Datenpaket: Simulated Day-Ahead Electricity Prices for 48 European Bidding Zones under Varying Geographic Simulation Scope

This dataset provides simulated hourly day-ahead electricity market prices for up to 48 European bidding zones, generated for the study "The impact of geographic scope on electricity market simulation". The prices were produced with the agent-based model PowerACE on the basis of ERAA 2025 data. They cover three bidding strategy groups (Base, Flex, and Welfare), three weather scenarios, and the simulation years 2028 to 2035. The geographic scope of the simulation was varied systematically: the number and combination of jointly simulated market areas range from a single zone up to the full scope of 48 zones. Reduced-scope simulations use fixed exchange flows derived from the corresponding full-scope simulation to represent the omitted market areas. The dataset comprises 570,028 hourly price series with 8,760 values each (EUR/MWh). It forms the empirical basis of the accompanying study, which analyses how the scope affects the average price and the intraday price variance of a target market area, whether results converge as the scope is extended, and how stable the ranking among bidding strategies remains across scope configurations. The data allow the reproduction of these results and can be reused for further analyses of geographic-scope sensitivity in electricity market simulations.

Content

Three Parquet files, one per bidding strategy group (Base, Flex, Welfare), generated for the study "The impact of geographic scope on electricity market simulation". Each file holds one row per hourly day-ahead price time series that was actually simulated (i.e. endogenously solved, with its own price formation -- see "Technical conditions" below), for every (scenario, weather scenario, simulation year, market zone) combination present for that strategy group within the study's scope. A "scenario" is one specific combination of jointly simulated market zones; n_zones records how many zones that combination contains. Across the three files: 570,028 time series, 48 European bidding zones, three weather scenarios, and simulation years 2028-2035. Combined size is approximately 8.6 GB.
Technical conditions under which the data were generated

Prices were produced with PowerACE, an agent-based day-ahead electricity market model (not included in this dataset; this archive contains only simulation output, no simulation code or model), using power plant fleet and demand assumptions based on ENTSO-E's ERAA 2025 (European Resource Adequacy Assessment) dataset. Each simulated scenario fixes a specific set of jointly simulated ("endogenous") market zones and a bidding strategy group (Base: a marginal-pricing; Flex: a variant with additional flexibility/storage representation; Welfare: welfare-maximising storage usage). Market zones outside a scenario's scope are not simulated endogenously; their effect on the scope is instead represented through fixed cross-border exchange flows taken from the corresponding full-scope (48-zone) simulation of the same strategy and weather/simulation year. For every scenario, PowerACE was run once per combination of weather scenario (driving renewable generation availability; three weather scenarios were used for this study) and simulation/target year, producing one hourly day-ahead price series (8,760 hours) per endogenously simulated market zone. This dataset collects exactly those raw hourly price outputs for the endogenous zones, restricted to the three weather scenarios and the scope variation used in this study.
File format and structure

Apache Parquet (columnar, Zstd-compressed, openly specified and language-independent). Columns: strategy (string, bidding strategy group), scenario (string, identifies the simulated zone combination), n_zones (integer, number of endogenously simulated zones in that scenario's scope), zone_set (string, comma-joined names of the endogenously simulated zones), wy (integer, weather scenario identifier), sim_year (integer, simulated calendar year), market (string, the zone this row's price series belongs to), price (array of 8,760 doubles, EUR/MWh, one value per hour of the simulated year). The data are intentionally row-per-series rather than a dense scenario x zone x hour array: not every scenario includes every zone, so a dense array would be mostly padding. np.stack(df["price"].to_numpy()) (Python/NumPy) turns any selection into a plain 2D time series matrix.
Software for viewing and reuse

No proprietary software is required. The files can be opened with any Parquet-capable tool, e.g.:

• Python: pandas.read_parquet(), pyarrow.parquet, or polars.read_parquet()
• R: the arrow package
• SQL-style ad-hoc querying without loading the full file into memory: DuckDB (duckdb.sql("SELECT * FROM 'dayahead_prices_Welfare.parquet' WHERE market = 'Germany_Luxembourg'"))
• General-purpose Parquet viewers (e.g. Tad, Parquet Explorer) for quick inspection without coding
  A companion open-source Python package, geoscope-analysis (MIT-licensed, published separately alongside this dataset), reads these files directly and reproduces the accompanying study's analyses -- the effect of scope on average price and intraday price variance, convergence as scope is extended, and the stability of bidding-strategy rankings across scope configurations; it is a convenience for reproducing the paper, not a requirement for using the raw data.
  Reuse possibilities

Beyond reproducing the accompanying study, the dataset supports any analysis of how simulated day-ahead price level, volatility, or cross-strategy differences change with simulated geographic scope -- e.g. benchmarking reduced-scope market models against full-scope references, studying neighbourhood/interconnection effects on price formation, or using the price series as input to downstream analyses (e.g. storage or flexibility valuation) under varying scope assumptions.
Known data gaps

Not every (strategy, scenario, simulation year) combination exists -- the different strategy groups covered different sets of scope combinations. this is expected and documented, not missing or corrupted data.
License

CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Attribution: Thorsten Weiskopf, Jonathan Stelzer Karlsruhe Institute of Technology.