when benchmarking.
Action Signal: Implement — The <1% curtailment against 90% availability figure is the empirical basis for treating curtailable interconnection as firm enough for most commercial load profiles, which is the assumption every deferral case rests on and which has until now been asserted rather than measured. Put it in the hosting-capacity business case, and pair it with the procurement consequence: PG&E can offer the deal only because it runs the program out of its DERMS function. Budget flexible interconnection and DERMS together — they are one investment with two line items, and a curtailment right you cannot enforce in real time is not bankable.

⚑ The ADMS Incumbent Is Absorbing DR Execution While the Independent Sells Interoperability — Settle Suite-vs-Best-of-Breed Before Requirements
Topic: DERMS/ADMS Procurement / Architecture Decision / Vendor Strategy
Relevance: Oracle added a DER-orchestration module that auto-builds demand-response and DER-event dispatch strategy templates to its Utilities NMS, alongside generative-AI storm forecasting and crew-dispatch automation — on a platform serving six of the top 10 U.S. utilities and >61 million customers, which makes template-driven DR dispatch inside the control room a de facto reference architecture. Survalent moved oppositely, extending SurvalentONE DERMS via IEEE 2030.5 and launching the Helix time-series historian for real-time DER behavior assessment and pre-emptive constraint-violation detection, explicitly aimed at cooperative and municipal utilities. The market frame: ADMS $3.85B (2025) → $19.80B (2035) at 20.2% CAGR, with DERMS ≈17% of it.
Action Signal: Implement — Decide the architecture question before writing requirements, because the two paths produce incompatible requirement sets: the incumbent-suite path optimizes for native integration and single-vendor accountability, the standards-based path optimizes for aggregator interoperability and exit. Whichever you choose, make data governance, open APIs and model-and-data exit terms commercial terms in the contract rather than schedule exhibits — a unified platform concentrates capability and dependency in the same place. For co-ops and munis, note that the interoperability tier is now actively courting your segment, which it was not two years ago.

⚑ Three Trackers, Three Counts of the Same Tariff Population — Cite the Right One
Topic: Large-Load Rate Design / Comparables / Filing Accuracy
Relevance: DELTa (NCCETC/SEPA) counts 104 large-load tariffs and service rules as of July 2026 — up from 41 a year earlier — across 70+ utilities and 37 states, 69 approved / 35 pending, with 45% using a ≥50 MW threshold aligned to FERC’s June definition. Halcyon, via the LBNL/Brattle brief reported last week, counts 264 tariff filings as of August 17. The earlier SEPA 77 count is a prior snapshot. These count different objects — instruments versus filings — and presenting them as a time series is an error an opposing expert will find. The substantive finding is stable across all three: one-quarter of tracked tariffs now carry a concrete dispatchable-flexibility option or codified curtailment pathway, typically optional interruptible riders (PA’s model 50 MW+ tariff tied to PJM’s ELRP; Xcel MN’s 100 MW+/3 MW-controllable rate).
Action Signal: Engage — Use 104 as the population of existing tariffs when defending your own terms, 264 as evidence of filing activity, and never blend them. More important, check which of the five flexibility pathways your state has already used — tariffs, wholesale-market rules (SPP CHILL, FERC §206), state legislation (TX S.B. 6, VA H.B. 284/S.B. 371 mandating load-flex programs for 25 MW+, ≥75% load factor), cross-agency working groups, or bespoke agreements (Georgia Power; Google with OPPD, TVA, I&M) — because the pathway your commission has already used is the one it will expect you to extend.

