The power flow solves the grid with the load in place (AC, generator reactive limits on). Then every line and transformer within three hops of the point of interconnection is taken out one at a time, plus any outage the statewide DC screen flags, and the grid is solved again each time. A failure is a line over its emergency rating or a bus outside its voltage band, under that outage.
A planned outage (one element out for maintenance, off-peak) followed by an unplanned one. N-2 takes both out at once; N-1-1 lets the operator re-dispatch plants between the two, as ERCOT Planning Guide 4.1.1.8 allows. Run on the system with the proposed upgrades applied. Adds 2 to 5 minutes.
NERC P3: the largest generating unit near the point of interconnection is out, then each single line or transformer outage on top. Tests whether the load leans on that plant. Run on the system with the proposed upgrades applied; no upgrade is proposed for a failure here.
NERC P6-2: a large (EHV) transformer near the point of interconnection is out, then each single outage on top. Run on the system with the proposed upgrades applied; no upgrade is proposed for a failure here.
PROXY: ACTIVSg2000 has no structure data, so this treats two circuits on the same bus pair (at least 0.5 miles) as sharing a tower; a single event, no N-1 stacked on top.
Base case: ACTIVSg2000, August 2018 MATPOWER release, representing a summer peak condition. This adds the off-peak case (pre-existing load and generation scaled down; this load stays at its full request), future-year cases (loads near the point of interconnection grown to the in service year and two years after), and a small sensitivity band on dispatch and the voltage criterion. Runs in the background after the report; the PDF is refreshed when it ends.
One point of interconnection, a load that steps up over several years (MW is the total at that year). The study runs at each step, and each upgrade gets a need-by year: the first step at which it is needed. The network is the same snapshot for every year; only the load changes. Firm capacity only, at the power factor above.
How the outages are chosen
Generator outages are not part of this step: the loss of the largest nearby plant is its own check (P3).
Buses are coloured by the MW they can take with no upgrade, by the DC screen (generator/load colour is off in this view, so free MW stays the one thing the colour means). In short: the largest new load before a monitored line or transformer hits its limit, intact or after any single outage (six ties whose loss islands a bus are left out) — thermal only, no voltage, rounded down to 5 MW.
Full methodology: how it's computed, the three screening rules, what it doesn't see, and how far to trust it (issue #149).
Ranked at the kV levels checked above; check or uncheck a level up there to change both the map and this ranking.
Each hour: the forecast sets renewable output caps and system load for that hour, a security constrained DC OPF clears the market against those and every monitored N-1 outage, and that hour's flows, LMPs and N-1 watch list fall out of the solve.
Open-Meteo's own seven day forecast, at its own sample points, for the hour on the scrubber above -- map overlays only (issue #186); the dispatch run itself solves against ERCOT's own published forecast instead. Off by default.
ERCOT's own published forecast is checked against what ERCOT actually recorded, for the same hours. Wind and solar are compared as a percent of ERCOT's own installed capacity. Load is compared by its shape (how it rises and falls through the day), not by megawatts. Lower numbers are better.
Until issue #186, this also scored a reconstruction of Ops's own former weather-curve model the same way, since dispatch used to run on it; that "model" column is retired now that dispatch runs on ERCOT's own forecast directly (see the Ops panel's own forecast status above).
Weather forecast in, hourly flows and an N-1 watch list out, over a window you pick above (the next 12 hours by default, from today's forecast) — extend to 24 h, 3 days or 7 days, or sample 4 stress hours a day across the full week. Branches on the map are coloured by loading for the hour on the scrubber, the same bands as Study a point.
The hour timeline below steps through the forecast: drag it, click a point on it, or focus it and use the arrow keys. Its track shows each hour's own status; branches on the map colour by that hour's loading (green/yellow/red) the moment it is solved.
The N-1 watch list is the binding and over-limit contingencies the security constrained dispatch found for the hour on screen.
A planned outage takes a line, transformer, or generator out of service for a range of hours; the hour on screen re-solves around it right away.
Binding: the line sits exactly at its limit and the dispatch was rearranged to keep it there. The shadow price is what 1 MW more room on it would be worth.
Relaxed: the dispatch could not keep it within its limit, so the flow goes over and the excess is priced at ERCOT's shadow price cap: $5,251 per MWh with every line in service, or $4,500, $3,500 or $2,800 after a trip, by voltage. That is a real violation.
Rows are the binding and relaxed contingencies the security constrained dispatch produces for this hour. Hover a row to find it on the map.
Simulated line, transformer and generator outages for a range of forecast hours. With Add outages by clicking the map on, click a line or transformer to add it right away, out from the hour on screen through the end of the run — adjust that window (or the label) in the list below afterward. Turn it off to inspect the map without adding anything; hovering a line or transformer always shows its popup either way. Type a name or id below to add a generator, or any element by hand (line:1234, trafo:12, gen:3, sgen:7); an unknown id is rejected with a reason. Adding, editing or removing an outage here takes effect on the hour above immediately; Rerun dispatch (using the run selector above) refreshes the precomputed run the status strip and summary read.
The network is synthetic; the weather and the dispatch method are real. Tracked as issue #81.
The same dispatch run Ops solves, read for prices: hub prices and spreads, binding constraints and their shadow prices, and renewable curtailment behind the lines that cause it. The grid and every constraint/hub name below are synthetic — the same solve on the real ERCOT model would produce the real rows.
