A Mexican plant that reports emissions from purchased electricity is reporting two numbers that behave in opposite ways. One is metered consumption in MWh. The other is the emissions content of the supply contract behind it. Software touches the first. Only a procurement decision touches the second.

That distinction decides whether an energy analytics platform belongs in this year's capital plan. If the target is the reported Scope 2 figure, a change of supply does more in one signature than three years of control tuning. If the target is the CFE bill, the control software has real work to do, and the work is mostly about demand and scheduling rather than about carbon.

Both cases share one precondition, and most plants in Mexico fail it. A CFE GDMTH invoice reports about eight numbers a month. No model builds a load curve from eight numbers. Before any of this is a software question, it is a metering question.

What the reported number is actually made of

The reporting duty comes from the Reglamento de la Ley General de Cambio Climático en Materia del Registro Nacional de Emisiones, published in the Diario Oficial de la Federación on 28 October 2014. An establishment emitting 25,000 tonnes of CO2 equivalent or more in a year has to report to the RENE, and emissions from purchased electricity count toward that threshold. SEMARNAT publishes the Factor de Emisión del Sistema Eléctrico Nacional each year for exactly this purpose, and the location based figure a plant reports is its consumption multiplied by the factor in force for the reporting year.

The arithmetic has one input a plant controls and one it does not. Consumption is measurable and reducible. The national factor is set by the generation mix CENACE dispatched, and no plant control system moves it.

The market based figure works differently. Under the Scope 2 Guidance issued by the GHG Protocol in 2015, a company may report the emissions attributes carried by its own supply contracts, provided the contractual instrument meets the quality criteria that guidance sets. In Mexico that route runs through the supply arrangement, so through Usuario Calificado registration, a qualified supplier contract with stated attributes, an onsite generation contract, or a PPA. It does not run through a building management system.

Certificados de Energías Limpias sit in a separate box and are commonly misfiled. CELs are a compliance obligation created by the Ley de Transición Energética, published in the DOF on 24 December 2015. SENER sets the annual requirement as a percentage of qualified consumption and publishes it by aviso in the DOF, and the applicable figure is the one in the aviso in force for the compliance year. That is an obligation on the load, not a corporate carbon instrument. Whether a CEL can carry an attribute claim in a market based inventory depends on the contract features the GHG Protocol requires, and it is a question for the contract, not for the software.

What a model can read at a Mexican plant

Start with what arrives in the post. A CFE GDMTH invoice reports energy in kWh split into base, intermedio and punta, the demand readings in kW behind the distribution and capacity charges, the power factor and the resulting charges. That is a monthly summary. It supports a tariff analysis and a billing audit. It does not support load disaggregation, anomaly detection or scheduling.

Anything beyond tariff work needs interval data at the point of use. That means sub-meters on the loads that matter, which in most Mexican industrial sites are compressed air, refrigeration, process heat, pumping and the drive fleet. It also means the data leaving those meters into a store the plant owns. Vendors who propose a model before the metering exists are proposing to fit a curve to a monthly bill.

The sequence is metering, then a measured baseline, then control. Reversing it produces savings claims that cannot be tested, because nobody can say what the site would have consumed without the intervention. A baseline built to a recognized measurement and verification method is what turns a vendor claim into a number a CFO can put in a budget. Read the proposal for the baseline before reading it for the algorithm.

Where the software actually earns money

On a GDMTH site the demand charge is the line most responsive to control. CFE bills GDMTH demand in pesos per kW and bills it twice, once for distribution and once for capacity. Both charges are set by demand the meter records during the month rather than by total consumption. A single unlucky coincidence of a compressor start, a furnace ramp and a chiller pull-down can therefore carry a charge the plant pays for the whole month.

Load scheduling and demand limiting act directly on that. So does moving discretionary load out of the punta window, whose hours CFE publishes by region and season and which shift with the horario de verano. Both are peso savings. Neither necessarily reduces kWh, and a saving that does not reduce kWh does not reduce reported emissions. That is the sentence most AI energy proposals leave out.

Power factor is the second line. CFE applies a surcharge below 90% under the Acuerdo que modifica las disposiciones complementarias a las tarifas para suministro y venta de energía eléctrica, published in the DOF on 31 October 2000, and a credit at or above it. The surcharge is capped at 120% of the invoice amount and the credit at 2.5%. Correction here is a capacitor bank and a controller, not a model.

