An optimal sizing model and post-optimal budget analysis framework for standalone PV-Battery-Hydrogen energy system

Budget-Constrained Sizing Optimization for Battery–Hydrogen Hybrid Systems (BHESS)

Optimal sizing and post-optimal budget-resilience analysis of a stand-alone PV–Battery–Hydrogen Energy Storage System (BHESS) across three European climate regions (Vaasa, Vienna, Naples), reproducing the methodology and results of:

T. Tran and J. Songok, "Budget-Constrained Sizing Optimization for Battery-Hydrogen Hybrid Systems Across European Climate Regions," Proceedings of SEEP2026, Tor Vergata University, Rome, Italy, 20–23 July 2026. Paper ID: SEEP-1063.

This repository contains the modeling framework only. The conference paper and presentation slides are not included here; see the Citation section below for how to reference the published work once available.

Problem statement

A stand-alone, fully islanded system (PV array, battery, PEM electrolyser, PEM fuel cell, H₂ tank) is sized to meet an hourly electrical and hydrogen demand over a full year (8,760 h), minimizing lifecycle Net Present Cost (NPC) subject to reliability constraints (LPSP, LOLE). After the unconstrained optimum is found, the available investment (CAPEX) is progressively reduced and the model is re-solved at each budget level to study:

• how PV / battery / electrolyser / fuel cell / tank capacities are reallocated as budget shrinks,
• how NPC, LCOE, and LCOH evolve,
• how far each location (Vaasa, Vienna, Naples) can tolerate budget reduction before becoming infeasible.

The optimization is formulated as an LP in Pyomo and solved with Gurobi.

Repository structure

.
├── Model_BHESS.py                                        # Standalone script (HPC/Puhti-ready)
├── configurable_pv_battery_hydrogen_budget_resilience.ipynb  # Notebook version (interactive)
├── data/
│   └── Input_sample.csv                                  # Hourly irradiance/temperature/load data (8,760 rows)
├── requirements.txt
└── README.md

The script expects Input_sample.csv next to itself; the notebook expects it under data/Input_sample.csv relative to the notebook's working directory. Adjust the path variables at the top of either file if your layout differs.

Input data

Input_sample.csv contains 8,760 hourly rows with the following columns:

Column	Description	Unit
Month, Season, Hour	Calendar/time index	—
Irradiance_<Vaasa/Vienna/Naples>	Global horizontal irradiance (PVGIS, 2023)	W/m²
Temp_<Vaasa/Vienna/Naples>	Ambient temperature (PVGIS, 2023)	°C
Case_1_Residential_Elec [kWh] / ..._H2toFuel [kg]	Case 1 (residential) demand	kWh, kg/h
Case_2_Hospital_Elec [kWh] / ..._H2toFuel [kg]	Case 2 (hospital) demand	kWh, kg/h
Case_3_Manufacture_Elec [kWh] / ..._H2toFuel [kg]	Case 3 (industrial, used in the paper) demand	kWh, kg/h

The electrical profile for Case 3 is adapted from the "Industrial-1-shift Fabricated Metals" facility dataset; the hydrogen profile is a constant 1.5 kg/h, representative of a medium-sized fleet of H₂-powered forklifts.

How to run

Option A — Script (Model_BHESS.py)

Intended for batch / HPC use (e.g. Puhti/CSC with SLURM). Edit the scenario selectors at the top of the file, then run:

pip install -r requirements.txt
python Model_BHESS.py

Set location to one of 'Vaasa', 'Vienna', 'Naples' and CASE to one of 'C1', 'C2', 'C3'. The script:

1. Solves the unconstrained baseline optimal sizing problem.
2. Prints NPC, CAPEX, LCOE, LCOH, LPSP, LOLE, and component sizes.
3. Runs a post-optimal budget sweep (CAPEX capped at 80%–100% of the baseline in 1% steps) and re-solves at each level.
4. Writes LCOH_Result_<location>_<CASE>.xlsx with a Results sheet (sizing + hourly dispatch) and a Budget_Sweep sheet.

Option B — Notebook (configurable_pv_battery_hydrogen_budget_resilience.ipynb)

Same model and workflow as the script, reorganized into cells for interactive exploration. Open with Jupyter, adjust location / CASE in the scenario cell, and run all cells top to bottom. Outputs are written to an outputs/ folder created next to the notebook.

Solver requirement

Both entry points call SolverFactory('gurobi') and require a working Gurobi installation and license (free academic licenses are available). See:

• Gurobi Optimizer installation guide
• Gurobi License Portal
• Academic program and licenses

A full 8,760-hour LP solve can take from a few minutes to several hours depending on hardware; the script sets an 8-hour wall-clock limit (TimeLimit = 28800) as a safety net for HPC job limits.

Output

Both the script and the notebook produce an Excel workbook LCOH_Result_<location>_<CASE>.xlsx with:

• Results sheet — baseline component sizing (PV, battery, electrolyser, fuel cell, tank), NPC, CAPEX, LPSP, LOLE, LCOE, LCOH, and the full 8,760-hour hourly dispatch table (PV generation/dispatch, battery SOC, H₂ production/consumption, tank SOH, curtailment).
• Budget_Sweep sheet — one row per budget factor (80%–100% of baseline CAPEX), with the re-optimized sizing, NPC, LCOE, LCOH, LPSP, and LOLE at each level.

Citation

If you use this code or data, please cite the SEEP2026 paper:

@inproceedings{tran2026bhess,
  title     = {Budget-Constrained Sizing Optimization for Battery-Hydrogen
               Hybrid Systems Across European Climate Regions},
  author    = {Tran, Tram and Songok, Joel},
  booktitle = {Proceedings of SEEP2026},
  address   = {Tor Vergata University, Rome, Italy},
  year      = {2026},
  note      = {Paper ID: SEEP-1063}
}

Acknowledgements

This research was funded by CETP, the Clean Energy Transition Partnership, under the 2023 CETP joint call for research proposals (GA N° Cetp-FP-2023-00567), co-funded by the European Union.