The 1/4-Hourly Day-Ahead Market & The Illusion of Better Forecasting

For those not deeply in the energy industry: Electricity from 🌬️ wind and ☀️ solar PV power plants is often traded in the short-term on the EEX – the European Energy Exchange. Think of it like a stock exchange, but for energy.

The Shift to 15-Minute Intervals

📊 Traditionally, short-term contracts were traded in 15-minute intervals (intraday), while the day-ahead market worked with hourly intervals. Now, the shift is happening: the day-ahead market is moving to 15-minute contracts.

✅ Advantage: This makes balancing easier in theory.

⚠️ But beware: It doesn't mean forecasts suddenly become more accurate.

Why Forecasts Don't Automatically Improve

Why? Because numerical weather prediction (NWP) models struggle to capture sub-hourly cloud development 🌥️, even in the short term. For very short horizons, we can use satellite images 🛰️ to improve solar forecasts (something we're working on at Alitiq with Ventus Technologies in the NowcastAI project).

👉 But for day-ahead forecasts, such real-time data isn't available. The expectation that balancing becomes much easier might therefore be an illusion.

The Technical Reality

I illustrated this using forecasts from the DWD ICON-D2 model. It produces 15-minute forecasts – but here's the catch:

• The model integrates weather equations over 1–2 minutes
• Results are aggregated into quarter-hourly steps
• This introduces "noise" 🌪️ – forecasts might capture the overall trend (e.g., decreasing afternoon solar production), but often not at the right time or location

Hourly vs. 1/4-Hourly Market Schedule

Figure 1: Actual solar irradiance vs. 1h day-ahead market schedule

Figure 2: The expectation - Actual solar irradiance vs. 1/4-hourly day-ahead market schedule

Forecast Accuracy Comparison

Figure 3: Actual solar irradiance vs. 1/4-hourly day-ahead forecast from ICON-D2

The Bottom Line

📉 Metrics actually look worse when comparing native quarter-hourly forecasts. Interestingly, accuracy improves if you just calculate hourly averages.

💡 Bottom line: "Skill is all you need." In forecasting, skill means one forecast performs better than another. What we really need now is skill in sub-hourly forecasting – not just finer time intervals.

---

At alitiq, we're working on advanced forecasting solutions that address these challenges, combining multiple data sources and AI-driven approaches to deliver truly accurate short-term predictions for the energy market.