A 100-millisecond voltage drop on a CFE line in Sonora is enough to trip a grinding mill. That is six cycles at 60 Hz. Six cycles, and a variable-frequency drive on a mill motor sees its DC bus collapse past the undervoltage threshold, drops out to protect itself, and the mill coasts to a stop mid-charge. Restarting a loaded ball mill is not a button press. You inch the charge back into motion, re-establish the grinding load, and wait for the concentrator downstream to stabilize. Hours of throughput gone, and the sag that caused it lasted less time than a camera shutter.
The mills in question are not small. Buenavista del Cobre at Cananea runs six ball mills driven by 14 MW GE Quadramatic systems adding 84 megawatts of grinding power, phased into commercial operation from 2013 according to GE’s September 2012 project announcement. A copper concentrator of that size is one of the largest single industrial loads hanging off the regional grid in northwest Mexico. Grupo México supplies most of that power from its own 516.2 MW Mexico Generadora de Energía combined-cycle plants and buys the balance from CFE (Southern Copper 10-K for 2025), but the power still reaches Cananea over CFE’s regional transmission, which was never built with much margin. When the line is weak and the load is heavy, power quality stops being an abstraction on a specification sheet. It becomes lost tonnes.
The binding reliability constraint at a remote Mexican mine is rarely a shortage of megawatt-hours. It is the quality of the power at the point of common coupling, the sags, swells, and harmonic distortion a long radial CFE feeder delivers and a large drive-heavy load makes worse. That is an engineering problem before it is a procurement problem. You solve it by measuring your own bus and sizing correction to your actual load curve, not by buying the largest battery a vendor will sell you.
What a weak line does to the power you pay for
Start with why a remote line sags in the first place. A voltage sag during a fault behaves like a voltage divider. The voltage that survives at your bus is set by the ratio of the impedance between you and the fault to the total impedance back to the source.
On a stiff urban bus the source impedance is small, so a fault a few kilometers away barely moves your voltage. On a long rural feeder in Sonora the source impedance is large, and the same fault drags your bus down hard. Distance is not your friend here. Every kilometer of conductor between the mine and a strong substation adds impedance and deepens every sag that a lightning strike, a line slap, or a distant fault produces.
The stiffness of a bus is captured by its short-circuit ratio, the available fault power over your load:
A high SCR means the grid can absorb your load swings without flinching. A low SCR means your own equipment moves the voltage. Mining loads are exactly the kind that punish a low-SCR bus. A mill motor, a large dewatering pump, a crusher, all of them draw heavy reactive current on start and swing load through the grinding cycle, and on a weak feeder that reactive draw shows up as a voltage depression the whole plant feels. The feeder voltage drop follows the reactive term directly:
The reactance term is the one that hurts on a long line, because transmission reactance is large and mining load runs at a lagging power factor. Poor power factor is not just a CFE billing surcharge and a Código de Red compliance gap. On a weak line it is a voltage-stability problem you create for yourself every time a big motor pulls reactive current the feeder cannot supply without sagging.
Swells are the same mechanism running the other way. Shed a large block of load, or leave a capacitor bank switched in when the mill drops off, and the lightly loaded feeder overshoots. IEEE 1159-2019 classifies a swell as 1.1 to 1.8 per unit lasting from half a cycle to a minute, with the 1.8 ceiling applying to instantaneous events of 30 cycles or less. Sustained overvoltage cooks insulation and shortens the life of every transformer and drive on the bus. A plant that fixes sags and ignores swells has solved half the problem.
Then there is harmonic distortion, which your own equipment manufactures. Every VFD and rectifier on the property is a non-linear load drawing current in pulses rather than a clean sinusoid. The distortion is measured against the fundamental:
IEEE 519-2022, which carries the 2014 limits forward unchanged, caps voltage total harmonic distortion at 5.0% for buses between 1 kV and 69 kV and 8.0% below 1 kV. The current limits are tighter on a weak bus, and this is the part operators miss. The allowable current distortion scales with SCR. A stiff bus with a short-circuit ratio above 1000 is permitted 20% total demand distortion. A weak bus below an SCR of 20 is held to 5%. The weaker your line, the less harmonic current you are allowed to inject, precisely because a weak line converts that current into voltage distortion everyone downstream has to live with. Remote mines tend to have the weakest buses and the most drive load. That is the worst combination the standard contemplates.
What a sag actually costs: lost throughput, stalled pumps, lost water control
The equipment does not read the standard. It reacts to physics. A drive-fed mill motor loses torque as bus voltage drops, and if the sag is deep enough the drive trips on undervoltage before the process even registers what happened. That 100-millisecond event is not hypothetical. A distant fault on a shared feeder produces it routinely, and it stops a mill that then costs hours to recover.
Dewatering is where the second-order damage hides. A mine that cannot keep its pumps running does not just lose pumping. It loses the water balance that keeps the pit or the workings dry. Centrifugal pump head follows the affinity law against speed:
A voltage sag that pulls pump speed down 20% does not cut head 20%. It cuts head by the square, to 64% of nominal, a 36% collapse in the pressure the pump can develop. On a dewatering circuit fighting an inflow, that is the difference between holding the water line and losing it. In a region where water is already scarce and the mine runs tight recirculation to conserve it, an unstable pumping system becomes a water-management problem the grid is quietly creating.
