Skip to content

Wind + solar daily capacity factor — year × day-of-year heatmap

Two stacked heatmaps (wind top, solar bottom) rendering every day of the observed record as a colour-coded cell. Blue stripes across years are multi-week droughts — dunkelflaute made visible to the naked eye.

Generation heatmap — wind and solar capacity factor by year × day

→ Interactive version

What the chart shows

Two stacked heatmaps sharing a common x-axis.

  • Top panel — Wind (Offshore + Onshore combined, capacity-weighted). Rows are calendar years (2017 at the top, descending to the latest complete year); columns are day-of-year (1 to 366). Each cell's colour encodes the daily fleet capacity factor for that date.
  • Bottom panel — Solar PV. Same axis layout, normalised to the installed solar CfD capacity on each date.

Colour scale runs from deep blue (0% CF) through light blue, pale red, and into deep red (100% CF). The shared 0–100% scale means the colour in one panel means the same thing as the colour in the other panel. Month labels (Jan, Feb, …, Dec) sit along the bottom x-axis; row labels on the y-axis are calendar years.

Droughts show up directly: a multi-week blue band crossing rows is a sustained fleet-wide low-output period. Solar's pattern is stark — every winter is a continuous blue band by construction. Wind's pattern is more scattered but economically more important because the CfD fleet is wind-dominant in both capacity and cost share.

The argument

Sustained low-output periods are structural, not exceptional. You can see them in the data with the naked eye.

Three paragraphs:

  1. The chart is deliberately direct-access. There is no aggregation, no rolling window, no smoothing, no summary statistic. Every cell is one day's raw fleet capacity factor, organised into a year-on-year grid so the reader can inspect the evidence without trusting any author's framing. The chart says "look for yourself" more emphatically than any summary chart on this site can.

  2. Blue stripes are the empirical record of droughts. Each horizontal blue band is a multi-day stretch where the CfD fleet produced far below its annual average. These are not modelled; they are every low-output event the fleet has lived through. The pattern repeats year-on-year: there is no year without substantial blue, and some years (notably summers of 2020 and 2021 for wind) are almost entirely in the cool half of the palette.

  3. Solar's pattern is starker; wind's is economically weightier. Solar's mid-winter blue band is unavoidable — daylight hours shrink, zenith angle shallows, cloud cover is high. Nobody argues otherwise. The important argument is about wind: wind droughts are scattered, irregular, and not concentrated seasonally. They hit spring, summer, autumn and winter alike. A fleet relying on wind for bulk-energy cannot predict when a drought will occur — it can only observe that droughts do occur, repeatedly, by looking at this chart.

This is the companion page to rolling-minimum, which quantifies the blue stripes into specific drought durations and flags the worst events. The heatmap delivers the unaggregated raw-evidence visualisation; rolling-minimum delivers the quantified policy-relevant summary.

Methodology

Source: LCCC Actual CfD Generation (daily generation per unit) + LCCC CfD Contract Portfolio Status (per-unit Maximum_Contract_Capacity_MW and Technology_Type).

Daily CF per fleet:

installed_MW(d)    = sum(unit_capacity_MW for units active on day d)
daily_gen_MWh(d)   = sum(CFD_Generation_MWh on day d for tech group)
CF(d)              = daily_gen_MWh(d) / (installed_MW(d) × 24)

Unit-active-on-day logic: a unit contributes to installed_MW(d) from its first observed generation date onwards (its Maximum_Contract_Capacity_MW enters the running total). This means pre-commissioning days where the unit was under construction are excluded from both numerator and denominator — the CF reflects the active fleet, not the paper fleet.

Heatmap assembly: pandas pivot with year on rows, day_of_year on columns, fleet_cf as the colour value. Plotly go.Heatmap with the custom colour scale (blue → red through a pale-neutral midpoint) and shared zmin=0, zmax=1 across both panels so colour semantics are absolute.

The two technology groups are computed independently from their own installed-capacity bases — wind's colour scale is normalised to wind's fleet, solar's to solar's fleet. See the Reliability methodology for the installed-capacity-over-time attribution rule and the rationale for splitting wind + solar rather than a combined fleet view.

Caveats

  • Pre-commissioning days are excluded. A unit is counted from its first observed generation date, not its contract-start date. This avoids the "ramp-up dip" that would otherwise bias the left edge of every row with a new build.
  • Leap-year day 366 is populated only for leap years (2020, 2024). Other years show column 366 blank by design — this is correct, not a rendering defect.
  • Colour scale is shared but perceptually independent per-panel when the reader compares rows within a panel. A "dark blue" in the wind panel and a "dark blue" in the solar panel both represent roughly 0% CF, but the absolute MWh represented differs because wind's installed capacity is much larger than solar's.
  • Fleet composition changes along rows. Early years are dominated by Investment-Contract offshore wind and Drax (biomass is not plotted here as its CF profile is dispatchable, not intermittent). Recent years include AR1–AR3 units. A single cell's CF reflects the fleet in place on that day, not a fixed reference fleet.

Data & code

To reproduce:

uv run python -m uk_subsidy_tracker.plotting.intermittency.generation_heatmap

See also