Grid Alpha

LYU LLC DBA Grid Alpha
  • 0,0 (0)
  • Tuto
  • Tous les jours

Grid Alpha turns real-time data from all nine North American power markets (ERCOT, PJM, CAISO, ISO-NE, NYISO, MISO, SPP, AESO, IESO) into short, trader-ready signal. Each episode reads the tape: fuel mix, LMP/DART spreads, congestion, storage response, and LinkedIn commentary from analysts, developers, and policy watchers who actually move size. No background music, no fluff, just the setups that matter this week for U.S. power and gas traders. Live dashboards at gridalpha.us.

  1. -7 h

    CAISO Prints a $1,945 Nodal Spread Before Sunrise

    CAISO's real-time market printed a $1,945.09/MWh nodal spread in the 13:55 UTC dispatch interval on Tuesday, August 4. Foster Plant's APND node cleared at $1,032.30/MWh while Bishop Alamo printed -$912.79/MWh, inside the same five-minute solution. Translated to local time, this was 06:55 PDT on a Tuesday morning, hours before solar output peaks. The decomposition tells the story. Foster's price carried zero congestion: a $1,000.00/MWh energy component plus $32.30 in losses. All five highest-priced nodes that interval, Foster, Grn Pete, Leaburg, Chandler, and Hills Cr, showed the identical structure: congestion of $0.00, energy of $1,000.00, with losses alone separating the prints. That is a system-wide energy print, not a local transmission constraint. Whether the $1,000 figure reflects CAISO's soft offer cap or an administered scarcity state is unconfirmed; no ISO notice in the 48-hour window explains it. The bottom of the stack is the mirror image. Bishop Alamo, Bishop Units, Contrl Poole, and Contrl QF all printed near -$912/MWh, each built from congestion components near -$944/MWh against a $27.82 energy component. Same interval, same market: one zone paying cap-level energy, another zone charged nearly $1,000 to generate. The negative leg was not isolated to that pocket. Forty minutes earlier, at 06:15 PDT, the Diablo Canyon-adjacent CSADIAB7N001 node printed -$183.67/MWh on a -$223.20 congestion component. The grid-wide curtailment backdrop is heavy: CAISO logged at least 1,632 curtailment events in the trailing seven days, with a single-interval solar sample of 920.6 MW curtailed on July 28. But those are CAISO-wide figures, not node-level data, and no fact in hand ties curtailment megawatts to the Bishop or Contrl pods at 13:55 UTC. The angle's "forced-curtailment zone" framing is a plausible read of the -$944 congestion print, not a confirmed one. What I'd watch: persistence first. The snapshot is a single interval; whether the $1,945 spread held across hours or collapsed by 07:05 determines whether this was a dispatch artifact or a tradable regime. Second, the constraint driver at Bishop and Contrl. A congestion component that large implies a binding element or derate, and the OMS or ISO market notice explaining it will also say how long it runs. Third, day-ahead behavior: if DA prices at these nodes anticipated nothing, real-time convergence trades and virtuals at the pods get interesting fast; if DA saw it coming, the money was already made. Fourth, any repeat of negative prints near Diablo, which would widen the geographic footprint of the negative leg beyond the eastern pods. > A $1,000 energy component at one end of the state and a $944 congestion penalty at the other is not one market clearing; it is two markets sharing a timestamp. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