The third line is consumption itself, and this is the one that moves both the bill and the emissions figure. Leak detection on compressed air, sequencing on chillers and compressors, setpoint control on process heat, and shutdown enforcement on idle equipment all cut kWh. This is also where demand forecasting at plant level stops being a grid concept and becomes a production scheduling input.

What it cannot do

Most vendor material on this subject describes work that belongs to a system operator. Renewable output forecasting, dispatch adjustment and grid stability are CENACE functions in Mexico. A plant does not perform them, does not buy them and does not report their results.

Four limits are worth stating plainly. Control software cannot change the national emission factor. It cannot create a contractual attribute claim where the supply contract carries none. It cannot see an unmetered load. And predictive maintenance detects the failure modes its sensors observe, which on a motor means bearing and alignment faults visible in vibration and current signature, and not the failure modes nobody instrumented.

There is also a reporting trap. A plant that cuts consumption by 8% while the national factor rises will report higher emissions from purchased electricity than the year before. The operational improvement is real and the reported number went the wrong way. Boards need that explained in advance, not in the footnote of an inventory.

The counter-case

The argument against spending on this in 2024 is not that the technology fails. It is that the payback is concentrated in a narrow set of sites. A plant with flat load, a single shift and a power factor already above 0.95 has little for a scheduling model to work on. Its money goes further in equipment replacement.

The second objection is data quality. Models trained on a plant's own history inherit that history, including the six months a meter drifted and the shutdown nobody logged. Cleaning that record is manual work and it is usually the largest hidden line in a platform deployment.

The third is lock-in. Where the analytics layer owns the data store, the plant loses the ability to change vendors without losing its baseline. Insist on export of raw interval data in an open format as a contract term, before signature.

What to do about it this quarter

Answer the boundary question first. Establish whether the emissions figure being reported today is location based, market based or an untraceable mixture, and whether the site sits above or below the 25,000 tonne RENE threshold. That determines which lever matters and it costs nothing but time. It is the same groundwork that sits behind a Mexican site's emissions baseline.

Then price the metering. Sub-metering the five or six loads that carry most of the site's consumption is a defined capital item with a known cost, and it is the precondition for everything a platform promises. Fund it separately from the software so that a failed platform trial does not strand the meters.

Run the supply question in parallel, because it is slower and it carries the larger emissions effect. Qualified supply migration and PPA negotiation take quarters, not weeks, and that is where AI touches an energy procurement decision in a way that shows up in the inventory. A plant that starts the software project and the supply review in the same quarter will have the metering in place by the time the contract options are on the table.

Get a defensible number for your purchased electricity emissions

Mexico Energy Partners reconstructs the emissions attached to purchased electricity across each of your Mexican sites. We state the figure on a location basis and on a market basis, then separate the part that moves through the supply contract from the part that moves only through consumption. The report gives the timing of each and the metering the second one requires. You supply twelve months of CFE invoices per site, the RPU numbers, and any qualified supply or CEL contracts in force. Where the figure you report today cannot be traced to a source document, the report says so.

Start on the request for quote page, or send the invoices through the form below.

Sources

• Reglamento de la Ley General de Cambio Climático en Materia del Registro Nacional de Emisiones, Diario Oficial de la Federación, 28 October 2014. Reporting obligation and the 25,000 tonne CO2 equivalent threshold.
• Secretaría de Medio Ambiente y Recursos Naturales, Factor de Emisión del Sistema Eléctrico Nacional, published annually for the Registro Nacional de Emisiones. Basis of the location based calculation. The applicable value is the one in the aviso in force for the reporting year.
• GHG Protocol, Scope 2 Guidance, World Resources Institute and World Business Council for Sustainable Development, 2015. Location based and market based reporting and the contractual instrument quality criteria.
• Ley de Transición Energética, Diario Oficial de la Federación, 24 December 2015. Legal basis of the Certificados de Energías Limpias obligation.
• Secretaría de Energía, Aviso por el que se dan a conocer los requisitos para la adquisición de Certificados de Energías Limpias en 2020, 2021 y 2022, Diario Oficial de la Federación, 31 March 2017. CEL requirement percentages for 2020, 2021 and 2022.
• Comisión Federal de Electricidad, published GDMTH tariff schedule, in force March 2024. Energy charges by base, intermedio and punta period, and the distribution and capacity charges billed in pesos per kW.
• Acuerdo que modifica las disposiciones complementarias a las tarifas para suministro y venta de energía eléctrica, Diario Oficial de la Federación, 31 October 2000. Power factor surcharge and credit and their caps of 120% and 2.5%.