A flat energy rate charges the same for a kilowatt-hour at three in the morning and at four on an August afternoon. Those two kilowatt-hours cost the utility very different amounts to supply, and a time-of-use schedule passes that difference on.
For a site with genuine timing flexibility, this is the most favorable structure available. For a site with none, it is a way of paying more for exactly the same operation.
The structure
The day is divided into periods, each with its own energy rate.
Three features of the structure matter more than the rates themselves.
The demand charge is frequently windowed. Many time-of-use schedules measure demand only inside the on-peak period. A spike at three in the morning then costs nothing at all, where on a facility-demand schedule it would cost full price. That single provision can be worth more than the entire energy differential to a site that can move load overnight.
Seasons change everything. Summer and winter definitions differ, sometimes with an on-peak window that moves by hours. A control schedule set once in June will be wrong in December.
Critical peak pricing may sit on top. Some schedules include a limited number of critical days per year, called with short notice, priced far above the normal on-peak rate. The exposure on those days can dominate the annual result.
Whether it suits you
The test is not a rate comparison. It is how much of your consumption can genuinely move, and at what operational cost.
Modeling a move to a time-of-use schedule
Same consumption, priced two ways, before any load is shifted.
- Annual consumption3,100,000 kWh
- Current flat energy rate6.6 ¢/kWh
- (Annual energy cost, flat rate)$204,600
- Consumption falling in the on-peak period, as operated today38%
- (On-peak kWh: 3,100,000 × 0.38)1,178,000 kWh
- On-peak rate12.4 ¢/kWh
- Off- and mid-peak rate, blended4.9 ¢/kWh
- (On-peak cost)$146,072
- (Remaining 1,922,000 kWh at 4.9¢)$94,178
Annual energy cost on the time-of-use schedule$240,250
Switching without changing anything costs this site $35,650 more. The schedule only becomes attractive once load actually moves — which is why the modeling has to be done on a shifted profile, not the current one. Rates illustrative.
That result is the standard trap. A site that switches to a time-of-use schedule intending to shift load later, and then does not shift it, is worse off than it started. The schedule rewards behavior, and the behavior has to be real.
Run the model twice: once on today's profile, and once on the profile you can credibly commit to. If the second one depends on operational changes nobody has agreed to, use the first.
Getting the shifting right
Know the window precisely. Start and end times, days of the week, named holidays, and both seasonal definitions. A schedule that is thirty minutes wrong runs expensive load in the expensive period, which is the exact failure the strategy exists to prevent.
Automate it. A period boundary that depends on somebody remembering will be missed, and the miss will happen in August.
Check what the demand determinant does. If the schedule bills a windowed demand charge, shifting load out of the window reduces it substantially. If it also bills a facility demand across all hours, the shifted load may create a new peak elsewhere and the saving halves: facility, on-peak and billing demand.
Watch the recovery. Load pushed out of the on-peak window tends to land immediately after it closes, and if enough of it lands in the same fifteen minutes the result is a new peak just outside the window. Staggering applies here as much as at a cold start: staggered startup.
Where the flexibility usually is
Sites consistently underestimate this, because nobody has ever been asked. The candidates worth examining:
- Battery and forklift charging, which almost always has latitude.
- Tank and vessel heating, where the requirement is a temperature by a deadline rather than continuous operation.
- Cooling, through pre-cooling or storage: pre-cooling and HVAC scheduling against the peak.
- Compressed air, where receiver capacity allows the compressors to be run outside the window.
- Non-urgent batch processes, cleaning cycles and testing.
- Water treatment and pumping where storage exists downstream.
What is rarely flexible: anything with material in process, anything with a quality or safety consequence, and anything customer-facing.
The way to establish which is which is to ask the people who operate the equipment, load by load, and to write the answers down with a number attached: how many kilowatts, movable by how many hours, at what cost. That inventory is the input to the model, and it is the same inventory needed for demand limiting, for demand response and for battery sizing. Building it once and keeping it current makes every subsequent question on this site answerable in an afternoon.
The relationship to real-time pricing
Time-of-use is a coarse, predictable approximation of the same idea that real-time pricing implements precisely. Time-of-use gives you fixed periods known a year ahead; real-time pricing gives you actual hourly prices with far more variation and far more risk.
A site that succeeds under time-of-use is a candidate for real-time pricing. A site that struggles under it is not: real-time pricing.
Before switching
Model it properly against your own intervals, on both today's profile and the committed one, and read the switching rules before electing — minimum stay obligations are common and a schedule that turns out to be wrong may not be reversible for a year: switching rate schedules and how to choose a rate schedule.