Each hub price is the simple average over its own 230 and 500 kV substations of each substation's average LMP (ERCOT prices a hub the same way, over its 345 kV Hub Buses only, never its local load buses), a hub's buses being those nearest one of ERCOT's 5 real hub cities (North/Dallas, South/San Antonio, West/Odessa, Houston, Panhandle/Amarillo) — a modelling convenience (A trading_hub_bus_assignment_rule), not ERCOT's own settlement points.
Binding: the line sits exactly at its limit and the dispatch was rearranged to keep it there. The shadow price is what 1 MW more room on it would be worth.
Relaxed: the dispatch could not keep it within its limit, so the flow goes over and the excess is priced at ERCOT's shadow price cap: $5,251 per MWh with every line in service, or $4,500, $3,500 or $2,800 after a trip, by voltage. That is a real violation.
Hover a bus for its LMP this hour. The colour toggle recolours every bus by price, instead of generator/load; the label toggle adds each hub's own average on top, at its anchor city, for a reading that survives a quick glance.
The network is synthetic; the weather and the dispatch method are real. Tracked as issue #89, part of issue #81.
Trip a line and watch where its power goes. DC power flow: MW only, voltages not modelled.
| Power flow (load flow) | Economic dispatch (ERCOT: SCED) | |
|---|---|---|
| The question | The plants are set: where does the power go? | Which plants run, and how much, at the lowest cost, keeping lines within their limits where it can? |
| A new MW comes from | A reference plant (the slack) or a fixed rule | The cheapest plant the lines allow |
| Outages | Run one by one, to see what overloads | Built in: the outages included in the model are checked before any of them happens |
| A line over 100% | Can happen after a trip: nothing moves to prevent it | Only on paper, when no plant can help; the overload pays a capped penalty, which sets the price |
| Prices | None | A price at every bus |
| Speed | Close to instant | Much slower: it re-solves as outages bind |
| Who runs it | Planners, for interconnection studies | The operator, every five minutes; traders read its prices |
| In the app | Power flows | Dispatch and prices |
Click a line to take it out or put it back. Click a bus, or search for one, to add a load, change a load or change a plant.
Two ways to solve, and how they differ
Plants stay at their outputs, except the ones you set. The plant picks up the difference. No prices.
The slack is the reference plant that picks up whatever the power flow does not assign. Run a power flow to see which plant it is.
Re-dispatch decides plant outputs; your plant edits become availability: out of service, or a lower maximum. Gives prices and the limits held at their max.
Add demand and see which plants change output, what limits the dispatch, and how local prices change. The calculation checks line flows with everything in service and after the outages included in the model.
A monitored line can read “N MW now · M MW under contingency (loss of X) · at its limit”. N is the flow now. M is the flow the model predicts on that line if X were lost: a prediction, not a flow happening now. The operator sets the plants before any outage so that these predicted flows stay within their limits, for the outages included in the model. When no plant can keep one there, the line is shown over its limit on paper and priced at a capped penalty; demand is still fully served.
The plants and their costs are synthetic: the fleet is a stand-in for Texas, with costs closer together than in the real market. A swap between two plants of similar cost is therefore close to arbitrary; the constraints and the shape of the prices are what carry the lesson.
A financial transmission right (FTR) pays, each hour, the price at its sink minus the price at its source, for every MW. In this view that difference comes only from the limits the dispatch holds at their max: a limit adds to the price on its high side and not on its low side, so a path pays when it runs from the low side to the high side of a limit, and costs money when it runs the other way. Priced for the case on screen, load level, line outages and added load included, for one hour.
Click any line for its numbers.
Click a line to take it out, or a bus to change its generation. Then Solve. To add a load and see who supplies it, use Dispatch and prices.
Generation follows demand in merit order: the most expensive plants back off first.
Click any line for its numbers.
Read this first. This is a screen of public data. Published hosting capacity is a utility model estimate, is not a commitment to serve, and does not show that any load can be energized. No candidate has been reviewed by the utility. Available MW, upgrade cost and energization time are unknown for all of them.
This tab shows data from the utility's published map and open sources. It has not been verified with the utility.
A 2,000-bus synthetic Texas power grid you can explore in the browser, released one chapter at a time. Real geography and real physics, a fictional network (Texas A&M ACTIVSg2000).
The chapters use different methods. Flows uses a DC power flow: MW only, voltages are not modelled. Study a point uses an AC power flow, with voltages, and tests single outages near the point of interconnection; it shows four preset studies, not a study of any point you click. On this site every result was solved ahead of time by this project's own engine; nothing is calculated in your browser. Every number that is not in the model file comes from a sourced assumption.
Model: Texas A&M ACTIVSg2000. Solver: pandapower with lightsim2grid.
This site counts visits anonymously (no cookies).
A demo that runs a utility large load interconnection study live on the Texas A&M ACTIVSg2000 synthetic Texas grid: 2000 buses, real geography and real physics, a fictional network. Pick a point of interconnection and a size, click through the five study steps or ask for the study in plain English, and get a study report as a PDF shaped like an ERCOT one.
Study a point runs the study at one bus. Find capacity colours every high voltage bus by the MW it can take with no upgrade, by a DC screen, and ranks interconnection points for a load near a pin. Ops is coming.
Every number that is not in the model file comes from a sourced assumption and carries the marker A; hover it for the source and the confidence. The report lists them all in an appendix.
Model: Texas A&M ACTIVSg2000. Solver: pandapower with lightsim2grid.
These public chapters use preset examples. In the private app, you can choose an outage, change the grid, and run the power flow again.
Get new chapters by email:
Subscribe on Substack