• Markets: ERCOT, Texas and NORD, Italy
  • Backtests on real prices, May 2025 to Aug 2026
  • Read-only. We never switch your equipment
International Batteries
Submit site data
  1. Home
  2. Platform

How the platform works

The console connects a site's data to market prices and rules, shows where the money goes, and plans every hour. This page shows the screens, the method and the data behind them.

DatasheetConsole
Markets
ERCOT (Texas), NORD zone (Italy)
Users
Energy managers, finance, plant operators, producers
Plan horizon
48 hours, solved again every hour
Solver
Mixed-integer program (HiGHS)
Mode
Advisory, read-only
Exports
CSV, JSON and a printable report

The screens

Six views of one site

All screenshots come from the working console, on the modeled reference AI campus and a modeled Texas solar plant. Prices are real; site figures are modeled and labelled as such.

Six figures for the reference AI campus on 16 May 2025: consumption 2,033 MWh, on-site production 1,147 MWh, average hub price $136.16 per MWh, total energy cost $86k, opportunity versus rule $6,654, grid imports 785 MWh.
Fig. 2The day's six figures. "Opportunity vs rule" is modeled, and the console says so under the number.

The overview answers three questions: what the site used and made today, what it cost, and what the plan did differently from the usual operating rule. A one-line brief sums up the day in plain words.

Stacked chart of supply by hour over 48 hours: on-site generation, CHP cooling, grid import, battery and curtailed load, with the grid price as a dotted line peaking near $1,570 at 19:00.
Fig. 3Supply and price for 16 and 17 May 2025. Right of the dotted line is the forecast.
Decision for 19:00: curtail 6 MW, with a 24-hour strip coloured by decision.
Fig. 4The decision for one hour, and the strip for the whole day.

Each hour gets one decision: run on site, buy from the grid, mix the two, or curtail flexible load. The plan shows why. At 19:00 the grid cost more than the $1,000/MWh a megawatt-hour of compute is worth to the site, and the engines were already at full output.

Waterfall of the yearly bill: grid power $7.00M, delivery charges $1.90M, transmission (4CP) $177k, engine gas $11.42M, O&M and starts $4.07M, battery wear $114k, lost compute $257k, total $24.95M.
Fig. 5How the yearly bill came to be, from grid power to lost compute.
Chart of what moves the bill: change in the annual energy cost by driver.
Fig. 6What moves the bill: gas, grid prices, peak charges and weather, each with its share.
Map of the ERCOT grid with load zones shaded by stress, power plants by fuel, large loads, and 144 small producers.
Fig. 7ERCOT by load zone: stress, plants by fuel, large loads and 144 small producers from EIA-860M.
Zone stress: each ERCOT load zone's load as a share of its 12-month peak.
Fig. 8Zone load against its 12-month peak.
Energy flow into the reference campus: grid, on-site generation and CHP cooling.
Fig. 9Where the site's power came from in the chosen hour.
Regulation view: Texas grid and interconnection rules for the site, each marked applies, check or pending, with the statute, effective date and a source.
Fig. 10Texas rules checked against the site's own numbers, each with its statute, date and source.

Texas changes its large-load rules often. The console tracks 15 of them, from SB 6 curtailment directives to ride-through requirements, and marks each one as applying, to check, or pending, using the site's own size and equipment.

What your MWh earn: capture price $22.62 per MWh against a flat average of $31.67, a shape discount of 29%, by month.
Fig. 11Capture price against the flat average for a modeled Texas solar plant.
Average modeled output by hour of day with the average price by hour.
Fig. 12Output and price by hour of day: why solar earns less per MWh.
Three ways to sell the same modeled output over a year, compared.
Fig. 13Three ways to sell the same output, compared over a year.

Method

How the plan is made

A forecast, an optimizer and a rule for re-planning. The same engine runs the backtests on this website.

  1. 00:00 daily

    Forecast

    Gradient-boosted models forecast 48 hours of real-time prices from data known at the time, including ERCOT's day-ahead price for the next 24 hours. Quantile models give P10, P50 and P90 bands.

  2. Each solve

    Optimize

    A mixed-integer program decides engine on and off, output, battery and curtailment to minimize energy cost plus lost production, within the grid limit and every unit's limits.

  3. Every hour

    Re-plan

    With the hour's actual price known and the forecast for later hours, it solves again and applies only the current hour.

  4. June to September

    Peak charges

    Expected transmission charges are added to grid imports in the hours most likely to set each month's ERCOT peak.

What every plan respects

Plans that ignore how plants actually run look good on paper and fail on site. These limits are in the model.

  • Start costs, minimum up and down times, and warm-up: a unit gives only part of its output in the hour it starts.
  • Minimum stable load, part-load efficiency and ramp limits.
  • Process heat from CHP or an auxiliary boiler, with dumped heat counted.
  • Permit hours and penalties on emergency-only units.
  • Battery power, energy and wear.
  • Carbon costs, white certificates and monthly peak power charges for Italian sites.
  • Units out for maintenance.

One limit of the current backtest, stated plainly: ERCOT's archived load forecasts were not available, so the peak-hour risk in the backtest uses actual load with 2% noise. A live service needs forecasts made only from data known at the time, and we will rerun the peak-charge results on those before we quote them to you.

Data

What the console runs on

Public market and plant data are real. Sites, their equipment and loads are modeled until we work with your own data.

DataSourceCoverageStatus
Hub prices, real-time and day-aheadERCOT, HB_HUBAVGMay 2024 to Aug 2026, hourlyReal
System load and generation by fuelEIA-930HourlyReal
Grid carbon intensityCalculated from EIA-930 fuel mixHourlyEstimate
Gas priceEIA, Henry Hub spotDailyReal
WeatherNOAA airport stations via Iowa Environmental MesonetHourly, 19 Texas stationsReal
Small producers in ERCOTEIA-860M via PUDL144 plants up to 100 MWReal
Italian prices and fuelsNORD day-ahead, PSV gas, EU carbon allowancesHourly and dailyReal
Sites, equipment and loadsInternational Batteries reference modelsBuilt from public specificationsModeled
Your meter, bills and equipmentYou, read-only15-minute or hourlyYours

Texas market prices: Source: ERCOT. ERCOT does not endorse this website. EIA data is public domain.

Next step

See the plan on your own site

Send a meter export and an equipment list. We build your site model and replay your history against your current operating rule.

Submit site data Ask for a call

Create a free website with Framer, the website builder loved by startups, designers and agencies.