NESP — No-Regret Investments for Nordic Resource Adequacy

This repository provides the open-access code and result datasets for a methodology that identifies no-regret transmission and generation investments — investments that remain near-optimal across a set of future scenarios to which probabilities cannot be assigned. The method was developed at RISE within the Nordic Energy Systems Programme (NESP) and demonstrated on the Swedish power system. Funded by the Swedish Energy Agency (Energimyndigheten), project P2025-01807.

Overview

Regret of applying one scenario's optimal portfolio in another scenario, relative to that scenario's own optimum. The definition and values match the regret figure in the project's final report.

The regret of a portfolio is the additional system cost incurred when it is committed and a different scenario is realised. A portfolio with low regret across all scenarios is robust. In the Swedish study, the EP and EF portfolios remain low-regret across scenarios (at most a few hundred MEUR/yr), whereas the SF and FM portfolios can incur regret of tens of thousands of MEUR/yr under an adverse realisation. The investments common to all scenario-optimal portfolios define a no-regret core.

Quick start

The result datasets are small and included in the repository, so the results can be inspected without installing the model:

pip install pandas matplotlib jupyterlab
jupyter lab notebooks/explore_results.ipynb

This notebook covers the regret matrix, the minimax-regret ranking, the no-regret core, the multi-criteria ranking, and the portfolios, using the shipped data. The figure above is regenerated with python scripts/plot_regret_matrices.py. A step-by-step treatment of the method is provided by the tutorial series in notebooks/ (see notebooks/README.md): five notebooks, each runnable on the shipped datasets.

Contents

Task	Location
Inspect the results	notebooks/explore_results.ipynb
Follow the method step by step	notebooks/ (tutorial series; notebooks/README.md)
Understand each dataset	data/README.md, config/data_provenance.yaml (data dictionary)
Locate a script	scripts/README.md
See what is and is not included	CURATION.md
Re-run the full study	"Reproduce the full study" below

Repository layout

• notebooks/ — results walkthrough and the tutorial series (notebooks/README.md)
• scripts/ — the methodology pipeline (case building → conflict/regret → no-regret core → MCDA → clustering → stability); see scripts/README.md
• src/ — openTEPES input/output wrappers and per-work-package notes (AP1/AP2/AP3)
• config/ — scenario registries and data_provenance.yaml (each input with its source)
• data/network_xeval/ — the curated result datasets
• figures/ — generated figures
• tests/ — unit tests

Relation to openTEPES

The capacity-expansion solves use openTEPES, an open-source stochastic MILP model (https://github.com/IIT-EnergySystemModels/openTEPES, GNU AGPL-3.0). Enhancements developed for this project — a Big-M Kirchhoff formulation for AC line expansion, per-technology real WACC, binary investment handling, and adequacy reporting (LOLE, ENS) — were contributed upstream and are available there. This repository depends on openTEPES as a package and does not bundle its source.

Data sources

Public national sources: Svenska kraftnät Long-term Market Analysis (LMA 2024; four scenarios; 2035 demand 176–270 TWh), the Nätutvecklingsplan 2024–2033 grid plan, ENTSO-E TYNDP 2024 candidate projects, and VRE profiles from the Bio-FlexGen project / ERA5. The adequacy standard follows EU Regulation 2019/943. Full provenance is in config/data_provenance.yaml.

Reproduce the full study (advanced)

The shipped datasets are the outputs of the method. Re-solving the twelve cases requires openTEPES and the LMA/grid input data (not shipped; ~105 GB of raw solves):

pip install -r requirements.txt          # openTEPES + the analysis stack
python scripts/create_scenario_cases.py  # build the 12 cases (requires the input data)
python scripts/run_experiment.py list    # list the pipeline stages

See CURATION.md for caveats and scripts/README.md for the stage order.

Acknowledgments

Akshaya Tammanur Ravi (formerly RISE Research Institutes of Sweden) contributed to the early stages of this project.

Licence and citation

• Code: GNU AGPL-3.0 (LICENSE). Data: CC-BY-4.0 (data/LICENSE).
• Cite via CITATION.cff (and its DOI once released) and the openTEPES SoftwareX paper (Ramos, Alvarez & Lumbreras, 2022).