Track Demand Peaks Before They Become a Cost Surprise

On a Malaysian large-supply tariff, maximum demand is billed on a 30-minute rolling window, and the time-of-use peak band runs from 14:00 to 22:00. One compressor start inside that eight-hour band can set the demand charge for the entire month. The bill that reports it reaches you about a month later — so a maximum demand figure on a bill is not a warning. It is a receipt.

A receipt is not a warning

A demand charge is unusual among operating costs, because it is decided in about thirty minutes and paid over thirty days.

That is the whole problem. By the time the figure is printed, the window that produced it has closed. Nobody was in the plant at 15:47 on a Tuesday, no alarm sounded, and the load that lifted the charge was a machine that had been running perfectly normally all month.

Facility teams usually respond to this by reviewing bills. That answers the wrong question. A bill tells you what the peak was, never what was about to happen — and the difference between those two questions is the entire cost of the charge.

The charge itself has become harder to ignore. Under TNB’s RP4 tariff structure, in force from 1 July 2025, maximum demand is no longer a single line. For a medium-voltage account it is billed as a Capacity Charge plus a Network Charge, and the two together land near RM89.27 per kW on the general scheme or RM97.06 per kW on time-of-use — every month, for the same peak.

Three clocks, and only one of them is still running

Demand is reported on three different clocks at once, and they do not have the same usefulness.

The third row is the only one worth wiring an alarm to. Everything a facility does about demand — staggering starts, rotating chillers, moving a flexible load out of the peak band — only works while the window is still open. After it closes, the number is a fact.

We have covered how the 30-minute window becomes the charge on a TNB bill and how to lay out a monitoring architecture in two earlier notes. This one is about the part those two could not cover: what happens while the window is still open.

So the design question changes. It is no longer how to read the bill better. It is how to see a peak forming, while it is still forming.

From a reading to a warning

An incoming meter can prove a spike happened. It cannot say which circuit caused it, and by the time the reading is looked at, the window is closed. Sub-metering moves the measurement to the circuits that actually set the peak.

The ADW300A wireless metering instrument is built for that position: three-phase measurement in a DIN-rail case, external split-core CTs so no cable has to come out of service, and a demand record the operations team can act on the same afternoon.

Where the meters sit matters as much as what they measure. On most sites the answer splits into four layers.

Which loads can actually be shed

A warning is only useful if something can be done about it. Before any metering is specified, it is worth writing down which loads a facility is genuinely willing to interrupt, and for how long.

The useful measure is not how large the load is. It is the cost of interrupting it, multiplied by how long it can be delayed.

The last column is the one that decides the project. A load that cannot report its own status cannot be counted on during a peak, and a control scheme built on assumptions about machine state will eventually trip production instead of a compressor.

Deciding before the spike, not after the bill

Three things have to exist before a peak can be caught.

What the first month looks like

One limit is worth stating plainly. ADW300A is a metering and recording instrument, not a protection device, and it does not open a breaker on its own. Its digital inputs report machine state and its digital outputs carry a signal to the control system the site already has. The decision stays with the plant; the meter makes sure the decision is made in time.

Questions we are asked