How different power plants trade in Spain's electricity markets

At 16:00 on December 4, 2024, a gas turbine scheduled 730 MWh in the day-ahead market — and delivered 1,526. A solar farm bid 87 and landed at 18. A wind farm went from 544 to 421. Same hour, same market, five completely different outcomes.

Spain’s electricity market isn’t one market — it’s a sequence of programs that reshape each power plant’s position between the day-ahead auction and real-time delivery. What determines whether a power plant earns or loses in these adjustments? We track five units across every program over a full year to find out.

Reading the waterfall

Every generating unit in Spain passes through a cascade of market programs. The first bar — PDBF — is the day-ahead auction: what the plant was scheduled to produce. Every bar after that is an adjustment. Green adds megawatt-hours, red removes them, and the blue bar at the end is what the plant actually delivered.

BAHIAB (gas) starts small and grows. It schedules conservatively in the day-ahead auction, then adds massive volume through PHF1 — the first intraday session. This isn’t a correction; it’s a strategy. Gas turbines exploit the spread between day-ahead and intraday prices, doubling their position before delivery.

COF1 (nuclear) is a flat line. It schedules 2,112 MWh and delivers exactly that. No corrections, no adjustments, no balancing. Nuclear runs at full capacity and the market works around it.

DUER (pumped hydro) tells a more complex story. It gains through intraday sessions but gives most of it back in replacement reserves. Hydro’s value isn’t in scheduling — it’s in being available to ramp when the system needs it.

FBRERAS (solar) shrinks. Its day-ahead schedule erodes through intraday corrections as forecasts change and the grid can’t absorb all the output. By delivery, it retains barely a fifth of what it was scheduled to produce.

EEGPEM (wind) loses across nearly every program — PDVP, PHF1, PHF3, BT. No other technology faces curtailment from so many directions simultaneously.

Program participation patterns

Over a full year, each technology distributes its volume across market programs differently. The heatmap shows the percentage breakdown.

Zooming out to a full year reveals the structural differences. Nuclear concentrates 95% of its absolute volume in the day-ahead market. It barely touches anything else. This is the purest baseload profile — schedule and hold.

Gas is the polar opposite. Only 23% of its volume comes from PDBF. More than half — 51% — sits in PDVP, where REE calls on the plant for upward redispatch. Another 11% comes from intraday corrections. Gas doesn’t compete on volume; it competes on flexibility.

Wind and hydro are mostly day-ahead — 91% and 82% respectively — with thin tails distributed across adjustment programs. They participate in corrections and balancing, but it’s not their primary business.

Solar stands out for a different reason. At 53% PDBF and 30% PDVP, it looks diversified. But unlike gas, where PDVP means upward redispatch and extra revenue, solar’s PDVP is curtailment. Nearly a third of its absolute market activity is the grid cutting its output.

Intraday correction patterns

How much does each technology change its position in the first intraday session? Green means energy added, red means reduced.

The PHF1/PDBF ratio shows how much each technology changes its position in the first intraday session, relative to its day-ahead schedule. The heatmap makes the answer obvious: gas is green and everyone else is white.

Gas doesn’t just make corrections — it uses intraday as a second market. Its PHF1 additions average 100% of its PDBF, meaning it routinely doubles its day-ahead schedule. In October, it peaked at 192%. The strategy is deliberate: schedule conservatively in the auction, then add volume when the plant has more operational certainty and prices may be more favorable.

Nuclear, hydro, and wind all hover near zero. They schedule once and don’t look back. This isn’t laziness — nuclear can’t ramp, hydro saves its flexibility for reserves, and wind can’t choose when to produce.

Solar is the wild card. Its mean is slightly negative, but the range swings from +20% in spring to -22% in autumn. These aren’t strategic adjustments — they’re forecast corrections. When cloud cover changes between the auction and delivery, the schedule has to change too.

Technical constraints (PDVP)

When REE intervenes for grid security, who benefits and who pays? Monthly PDVP volumes by technology — green is upward redispatch, red is curtailment.

Intraday corrections are voluntary. Technical constraints are not — REE imposes them to maintain grid security. This heatmap is the clearest picture of who the system rewards and who it penalizes.

Gas is entirely green. Every single month, REE calls on BAHIAB for upward redispatch — over 1,600 GWh across the year. When a constrained zone needs more power, gas is the first call. This is pure revenue, compensated at marginal cost plus a regulated premium.

The rest of the heatmap is red. Hydro bears the deepest absolute curtailment, with December alone reaching -77 GWh. Solar’s curtailment is smaller but relentless — every month negative, worst in spring when production peaks and the grid is saturated. Nuclear gets hit too, particularly in April when excess baseload collides with high renewable output. Wind’s curtailment is the lightest in absolute terms, but it’s present every single month without exception.

The asymmetry is stark: one gas plant earns more from technical constraints than all four other technologies lose combined.

Balancing (RR/BT)

Replacement reserves (RR) and tertiary reserve (BT) are where generators volunteer to provide frequency support. The split panels show who participates.

Constraints are imposed. Balancing markets are different — generators volunteer to provide reserves, and the TSO activates them when the system needs frequency support. The chart splits into replacement reserves (RR) and tertiary reserve (BT).

Pumped hydro dominates both. When there’s excess energy on the grid, DUER absorbs it by pumping water uphill. That’s its core value proposition — it earns by consuming, not producing. The scale is massive: hydro dwarfs every other technology in both panels.

Wind is the second-largest downward provider. Wind farms can reduce output quickly when the TSO signals excess, making them useful for downward balancing even though they can’t ramp up on demand. This sets wind apart from solar, which is nearly invisible in both panels.

Gas barely participates. This is surprising until you look at the PDVP heatmap — BAHIAB earns its flexibility premium through constraints resolution, not balancing markets. Different gas plants make different choices about where to sell their flexibility; this one chose PDVP.

Nuclear and solar are negligible. Nuclear can’t ramp fast enough. Solar can’t guarantee it will be producing when the TSO needs it.

The five market fingerprints

Every technology leaves a distinct signature across Spain’s electricity markets. The same I90 data that records a single 15-minute interval also tells a year-long story about what each power plant is built to do — and what the system asks of it.

Gas is the market’s flexibility engine. Most of its volume isn’t from the day-ahead auction — it’s from constraints resolution and intraday corrections. Gas earns by being available when the system needs it most.

Nuclear is the opposite. Almost all its volume is day-ahead. It schedules once and holds. But inflexibility has a cost: when spring brings excess baseload and high renewables, nuclear gets curtailed too.

Pumped hydro absorbs what the system can’t use. Heavy PDVP curtailment, heavy downward balancing — when there’s excess energy, hydro pumps. Its value is in consumption, not production.

Solar bears consistent curtailment, every month negative. Nearly a third of its absolute market activity is the grid cutting its output. Solar’s fingerprint is shaped more by what the system takes away than by what it schedules.

Wind contributes significant downward balancing — something solar can’t do. Its PDVP curtailment is modest, but its ability to respond to TSO signals sets it apart as a participant, not just a price taker.

These five fingerprints aren’t anomalies. They’re the structural incentives of each technology, visible in every settlement file REE publishes. At 16:00 on December 4, BAHIAB doubled its schedule because that’s what gas does — the system needs flexibility, and the I90 records every megawatt-hour of it. Now you can read that waterfall.

---

All data in this article comes from REE’s I90 settlement files, queried via the datons Python library. You can reproduce every chart with a few lines of code.