What is peak shaving? Peak shaving vs load shifting

Peak shaving reduces a site's highest metered demand, the average kW or kVA over one billing interval, so that demand, capacity and excess capacity charges fall. Load shifting moves consumption, in kWh, to a cheaper or less constrained time: it changes when energy is used, not necessarily how high the peak is.

Two sites can cut the same kWh and see very different bills, because peak shaving and load shifting reduce different lines on it. One action can do both, but each one reduces a different line on the bill. Peak shaving reduces the charge per kW or kVA: a demand charge, a capacity charge, or the penalty for exceeding the connection limit. Load shifting reduces the charge per kWh where the unit price changes through the day. Before you choose, find the larger of the two lines on the site's bill.

Side by side

What is a demand charge?

A demand charge bills the highest average demand in one metering interval of the billing period, in €/kW, £/kVA or $/kW per month. The meter records energy per interval, and the supplier or network divides it by the interval length to get demand. At €10 per kW per month, a site whose highest half hour averages 180 kW pays €1,800 for that month, even if every other half hour is below 120 kW.

The name and the unit differ by market:

Peak shaving pays wherever one of these lines is large. A flat energy-only tariff gives it no value unless the site is close to its connection limit, where a lower peak can avoid a connection upgrade.

Worked example: moving a process

A site draws a steady 100 kW in the day and 40 kW at night. A batch process adds 80 kW for 30 minutes at 17:30 every working day. The process is moved to 23:30:

Moving the process shifts 40 kWh and also shaves the peak by 60 kW. For illustration, take a demand charge of €10 per kW per month, a day rate of €0.25/kWh, a night rate of €0.15/kWh and 22 working days:

The demand saving is nearly seven times the energy saving, because one 30-minute event sets the demand charge for the whole month. In the 2022/23 ESB Networks schedule for maximum demand groups DG7 to DG9b, a 17:30 start also falls in the 17:00 to 19:00 peak unit rate. There, night rates start at 23:00, so a move to 22:00 would still pay the day rate.

Worked example: a battery

The process stays at 17:30. A battery controller holds site import at 130 kW. For 30 minutes the site needs 180 kW, so the battery supplies 50 kW, which is 25 kWh. With 90% round-trip efficiency at the AC terminals, the battery takes about 28 kWh back from the grid. The auxiliary and HVAC loads of the battery add to this, most of all when it cycles lightly.

The recharge is part of the design. At 50 kW, 28 kWh takes about 34 minutes. At night the site load is 40 kW, so import during the recharge is 90 kW, under the threshold. If the battery recharges at 14:00 on the 100 kW daytime base, import rises to 150 kW. That sets a new peak and gives back 20 kW of the 50 kW saving.

With the same illustrative tariff, the demand saving is 50 kW × €10 = €500 per month. The energy line changes by only 25 × €0.25 − 28 × €0.15 = €2.05 per day, about €45 per month. The battery earns its money on the demand line.

Size the energy from the area of the load above the threshold, over the longest peak you expect. If the process overruns to 45 minutes, the battery must supply 50 kW × 0.75 h = 37.5 kWh, not 25 kWh. Add a margin for forecast error and for a state of charge below full at the start of the peak, and subtract any energy reserved for other services. Once the month's peak is set, raise the threshold to that value for the rest of the billing period. Shaving below it saves nothing on the demand line and costs battery cycles. The BESS guide explains which system owns each limit.

The demand interval sets the bill

The billed demand is the average over one metering interval, not the instantaneous peak. The interval is usually 15 or 30 minutes. A 60 kW spike that lasts 2 minutes on a 100 kW base adds 60 × 2 / 30 = 4 kW to a 30-minute average. The same 60 kW held for the full 30 minutes adds 60 kW. The interval demand calculator converts the energy in one interval to average kW. For the spike, the half hour holds 50 + 2 = 52 kWh:

Because the bill counts the average, a controller can make up for a late response in the same interval, if the battery has spare power. The next section works through it.

Some tariffs keep a peak on the bill for longer than one month. The U.S. Department of Energy describes ratchet clauses, where the billed demand is the higher of this month's peak and a percentage of the peak in the previous 11 months. Other tariffs charge demand only in set hours of the day, so a peak outside those hours costs nothing.

The load factor calculator shows how peaky a profile is:

A load factor of 50% says that the average demand is half the peak. It does not say whether shaving is cheap. That depends on the shape of the peak. A 30-minute peak each day above a flat base needs 25 kWh of storage to remove 50 kW. A plateau from 08:00 to 18:00 at the same load factor needs 500 kWh to remove the same 50 kW. The electricity cost calculator puts a value on the demand line.

When the battery responds late

A 130 kW target over a 30-minute interval allows 130 × 0.5 = 65 kWh of import. Suppose the battery in the example starts 3 minutes late. For those 3 minutes the site imports the full 180 kW, which is 9 kWh. That leaves 56 kWh for the remaining 27 minutes (0.45 h), an average import of 56 / 0.45 = 124.4 kW. The battery must therefore discharge 180 − 124.4 = 55.6 kW, not 50 kW, for the rest of the interval.

The recovery is arithmetically possible only when the battery can deliver more than the planned power. With a 50 kW limit, the import for the interval is 9 + (180 − 50) × 0.45 = 67.5 kWh, an average of 135 kW. A controller that acts only on the instantaneous reading gets the same 135 kW.

