Demand Charges Explained

Two facilities on the same street, on the same tariff, each consuming 400,000 kilowatt-hours in a month. One receives a bill substantially larger than the other.

Neither bill is wrong. The difference is entirely in when the energy arrived, and the mechanism that prices it is the demand charge.

What is being measured

The demand charge prices the highest rate of consumption your facility reached during the billing period, rather than the total amount consumed.

The meter does not record an instantaneous maximum. It records the average load over each successive interval — most commonly fifteen minutes — and the demand determinant is the highest of those averages across the period. A momentary inrush current when a large motor starts is spread across the interval and usually leaves no mark. A sustained elevated load lasting the whole interval leaves a very clear one.

The averaging is worth working through with numbers, because it is the source of most of what seems counterintuitive later. A 300 kW load that runs for three minutes of a fifteen-minute interval contributes 300 × (3 ÷ 15) = 60 kW to that interval's average. The same load running for the full fifteen minutes contributes 300 kW. Identical equipment, five times the determinant, purely because of duration.

Three consequences follow directly:

• Starting inrush rarely matters. It lasts seconds and is diluted almost to nothing.
• Simultaneity matters enormously. Three 200 kW loads that run one after another never produce an average above 200 kW. The same three running together produce 600 kW.
• Brief interruptions help less than expected. Shedding a load for two minutes inside a fifteen-minute interval removes about 13 percent of its contribution to that interval, not all of it.

The consequence, stated plainly: a single quarter of an hour can set a charge you pay for the entire month. And where a ratchet clause is in force, for much of the year after that.

Why utilities charge for it

The rationale is cost causation, and it is worth understanding rather than merely accepting, because it predicts how the charge behaves.

A utility incurs two structurally different kinds of cost. The first scales with volume: fuel burned, power purchased on the wholesale market, energy lost as heat in the conductors. Every one of those moves with the number of kilowatt-hours delivered, and every one of them is naturally recovered through a rate per kilowatt-hour.

The second scales with peak. The generation capacity that has to be available at the worst moment of the year, the transmission lines that have to carry the maximum flow, the substation transformer sized for the largest load its feeder will ever see, the service conductors to your own building — all of these are sized for the maximum and cost essentially the same whether that maximum persists for a season or for a quarter of an hour. Once built, they are fixed costs, and they were built to a specification set by somebody's peak.

Recovering that second category through a per-kilowatt-hour rate works acceptably when customers have similar load shapes, because the averaging is roughly fair to everyone. Commercial and industrial customers do not have similar load shapes. Consider two accounts consuming exactly the same energy in a month: one running continuously at a steady load, the other running two intense production shifts a week and idling the rest of the time. The system has to be built to serve the second customer's peak, and that capacity sits underused for the remaining days. Under a purely volumetric rate both pay identically, which means the steady customer is paying part of the cost of capacity that exists solely because of the spiky one.

Eliminating that kind of cross-subsidy between customer classes is one of the things a state regulator exists to do. So the tariff unbundles the two. Volume is charged per kilowatt-hour; capacity is charged per kilowatt of the peak that made the capacity necessary. Whatever else is true of the demand charge, it is an attempt to bill people for what they cause.

Why utilities bill for demand at all takes the argument further, including the places where it is genuinely contested in rate cases — because the cost causation story is strongest at the distribution level and gets noticeably weaker higher up the system.

The definitions that decide the money

The concept is simple. The definitions are where bills are actually decided, and they differ from tariff to tariff. Four of them matter.

The interval

Fifteen minutes is the common standard; thirty appears in some tariffs. The length matters enormously for a spiky load, because a shorter interval gives a spike less time to be averaged away. It is stated in the tariff, and any analysis of your own data has to use the same interval the tariff bills on. See the demand interval.

The window

Some tariffs measure demand across every interval in the month — facility demand. Others measure only intervals inside a defined on-peak window. Many bill both, as separate determinants at separate prices.

This decides which strategies work. Moving load out of the afternoon reduces on-peak demand and can leave facility demand exactly where it was, because the new peak just appears at a different hour. See facility, on-peak and billing demand.

Coincidence

A non-coincident charge bills your own maximum, whenever it occurred. A coincident charge bills whatever you happened to be drawing when the system peaked — a moment you do not control and cannot know in advance.

These are entirely different problems. Managing the first is an internal exercise; managing the second means forecasting somebody else's peak. The distinction, and the two best-known regional implementations, are covered in coincident and non-coincident demand.