  2. -1 j

    Long Island Carries the Widest Hub Premium on a Quiet Saturday Tape

    LONGIL printed a $112/MWh real-time peak at the 00:00 UTC bar, 8pm Saturday local, against a 24h mean of $62. That is a $50 spread above the average on a weekend session, with the day's low of $45 set just twelve hours earlier at the 12:00 UTC bar, 8am Saturday morning. The $67 peak-to-trough range was the widest of the five NYISO hubs on the board. The ramp was system-wide in timing, if not in magnitude. All five listed hubs printed their 24h peaks at the same 00:00 UTC bar and their lows at the same 12:00 UTC bar, which reads as a clean evening pickup off a Saturday morning floor rather than anything localized. Long Island simply sat on top of the stack. Its peak edged N.Y.C. by $1, CAPITL by $2, CENTRL by $6, and WEST by $8. The same ordering held in the averages: LONGIL's $62 mean ran $1 above N.Y.C., $3 above CAPITL, and $5 above both CENTRL and WEST. The peak-to-mean markup of $50 matched N.Y.C. and CAPITL, ahead of CENTRL at $48 and WEST at $47, so the Long Island premium was a level shift across the whole curve rather than a one-bar spike. N.Y.C. itself peaked at $111 with a $61 mean and a $66 intraday spread, keeping the two downstate hubs within a dollar of each other on both peak and average. What the tape does not say is why. The source data is hub LMP series only: no load, no weather, no outage, no transmission data, so no driver for the LONGIL premium can be established from this window. The day-of-week reading is at least confirmed; derived timestamps put the peak at Saturday evening and the trough at Saturday morning local time, consistent with a muted weekend shape. A $112 Saturday evening print is a modest ramp by summer standards, and the tight $1 spread to N.Y.C. at the top of the stack argues against any acute downstate separation. But without constraint or load data, the persistent $3 to $8 premium over the upstate hubs is a pattern, not an explanation. What I'd watch: whether the LONGIL-to-WEST mean spread, $5 on this window, holds into the Sunday and Monday sessions or compresses back toward the N.Y.C. parity it nearly touched at the peak. If the premium persists on weekday load, the case for a structural downstate constraint strengthens; if it fades with the weekend, this was just Long Island sitting at the expensive end of a quiet stack. The DA-RT spread at the 00:00 UTC bar is the other open item, since a $50 peak-to-mean markup that day-ahead already priced carries very different information than one it missed. > When every hub peaks in the same bar and Long Island still finishes on top, the premium is in the geography, not the hour. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

  3. -1 j

    Long Island's $291 Friday Ramp Sets the Weekend Basis Map

    LONGIL printed $291/MWh at 21:00Z Friday, 17:00 local, against a 24h mean of $84.26/MWh; the snapshot series shows $290.59/MWh at the same timestamp, so treat the handle as rounding, not disagreement. The low came at $44.80/MWh at 10:00Z Saturday, leaving a $246/MWh intraday spread across the 24-hour window. That is a ramp print, not a plateau: a tight evening peak sitting on an otherwise ordinary mean. The machinery matters more than the headline handle. A peak at 21:00Z with the trough the next morning says scarcity arrived when downstate demand was still working through the evening and relaxed before the Saturday morning shoulder. In NYISO, that pattern usually points to interface pricing into Long Island, local reserve pressure, or shortage adders rather than a clean statewide energy move. The fact pack cannot prove it: the recomputation explicitly carries no load, weather, outage, or transmission data, and the cadence caveat is real because the snapshot and recomputation describe the window with different bar treatment. What the pack can prove is relative stress. LONGIL's $291 peak and $246 spread were the widest among the hubs tracked, above PJM DOMINION's $250 peak and SPP SPPSOUTH_HUB's $238 peak. That keeps the story regional and downstate, not a broad Eastern bid. The missing leg is Zone J. If NYC printed with LONGIL through the same evening block, the premium reads as downstate capacity scarcity and the weekend risk sits in both zones. If NYC stayed near the mean while LONGIL ripped, the cleaner read is congestion across the interface and the tradeable variable is basis, not outright energy. Day-ahead behavior is the second leg: if day-ahead LONGIL failed to clear anywhere near the real-time peak, Friday punished weak ramp anticipation and puts DART around the evening boundary on review. If day-ahead did clear elevated and real-time still separated, the issue was not forecast error alone but deliverability into the evening constraint. For the weekend tape, watch repetition before narrative. If Saturday and Sunday hold a LONGIL premium into the same 21:00Z block while upstate remains anchored near the mean, downstate basis stays convex and NYC-zone spreads can gap at the evening boundary even without a heat headline. If the evening print fades back toward the $84.26 mean, Friday becomes an isolated scarcity event and the carry argument weakens fast. Either way, the relevant mark is not the $291 print by itself; it is whether the market keeps paying for the same hour, the same interface, and the same uncertainty after the first warning. As of the Monday afternoon snapshot clock, treat any fresh reading as confirmation or refutation, not as a fresh signal. > A $246 LONGIL range with no confirmed driver is where weekend basis risk hides: scarce enough to respect, unexplained enough to fade only with evidence. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