If nobody has logged sag depth and duration at your point of common coupling, that measurement costs less than one lost grinding shift and it settles whether the losses are electrical at all. Talk to an advisor.
None of this shows up as an outage on a CFE bill. The meter records energy delivered, not the sub-second events that scrapped a grinding shift or stalled a pump. That is the trap. A site can go years without a recorded outage and still bleed throughput to sags and harmonics nobody is measuring, because the only instrument watching the bus is a revenue meter that averages the very events doing the damage.
What to buy, and the order to buy it in
Measure first. You cannot correct a disturbance you have not characterized, and a revenue meter will not characterize it. A Class A power quality analyzer, permanently installed at the point of common coupling and logging at the sampling rate needed to catch a six-cycle sag, is the first piece of capital, not the last. It tells you your real SCR, your sag frequency and depth, your harmonic spectrum, and your power factor. The Código de Red power-factor requirement for high-voltage load centers and medium-voltage load centers at 1 MW and above stepped up from 0.95 to 0.97 lagging on April 8, 2026, a change scheduled in the original 2016 code (RES/151/2016) and carried into Código de Red 2.0 (RES/550/2021), measured in five-minute intervals over the billing month and now administered by the Comisión Nacional de Energía. Without that data you are sizing correction by guesswork, and guesswork on a weak bus tends to buy too much of the wrong thing.
Power factor correction comes next, because it is the cheapest lever and it does double duty. Sizing a capacitor bank is straightforward once you know your load:
Moving a load from a 0.90 to a 0.97 power factor needs , roughly 234 kVAr for every megawatt of real load. That correction pulls you into Código de Red compliance and cuts the reactive current on your feeder, which directly reduces the voltage drop that deepens every sag. One investment, two returns. The caution is real. A fixed bank left switched in at light load is how you manufacture a swell. On a mine whose load swings between full grinding and idle, use switched or automatic banks that follow the load, not a fixed block sized for peak.
Harmonic filtering is the third layer, sized to bring your injected current distortion back under the IEEE 519 limit your specific SCR allows. On a weak remote bus that limit is strict, and tuned passive filters or active filters on the worst drive groups are usually cheaper than the transformer life you lose running a distorted bus for years.
The heaviest tool is a hybrid microgrid, solar paired with battery storage, and its value on a remote mine is not only the diesel it displaces. It is isolation. A battery inverter sitting between the CFE feeder and your critical extraction loads can ride through a sag the grid throws at you. It holds voltage on the mill and pump bus for the fraction of a second the disturbance lasts. The drives never see the event. The same asset shaves the peak-demand charges that punish a spiky mining load and displaces diesel during daylight, and diesel is not cheap. Mexican diesel averaged 26.99 pesos per liter nationally in the week to October 9, 2026, per Profeco, after a spring and summer in which it ran above 28 pesos, and at a remote site the delivered cost runs higher once you truck it in over the same bad roads that make everything else expensive. Grid-connected mining solar already pencils out. Minera Frisco’s 450 kWp rooftop array at Real de Ángeles in Zacatecas supplies about 65% of facility demand and cuts electrical costs 17% under a 10-year power purchase agreement, according to the developer, Energía Real. Off-grid, Resolute Mining told the ASX that Aggreko’s hybrid contract for the Syama mine would cut power costs by about 40%, and Aggreko’s own case study adds 10 million dollars in first-year savings, in Mali and not Mexico, so read the direction and not the exact figure.
A word on friction, because the brochures leave it out. Interconnecting anything to CFE in a constrained region is slow. Permitting drags, and the transmission the mine hangs off is part of a network where roughly 60% already runs near maximum capacity, per Eaton analysis reported by Mexico Business News on June 26, 2026. National demand grew 3.4% and 3.5% in 2022 and 2023 while transmission expanded 0.09% and 0.10% in the same years, according to the Mexican Institute for Competitiveness (IMCO, September 2024), so the congestion is structural and it is not easing on CFE’s current budget. In parts of the northwest the grid is weak and effectively islanded. Baja California still runs a system that is not synchronously tied to the national grid as of November 2026; the 300 MW, 400 kV line from CFE’s Puerto Peñasco solar plant that the federal government presented in June 2026 as the state’s first link to the rest of the country is a radial tie into the Cucapah substation, and the Puerto Peñasco to Hermosillo line that would close the loop was still at the authorization stage in 2026. A microgrid that can island cleanly and hold your critical load through a grid event is worth more in that setting, not less, and the permitting timeline is a reason to start the interconnection paperwork before you need the asset, not after.
Why we size the fix instead of selling you hardware
Here is the part most vendors will not say out loud. A company that sells battery hardware is paid to sell battery hardware, and the sizing tool it runs tends to arrive at a number that looks a lot like the largest system it stocks. That is not analysis. That is a quote wearing an analysis costume.
