10 checks before you specify a battery for peak shaving

1. Read your own demand profile first

Your TNB statement reports maximum demand for the period. Track it across twelve months. Stable month to month means peaks are structural — a recurring operational pattern, the case most amenable to peak shaving. Highly variable means something episodic is setting it, and displacing an unpredictable peak requires the system to stand ready continuously. Steadily rising is a capacity trajectory question, not a peak-shaving one.

2. Rule out the operational fixes

Load sequencing — staggering start-up so large loads energise in sequence often produces a meaningful reduction at zero capital cost. Soft starting — where motor inrush contributes, soft starters or variable speed drives address it directly with running-efficiency benefits too. Process scheduling — moving a discretionary high-demand process out of the peak window reduces coincidence. A supplier proposing storage without asking whether these were considered is selling equipment, not solving your problem. Any storage added must still meet the electrical installation requirements administered by the Energy Commission, and its Guidelines on Battery Energy Storage System (BESS) Safety.

3. Work out the power and the duration separately

Take a plant with an 800 kW baseline rising to 1,100 kW for twenty minutes at each of two shift changes. Displacing that needs 300 kW for 20 minutes — around 100 kWh per event, or 200 kWh a day. High power, modest energy: the configuration where peak shaving is most cost-effective.

Change one variable. If the same 300 kW excess persists for three hours, the requirement rises to roughly 900 kWh per event — nine times the capacity for the same demand reduction. The power rating is unchanged; the cost is not.

4. Insist on interval data before any proposal

Monthly bills report the 1,100 kW figure but say nothing about whether it lasted twenty minutes or three hours. Sizing needs demand recorded at regular intervals across at least a month, ideally longer to capture seasonal and production variation. Any battery proposal issued without interval data is an estimate presented as a specification — including ours.

5. Check the cycling assumption against your duty

A system operating twice daily accumulates cycles far faster than one operating twice weekly, and warranties are typically expressed in both years and throughput. A proposal quoting design life without reference to your duty cycle has not engaged with how the system will actually be used.

6. Separate the solar case from the storage case

Solar reduces the energy charge by generating kWh you would otherwise buy. Storage configured for peak shaving reduces the demand charge by lowering the recorded maximum. They are complementary, and on some sites the combination is what makes either viable — but they are separate investment cases, and a proposal blending them into one figure is obscuring which component earns the return.

Run your own arithmetic: take the capacity and network charges on your most recent statement — since July 2025 these replace the former single maximum demand line —, multiply by the kW you could realistically displace, and annualise. That number against installed cost is the whole case.

What a defensible study produces

A proposal worth acting on states the measured demand profile it was calculated against, the proposed threshold and resulting reduction, the power and energy capacity required, the expected number of operating events per year, the depth of cycling and its implication for battery life, and the projected annual saving with assumptions visible.

If any of those are missing, the proposal is a product recommendation wearing the clothes of an engineering study.

Combining objectives without over-promising

Storage sized for peak shaving can often serve a secondary purpose — riding through short interruptions, or shifting solar generation into the evening. Whether it can do so simultaneously depends on the control strategy and whether capacity reserved for one function remains available for another.

This is worth resolving explicitly. A system marketed as delivering demand reduction, outage resilience and solar time-shifting may deliver all three, or may deliver whichever the controller prioritises at the moment you need the others. Ask how the control strategy arbitrates when objectives conflict.

What creates a demand peak in the first place

In manufacturing, peaks are usually coincidence rather than any single machine: shift start-up energising multiple large loads within a short window, simultaneous motor starting where inrush is several times running current, and batch equipment — furnaces, ovens, compressors, chillers — cycling into phase with one another.

Frequently no individual load is problematic; the peak exists because several arrive together. That is also why peak shaving can work, because the excess above baseline is often brief.

7. Check whether your peak is seasonal

A month of interval data captures the pattern; twelve months captures the exceptions. Production peaks before a shutdown period, seasonal cooling load, or an annual process run can set the maximum for a period without appearing in a short study. Size against the profile you actually operate, not the quietest month you happened to measure.

8. Confirm how the threshold is set and who can change it

The discharge threshold determines the reduction achieved. Establish whether it is fixed, scheduled, or adaptive — and who can change it after handover. A threshold set conservatively at commissioning and never revisited leaves saving on the table; one set aggressively without headroom will occasionally fail to hold the peak, which loses the entire month’s benefit.

9. Ask what happens when the battery is unavailable

Maintenance, a fault, or a state of charge too low to cover an unexpected peak all mean the system is not there when the peak arrives. Because maximum demand is recorded across the whole period, a single missed peak can undo a month of successful shaving. Ask what the availability commitment is and how the system behaves when it cannot meet the threshold.

10. Separate the guaranteed saving from the modelled one

A proposal may present an annual figure that assumes every peak is successfully shaved. Ask what proportion of events the model assumes are captured, and what the figure becomes at ninety per cent. If the supplier will not put a performance commitment behind the number, treat it as a projection rather than a saving.