Planned outage

What actually gets scheduled on a battery plant

The capacity test is the headline event. A controlled full discharge at defined temperature, SOC window and C-rate, run to prove retained energy against the warranted degradation table, and depending on the contract it is annual or on a longer cycle.

It takes the tested capacity offline for its duration and it is not optional — the test result is what triggers an augmentation tranche or a liquidated-damages claim. Preventive maintenance follows a defined calendar and, on a BESS, is light on wrenches: thermal imaging, torque and connection checks, transformer oil sampling, protection-relay testing, HVAC service, and keeping vegetation off the fence line.

The rest of the list is more disruptive than the maintenance schedule suggests. Firmware and BMS, EMS and PCS software campaigns, increasingly arriving as fleet-wide advisories rather than local upgrades. Augmentation, which is a construction project inside an energised plant, commonly on a 2-5 year cadence or as one or two larger mid-life events. Utility-driven work on the gen-tie, the substation or the MV feeder, scheduled by someone else and imposed on the site. And statutory or AHJ re-inspections. Only some of these are within the operator's gift to move.

The availability clause is where the value sits

Three structures appear. Scheduled work can be excluded from the availability calculation, capped at a stated number of hours per year with the excess counted against the guarantee, or counted in full with the supplier expected to have priced its own maintenance windows. Exclusion itself has two mechanics that do not give the same answer, as the availability entry sets out: the hours can be removed from the denominator, or the plant can be deemed available in the numerator. Read which one the contract uses before you accept that planned work is free.

Scale it against the guarantee to see why this is not a drafting detail. Utility-scale availability guarantees commonly land at 95-98 percent measured annually, and 97 percent allows roughly 263 hours of unexcused downtime across the year. A capacity test, a couple of firmware campaigns, an HVAC service round and one augmentation window can consume a meaningful share of that if none of it is excluded. Then read the surrounding mechanics: notice periods, whether the owner can refuse or defer a proposed window, what happens when a window overruns, and whether the cap is per event or annual.

When the outage happens is a market decision

Outage hours are not fungible, because the hours themselves are not worth the same. Take a plant offline during a capacity-market performance window or a scarcity event and the cost is not the lost energy margin — it is the non-performance exposure. The capacity-market entry on this site notes that PJM's Capacity Performance regime charges non-performance during emergency events at a steep per-MWh rate, with an annual stop-loss commonly near 1.5 times the year's capacity revenue. The same hours moved into a shoulder month cost a fraction of that.

So the outage plan has to be built against the revenue calendar and the market's own rules, not just the maintenance calendar. Check whether a registered outage relieves the obligation in your market, what notice that registration needs, and whether the LTSA's window rights are compatible with it — those two documents are usually drafted by people who never spoke.

Seasonality helps sometimes and fights you at others: HVAC work wants a mild ambient, and mild months are also when spreads are thin, which is the convenient case; work that can only be performed at high load or high temperature is the awkward one.

Running the outage plan

The practical instrument is a rolling annual outage plan agreed between the owner, the LTSA supplier and the O&M contractor, reconciled against the utility's own outage schedule. A utility feeder outage is an opportunity, not just an imposition — it is a window in which owner-side work can be done for no incremental availability cost, and coordinating into it is one of the cheapest optimisations available.

Modularity is the other lever. A plant built as many blocks can take one block at a time and keep the rest earning, which under a capacity-weighted metric turns a site outage into a pro-rata derate. It costs more calendar time and more mobilisations, and it is almost always the right trade when the alternative is a full-site window during a period that carries a capacity obligation.

Where planned outages turn into arguments

The overrun is the classic. A window agreed for eight hours that runs to thirty — is the excess planned, or has it become a forced outage from the moment the agreed window closed? Contracts that do not answer that produce a dispute at the first overrun, and the honest answer is that the clause has to say so explicitly. Second, the safety advisory.

A fleet-wide firmware or hardware campaign issued by the OEM is scheduled work, but the availability entry warns specifically about exclusion language covering response to safety advisories, which can quietly excuse a corrective campaign the supplier caused. Third, augmentation: the availability guarantee normally caps how much outage time an augmentation installation may consume, and that cap has to be checked against a realistic installation and recommissioning duration before it is agreed, not after.

The last one is scope. When a scheduled window belongs to the LTSA supplier but the work touches balance-of-plant equipment under the O&M agreement, both contracts have to recognise the same window and the same clock. The interface matrix that assigns each maintainable item to exactly one contract is what makes that possible, and it is cheaper to write before either contract is signed than to reconstruct in year six.