Refueling Outages and the Rhythm of Operations

A nuclear plant cannot refuel on the fly. The fuel sits inside a sealed pressure vessel inside containment, so replacing it means shutting the chain reaction down, cooling and depressurizing the primary system, unbolting the vessel head, and moving assemblies one at a time under water — a process with its own name and culture in the industry: the refueling outage.

US plants run on an 18-month or 24-month fuel cycle. At each outage, roughly one-third of the core — the assemblies that have spent three cycles in the reactor — moves to the spent-fuel pool, the remaining assemblies are shuffled to flatten the power distribution, and fresh fuel fills the vacancies. But refueling is almost the side show: the outage is the only window when the plant's normally-inaccessible systems can be inspected, tested, and overhauled, so operators pack it with thousands of coordinated maintenance jobs — vessel and steam-generator inspections, valve and pump overhauls, turbine work, safety-system surveillance tests required by the operating license. A modern outage is planned for years in advance, staffed by a thousand or more supplemental workers, and executed around the clock. The US fleet has driven the average duration down from over 100 days in the early 1990s to the low 30s today; the best plants complete one in under three weeks.

Outages cluster in spring and fall, and the reason is the grid: electricity demand peaks in summer (air conditioning) and winter (heating), so plants schedule their downtime for the mild shoulder seasons when their output is least needed and power prices are lowest. This is why the fleet-history chart on this map has its characteristic shape — a high summer plateau with soft dips in March–May and September–November. That seasonal scallop is the entire US nuclear fleet breathing in unison.

Not every zero on a sparkline is a refueling, and this site is careful not to guess. The NRC's daily status report — the source of the map's live layer — publishes each reactor's power level, not the reason for it. A plant can also be down for an unplanned repair, a regulatory issue, or something rarer. The map therefore reports exactly what the data shows: power levels and their changes. A months-long outage usually has a story behind it, and the news layer will often supply it — but the distinction between "what the data says" and "what we infer" is a line this site deliberately doesn't cross. (The most dramatic example in the current data: Palisades in Michigan, whose history shows an outage of well over 200 days — not a refueling but a first-of-its-kind story, the plant having been fully shut down in 2022 and brought back as the first restart of a retired reactor in US history.)

Between outages, what does operation look like? Mostly: nothing happening, on purpose. A control-room crew of licensed operators monitors a machine designed to sit at 100% for months. Brief downpowers appear in the data — a few days at 60–80% for condenser cleaning, grid requests, or minor repairs — and the sparklines render them as shallow notches. The deepest skill in the industry is making the plots boring.