+60.1%
vs persistence baseline
+4.5%
vs persistence baseline
57.1k MW
next 72h
0%
0 critical hrs
20
detected next 72h
€16.4M
12h Solarspitzen risk
German Day-Ahead Market Sequence
Every hour of the next-day forecast must clear these checkpoints. After the schedule Cut-Off Time at 15:30 CET, any forecast error is settled in real-time at the reBAP imbalance price.
Next checkpoint
7h 21m
Delivery & 15-min ISP settlement
08:00
D-1
12:00
D-1
12:45 – 13:05
D-1
14:30
D-1
15:30
D-1
16:00
D-1
00:00 – 23:45
D
TSO renewable forecast published
All four German TSOs publish their day-ahead wind & solar forecast (15-min granularity) per § 1 EEG transparency rules. Direct-marketers and traders feed this into their bidding strategy.
netztransparenz.de · EEG transparency
EPEX day-ahead order book closes
Gate Closure Time of the Single Day-Ahead Coupling (SDAC) auction on EPEX SPOT. Hourly products for the next day are locked at the cleared market price.
EPEX SPOT MRC Day-Ahead Timings
Day-ahead market results published
SDAC clearing prices and volumes published. BRPs now know their day-ahead position; the difference vs. their physical forecast is what they must close on intraday.
EPEX SPOT MRC
Schedule nomination GCT
Gate Closure Time per StromNZV §5(1): every balancing-group manager (BKV) must have nominated its day-ahead schedule to the relevant TSO. Cross-border schedules are matched immediately after.
Amprion AG-FPM v4.6 · StromNZV §5(1)
Schedule Cut-Off Time (COT)
Last opportunity to correct a nominated schedule after the TSO matching process flags mismatches. After 15:30 the day-ahead schedule is final.
Amprion AG-FPM v4.6
Continuous intraday opens
EPEX intraday continuous trading opens for German 15-min, 30-min and hourly contracts. BRPs trade here to close forecast errors as the delivery hour approaches.
EPEX SPOT trading brochure
Delivery & 15-min ISP settlement
Physical delivery. Every 15-min ISP, the TSO measures the balancing-group imbalance and settles it at the reBAP. Intraday continuous remains open until 5 min before delivery within the same TSO zone.
Netztransparenz reBAP · TransnetBW MaBiS
GridSight's 72-hour forecast is published before 12:00 CET D-1 so day-ahead bids and TSO schedules can be priced against it — every minute earlier than the schedule Cut-Off Time ( 15:30 CET ) reduces the volume that has to be closed expensively on intraday or settled at the reBAP.
TSO Region Selection
Zone Capacity
92.0k MW
Peak Utilization
62%
Imbalance Exposure (reBAP Risk)
Forecast uncertainty translated into € exposure at the German cross-zone imbalance price (reBAP). Quarter-hour settlement: a balancing group that overproduces while the system is long pays the negative reBAP; underproducing while the system is short pays the positive reBAP.
Expected 24h cost
€16.4M
1076k MWh total band
Worst-case 72h
€133.5M
at €250/MWh scarcity
15hrs
System surplus expected → overproduction = negative reBAP
5hrs
System deficit expected → underproduction = positive reBAP
12hrs
Midday solar peak quarters at risk of €0 EEG remuneration (573 hrs of negative DA prices in 2025)
06-13 13 CET
€3.5M worst-case at this quarter
reBAP is symmetrical and quarter-hourly. Magnitudes used here: base €80/MWh (median 2024 magnitude per 50Hertz Almanach), scarcity €250/MWh. Sign convention follows § 8 EBGL: negative when system is long, positive when system is short. Solarspitzengesetz: new PV gets €0 EEG remuneration during any quarter-hour with a negative day-ahead price (573 such hours in 2025, source: EEX / pv-magazine).
Grid Alerts
10 criticalModel vs Baseline
Our MAE
984 MWh
Baseline MAE
2468 MWh
Band Coverage
65%
Our MAE
781 MWh
Baseline MAE
818 MWh
Band Coverage
48.7%
Methodology
Algorithm
LightGBM GBDT
Quantiles
P10 / P50 / P90
Training Data
182k hours
Test Data
8.8k hours (92 days)
Feature Importance
Wind Model Drivers
Solar Model Drivers
Problem
German TSOs spend EUR 4.2B/year on redispatch. Inaccurate renewable forecasts force expensive last-minute grid interventions.
Solution
LightGBM quantile models trained on 5 years of SMARD + Open-Meteo data reduce forecast error by up to 60%, enabling proactive congestion management.
Data Foundation
5 years of hourly generation data
4 TSO zones × 6 grid points each
16 weather features from Open-Meteo
SMARD (Bundesnetzagentur, CC BY 4.0)
Economic Impact
€616.9M/yr
Imbalance cost reduction
6260 MWh/hr
Forecast error eliminated