Value the result conditionally. If this interval set the month's 180 kW maximum and no other interval exceeds 135 kW after the action, the billed maximum falls by 45 kW, worth €450 at the illustrative €10 per kW per month, instead of €500 for the on-time case. If another interval in the month exceeds 135 kW, the billed peak falls by less, or not at all. A capacity charge on an agreed MIC does not change unless the MIC is reduced.

Recharge the energy actually discharged. With the idealised 90% round-trip efficiency at the AC terminals and the same state of charge at the start and the end, 22.5 kWh needs 22.5 / 0.90 = 25 kWh from the grid. At a constant 50 kW that takes 30 minutes, and on the 40 kW night base it raises import to 90 kW. The real figure also includes battery auxiliaries and depends on temperature and where efficiency is measured. Check every interval the recharge crosses: a 30-minute recharge starting at 23:45 touches two half hours.

The 3 minutes is illustrative. The complete response time is the sum of measurement availability, the decision delay, command delivery, the asset's ramp and the time until the change shows at the connection point. A command acknowledgement is not a physical response, so measure each stage at commissioning. The controller's meter is not the billing meter either: reconcile the two over several weeks and set the threshold below the target by the measured difference plus the documented uncertainty of both metering chains.

Capacity charges in Ireland and Great Britain

Irish and GB network tariffs for larger business sites include a capacity charge on the Maximum Import Capacity (MIC) that the site agreed with the network, in kVA. ESB Networks states that a MIC set too high means paying for capacity the site does not use, and a MIC set too low brings an excess capacity charge.

A lower peak therefore reduces the capacity charge only when you also reduce the MIC. UK Power Distribution, a GB network operator, allows one reduction in 12 months, and warns that the old capacity may not come back without reinforcement (charging statement, paragraphs 2.40 and 2.41).

The excess charge is where one peak costs most. UK Power Distribution calculates the kVA for every half hour from the active and reactive energy, and applies the highest value in the billing period. It charges each kVA above the MIC per day, for the whole billing period in which the breach occurs (paragraphs 2.43 to 2.45). In Ireland, the 2022/23 ESB Networks schedule sets the surcharge for interval-metered sites at 5 times the capacity charge rate on each excess kVA, for the billing period (note 5 for groups DG6 to DG9b). In both cases a single metering interval behaves like a ratchet for one billing period.

Because the limit is in kVA, power factor matters. At a power factor of 0.9, 180 kW is 200 kVA. Both networks also charge for reactive energy above one third of the active energy, which is an average power factor of about 0.95. A battery inverter that supplies reactive power, or power factor correction, can reduce the kVA peak when a kW reduction alone is not enough.

Peak shaving vs demand response

Peak shaving and demand side response use the same loads, generators and batteries, but they answer to different parties.

The two interact. Energy reserved for a demand response contract is not available to shave the site's own peak, and the reverse. A site that shaves its peak every day also lowers the recent-day average that many demand response baselines use, so less of the next event counts as delivered. The demand response baseline guide shows how each baseline method treats it. Decide the priority order in the controller before the site joins a programme.

Peak shaving vs load levelling

Load levelling flattens the whole profile towards its average. Storage charges in the low periods and discharges in the high periods, so the site draws close to a constant power. Peak shaving acts only above a threshold and leaves the rest of the profile alone.

The difference is mostly one of energy. In the load factor example above, a site averages 50 kW against a 100 kW peak. Shaving a 30-minute peak of 50 kW above a flat base takes 25 kWh. Levelling a site that sits at 100 kW from 08:00 to 18:00 down to its 50 kW average means moving 50 kW × 10 h = 500 kWh every working day, and charging it back in the night valley. Levelling suits a site that pays mainly for energy at time-of-use rates, or a network operator relieving a constrained feeder. Peak shaving suits a site that pays mainly for its single highest interval.

Watch for the rebound

Most shed or shifted loads come back later. The exception is a load that is simply not needed, such as lighting in an empty area. A battery must recharge. EVs that were held back want more power later (the EV smart charging guide covers this). Thermal plant makes up the energy it missed.

A simple energy balance gives the size of a thermal rebound. A chiller that runs at 60% duty and stops for 30 minutes owes 0.6 × 30 = 18 minutes of full-capacity cooling. When it restarts at full capacity, only the 40% above its normal duty goes to pay this back, so it runs flat out for about 18 / 0.4 = 45 minutes. This is a lower bound: restart delays and any overshoot below the set point add to it. If the chiller draws 50 kW at full load and 30 kW on average, it adds 20 kW to the site for one and a half intervals after the event. If that lands on the batch process, the site sets a new peak.

Check the 30 to 60 minutes after each action as carefully as the peak itself. The EV charging capacity calculator checks whether chargers fit under the site limit in each interval.

Measure before and after

Record at the tariff's demand interval:

Compare equivalent periods: the same weekdays, similar weather in degree days, and similar production output. Report the measured peak reduction separately from the financial value, which depends on the actual contract. The savings guide explains how to adjust a baseline.

Peak shaving with EpiSensor

ZEM electricity monitors on the incomer and the large circuits show which loads make the peak and how long it lasts. Edge on the Gateway runs the threshold as a local Automation rule: a data trigger on site import, an open delay to ignore short spikes, and a device command that switches a relay or writes a set point. The rule runs on the Gateway, so it does not depend on an internet connection. Edge checks duration conditions once a minute, so a rule with an open delay acts a minute or more after the threshold is crossed. Set the threshold with that margin.

The same loads can also be sold to the grid through an aggregator. The ZDR demand response controller combines metering with a local response to grid frequency, through a relay for a load or generator, or a set point for a battery or UPS. A virtual power plant is built from these actions, and the demand response solution shows the full system.