CENACE publishes hourly Precios Marginales Locales for the Mercado del Día en Adelantado and the Mercado de Tiempo Real through its Sistema de Información del Mercado, on a network model that runs from 69 kV upward and covers more than 6,000 nodes. The market side of any analytics project in Mexico is already public and free. What is usually missing sits inside the plant, in the interval demand history its own meters were never set to retain.

That gap decides which of the uses claimed for artificial intelligence a Mexican industrial site can actually buy. Two of them are deployable by a private plant today. The rest belong to CENACE and CFE, and no purchase by a load center changes that. For a plant manager or finance director, the decision this article helps evaluate is narrow and immediate. Spend on analytics, or spend first on the metering that analytics needs.

What a Mexican site can deploy now

Demand forecasting is the first. It pays through the tariff rather than through the meter total. On CFE's Gran Demanda en Media Tensión Horaria schedule the site is charged for energy in three time bands, base, intermedio and punta. It is charged for demand on two separate per kilowatt charges, distribución and capacidad. Those demand charges do not fall because the plant used fewer kilowatt hours. They fall when the peak moves.

That distinction is where most vendor cases go wrong. A model that predicts monthly consumption accurately and says nothing about which quarter hour sets the maximum leaves the demand charges untouched. A forecast worth paying for predicts the coincidence of compressors, chillers and process restarts, close enough ahead that someone on shift can resequence them. This is the practical core of forecasting and managing energy demand, and it needs interval data, not monthly bills.

Consumption analytics is the second. Once a site records interval data at the main incomer and at its largest feeders, anomaly detection finds the drift that no one sees on a monthly invoice. A compressor running unloaded through the night shift, a chiller short cycling, a line left energized through a changeover. None of these requires a model that is more sophisticated than the data underneath it, which is the mistake behind most disappointed pilots. The same logic applies across AI driven savings at Mexican industrial sites.

What belongs to CENACE, not to your plant

Grid stress prediction and automated outage response are the applications most often listed and the least available to the reader. Under the Ley de la Industria Eléctrica, published in the DOF on 11 August 2014, CENACE operates the Sistema Eléctrico Nacional and the Mercado Eléctrico Mayorista. A Centro de Carga is not a party to that. It cannot buy grid level analytics, cannot deploy them, and does not choose whether they are used.

Mexico Energy Partners could not identify a published CENACE program for automated grid analytics in force as of May 2024. That is our own search result, not a statement that none exists internally. Until CENACE publishes one, a vendor claim that your site benefits from grid level AI should be treated as unsupported, and it should not appear in a business case.

What a site can do with the same subject is read what CENACE already publishes. Nodal prices, load zone prices and system demand are posted on a fixed schedule. For a plant exposed to those prices, that feed is an input. For a plant on CFE basic supply it is context and nothing more.

The prerequisite is measurement, and it is already regulated

Most Mexican industrial sites do not hold enough interval history to train anything. The first project is therefore measurement. Part of that spend has already been mandated, which is the point plant teams tend to miss.

CRE issued the Código de Red as RES/151/2016, published in the DOF on 8 April 2016, and replaced it with RES/550/2021, published in the DOF on 31 December 2021 and in force from 1 January 2022. It applies to Centros de Carga connected at media tensión, above 1 kV and up to 35 kV, and at alta tensión, 35 kV and above, regardless of contracted demand or supply scheme. Its requirements include information exchange, and the studies behind them must be taken with Class A measurement instruments under NMX-J-610/4-3-ANCE or IEC 61000-4-30. Load centers already connected on 8 April 2016 had until 9 April 2019 to comply.

A site that complied bought Class A instrumentation and a measurement discipline at the same time. That is the same class of instrument a forecasting project wants on the incomer. Two corrections are worth carrying into the next vendor conversation. CRE holds compliance and sanction authority, not CENACE, which operates the system and the market. And CRE does not authorize or certify inspectors for these studies, so a load center may run them itself or contract a specialized firm.

Where procurement analytics actually applies

Analytics is often sold as a route to better energy procurement. That only holds where a procurement decision exists. A site taking CFE basic supply pays a published tariff and has no counterparty to negotiate with, so a market price model changes nothing on its bill.

The claim becomes real for a load center registered with CRE as a Usuario Calificado, or one supplied by a qualified supplier under contract. There the site is exposed to market prices and to the shape of its own load, and a forecast that is wrong costs money in both directions. Before commissioning any procurement analytics, establish which of those two positions the site is in. The answer sits on the supply contract, and it decides whether the project has a payback at all.

What analytics does not change

Two constraints sit outside the reach of any model. The first is the exempt generation ceiling. CRE set the administrative provisions for distributed generation in RES/142/2017, published in the DOF on 7 March 2017, covering plants with capacity below 0.5 MW. An onsite array above that limit follows a different route, and no forecasting tool moves the threshold.

The second is the carbon arithmetic. SEMARNAT publishes a Factor de Emisión del Sistema Eléctrico Nacional each year for Registro Nacional de Emisiones reporting, and every megawatt hour the plant draws from the grid carries that factor as Scope 2. Analytics does not alter the factor. It alters how many megawatt hours the plant draws, which is the only lever it has on that line.

Find out what your plant's data can support

Send Mexico Energy Partners the last twelve CFE invoices for the site and whatever interval data the meters or the building system already hold. We report back on the interval and the length of history you have, what a forecast built on that record could realistically predict, and whether the first useful project is analytics or more metering. No software purchase is needed to find out, and the review states which conclusions rest on measured data and which rest on assumption.

Request an energy data readiness review

Sources

• CENACE, Sistema de Información del Mercado, Precios Marginales Locales for the Mercado del Día en Adelantado and the Mercado de Tiempo Real, network model from 69 kV with more than 6,000 nodes.
• Congress of the Union, Ley de la Industria Eléctrica, DOF, 11 August 2014.
• CRE, RES/151/2016, Código de Red, DOF, 8 April 2016.
• CRE, RES/550/2021, Código de Red, DOF, 31 December 2021.
• CRE, RES/142/2017, disposiciones administrativas de carácter general para centrales eléctricas de generación distribuida menores a 0.5 MW, DOF, 7 March 2017.
• CFE, Gran Demanda en Media Tensión Horaria tariff schedule, published monthly by distribution division.