The cost of fully firm solar

The cost of fully firm solar

Cheap solar is only half the story. This map shows what it costs to run on solar around the clock — not just the raw resource, but the price of surviving the night and multi-day cloudy spells to a chosen reliability. Two tools do the work and they are complements, not substitutes: overbuild the array to make enough energy, and add a battery to shift it in time. You can never firm with storage alone.

What the model does: for each of 236 cities and countries we simulate 20 years of hour-by-hour solar output — calibrated to satellite data, including realistic multi-day runs of cloud — and set it against the country's real measured hourly demand. The model then finds the cheapest mix of extra panels and battery that serves your chosen share of demand every hour, and prices that mix at today's benchmark costs. Map colours are a fast approximation of those runs; the side panel shows the exact modelled number, broken into its parts.

The overbuild and battery are both sized by the model — the total array typically lands 40–60% above the simple worst-month ratio, with just enough battery for the residual timing. The battery runs roughly 10–15 hours of storage where sun is steady, 40–60+ hours where cloudy spells must be ridden through. The side panel shows each place's total overbuild, battery size, duty and cost share. How the sizing works →

Set the firmness target (95 / 97 / 99%) — the single biggest cost lever — then hover any city or country for its base cost, its overbuild + battery, and the full firm price.

Toggles: Surplus credit values curtailed summer output at $15/MWh; Real demand (on by default) uses each country's measured hourly load shape (PLEXOS-World) — untick it for a demand-neutral flat 24h target. Both the heat layer and the panel follow the toggle.

Illustrative, not bankable: cloudy-winter and high-mountain values carry roughly ±30% model spread, and far-northern points need seasonal-scale storage and are shown for completeness only.

How the model sizes the array and battery

For every place and firmness target the model searches over array sizes, lets a dispatch simulation size the battery at each step, and keeps the cheapest combination per delivered MWh. Step by step:

1. Build the weather. 20 years of hour-by-hour solar output: monthly satellite yield sets each month's energy, and a variability model calibrated to real hourly data draws day-to-day cloudiness — with persistence, so cloudy days arrive in realistic multi-day runs. That tail is what matters at 99%.
2. Try an array size k. k = 1 is an array whose average output equals average demand. Dispatch hour by hour: solar serves demand first; surplus charges the battery (88% round-trip) or is curtailed once it's full; the battery covers night and cloud; if it runs dry, those hours go unserved.
3. Size the battery for that k. Find the smallest battery that serves the target share of demand (95 / 97 / 99%) across all 20 years. Bigger arrays leave shallower, rarer deficits, so they need smaller batteries. A battery only shifts energy — losses and dark weeks are an energy problem only panels can solve, so below some k no battery is big enough.
4. Cost the pair. Panels: base LCoE × k — cheap, and the cost scales linearly. Battery: benchmark storage cost corrected for duration and, dominantly, for utilisation — sized for the rare worst spell it may cycle ~65 rather than ~300 times a year, making each stored MWh several times dearer.
5. Keep the minimum. Total cost over k is U-shaped: at low k a huge battery sits idle (ruinous); at high k mountains of summer surplus are curtailed for nothing. The bottom of the U is the total overbuild and battery shown in the side panel.

Worked example — London at 97%: total overbuild ×2.8, a 46-hour battery cycling ~65 times a year; the battery alone contributes $129 of the $327/MWh.