Demand charges are a different bill

An energy charge bills the kilowatt hours a site used. A demand charge bills the single highest interval it reached. They respond to different work, and an array that lowers one can leave the other exactly where it was.

A commercial electricity invoice on a demand tariff contains two charges that behave nothing like each other, and much of the confusion about solar returns starts here.

The energy charge bills consumption. It is kilowatt hours multiplied by a rate, it accumulates through the month, and every kilowatt hour avoided reduces it by the same amount.

The demand charge bills a single measurement. It is the highest demand the site reached inside a defined window during the period, and the rest of the month has no effect on it at all.

The units give it away before anything else does. An energy line is priced per kilowatt hour. A demand line is priced per kilowatt or per kilovolt ampere, with no hour attached, because it is not measuring a quantity of energy at all.

That is the difference in one sentence. The energy charge is a total. The demand charge is a maximum.

A site can cut its consumption substantially and see the demand charge unchanged, because the interval that set it was one afternoon in the third week and nothing that happened around it mattered.

Understanding which of the two a project is aimed at is the difference between a result and a disappointment, and it is established by reading an invoice rather than by modelling anything.

The reason the charge exists is physical rather than commercial. A network has to be built for the maximum its customers might simultaneously draw, not for the average they actually draw.

Conductors, transformers and switchgear are sized for peak. That capacity has to exist whether it is used for six hours a year or six thousand, and the demand charge is how the cost of holding it is allocated.

Which is why the charge responds to shape rather than to volume. Two sites consuming identical annual totals pay different demand charges if one draws steadily and the other runs a plant load in short, sharp blocks.

It is also why reducing the peak has a value beyond the invoice. A lower peak is less switchboard capacity, less cable, and sometimes the difference between a connection that works and one that needs upgrading.

The published tariff structures and the prices attached to them are set annually and are public¹. The structure is the part worth reading. The price is the part that changes.

There is a second consequence worth holding on to. Because the charge is set by a maximum, the work that reduces it is concentrated rather than continuous. A site does not have to run differently all year. It has to avoid a small number of intervals.

Three things have to be known before a demand charge can be reasoned about: what is measured, when it is measured, and how long a period the maximum is held over.

The window is the part that catches people. Several commercial demand tariffs measure only within a defined period, commonly a weekday afternoon and evening block, which means demand at ten in the morning does not set the charge however high it goes².

The averaging period matters just as much. Demand is measured over an interval rather than instantaneously, so a brief inrush as a motor starts is not the same thing as a sustained interval of high load.

And some tariffs hold the maximum for longer than the month it occurred in. Where that applies, a single event can set the charge for several billing periods, which changes the value of avoiding it considerably.

None of these are difficult to establish. All three are on the invoice or in the published price list for the tariff the site is assigned to.

Many commercial demand tariffs charge in kilovolt amperes rather than kilowatts, and the difference is money.

Kilowatts are real power, the part doing work. Kilovolt amperes are apparent power, the total the network has to deliver, including the reactive component that motors and transformers draw and give back without consuming.

The ratio between them is the power factor. A site with a poor power factor draws more apparent power for the same useful work, and on a kilovolt ampere tariff it is billed for the difference.

Motors are the usual reason. An induction motor draws magnetising current that does no useful work and is returned each cycle, and a site full of pumps, fans and compressors running lightly loaded is a site with a power factor problem it cannot see on a consumption figure.

Correction is an established piece of electrical work rather than a novelty. Capacitors, or an active unit where the load contains significant harmonic content, sized against the site's own measurements.

It is worth checking before anything more ambitious is considered, because it is one of the few interventions that reduces a bill without changing anything the business does.

The related subject, and the one that decides whether correction is straightforward or complicated, is what the site's loads are doing to the waveform in the first place: see "Power quality is the other half of the connection".

One caution. Correction equipment that is oversized or left connected when the load has gone can over-correct the site to a leading power factor, which is not an improvement. It is sized against measurements and it is checked afterwards.

Solar reduces the energy charge reliably. Its effect on the demand charge depends entirely on when the site's maximum occurs.

Where the peak lands in the middle of the day, generation is offsetting it at the moment it happens, and the measured demand falls. Where the peak lands at six in the evening, generation has largely finished and the measured demand is untouched.

The two cases can be told apart in an afternoon. Overlay the generation profile the model predicts on the interval data the site already has, and look at the intervals that set the charge. Either generation covers them or it does not.

This is a common gap between a modelled saving and an actual one on a commercial site. The model reduced consumption correctly and left the demand line alone, and on a site billed this way the demand line is a substantial part of the invoice.

There is a further wrinkle on sites with a demand window that runs into the evening. An array can shift the shape of the day without moving the peak, because the peak was never inside generating hours to begin with.

The instruction that follows is simple enough. Before assuming an array will reduce a demand charge, look at the interval data and find the intervals that set it over the last twelve months. If they are after four in the afternoon, the array is an energy project.

How the load shape is read, and what it settles about array size, has its own post: see "Size it from the load, not the roof".

Three things reduce a measured maximum, and they are worth trying in this order because the cost rises steeply across them.

First, sequencing. Where several large loads start together, staggering them lowers the coincident maximum without reducing anything the site does. Compressors, chillers, ovens, pumps and charging equipment are the usual candidates.

Second, control. A demand controller watches the running interval and sheds or delays nominated loads as the measurement approaches a set threshold. It is well-established technology and it works because the charge is set by a small number of intervals each month.

Third, storage. A battery discharges into the peak, lowering the measured demand at the meter regardless of what the site is doing. It is the most capable answer and the most expensive, and it should follow the first two rather than replace them.

In New South Wales, demand reduction activities sit under a certificate scheme, and eligibility should be established before equipment is ordered rather than after it is installed³.

Whether storage belongs in the scope at all is a separate design question: see "Sizing solar and storage together".

Two cautions on storage. It has to be sized against the depth and duration of the actual peaks rather than against the site's total consumption, and it has to be controlled to discharge into the measured window rather than whenever its state of charge suits.

Everything above is conditional on which tariff the site is actually assigned to, and that is a question with a definite answer.

It is also a question worth revisiting periodically. Tariff assignments change as a site's demand crosses thresholds, and a business that grew into a demand tariff two years ago may never have been told in those words.

Start with the invoice. Identify each line, and separate the energy charges from the demand or capacity charges. Note the unit each is billed in, kilowatt hours, kilowatts or kilovolt amperes.

Then find the tariff code and read its structure in the published price list¹. That is where the window, the averaging period and the duration the maximum is held for are defined.

Then look at twelve months of interval data and mark the intervals that set the charge in each period. That short list is the entire target, and it is usually a handful of afternoons.

Do that before commissioning any work, because it tells you whether the site has a demand problem, an energy problem or both, and those three situations call for entirely different projects.