Published by the official operator or scientific mission named, and reproduced unchanged.
METHODOLOGY AND SOURCES
Every number on this site can be taken apart.
This page is the long form of the labels beside each figure: what Irven measured, what it calculated, what it estimated, at what resolution, under which model version, and how far each of those can be trusted.
MAP DOCUMENTATION
The method behind each tool.
Initial map coverage focuses on Hermosillo. These documents describe sources, dates, assumptions and limits for each analysis. The electrical inventory does not confirm capacity for a new load; consumption and loss scenarios are not field measurements.
- Industrial shortlisting
- Urban planning and constraints
- Residential consumption and coverage
- Circuit energy balance and losses
- Transformer reference comparisons
- Radial power flow scenarios
Technical documents are available in Spanish, in Markdown format.
01THREE TIERS
A measurement and an estimate are different claims.
Arithmetic on observed values under assumptions stated next to the figure. No fitted coefficients.
An Irven estimate. Directional, carries an uncertainty range, and is never a measurement.
Every published estimate carries seven things: the value, its interval, the source, the date, the resolution, the model version and the assumptions behind it. Trust goes up when a system admits uncertainty instead of performing precision.
02METHOD
Six rules the analysis is held to.
Baseline before saving
A saving only exists against a documented baseline. Irven builds one from metered data and states the period, the independent variables and the adjustments before any measure is proposed.
Uncertainty is published
Every estimate carries a range, a confidence level, a source, a date, a resolution, a model version and its assumptions. Where data is thin the range widens and the interface says so rather than hiding it behind a single number.
Models are inspectable
Inputs, coefficients, weights and assumptions sit next to the result. A weight nobody can see is a weight nobody can argue with.
Observed, derived, modelled
Three tiers, labelled everywhere. A measurement, arithmetic on a measurement and an estimate are different claims and are never shown as the same thing.
Savings are evidenced
Savings are compared with a documented baseline and adjusted for weather, production and occupancy, so the reported result can be traced to its data and assumptions.
Read-only by default
Measure, recommend, obtain authorisation, then execute. Autonomous control of HVAC, pumps, storage or critical process comes later, if at all — not in a first version.
03HOW THE MODEL WORKS
The demonstration index, in full.
The opportunity index in the demonstration dashboard is deliberately simple. It exists to show what an inspectable model looks like — not to value a site, which no national dataset can do. Its four inputs, their weights and its limits are published here in the same form Irven publishes a client model. Current version 0.3.0.
| Component | Weight | Basis |
|---|---|---|
| Solar yield | 0.30 | NASA POWER global horizontal irradiance at the state centroid, converted to specific yield. |
| Cooling load | 0.30 | NASA POWER cooling degree-days above 18.3 °C: the share of the year a building must reject heat. |
| Energy price | 0.25 | Mean day-ahead zonal energy price in the state’s CENACE control region. |
| Grid dependence | 0.15 | Regional net imports as a share of regional demand: load a region cannot serve locally. |
Normalisation and uncertainty
Each component is min–max normalised across the 32 states, so the index ranks states against each other within Mexico and carries no absolute unit. The score is the weighted mean of the available components, renormalised over the weight that could be evaluated.
The uncertainty band starts at ±8 points — the irreducible doubt in a four-input index built from point samples — and widens by 0.4 points for every percentage point of weight that could not be evaluated. Confidence is reported as high when all four inputs are present, medium above 70% of weight, and low below it.
The load-shifting calculation
The peso figure beside the price chart is arithmetic, not a model. A flat 1000 kW reference load is priced across the published day-ahead zonal curve for 23 Sept 2026. Then 20% of that load is moved out of the 4 dearest hours into the 4 cheapest, and the difference is priced at the same published rates.
A flat 1 MW load over one market day, settled at the published zonal price. Real sites are not flat and most pay a regulated tariff instead; this isolates the price signal, nothing more.
What this model cannot tell you
- Climate is a 20-year climatology sampled at one point per state, not a state average and not this year’s weather.
- Demand, generation and interchange are published per CENACE control region. Regions follow transmission topology, not state borders, so a state coloured by a regional figure carries its region’s value.
- Prices are one published market day for one load zone per region. A single day is not a seasonal average.
- The index is min–max normalised across the 32 states: it ranks states against each other and says nothing about absolute return.
- Nothing here describes any particular site. Site-level work starts from the site’s own meters, invoices and equipment.
Solar potential is four different quantities
Geographic potential is the resource falling on a place. Technical potential applies usable area, orientation, shading and a performance ratio — here 0.78, covering temperature, soiling, wiring, inverter and availability losses, against reported real-world values of roughly 0.75 to 0.85. Economic potential adds tariff, capital, cost of money and self-consumption. Executable potential adds structure, electrical capacity, interconnection and permits.
An algorithm can identify a promising roof. It cannot sign off the structure, the switchgear or the interconnection, and a model that reports geographic potential as though it were executable is how oversized projects get financed.
04EARTH OBSERVATION
What satellites can settle, and what they cannot.
NASA POWER
Irradiance, temperature, humidity and wind, as 20-year climate normals and historical series. Used to normalise a weather year out of an operating problem, and to size solar resource.
Used on this site.ECOSTRESS
Land surface temperature at high resolution. Used for heat stress, urban heat island and the territorial side of cooling analysis.
Client work.NASA Black Marble
Nighttime radiance as a proxy for human activity, urban expansion and change detection — supporting the spatial downscaling of demand.
