What is LCOE and where does it mislead?

LCOE, the levelized cost of electricity, divides a power plant's lifetime construction and operating costs by the total electricity it generates, producing a cost per megawatt-hour that lets technologies be compared on the same basis. It misleads because it counts only the plant's own costs and ignores system costs — the backup storage and peaking capacity that intermittent generation requires.

By 2022 the LCOE numbers described a complete reversal. Utility-scale solar PV reached USD 24–41/MWh and onshore wind USD 26–50/MWh, both unsubsidised, against USD 68–166/MWh for coal, USD 131–204/MWh for nuclear and more for gas peaking plants (Lazard's Levelized Cost of Energy Analysis, Version 16.0). Since 2009 solar PV construction costs have collapsed by roughly 90% and onshore wind by roughly 70% (IRENA, Renewable Power Generation Costs in 2022).

The blind spot is that LCOE treats every megawatt-hour as equivalent. A solar plant generates only when the sun is shining, so keeping the grid stable requires backup storage or peaking units — integration costs that never appear in its LCOE. Nuclear runs around the clock and avoids those costs, yet looks expensive on an LCOE basis because of a ten-year build and billions in up-front financing. The industry therefore also uses LACE, the levelized avoided cost of energy, which measures what the grid saves by having reliable baseload. Read together, they show that once renewable penetration passes 30%, the cost of balancing intermittency grows exponentially.

California's duck curve gives system cost a physical shape. On one clear day in April 2021, California's net load fell 9,000 MW in two hours and then climbed 16,000 MW in the three hours after sunset (CAISO). Riding swings of that magnitude takes storage, peaking plant and grid reinforcement — every dollar of it outside the LCOE calculation.

The right way to read the metric, then, is this: LCOE compares what it costs to generate a megawatt-hour, while what a power system actually buys is a megawatt-hour delivered reliably at the moment it is needed. The two are close at low penetration and diverge sharply at high penetration. Ranking technologies by LCOE is a reasonable first step; deriving total system cost from it directly is not — least of all when comparing dispatchable generation with intermittent generation.

Once renewable penetration passes 30%, the cost of balancing its intermittency grows exponentially — and the economic case for nuclear as a grid anchor becomes more persuasive.

This question is covered in depth in The Full Spectrum Every Energy Source Explained — A Singapore Perspective，第一章 1.8 + 第六章 6.5