Sizing a project — the method

Status: living draft. This page describes how we think the feasibility stage should guide you from a site to a defensible plant. Some of it is fully built today; a few parts (choosing your commitment, the episode-led reliability headline) are being wired in. Where that matters, the text says so.

The one idea

There is no single “right” plant — there is the plant you can defend operating, and the plant that is theoretically possible. This stage helps you find the first and measure it against the second. You size step by step, read one plain takeaway at each step, and never lose sight of which number is honest and which is optimistic.

Two engines, two meanings

Every result here comes from one of two engines. The badge in each room tells you which.

• Base case — rule-based. The plant replayed hour by hour with no foresight, following simple dispatch rules and your storage strategy. It cannot see a lull coming, so it behaves like a real operator would. This is the number you defend.
• Ideal — LP solver. The same plant solved with perfect foresight — the best it could possibly run. This is a ceiling, not a promise.

The gap between them is itself a result: it tells you how much good control (or better automation) is worth on this site. A small gap means operating strategy barely matters here; a large gap means it matters a lot, and building to the lean ideal is a trap because you will not actually operate that well.

The steps

Each step ends in one takeaway. You will move between them, not march through once.

1. Goal & commitment. What are you sizing toward — a production target (I need X tonnes a year) or your site’s resources (given this wind and sun, what can I make)? And what have you committed to deliver — a yearly total, or a plant that must run continuously? (Choosing the commitment is the part still being wired in; see “What counts as falling short” below.) The yearly number is saved as each product’s promise in the Reliability tab, which judges every result against it; the Design lab’s Find a design and Least cost (LP) start from it — see How feasibility works, “The promise”.
2. Power supply. Size wind and solar (and, if you use it, grid back-up). → “Your supply delivers X GWh a year.”
3. Conversion. Size the electrolyser against that supply. → “It captures Y% of your power; the rest is spilled.”
4. Storage & how you run it. Size the battery and the H₂ buffer, and pick your storage strategy. → “With this storage, you cover the gaps except N episodes a year.”
5. Product. Size the derivative unit(s). → “You make P tonnes a year.”
6. Check it. Read how it runs (Operations), where it falls short (Reliability), and what it costs (Results). → “P t/yr at C $/t, N episodes short.”
7. Compare to the ideal. The LP shows the best this exact plant could do; the Optimizer shows the cheapest sizing that hits your target. → “Perfect control would make more, for less — the gap is what better operation is worth.”
8. Refine. Adjust and re-check until the plant is balanced (below).

The loop — why sizing goes in circles

Steps 2–5 are coupled. How much storage you need depends on the wind-to-electrolyser ratio and on how flexible the product unit is. The two stores also trade off against each other: the battery smooths the short power swings (hours), the H₂ buffer smooths the longer gaps in hydrogen supply to the synthesis unit. Change one size and you change how much of the others you need.

So you loop: size a first pass → simulate → read the signals → adjust the coupled pieces → simulate again. You stop when the plant is internally balanced:

• the energy balance closes (no large unexplained spill);
• both stores actually cycle — if the battery or the H₂ buffer sits full or empty all year, it is mis-sized (money doing nothing);
• curtailment versus shortfall is a trade you accepted, not an accident;
• no stage sits mostly idle.

This loop is a base-case (rule-based) activity — you converge the config from the outside by simulating and adjusting, and the sizing tools (scale, sweep, right-size) are the instruments you turn. The LP / Optimizer does the same convergence internally, in one shot — which is exactly why it lands on a leaner plant than the cautious base case.

What counts as “falling short”

“Falling short” only means something against a commitment, and different projects commit to different things:

• An annual production target — you promised X tonnes a year. Scattered idle hours are fine as long as the yearly total lands. Shortfall = missing the annual number.
• Firm, always-on — the plant (or specific units) must keep running above a minimum rate. Shortfall = any time a firm unit drops below that rate.

(Choosing which commitment applies — and, under “firm,” which units are firm — is being wired in. Today the stage reports hours below minimum run; that is the right signal for a firm plant and an over-strict one for a plant allowed to idle.)

Read shortfall as episodes, not a raw hour-count. One 100-hour outage and 100 scattered one-hour dips add up to the same number but are completely different for an operator or a lender. The Reliability room groups shortfall into episodes — when they hit, how long they lasted, how deep they went — which is the honest way to read it.

Where each tool fits

• Set up — steps 1–5 (resources, capacities, run settings).
• Operations — step 6, how it runs hour by hour.
• Reliability — step 6, each product delivered against its promise, and where and how it falls short.
• Results — step 6, what it costs (LCO, CAPEX, OPEX).
• Storage lab — step 4, storage behaviour and the cost-effective size.
• Optimizer (Least cost (LP)) — step 7, the cheapest sizing that makes what you ask of each product (it starts from the promise).
• Scale the plant — step 8, grow or shrink the whole plant on fixed storage to see how far it holds up.
• Grid & back-up — a lever that runs through the loop: it substitutes against storage and changes both cost and the green-hydrogen share, so it is not a separate step but a choice inside supply and refinement.

Where this is going. These tools overlap today (several ways to “hold some components and size the rest”). They are being consolidated into one clear sizing instrument so the method above reads as a single, guided flow rather than a set of separate rooms.