Client work, with the caveat below.Overture Maps
Building footprints as a base layer, complemented with OpenStreetMap, authorised cadastre, LiDAR, aerial imagery and the client’s own records, to establish usable roof area and context.
Client work.CENACE
Demand, generation, interchange and market prices. Aggregate data calibrates territorial models; a specific installation still needs its own invoices and meters.
Used on this site.
METHODOLOGICAL CAVEAT
Nighttime radiance is not an electricity meter. Black Marble and VIIRS measure light, not load. Radiance is a proxy variable and only carries information inside a multimodal model combining buildings, use, population, industry, climate and observed consumption, with the estimate constrained so that the sum across cells reconciles with observed system demand. Irven does not use it as a substitute for measured data, and no figure on this site is derived from it.
05SCIENCE
The work this rests on.
Causal-effect inference on the relationship between nighttime radiance and electricity consumption.
Why a single national relationship between light and load does not hold across city types.
Downscaling behaviour and error at the scale a portfolio screening actually operates on.
The reference method for separating geographic, technical and economic rooftop potential.
Assessing rooftop development potential from imagery at regional scale.
The urban metabolism frame behind Irven’s long-range view of energy, water, carbon and money.
Climate normals matter to a financial model for one reason: comparing this year’s bill to last year’s without adjusting for weather and production is the most common way a saving gets claimed that never existed. Cooling degree-days above 18.3 °C are the standard way to express how hard a building must work to reject heat, and they make sites in different climates comparable.
06DATA SOURCES
Everything this site cites.
The site is a static export: these datasets were captured when it was built and are served as plain JSON, not fetched from a visitor’s browser. The snapshot is regenerated by npm run data:refresh, which also rewrites its own provenance document.
Instantaneous demand, generation and interchange for the nine control regions and the national interconnected system.
- Captured
- 24 Sept 2026, 20:53 UTC
- Cadence
- Published roughly every 5 minutes.
- Terms
- Public information under Mexican transparency obligations.
Hourly settled demand and generation against the day-ahead forecast, nationally and per control region.
- Captured
- 24 Sept 2026, 20:53 UTC
- Cadence
- Hourly, over the operating day.
- Terms
- Public information under Mexican transparency obligations.
Precios de energía por zona de carga, Mercado del Día en Adelante (opens in a new tab)
CENACE — Sistema de Información del Mercado
Hourly zonal energy prices in MXN/MWh, decomposed into energy, losses and congestion.
- Captured
- 24 Sept 2026, 20:53 UTC
- Cadence
- One market day per run, published the afternoon before.
- Terms
- Public area of the market information system.
POWER climatology (opens in a new tab)
NASA Prediction Of Worldwide Energy Resources
Solar irradiance, air temperature and cooling degree-days at each state centroid.
- Captured
- 24 Sept 2026, 20:53 UTC
- Cadence
- 20-year climatology, revised with each POWER release.
- Terms
- Freely available; NASA requests acknowledgement of the POWER Project.
State boundaries, simplified and projected for the national map.
- Captured
- 24 Sept 2026, 20:53 UTC
- Cadence
- Versioned releases.
- Terms
- Public domain.
Households, electricity access and appliance ownership at AGEB level, used in the Hermosillo map.
- Captured
- 2026-09-13
- Cadence
- Decennial.
- Terms
- Open data under the Ley del Sistema Nacional de Información Estadística y Geográfica.
NASA POWER release v2.10.0, 20-year Meteorological and Solar Monthly & Annual Climatologies (January 2001 - December 2020). Source models: MERRA2. Only each publisher’s own public endpoints are used, once per build, at the cadence they publish. No page is scraped for values where a dataset or endpoint exists.
07DATA HANDLING
Client data is the one dataset that never appears here.
Least data that answers the question
Irven asks for the invoices, interval data and equipment records a diagnostic needs, and no more. A site’s process data is not collected because it might be interesting later.
Client data never leaves the engagement
Contracts define confidentiality, data ownership, use of derived data, model training, anonymisation and aggregation, encryption, subprocessors, retention, deletion and incident handling. Site data is not pooled across clients and appears in no public demonstration, including this site.
Read-only by default
Measure, recommend, obtain authorisation, then execute. Where Irven connects to a BMS, SCADA historian or metering platform it asks for read access to an export, not control of the system. Autonomous control of HVAC, pumps, storage or critical process is a later decision, not a first version.
08THE DATA LAYER
What a site model is assembled from.
Invoices and tariff
12–24 months of CFE billing, RPU, billed demand, power factor and the tariff route the site is actually on, read line by line.
Meters and submeters
Interval data at the point of supply and, where it exists or is temporarily installed, below it — the difference between a total and a diagnosis.
BMS and SCADA
Setpoints, run hours, sequences and alarms: what the plant was told to do, next to what it did.
Market and system data
CENACE demand, day-ahead and real-time zonal prices and interchange between control regions, for territorial calibration.
Climate
NASA POWER irradiance, temperature, humidity and wind, to normalise a weather year out of an operating problem.
Earth observation
ECOSTRESS land surface temperature for heat and cooling analysis, and NASA Black Marble nighttime radiance strictly as an activity proxy, never as a meter.
Buildings and territory
Overture Maps building footprints with OSM, authorised cadastre, LiDAR and aerial imagery, to establish usable area and context.
Financial models
Tariffs, capital cost, cost of money, financing instruments and risk, so an engineering finding arrives as a financial one.