  4. -1 j

    POD_GALE Pays $649 as CAISO Real-Time Nodal Spread Hits $680

    CAISO's real-time market printed a maximum nodal LMP spread of $679.70/MWh in the 09:55 PDT snapshot Monday morning. At the top of the stack, PODGALE1SR3SR3-APND cleared $649.38/MWh; at the bottom, HPLNDJT6_N001 printed negative $30.31/MWh in the 09:15 interval. Same system, same energy component near $35/MWh, a $680 gap between them. The decomposition tells the story. At PODGALE, $615.24/MWh of the $649.38 print was congestion; the energy component was $34.97. PODSEGS1SR2SL2-APND showed the same anatomy: $647.35/MWh LMP on $613.49 of congestion. On the other side, HPLNDJT carried a congestion component of negative $64.84 against $34.55 energy, and the solar nodes confirmed the pattern: PNCHVS2SOLAR printed negative $29.47 and negative $29.42 with congestion of negative $63.45, while both KRAMER_1 nodes sat at negative $20.02 on negative $52.48 of congestion. When the entire spread lives in the congestion component and the energy price barely moves, the ISO is not short of power; the transmission system is short of paths. Curtailment data backs the supply-glut side of that read: CAISO logged at least 1,596 curtailment events over the prior seven days, with one sample interval showing 1,421 MW of solar curtailed. Two caveats before sizing anything. The high and low prints come from different intervals, 09:55 versus 09:15 PDT, so the $679.70 figure is the market's reported maximum for the snapshot, not necessarily a single matched pair. And the binding constraint behind the roughly $615/MWh congestion at PODGALE and PODSEGS is unidentified in the data; no outage or flow-gate cause is confirmed. Whether the $649 print has any day-ahead counterpart is also unknown, so treat the DA/RT basis at these nodes as an open question, not an input. That leaves the trade framed by the angle: a spread this wide, this early in the morning, is a setup rather than a conclusion. If solar output keeps climbing into midday, expect the negative prints at PNCHVS and Kramer to deepen or curtailment to absorb the surplus; either way the southern congestion components stay pressured. The real test is the evening ramp: if the POD_GALE congestion persists past solar-down, the basis trade is structural for the day and the evening hours reprice higher; if it collapses once curtailment relief arrives, the morning print was a transient and the wide nodal basis is a fade. Watch the congestion components, not the LMPs. The energy leg at $35 is inert; everything that moves today moves through the constraint. The $470 million in Q2 gross benefits CAISO reported for EDAM and WEIM participants is the reminder that spreads like this are exactly what the expanded western footprint is supposed to arbitrage away, and has not yet. > A $680 spread with a $35 energy component is not scarcity; it is a grid telling you where the wires end, and the evening ramp decides whether that map is worth trading. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

  5. -4 j

    AEP's 45 GW Queue Bet Concentrates ERCOT Basis Risk

    American Electric Power carried 45 GW into the ERCOT batch interconnection process, the largest single-utility block in the cycle. The number landed the same day AEP and Xcel reported second-quarter earnings into a down tape for utility equities. What the source material does not carry is any breakdown by fuel, by load versus generation, or by zone, which is exactly why 45 GW is hard to price. Today's grid offers no read-through. The binding-constraint sample from the past seven days is dominated by low-voltage elements with a thin tail: LAKENASAMATH11 at 69 kV posted the largest sampled shadow price at 36.478, SANAFMR1 at 138 kV printed 5.245, and the remainder of the sample collapsed to 0.272 at VALEXP and 0.006 at both OLINGRFMR1 and DIESELFMR1. Units are not stated in the feed, so treat the ratio rather than the level. Supply-side attrition was similarly quiet: MLSES lignite ran forced against 805 MW max capability, though the feed carries two conflicting records for the same 2026-07-27 05:02 stamp, one showing a 65 MW effective reduction and one showing 45 MW, and LACYCRK wind logged reductions of 0 MW and 5 MW. ERCOT booked at least 1,716 unplanned outage events and at least 7,847 SCED shadow-price records on the week, both lower bounds because the queries returned truncated samples. A quiet tape with a data-quality asterisk is not the same as a clean one. The queue is where the risk accumulates, and it accumulates unevenly. If AEP's block skews toward large load rather than generation, the exposure shows up as import-constrained pockets inside one utility's Texas footprint and as basis that widens against the hub during transmission build-out lag. If it skews toward generation, the same footprint becomes export-constrained and the basis sign flips. Texas regulators' 30 July approval of AI data center co-location beside a wind farm is the variable that decides which: the order requires rapid curtailment during grid emergencies while limiting participation in demand response, so co-located load enters the picture as something between firm nodal demand and an uncompensated interruptible. Replicate that template across a few gigawatts of the 45 and the nodal load shape becomes state-dependent rather than deterministic. Two adjacent data points frame the cost side. Entergy's $1.8B Cottonwood purchase in Texas is described as increasingly in doubt, which says something about what the market will pay for existing steel versus queue position. And gigawatts of data center load tripping offline in Dominion's territory, producing over-frequency and voltage problems, says something about what happens when the load side of an interconnection queue actually energizes. What I would watch: whether ERCOT publishes the batch composition by zone and fuel, and whether the 69 kV and 138 kV constraints that currently price near zero start clustering in the AEP service territory. That clustering, not the headline gigawatts, is the first tradeable signal. Until then the 45 GW is a denominator problem, since the total batch size is not disclosed and the "share" cannot be confirmed. > Forty-five gigawatts of queue position is simultaneously a claim on future congestion rent and an invoice for the transmission that relieves it; without fuel, zone, and node, it prices as neither. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