⚑ Computational-Load Ride-Through Standards Will Be Written Before Anyone Measures Flexibility
Topic: Large-Load Reliability Standards / Interconnection Terms / Timing Risk
Relevance: Roughly 3,800 MW of northern Virginia data-center load tripped offline on July 22 on a normally cleared 230-kV fault — PJM load fell 3.8% (99,984 → 96,205 MW), the largest such event in PJM history, following ~1,500-MW events in the same zone in 2024 and 2025. PJM’s operating committee: the data centers “are disconnecting too early.” NERC must file computational-load registry criteria and initial standards with FERC by December 31, with a March plan for more — but voltage/frequency ride-through is deferred to next year, which PJM warns is too late. Meanwhile the 96-MW Aurora reference design that would supply demonstration data on useful flexibility is still targeting late 2026, after the filing date. The only large-scale empirical record is a mass disconnection.
Action Signal: Watch — Assume large-load ride-through obligations are coming and write them into interconnection agreements for new computational load now, while the terms are negotiable and before a standard forces a retrofit at the customer’s expense and your schedule. Track the CLO-001-1 comment record (closed September 18) and the fall ballot — and note the detail this series flagged when the draft posted: CLO-001-1’s modeling-data obligations land on the utility, not the data center. Budget the modeling work.

⚑ The Flexibility-Platform Layer Consolidated — Treat It as a Contract-Review Trigger
Topic: Vendor Risk / DSM Outsourcing / Contract Terms
Relevance: Octopus Energy closed its majority investment in Uplight on September 1 (agreement signed March 24), with Schneider Electric retaining a significant minority stake and contributing One Digital Grid DERMS and DER-orchestration capabilities; Uplight remains nominally independent. Targets: $1B in customer savings and more than doubling flexible capacity to >20 GW in five years. Uplight is the customer-facing enrollment and engagement layer behind a large share of U.S. residential and small-C&I DR programs, so a change in its ownership and capitalization reaches directly into utility DSM operations. The 20-GW target is enrollment-dependent and exposed to the early-adopter plateau analysts have been flagging all year.
Action Signal: Watch — At next renewal, confirm three things specifically: data-portability and model-export terms (who owns the enrollment and performance history); whether the Schneider relationship creates a preferred-integration path that narrows your independent DERMS options; and which entity holds the customer relationship of record. Better capitalization in this layer is on balance good news for DR tooling — but it concentrates vendor risk in a tier that sits between the utility and its own customers, and that is a contract question rather than a technology one. Do not model the 20-GW target as committed capacity.

⚑ Two States Now Owe a Regulator a Number for Avoided Distribution Investment
Topic: DER Valuation / M&V-to-Procurement / Evidence Pipeline
Relevance: This series has argued for months that the crux of the DER business case is putting a defensible number on avoided or deferred distribution investment — the value stream utilities have historically been unable to quantify. Two commissions have now ordered it produced. Minnesota requires Xcel to establish specific grid-benefit estimates for Capacity*Connect by its November 2027 Integrated Grid Plan, with quarterly reports before then, for up to 200 MW / $430M of batteries sited at strategic locations to relieve local constraints (with MISO revenues repaying nearly the full cost and a $50M Google contribution). Virginia requires the SCC to assess its 450-MW pilot’s peak-demand grid-service value after July 1, 2028 and to develop a permanent program with procurement targets and performance metrics. Massachusetts adds a third data point of a different kind: a September baseline inventory against a 3.5-GW by 2035 demand-management target (~13% of New England’s 26.1-GW peak).
Action Signal: Watch — These are the two dates that will eventually settle whether locational DER siting earns its premium, and they are worth calendaring now: November 2027 (Minnesota) and post-July 2028 (Virginia). In the interim, structure your own DER pilots so they can answer the same question — that means recording the deferred project, its cost and its planned in-service date at enrollment, not reconstructing it afterward. A utility that can produce a distribution-deferral number before a commission orders one has an argument nobody else has.

📌 Sources

September 14, 2026 Entry
– GlobeNewswire / SNS Insider — Advanced Distribution Management Systems Market Size to Hit USD 19.80 Billion by 2035, at 20.2% CAGR (September 9, 2026)
– Renewable Energy World — ComEd Drives Utility Affordability With Flexible DER Interconnections (2026)
– New York Department of Public Service — Commission Improves Customer-Centered Electric Demand Response Programs (April 2026) (DLM program suite previously reported in the August 28 digest; carried here for the Con Edison / Central Hudson Bring-Your-Own-Battery rollout only)
– Mass.gov — Executive Order to Secure Massachusetts’ Energy Future (Commonwealth of Massachusetts; March 13, 2026 order, September 2026 baseline report)

