How to Size a Commercial Battery Storage System for Peak Shaving: Demand Limit, Load Spikes, and Usable kWh

A 1 MWh Battery Can Still Be Too Small for the Peak

For peak shaving, size a commercial BESS from the load above the target demand limit—not from annual electricity consumption or a fixed kWh/kW ratio. PCS power must cover the height of the peak, while usable battery energy must cover the area of the load curve above the demand limit. Then correct for efficiency, usable SoC window, degradation, recharge constraints, and the utility's demand-measurement interval.

NREL's analysis of commercial buildings reaches a similar conclusion: load-profile shape is a major predictor of required BESS size , while demand-charge structure strongly affects whether the project is economically worthwhile.

When a factory asks Ruibit/Dawnice for "500 kWh for peak shaving," I would ask for the interval load file before choosing the cabinet.

Start With the Demand Limit

Consider a hypothetical factory with 15-minute meter data.

Maximum measured demand: 760 kW

Target grid demand: 600 kW

The maximum required BESS discharge power is initially:

Peak-shaving power = 760 − 600 = 160 kW

That immediately tells me something a monthly electricity bill cannot.

A 100 kW PCS cannot fully enforce the 600 kW limit, regardless of whether the battery contains 200 kWh or 2 MWh.

But 160 kW does not tell me battery capacity .

For that, I need the shape of the peak.

One Spike and One Plateau Need Different Batteries

Suppose the load around the peak looks like this:

Required AC-side energy is approximately:

E = (20 + 90 + 160 + 120 + 50) × 0.25 = 110 kWh

So this peak requires roughly:

160 kW maximum discharge power

and

110 kWh AC-side energy

before design margins and losses.

Now imagine the factory stays at 700 kW for three hours .

Required reduction:

700 − 600 = 100 kW

Required energy:

100 × 3 = 300 kWh

The second factory needs less PCS power but substantially more battery energy.

This is why I avoid rules such as:

"Peak shaving requires a two-hour battery."

NREL models commercial storage across multiple energy-to-power ratios and emphasizes that accurately estimating duration is important to total system cost.

The load curve should choose the duration.

Usable kWh Is the Number the Peak Actually Sees

Suppose the worked example requires 110 kWh AC .

For preliminary sizing, assume:

Discharge-path efficiency: 94%

Usable battery fraction: 90%

Then:

Battery energy before usable-capacity allowance = 110 ÷ 0.94 ≈ 117 kWh

Preliminary nameplate energy = 117 ÷ 0.90 ≈ 130 kWh

A theoretical minimum is therefore around 130 kWh under these assumptions.

I would not order a 130 kWh system.

We still need to consider:

future degradation

forecast error

peak variability

temperature derating

minimum reserve SoC

recharge opportunity

and whether another peak occurs before the battery has recovered.

The assumptions above are a worked example, not Ruibit/Dawnice product specifications.

NREL's 2024 commercial BESS benchmark, for example, uses 85% round-trip efficiency as a representative modeling assumption, illustrating why the efficiency boundary must always be stated rather than silently assumed.

The Highest Peak Is Not Always the Hardest Peak

This is where I stop looking only at the annual maximum.

Suppose:

Peak A: 760 kW for 15 minutes

Peak B: 680 kW for 2.5 hours

With a 600 kW target:

Peak A requires:

160 kW × 0.25 h = 40 kWh

Peak B requires:

80 kW × 2.5 h = 200 kWh

Peak A determines more of the PCS requirement.

Peak B is harder on battery energy.

So I normally search the interval data for at least:

highest peak

longest peak above target

repeated peaks

seasonal peaks

back-to-back events

NREL's peak-shaving work explicitly models both available battery energy and maximum inverter power relative to facility load because both constrain achievable peak reduction.

One record-high kW value is not a sizing study.

The Demand Limit Is an Economic Variable Too

Why 600 kW?

Why not 550 kW?

The deeper the target, the more of the load curve moves above the threshold.

Suppose reducing the limit from 600 to 550 kW doubles the required battery energy but saves only another $4,000 per year in demand charges.

That extra battery may never pay for itself.

NREL found demand charges to be a particularly important predictor of commercial BESS economic viability.

So I would test several demand limits:

The correct target is not necessarily the lowest technically achievable demand.

It is often the point where the next increment of battery costs more than the peak reduction is worth .

Recharge Can Break a Good Peak-Shaving Design

After the battery removes the afternoon peak, when does it recharge?

If the factory immediately charges at 160 kW while production remains high, the BESS can create another demand peak.

The EMS therefore needs to coordinate:

site load

battery SoC

charging power

next expected peak

tariff window

NREL's storage dispatch work distinguishes peak-shaving control from price-signal dispatch because different objectives can produce different charge/discharge behavior.

For a real Ruibit/Dawnice C&I proposal, I would therefore provide the engineering team with the full interval load profile, tariff, transformer capacity, demand target and available charging window , not just the desired battery kWh.

The Sizing Rule I Would Keep

For a first-pass peak-shaving design:

Required PCS Power = Maximum (Site Load − Demand Limit)

For every interval above the target:

Peak-Shaving Energy = Σ[(Site Load − Demand Limit) × Interval Duration]

Then convert required AC energy into battery nameplate capacity using the project's actual:

efficiency

usable SoC window

degradation allowance

temperature conditions

reserve requirement

Finally, simulate the design across the full load dataset.

That last step matters.

A commercial BESS is not correctly sized because it survives the worst 15 minutes in Excel.

It is correctly sized when it can repeatedly hold the chosen demand limit through the real sequence of peaks, recharge without creating another one, and still meet the economic objective that justified buying it.

Size PCS power from the height of the load above the demand limit. Size usable kWh from the area above that limit. Let the factory's load curve—not a standard battery duration—decide the rest.

FAQs

1. How do you size a commercial BESS for peak shaving?

Calculate PCS power from the maximum load above the target demand limit , then calculate battery energy from the duration and shape of all load intervals above that limit.

2. Why can't peak-shaving BESS capacity be sized from the highest kW peak alone?

The highest peak determines power, but not necessarily energy. A short 160 kW spike may require less battery capacity than an 80 kW reduction sustained for several hours.

3. How do you calculate peak-shaving energy from interval load data?

For every interval above the demand limit, calculate:

Energy = (Site Load − Demand Limit) × Interval Duration

Then add the energy requirements across the complete peak event.

4. What is the difference between usable kWh and battery nameplate capacity?

Usable kWh is the energy actually available for peak shaving. Nameplate capacity must also account for SoC limits, conversion losses, degradation allowance, temperature conditions, and reserve requirements .

5. Can an oversized BESS improve peak-shaving economics?

Not necessarily. A lower demand target requires more PCS power and/or battery energy, but the additional demand-charge savings may not justify the extra system cost. The economic target should be optimized alongside technical sizing.