Fifteen minutes set your demand charge for the month.
The demand charge is billed on your single highest 15-minute interval of on-peak demand, not on what you consumed. A battery that covers that interval brings the charge down for the whole month. It needs no solar. It does need to know the interval is coming, and that is what CESS is for.
Modelled at the 120 kWh / 60 kW system our engine sized for the pilot site, from the site's own 15-minute metered load.
A demand charge worth cutting, and a load that spikes.
This page is for a factory on a time-of-use tariff whose demand charge is a line worth attacking. You do not need solar: peak shaving works on grid supply alone.
Billed on one interval
The charge is priced per kilowatt of your highest 15-minute average in on-peak hours. What the rest of the month looked like does not change it. How the tariff sets it.
Spikes, not a plateau
A compressor starting while the line is already running, or two shifts overlapping: short peaks well above your usual draw. A short peak takes little stored energy to cover, so it is the cheapest kind to shave.
Solar or not
Peak shaving and arbitrage need no array. If you have one that makes more than your plant can use at midday, solutions for solar producers is the better page: there, recovering that solar carries the case.
Keep the charge for the interval that counts.
A battery lowers the demand charge only if it still holds charge when the month's peak arrives. Spend it on the way up and it is empty at the crest, and the crest is the interval you are billed on.
Your load predicted up to 24 hours ahead from your site's own history, so the likely peak is known before the shift that causes it.
Plans the day with model-predictive control, keeping the battery for the crest rather than spending it the first time load crosses a line.
The battery supplies the top of the peak, so the grid sees a lower maximum and the charge for the month comes in lower.
Monitors cell health and degradation continuously, and runs the electro-healing routine built into CESS.
PJJ Dispatch shows one afternoon under a rule-based controller and under MPC, with the same battery. The rule-based controller aims lower and is billed higher.
Off the peak, the same battery earns the spread.
A time-of-use tariff charges more on-peak than off-peak. The battery can charge off-peak and cover part of your on-peak load, earning the gap between the two rates, less what the round trip loses. Every kilowatt-hour in the battery can do only one job, so PJJ Dispatch decides which, hour by hour.
What the two were worth at our pilot.
Our pilot is one site in Bangkok, running since May 2025. It has solar, and recovering it is its largest driver. The two below are the ones that need no solar, modelled separately so you can see each on its own. It is one site and a pilot, and the figures say so.
Food manufacturing · Bangkok
Commissioned May 2025 · client anonymised at their request
CESS runs against the site's existing Huawei FusionSolar infrastructure alongside a Deye inverter. Roughly 30% of what the array generated was being curtailed before we connected. The installed battery is a sixth of what our engine sized for this load, so the figures on this page are modelled at full size rather than measured at the pilot's.
Modelled at the 120 kWh / 60 kW system our engine sized for this site, from the site's own metered load. Our fee is a share of measured savings, billed monthly, so the line item only exists in a month where the saving did. Savings do not scale linearly with battery size, so we will not multiply these up for you — your figures come from your own model.
At the pilot, peak shaving is the second of three drivers.
Solar recovery is larger there because the site has an array to recover. On a site without one, the case rests on peak shaving and arbitrage, and whether that pays depends on the shape of your peaks. The investment case shows how each driver was modelled.Modelled on the interval you are billed on.
The demand charge is set by a 15-minute average, so we model it from 15-minute data. An hourly series averages a short spike in with the quieter minutes around it, and hides the very peaks that set the bill.
Peak shaving pays least on a flat load.
If your demand stays near its peak for hours at a time, there is no short spike to take off the top, and covering the whole plateau takes a much larger battery. The model tests that before you spend anything, and a flat load is one of the ways the answer comes back no.Find out what your peak is costing you.
Twelve months of billing data, and we will model your load, size the system and show you the whole financial case, including when the answer is not to build.
Or reach us at solarbuddies@pjjsolutions.tech