Mexico Energy Partners does not sell the hardware. Our engineering team instruments your bus, pulls your real load curve at the resolution that shows the sags and the reactive swings, and sizes correction against what your plant actually does through a grinding cycle. Sometimes the answer is a switched capacitor bank and a tuned filter that costs a fraction of a microgrid and fixes 80% of the problem. Sometimes the mine has three years of reserve left, and the right move is to condition the power you have and skip the capital project. Sometimes the load curve justifies a full solar-plus-storage island, and then we size it to the critical load and the sag profile, not to a catalog. The recommendation comes from your data. It does not come with a commission riding on the answer.
When the fix is not worth the spend
Not every remote mine has a power quality problem worth spending on. If your bus turns out to be stiffer than assumed, a short line to a strong substation with a high SCR, your sags may be shallow and infrequent, and the case for heavy correction weakens. Measure before you conclude. A mine near the end of its reserve life has a different answer again, because a multi-year payback on a microgrid does not close if the pit has two years left, and the right call there is cheap conditioning or nothing at all.
Your losses may turn out to be mechanical or operational instead of electrical. A power quality audit that finds a clean bus has still earned its cost by ruling the grid out and pointing you back at the process. Measuring tells you which problem you actually have before you spend against it. Justifying a purchase is not the point.
Power Quality and Grid Code Risk Review
Mexico Energy Partners logs your point of common coupling for sag depth, harmonic current against the IEEE 519 limit your short-circuit ratio allows, and power factor against the 0.97 Código de Red threshold. It needs a metering window on the bus and your single-line diagram, and the output is a written read on whether the answer is a capacitor bank, a filter, or an islanding microgrid.
Talk to a Mexico Energy Partners advisor or email info@mexicoenergypartners.com. Mexico Energy Partners sells no equipment and is compensated only by the client.
Sources
- Voltage sag and swell classification, magnitude 0.1 to 0.9 per unit and 1.1 to 1.8 per unit, durations from 0.5 cycle to 1 minute, with the 1.8 per unit ceiling for instantaneous swells. IEEE Std 1159-2019, Recommended Practice for Monitoring Electric Power Quality.
- Voltage total harmonic distortion limits of 5.0% for 1 kV to 69 kV buses and 8.0% below 1 kV, and current total demand distortion limits scaling with short-circuit ratio from 5.0% below an SCR of 20 to 20.0% above an SCR of 1000. IEEE Std 519-2022, Standard for Harmonic Control in Electric Power Systems (limits unchanged from IEEE 519-2014).
- Six 14 MW GE Quadramatic systems adding 84 megawatts at the Buenavista ball mills, Cananea, Sonora, commercial operation phased in from the second quarter of 2013. GE, “GE Motor Technology to Drive Mexico’s Largest Ball Mill Copper Mine,” September 26, 2012. ge.com.
- Southern Copper Corporation, Form 10-K for the year ended December 31, 2025, “Fuel, Electricity and Water Supplies” (Mexico Generadora de Energía, 516.2 MW net, supplies La Caridad and Buenavista; balance purchased from CFE). sec.gov.
- CRE, RES/151/2016, Código de Red, DOF April 8, 2016 (power factor 0.95 lagging to 1.0 for ten years from publication, then 0.97 lagging to 1.0, five-minute measurement). sidof.segob.gob.mx. CRE, RES/550/2021, Código de Red 2.0, DOF December 31, 2021, in force January 1, 2022; applies to high-voltage load centers and medium-voltage load centers with contracted demand of 1 MW and above. sidof.segob.gob.mx. Administered since 2025 by the Comisión Nacional de Energía.
- Profeco, national average diesel price 26.99 MXN per liter, data to October 9, 2026, as reported by La Crónica, October 12, 2026. cronica.com.mx.
- Energía Real, Minera Frisco case study, Minera Real de Ángeles, Zacatecas: 0.45 MWp rooftop PV, 65% of demand, 17% cut in electricity costs, 10-year solar PPA. energiareal.mx.
- Resolute Mining, ASX announcement, November 25, 2019 (Syama hybrid power plant expected to reduce power costs by approximately 40%). clients3.weblink.com.au. Aggreko, Resolute Mining Syama case study (USD 10 million first-year savings, 16-year agreement). aggreko.com.
- Approximately 60% of Mexico’s transmission network operating near maximum capacity. Eaton analysis, reported by Mexico Business News, “Eaton: Mexico’s Industrial Sector Cannot Wait for Grid Reform,” June 26, 2026. mexicobusiness.news.
- IMCO, “Sin transmisión eléctrica no será posible acelerar la transición energética,” September 19, 2024 (demand growth 3.4% in 2022 and 3.5% in 2023; transmission network growth 0.09% and 0.10%). imco.org.mx.
- La Crónica, February 27, 2026 (Baja California isolated system; Central Fotovoltaica Rafael Galván Maldonado, Puerto Peñasco) and June 24, 2026 (300 MW Puerto Peñasco line presented as Baja California’s first link to the national system). cronica.com.mx. CENACE, 2026 Annual Progress Report to WECC (Puerto Peñasco sequences radially interconnected to Cucapah). wecc.org.
The MEP newsletter covers CFE tariffs and industrial energy every two weeks, in Spanish.