Critical spare

Rank by lead time against consequence

Two questions decide the list, and neither is what the part costs. How much of the plant stops when this item fails, and how long until a replacement is physically on site? Multiply them and you have the exposure the spare exists to buy down.

Cheap and slow beats expensive and fast every time: an item with a two-week lead is a stores problem, while an obsolete board with a twelve-month lead on a discontinued platform is a critical spare regardless of its invoice value. Run the screen against the single-line diagram rather than against the asset register, because the register lists what you own and the single-line shows what is in series with the revenue.

Architecture changes the answer. A plant assembled from many modular blocks bounds its per-fault exposure by design — one of ten power blocks down for a month costs on the order of 0.83 points of annual energy availability, and no single block failure strands the site. Items inside that repeated structure sit lower on the list because redundancy is already doing part of the work. Items at an interface have no such bound: everything the plant produces passes through one main transformer, one set of HV switchgear, one metering point, and there is nothing behind them to pick up the load.

The mechanism the whole page turns on is repair time. Availability is A ≈ MTBF / (MTBF + MTTR) for a repairable system, and logistics delay — the wait for the part — is commonly excluded from MTTR and tracked separately, which is why it is invisible in a healthy repair-time statistic and fully visible in a poor availability number. Stocking the part does not make the plant fail less often. It shortens the second term, and the second term is where a battery site's worst availability years are actually spent.

The list, roughly in order

The main power transformer sits at the top for reasons the transformer entry works through in full: a lead time in years, no second unit behind it, and a failure that strands the entire site. The available answers are a purchased spare, an agreement covering a mobile or emergency unit, or two half-rated units from the start, which buys partial output through a failure as well as a shorter road to full output.

Shared and pooled reserves exist for exactly this item, run by utility groupings and by commercial reserve providers, and both are membership arrangements — so the questions for a project company are whether it is eligible to join, what its call right is worth against other members' claims, and whether the pooled unit's impedance and vector group actually match its specification.

Medium-voltage switchgear comes next, and its spare has a shape people get wrong. The assembly carries the rating as tested, so substituting a component or omitting an accessory changes the rating rather than the parts list — which means the spare is the cubicle, not the breaker inside it. Sealed gas-insulated tanks cannot be opened in the field at all. Below that sit the PCS power stacks, which are the modular replaceable unit on the conversion side and should be stored with the site's parameter and firmware file rather than the factory default.

Then the mundane items that actually generate the tickets. HVAC compressors, circulation pumps, fan assemblies and the chiller's control board have the highest failure rate on the site by a wide margin, and a container derates within hours of losing its cooling loop rather than days. Battery modules, in whatever quantity the failure history and the OEM's response time justify.

And last on the invoice, high on the list: the communications hardware. Media converters, the exact SFP the plant was built with, managed switches, the RTU card, the BMS rack controller. A plant that cannot be dispatched is unavailable with every cell in perfect health, and that outage is caused by a part somebody's stores would not normally bother tracking.

Consignment, owned stock, and who holds title

Owned stock is the simple version: the project buys the parts, holds them, and capital sits in a warehouse doing nothing until the day it does everything. Title is unambiguous, the count is yours, and nobody else can draw on it. Consignment inverts that — the supplier stages stock at or near the site and retains title until it is drawn, and the project pays on use. The holding cost moves to the supplier and so does the risk of buying the wrong thing, which is genuinely valuable early in life when the failure history does not exist yet.

Consignment raises three questions that need answering before the parts arrive rather than after. Whose insurance covers stock sitting on your site under someone else's title, and who counts it. Whether the supplier may pull consigned stock to serve another customer's emergency, which is the clause that decides whether you have a spare or a promise.

And what happens to the stock on supplier insolvency or on termination of the service agreement — a real question on an asset class where suppliers have changed hands and exited markets mid-contract. Title determines the answer at all three moments, so the arrangement is a legal structure rather than a logistics preference.

The third option is to hold no metal and buy a response time instead: a contractual commitment that a replacement is on site within a stated period, backed by liquidated damages.

That transfers the holding cost to the supplier and works exactly as well as the supplier's own stock position, so it is worth verifying rather than assuming — ask where the stock is held, whether it is dedicated or pooled across their fleet, and what the damages actually pay relative to a month of stranded revenue. A response guarantee with damages far below the revenue at risk is a discount on the outage, not a defence against it.

A stored battery module is still aging

Calendar aging runs on elapsed time regardless of throughput, so a module bought at commercial operation and installed in year eight is not a year-zero module — it has spent eight years on the same clock the fleet was on, just without the cycling. That has one modest upside and one real obligation. The upside is that a stored spare sits closer in state to the aged string it will join than a factory-fresh unit would, which slightly reduces the spread the balancing system has to carry.

The obligation is that storage is a maintenance task with its own schedule: modules are commonly required to be held at around 30-50% state of charge with recharge intervals specified by the manufacturer, and neglecting those intervals can push a cell into deep discharge and void whatever warranty came with it.

Storage temperature matters for the same Arrhenius reason everything else on the site does, which means an unconditioned container in a hot climate is not storage, it is accelerated aging with a roof. And a quantity of charged lithium modules held in a building is a fire-code question in its own right — the authority having jurisdiction that permitted the plant may have a view about the warehouse, and the insurer certainly will.

Budget conditioned, separated, monitored space for the battery spares, and put the recharge intervals on the same preventive-maintenance calendar as everything else, because a spare that failed its own storage regime is discovered at the worst possible moment.

Obsolescence: the OEM revises the module mid-life

The spare you hold matches the plant you built. The spare you can buy in year seven matches the plant the OEM is building now, and cell formats, module mechanics and rack designs move quickly enough that a mid-life replacement can be a different product wearing the same product name.

The mismatch is rarely only mechanical. A revised module with a different cell capacity, a different sense-harness pinout or a newer BMS protocol version changes how the string behaves and can force a firmware campaign across the whole block just to accept it, which converts a one-module corrective job into a fleet change.

The contract is where this is managed, and the hooks are specific. A parts-availability commitment for a stated period, a last-time-buy notification obligation so the project can place a final order before a line closes, an agreed form-fit-function substitution process with the owner's right to test the substitute, and escrow of drawings, firmware and configuration files where the supplier's balance sheet is thin relative to the twenty-year obligation.

Where none of that exists, the practical hedge is buying the initial spares deeper than the failure statistics justify — which costs capital immediately and is still cheaper than discovering in year nine that the only compatible module in the world is inside another one of your containers.