The ratchet

A ratchet clause sets a floor under billing demand based on a peak from an earlier period — typically a percentage of the highest demand in the preceding eleven or twelve months.

The effect is that one bad afternoon is not billed once. It is billed repeatedly, in every subsequent month where actual demand falls below the floor. This is the mechanism most likely to be present on a bill and absent from the reader's understanding of it: ratchet clauses.

What it costs

The headline rate in the tariff is not the number to use.

Demand-based riders — transmission cost recovery, distribution investment recovery — stack on top, and percentage-based taxes stack on top of those. The value of avoiding one kilowatt is the sum of all of them.

Where a ratchet is in force the figure can be considerably higher, because avoiding a peak avoids not only that month's charge but the floor it would have set. That interaction is worked through in what a kilowatt of avoided peak is actually worth.

Two tariffs, one plant

The same consumption, billed under two structures, produces two very different splits.

Which structure suits a site depends on its load factor — energy divided by peak times hours in the period. A high load factor site runs steadily, uses the capacity it pays for, and is comparatively indifferent to demand charges. A low load factor site is punished by them, and may be substantially better off on a schedule that puts more weight on energy, if it qualifies for one.

That is a tariff selection question rather than an engineering one, and it is frequently the highest-return action available: demand-heavy versus energy-heavy tariffs and how to choose a rate schedule.

It is also the reason two apparently similar plants can reach opposite conclusions about the same investment. A battery that pays back in four years at a site with a sharp, predictable, once-a-day peak may never pay back at a site whose load is already flat, because there is nothing above the average to shave. Nothing about the equipment differs. The load shape does, and the load shape is what the charge prices.

The one distinction that governs everything else

Before any measure is considered, one question has to be settled: is the charge measured against your own peak, or against the system's?

A non-coincident charge bills your highest interval, whenever it occurred. It is visible in your own data, attributable to specific equipment, and controllable with sequencing, scheduling or storage. The feedback arrives on the next bill.

A coincident charge bills what you were drawing during somebody else's maximum. You cannot see it at the time, you have to forecast it, and the feedback may not arrive for a year. Managing it means acting on predictions and accepting that some curtailments will turn out to have been unnecessary.

Many bills carry both, on separate lines, at separate rates. Spending money to flatten a peak that was never being billed is a real and avoidable mistake, and it starts with reading which determinant is which: coincident and non-coincident demand.

What actually reduces it

The demand charge responds to the shape of the load curve, not its area. That distinction is the whole of demand management, and it has a useful corollary: shifting a load is often much cheaper than eliminating it.

The levers, roughly in order of cost:

1. Sequencing and staggered startup. Usually free, frequently effective, and the single most common finding at sites with a sharp morning spike: staggered startup.
2. Scheduling and setpoint strategy. Pre-cooling, batch timing, avoiding coincident operation of large loads: HVAC scheduling against the peak.
3. Demand limiting controls. Automatic shedding when a predicted interval average approaches a target: demand limiting controls.
4. Thermal storage. Making cooling at night and using it during the day: thermal energy storage.
5. Batteries. Discharging into the peak. Effective, and capital-intensive enough that the sizing arithmetic matters: sizing a battery for peak shaving.
6. On-site generation and curtailment agreements. Including selling the capability back through a program: demand response programs.

And the measures that do not work, which are worth knowing before you buy one: six efficiency measures that do not cut your demand charge. The theme running through them is that saving kilowatt-hours at hours other than your peak has no effect whatever on this charge.

Where to start

Not with equipment. Buying a measure before establishing whether the site has a shape problem is how demand projects acquire a reputation for underdelivering, and the diagnosis costs nothing but an hour with a spreadsheet.

Start with three numbers.

Your load factor, from figures already on the bill: energy divided by billed demand times hours in the period. Under about 40% and there is usually shape worth fixing.

The gap between your highest interval and your tenth-highest, from interval data. A wide gap means very few intervals set the charge and a cheap occasional intervention captures most of the value. A narrow gap means the peak is the whole load, and you are looking at storage or a different tariff.

The all-in value of a kilowatt, including riders, as calculated above.

Those three tell you whether this is your highest-return project or a distraction. Getting the data to compute them is covered in how to get your interval data. The document that governs every definition above is the filed tariff, and it is a public one — obtain it before spending money on the strength of anything written here.