CoolFace
Datasetpublic

Tristanchou/Preference-Conditioned-Heterogeneous-MARL-Microgrid

SEGAN OPSD-Derived Microgrid Multiyear Benchmark This repository contains the processed multiyear microgrid benchmark used for the study “Preference-Conditioned Heterogeneous Multi-Agent Reinforcement Learning for Safe Microgrid Energy Management.” Files microgrid_opsd_multiyear.csv — processed hourly benchmark data. opsd_multiyear_metadata.json — provenance, selected OPSD nodes, source-column mapping, scaling notes, and processing metadata.… See the full description on the dataset page: https://huggingface.co/datasets/Tristanchou/Preference-Conditioned-Heterogeneous-MARL-Microgrid.

sourceHugging Faceotherupdated 13d agoView on Hugging Face
0likes36downloads
Dataset Card

SEGAN OPSD-Derived Microgrid Multiyear Benchmark

This repository contains the processed multiyear microgrid benchmark used for the study “Preference-Conditioned Heterogeneous Multi-Agent Reinforcement Learning for Safe Microgrid Energy Management.”

Files

  • —microgrid_opsd_multiyear.csv — processed hourly benchmark data.
  • —opsd_multiyear_metadata.json — provenance, selected OPSD nodes, source-column mapping, scaling notes, and processing metadata.

Provenance

The exogenous load, solar generation, wind generation, and day-ahead price traces are derived from the Open Power System Data (OPSD) Time Series package, Version 2020-10-06.

Upstream dataset DOI: 10.25832/time_series/2020-10-06

The metadata file records the original source file as time_series_60min_singleindex.csv, covering 2014-12-31 23:00:00+00:00 to 2020-09-30 23:00:00+00:00, with 50401 hourly rows and 2100 complete 24-hour windows.

Selected OPSD regions/nodes: DK1, DK2, AT, GBGBN, ITNORD

Important interpretation

This is an OPSD-derived, trace-driven benchmark, not a claim that every column is a directly measured microgrid variable. The load, renewable-generation, and price traces are derived from OPSD time series after documented preprocessing and scaling. Additional operational variables in the processed benchmark may be derived or model-constructed.

In particular, the metadata explicitly states that carbon_factor_kg_kwh is a dispatch proxy derived from renewable penetration, seasonal stress, and risk because the OPSD source file does not contain measured hourly carbon intensity.

Dataset shape

  • —Rows: 50,401
  • —Columns: 19
  • —Frequency: 1 hour
  • —Columns: timestamp, pv_kw, wind_kw, load_kw, critical_load_kw, flexible_load_kw, price_cny_kwh, carbon_factor_kg_kwh, risk_level, reserve_ratio, island_flag, hour, month, dayofyear, node_renewable_ratio_mean, node_renewable_ratio_std, node_deficit_pressure_mean, node_deficit_pressure_std, node_price_signal_mean

Scaling and conversion notes

  • —Target mean load: 260.0 kW
  • —Renewable multiplier: 3.0
  • —Price conversion: pricecnykwh = day-ahead EUR/MWh * 7.8 / 1000; values clipped to [0.02, 2.50].
  • —Missing values: all rows retained after interpolation according to the metadata.

Citation

Please cite the upstream OPSD dataset:

Open Power System Data, Time Series Package, Version 2020-10-06, 2020. DOI: 10.25832/time_series/2020-10-06

A paper-specific citation can be added after publication.

Reproducibility

For full reproduction of paper results, this dataset should be used together with the corresponding preprocessing, training, evaluation, and plotting scripts/checkpoints from the project repository.