Why half-hourly demand data changes the system size answer

Headline annual electricity consumption tells you almost nothing about the right solar system size. The half-hourly profile changes the answer dramatically — and the size that maximises NPV is rarely the size the installer initially proposes. Half-hourly data is worth insisting on in discovery for any project above 100 kWp, because system designs change when it is used.

Why annual consumption is misleading

A commercial site consuming 800 MWh per year could be: a 24/7 logistics operation with flat overnight load (suggests 1.2–1.5 MWp solar with significant self-consumption), a daytime-heavy manufacturer (suggests 700 kWp at very high self-consumption), a school with summer shutdown (suggests 350 kWp because excess summer generation has nowhere to go), or a hotel with seasonal variation (suggests battery-storage-paired sizing). Same annual consumption, four different optimal answers.

The unhelpful default is sizing solar to "60% of annual consumption" or some similar rule-of-thumb. This works on average but is wrong on most specific sites. The right approach: model self-consumption percentage at a range of system sizes against half-hourly demand, and pick the size that maximises NPV.

What the data shows

In our recent reviews of half-hourly profiles for commercial sites considering solar, the optimal system size has differed from the "60% rule" answer by an average of ±18%. Some sites should be sized substantially larger than rule-of-thumb (24/7 cold storage, food production with continuous refrigeration); others substantially smaller (offices with single-shift operation, summer-shutdown education sites).

The economic implication is meaningful. Oversizing reduces project IRR by 1.5–3 percentage points because the marginal kWp at the high end of the size curve has lower self-consumption (more export, lower per-kWh value). Undersizing leaves project NPV on the table by limiting the cost-effective scale.

How to get the half-hourly data

For sites on a half-hourly settled tariff (most commercial sites above ~100 kW peak demand), the supplier holds the data and is required to provide it on request. The relevant request term is "HH data" or "AMR data" — the supplier issues a CSV file with the past 12–24 months of consumption at 30-minute intervals.

For smaller commercial sites on monthly-settled tariffs, half-hourly data may not be available. Two workarounds: (1) install a temporary smart meter or current transformer with a logger for 4–6 weeks across a representative period; (2) use industry-standard load profile estimates calibrated to monthly consumption — workable for offices and retail but inadequate for industrial or seasonal businesses.

The modelling approach

Model self-consumption as: for each half-hour interval, take min(solar generation in that interval, demand in that interval). Sum across the year to get total self-consumed solar. Generation in each interval comes from a NASA POWER irradiance model for the specific lat/lng, adjusted for orientation, tilt, and shading. Demand in each interval comes from the half-hourly data file.

Run this calculation for a range of system sizes (e.g. 100 kWp to 1.5 MWp in 50 kWp increments). Plot self-consumption percentage against size — typically a curve that's flat at small sizes (high self-consumption, but small absolute generation) and falls as sizing increases (more export, lower self-consumption). The optimal size sits where marginal generation × marginal value-per-kWh equals marginal capex per kWp.

Marginal value-per-kWh is the deciding variable: it's avoided-cost (typically 14–28p) for self-consumed energy, and SEG-tariff (typically 4–8p) for exported energy. The optimal size is generally where the marginal kWp delivers ~70% self-consumption — at higher self-consumption you're leaving capacity on the table; at lower, you're discounting capex against export value.

Battery storage in the same model

Adding battery storage to the half-hourly model is straightforward: at each interval, solar generation in excess of demand goes to battery (up to capacity). Battery discharges at intervals where demand exceeds solar generation (subject to round-trip efficiency). The optimal battery size emerges from the same marginal-value calculation.

In practice: for daytime-heavy demand profiles, battery rarely justifies marginal capex. For night-heavy or evening-spike profiles, battery can move project IRR upward by 3–8% on the storage capex alone (per the storage piece earlier).

Where this changes our recommendations

A live example from recent work: a Yorkshire food production facility with 1.4 GWh annual consumption was originally proposed by a quoting installer for 700 kWp solar. Half-hourly modelling showed the site has heavy 24/7 cold-storage demand with overnight load consistently above 80 kW. Optimal size landed at 1.1 MWp — the additional 400 kWp returned a 16% IRR on incremental capex versus the 12% on the base case, because overnight self-consumption supports the larger system.

Conversely, a Norfolk distribution warehouse with the same 1.4 GWh consumption but single-shift operation (7am–7pm, weekday-only) returned an optimal size of 600 kWp, not the 700 kWp originally proposed. The marginal 100 kWp had self-consumption below 50% and didn't justify the capex.

Two sites, identical headline consumption, optimal sizes 1.1 MWp and 0.6 MWp. The half-hourly profile was the only thing that distinguished them — and getting it wrong means leaving 5–10% of NPV on the table.