September 15, 2026 Entry
– Utility Dive — Court Rejects DOE ‘Emergency’ Order Delaying Coal Plant Retirement as Overstep (Ethan Howland; September 11, 2026)
– Utility Dive — PG&E Sees Rising Interest in ‘Customer-Driven’ Flexible Interconnection Pilot (Brian Martucci; September 2, 2026)
– DSIRE Insight — Stretching the Possibilities: Where Large-Load Tariffs Fit in the Future of Data Center Flexibility (Justin Lindemann, NCCETC & Ann Collier, SEPA; September 1, 2026)
– Utility Dive — PJM Eyes Data Center, Crypto Reliability Requirements After 3.8 GW of Load Trips Offline (Ethan Howland; August 12, 2026)

September 16, 2026 Entry
– Utility Dive — PJM’s Reliability Could Start Failing by 2030 if Data Center Growth Continues: Report (Diana DiGangi; September 15, 2026)
– Canary Media — Xcel Minnesota Is Building a First-of-its-Kind Virtual Power Plant (Jeff St. John; April 8, 2026) (previously reported in the September 4 and September 11 digests; carried here as a standing benchmark, not re-reported)
– New Jersey Board of Public Utilities — NJ Proposes “Virtual Power Plant” Plan to Save Ratepayers Millions Annually, Docket QO26030099 (July 27, 2026) (fourth consecutive digest; carried in one line of context)
– Utility Dive — PJM Board Proposes Backstop Capacity Auction, Data Center Curtailment Plans (Ethan Howland; July 28, 2026)

September 17, 2026 Entry
– PR Newswire — Survalent Launches Helix Time-Series Historian and Expands DERMS Capabilities to Enable Unified Operational Insight and DER Coordination (September 2026)
– PR Newswire — Oracle Empowers Utilities With New Advanced Distribution Management System Enhancements (September 2026)
– FERC — FERC Launches Aggressive Targeted Action to Speed Large Load Integration (June 18, 2026) (referenced as context in prior digests; reported here from the Commission’s own notice for the first time)
– Canary Media — Tesla, Sunrun, Renew Home Team Up on Massive 16GW Virtual Power Plant (June 24, 2026) (carried as an open watch item since June; first full treatment)

September 18, 2026 Entry
– Discovery Alert — DOE Issues 7th Emergency Grid Order to PJM in 2026, Order No. 202-26-45 (September 2026)
– California Public Utilities Commission — DER Cost-Effectiveness / Avoided Cost Calculator (Decision D.26-09-007, September 3, 2026)
– Utility Dive — Virginia Utility-Scale VPP Pilot Mandate Is First Amid National Push (2026)
– GlobeNewswire — Octopus Energy Completes Investment in Uplight, Accelerating Plans to Save US Customers $1 Billion and Double Grid Flexibility (September 1, 2026) (same transaction reported in the September 11 digest from a different outlet; carried for the closing mechanics only)

Verification Sources (targeted search, not from the daily research log)
– U.S. Department of Energy — Federal Power Act Section 202(c): PJM Interconnection, L.L.C., Order No. 202-26-41 (202-26-45 confirmed by direct search as issued September 17, 2026, effective through September 18; DOE’s own order page for -45 not yet indexed at time of writing)
– FERC — PJM Governance and Stakeholder Processes (proceeding page; end-of-September deadline still open, Dispute Resolution Services forum convened September 1)
– McGuireWoods — FERC Issues Section 206 Show Cause Orders Directing All Six RTOs/ISOs to Justify or Reform Large Load Integration Rules (June 2026; 60-day response, 30-day resource-adequacy report)
– EPRI — DCFlex Initiative Expands to Nine Demonstration Sites Across U.S. and Europe (most recent substantive EPRI release; tenth dry week confirmed by direct check)

Compiled from the DER Research Log, Monday September 14 – Friday September 18, 2026. Twenty daily items consolidated; four cross-week duplicates carried as movement or standing benchmarks rather than re-reported; three verification items chased by targeted search.