  6. -5 j

    The Loudest Number on CAISO This Week Was 1.686 MW

    The largest curtailment in the sampled CAISO window was 1.686 MW of wind, 1.538 MWh, at 05:00 PDT on Thursday, July 23. Solar entries in the same sample round to zero. The system logged at least 1,668 curtailment events over the trailing seven days, a count that is sampled and probably truncated, but the magnitudes tell the story: nothing on this grid is straining. One caveat on the calm. No CAISO LMP, congestion, or ancillary-service price data accompanied this run, so treat today's quiet as inferred from curtailment prints rather than measured off the spread. That pushes the tradable content into the rulebook. On July 29, CPUC staff recommended the commission reject proposed changes to a CAISO demand response program, arguing the revisions would leave ratepayers "worse off." The public headline does not name which program: PDR, RDRR, DRAM, and the DR ancillary-services work all sit under overlapping CPUC and CAISO jurisdiction, and neither the MW of price-responsive load nor the RA credit at stake is disclosed. Staff recommendations are non-binding, and no vote date is public. Note the direction of travel against the rest of the continent: New Jersey's BPU is pushing a $68 million virtual power plant program aimed at cutting peak demand 3% by mid-2027, and an E3 analysis pegs bidirectional EV charging at 15 times the value of one-way managed charging, roughly $7 billion in unlocked value. California is litigating whether its DR fleet is worth expanding while other jurisdictions are writing checks for the same megawatts. The load forecast is why this matters more than a July Thursday suggests. BNEF's chip-based model puts US data center demand near 118 GW by 2030 and as high as 207 GW by 2033. PJM has been directed to file a large load registry and a Reliability Backstop Procurement auction proposed for October 2026 precisely because supply-side capacity procurement does not close that gap on its own, and Brookfield and NextEra just announced a $100 billion campus at DOE's Paducah site with up to 4.6 GW of dedicated generation attached, which is the industry conceding that grid-supplied firm capacity is the binding constraint. Price-responsive load is the cheapest instrument that shows up inside an operating hour. Rules that shrink it shift more of the tail onto the energy price. What I would watch is two-sided and slow. If the commission adopts the staff position and the revisions die, the question is whether CAISO's enrolled DR capacity contracts outright or simply stops growing, and the tell is RA counting and enrollment MW in the next compliance filings rather than anything that prints in the day-ahead. If the commission departs from staff, the read-through is a wider DR-eligible pool and a marginally softer upper tail on CAISO summer peaks, worth a few basis points of implied vol on on-peak forwards at most, but structural. Until the docket names the program and the MW, this is a directional flag, not a position: the missing numbers here are the whole trade. > Nobody trades a demand response rulemaking on a 1.686 MW day, and that is exactly when the next scarcity print gets written. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

