Deliverable energy

What it is (precise)

The calculation starts where usable energy stops. Usable energy is the swept SOC window times present capacity — a DC quantity at the racks. Deliverable energy takes that quantity and walks it down the one-line diagram: through the PCS, where conversion loss is taken; through the medium-voltage step-up transformer, with its no-load and load losses; through the collection cables to the substation; and past the auxiliary load — HVAC, controls, thermal management — that runs while the plant discharges.

What arrives at the named bus is what the plant can deliver there. The measurement-boundary entry covers which bus a number is defined at; this entry is about what happens to the energy between buses.

Three quantities sit on the same axis and get conflated constantly. Usable energy is the DC-side starting stock. Deliverable energy is the capability at a boundary — what the plant could meter there if dispatched for a full discharge right now, under stated conditions. Delivered energy is what the revenue meter actually recorded after the fact, which folds in dispatch instructions, availability and curtailment on top of capability.

The alias trap is that 'delivered energy' gets used loosely for both of the last two: a delivered figure is evidence about deliverable capability only when the dispatch that produced it swept the full window at a known rate and temperature — which is exactly what a capacity test is.

The loss chain from usable to deliverable

Conversion first. Modern central PCS datasheets put maximum one-way efficiency around 98.5-99.2 percent; quoted together with the MV transformer, the same class of hardware reads nearer 98 percent. Two habits keep those figures honest. Read the weighting: the peak number holds at one favourable operating point, while CEC- or Euro-weighted figures — averages over a standardized load profile — run a few tenths lower, and real part-load operation sits below the headline.

And read the footnotes, because they move the boundary inside a single component: one vendor's maximum efficiency is measured without the internal power supply, another's includes the MV transformer. That is the measurement-boundary discipline applied at component scale, and it decides whether the datasheet number already contains losses you were about to subtract again.

Then the passive chain and the parasitics. The step-up transformer passes roughly 99-99.5 percent each way — slightly better at partial load, where load losses fall away faster than the constant no-load loss — and the collection cables take their I²R share on top.

Auxiliary consumption commonly amounts to about 1-3 percent of annual throughput depending on climate and thermal design, and where it is fed from matters as much as its size: station service drawn from the battery side of the meter reduces deliverable energy directly, while a separate upstream feed leaves the metered figure alone and shows up in the operating cost line instead.

Multiply the discharge leg through and the familiar anchor appears: a usable window around 90-95 percent of nameplate for grid-scale LFP, PCS and transformer conversion, cables and auxiliaries — and deliverable AC energy at the POI lands around 85-92 percent of DC nameplate at beginning of life. Treat that band as a sanity check on a project-specific calculation, never as a value to copy: each factor in the chain has a range, and a hot site with heavy HVAC duty and long collection runs sits at the bottom of it.

The conditions that move it

Deliverable energy is not a constant of the plant. Temperature moves both ends of the chain: cold shrinks the extractable band and forces the BMS to derate power near the window edges, while heat raises HVAC load and can derate the PCS. The SOC window in force moves the starting stock before any loss is counted — a firmware update, a warranty renegotiation or an EMS margin release changes deliverable energy without touching a single piece of hardware, which is why the window that applies today, not the one on the datasheet, belongs in the calculation.

Current does the quiet damage. At a higher discharge rate, I²R losses in cells, busbars and cables grow with the square of current and the voltage sag deepens, so the weakest rack reaches its voltage floor earlier and the discharge terminates above the nominal SOC floor. The same window therefore yields fewer MWh at rated power than at a gentle test rate — a capacity test at a low C-rate and a full-current market dispatch are two different deliverable-energy measurements of one plant.

Resistance growth compounds this with age, which is why deliverable energy at rated power fades faster than a capacity-only view of degradation suggests. A complete statement of deliverable energy carries five stamps: the boundary, the temperature, the rate, the window in force, and the date.

Deliverable versus contract energy

Contract energy is the guaranteed MWh — fixed by a meter, a test condition set and a year, and enforceable by a counterparty. Deliverable energy is what the plant can physically do at that meter today. The two are designed never to meet: the contract number sits below deliverable capability by a margin the project pays for, first as DC overbuild of roughly 10 to 25 percent above the day-one contract quantity, later as augmentation.

Deliverable energy fades with the fleet; the contract number sits still or steps down its agreed schedule; the vertical distance between the two curves is the project's risk buffer, and managing it is what the augmentation program is for.

The capacity test is where the two quantities touch. A test is nothing more than a witnessed measurement of deliverable energy at the named meter under the contractual condition set, compared against the guaranteed number — pass or fail is simply whether deliverable still clears contract on the day.

That makes trend-tracking the practical discipline: measured deliverable energy at a consistent boundary and condition set, plotted test over test against the guaranteed curve, shows the crossing point years before it arrives. An operator who only looks at pass/fail discovers the shortfall at the test that fails; one who tracks the gap orders racks while there is still schedule to install them.

Common pitfalls

The recurring error is boundary slip between documents: a nameplate DC figure or a usable DC figure standing in for deliverable AC at the meter. Every handoff — datasheet to bid, bid to financial model, model to offtake exhibit — is a chance for the number to silently change buses, and the loss chain between the racks and the POI is worth several percent, which is real MWh at settlement.

The counter is mechanical: build the loss chain once, from cells to meter, label every intermediate value with its boundary and conditions, and check what any vendor 'usable' or 'deliverable' figure already includes before subtracting anything from it — some quoted figures have netted off part of the chain, and deducting the same loss twice understates the plant as badly as skipping it overstates it.

The subtler trap is confusing capability with history. A month of low delivered energy may mean curtailment or a market that never called for full duration — not a plant that lost capability; a year of comfortable delivered numbers can mask fading deliverable energy if no dispatch ever swept the whole window.

Delivered energy is evidence, not measurement, and the only clean read on deliverable capability is a full-window discharge under known conditions. Treating one beginning-of-life deliverable figure as a constant in a spreadsheet makes the same mistake in the other direction: the quantity moves with season, rate and age, and the year-by-year curve — not the day-one number — is what the contract will eventually be tested against.