  7. -6 j

    Calm Grids, Loud Rulebooks: Large Loads Set the Trade

    IESO's real-time shadow prices printed zero across every sampled 5-minute interval on July 22, the signature of a grid with nothing binding. Nothing on the physical side handed traders a dislocation this week; the action sits in the rulebooks that will price the next decade of demand. PJM's Board directed the RTO to file two large-load proposals at FERC, one laying out the structure and schedule for a Reliability Backstop Procurement auction that PJM proposes to hold in October. Ontario is moving faster on process than on paper: IESO has begun imposing reliability standards on large data centers through connection assessments, months before it expects to fold them into formal market rules. A single sampled PJM daily record on July 23 showed 14,726 MW forced out alongside 13,561 MW in maintenance, one entry in a 147-record set whose full distribution is not yet public. The demand these regimes are built to absorb is not small: BNEF's chip-based base model puts U.S. data center load near 118 GW by 2030 and 207 GW by 2033. In Texas, CenterPoint expects 50% load growth by the end of 2029 and flagged 14 GW as likely eligible for ERCOT's Batch Zero large-load interconnection, while New Jersey works the other end with a $68 million virtual power plant plan targeting a 3% peak-demand cut by mid-2027. Watch FERC's handling of the PJM filings against the October backstop timeline: approval on schedule pulls a new capacity procurement into the forward frame, while a slip leaves the large-load registry as the only near-term lever on interconnection risk. If IESO's connection-assessment standards harden into explicit curtailability terms, the marginal Ontario data center reads as a dispatchable demand block rather than firm load, and that reshapes how you carry peak risk north of the border. Between BNEF's 2033 trajectory and CenterPoint's 2029 read, the open question is not whether the load arrives but which market fixes the price of its reliability first. > With no spread to trade today, the position is a bet on which regulator's pen clears first. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

  8. 28 juil.

    IESO Moves to Codify Its 100-MW/Minute Battery Ramp Cap

    Nothing on the nine grids cleared a volatility threshold this week: IESO's real-time shadow-price feed printed near-uniform zeros across the trailing seven days, with a lone -40.55 excursion (units unconfirmed in the feed) on the morning of July 21. On a tape this quiet, the actionable item is regulatory. IESO plans market rule amendments in 2026 to codify a 100-MW/minute ramp limit on battery facilities, formalizing a constraint it has imposed on batteries registering with the market since roughly the second quarter of 2025. The sequencing matters as much as the number. The limit has operated for over a year as an administrative condition attached at registration rather than as a market rule; the 2026 amendments would retrofit the rulebook to match existing practice, and RTO Insider reports the move is already raising market-governance questions in Ontario. Mechanically, a ramp cap targets the one attribute that separates storage from a fast thermal peaker: near-instantaneous response. At 100 MW/minute, most facilities can still traverse their full operating range inside a five-minute dispatch interval, so the cap bites hardest on sub-interval response, on speed-sensitive ancillary products, and on any future project sized in the several-hundred-megawatt range. The operability logic is not mysterious; a large battery swinging its full output in seconds is a step change the balancing authority has to absorb through AGC and area control error, and IESO evidently decided it would rather cap the swing than carry the exposure. The unresolved details will do the real work: whether the limit applies per facility or across a portfolio under common control, whether batteries registered before mid-2025 are grandfathered, and how many megawatts of Ontario storage sit under the constraint today. None of that has surfaced yet, and neither has a quantified revenue impact. The governance question runs alongside the technical one. A constraint imposed at the registration desk and codified afterward inverts the usual rule-change sequence, and how Ontario's stakeholder process handles that inversion will tell other ISOs something about the path of least resistance for constraining storage. The amendment text is the watch item. If the codified limit reaches ancillary offer quantities as well as energy dispatch, Ontario's storage revenue stack compresses relative to markets that leave ramp rate unconstrained, and strategies built on beating slower assets to a scarcity print lose their edge exactly where that edge is worth the most. If the rule instead lands per-facility with grandfathering, the effect migrates upstream into new-build sizing: developers will scale units to the point where 100 MW/minute stops binding within a dispatch interval, and the constraint will show up in project pro formas years before it shows up in an LMP. Either way, watch the spread between how Ontario storage prices in capacity and financing markets versus comparable assets in jurisdictions that let batteries move as fast as their inverters allow. > A battery sells speed; codify a speed limit and you have repriced the asset class, not just the megawatt. Not investment advice. For informational purposes only. Grid Alpha — daily briefs · live dashboards, all nine markets · public call record

Tout afficher (41)

À propos

Grid Alpha turns real-time data from all nine North American power markets (ERCOT, PJM, CAISO, ISO-NE, NYISO, MISO, SPP, AESO, IESO) into short, trader-ready signal. Each episode reads the tape: fuel mix, LMP/DART spreads, congestion, storage response, and LinkedIn commentary from analysts, developers, and policy watchers who actually move size. No background music, no fluff, just the setups that matter this week for U.S. power and gas traders. Live dashboards at gridalpha.us.

Informations

  • Création
    LYU LLC DBA Grid Alpha
  • Années d’activité
    2026
  • Épisodes
    41
  • Classification
    Tous publics
  • Copyright
    © LYU LLC DBA Grid Alpha
  • Site web de l’émission
    Grid Alpha