API

All the modelling code lives in the src package. Running the study does not require it, because python workflow/main.py calls everything. This section is for reusing one step of the method in other work, or for seeing exactly how a step in the paper is written in code. The cards lead to the reference; the guide below explains how it can be put to use.

Overview Module overview What each module does, with examples and the full workflow in one script. Method Representative days src.cluster: k-means clustering that picks the typical days. Study Reproduce helpers src.reproduce: run scenarios, load results and rebuild the paper's tables. Reference Full reference Every module and function in src, generated from the source code.

How to use this section

1. Locate the step of interest. Each module does one step of the method, so the reference reads like the method section of the paper:

Step in the study	Module	What it provides
Pick representative days	src.cluster	cluster_data() groups the 365 days with k-means and returns the chosen days plus a day-by-day map of which representative day stands in for each real day
Shrink the hourly data	src.update_CFandSDP	CFandSDP() keeps only the representative days and averages (or sums) hours into blocks
Build the time structure	src.update_yearsplit, src.update_conversion*, src.update_daysindaytype	The OSeMOSYS time-slice sets and weights each storage formulation needs
Rebuild the intra-day SoC	src.Kotzur_intraday, src.Welsch_intraday2	Turns the solved Kotzur and Welsch storage levels into a SoC series for every time block of the year
Solve and collect	src.simulation, src.table_generator	Runs GLPK and writes cost, capacity and runtime to Simulation_Results.xlsx
Reproduce the paper	src.reproduce	run_case(), load_results(), add_errors(), table4() for scripted scenario sweeps

2. Call it from a script. The functions take file paths and plain Python values, so any one of them can be used without the rest. For example, the clustering can be applied to other wind and demand profiles to see which representative days it picks:

from src.cluster import cluster_data

days, sequence = cluster_data("my_wind_cf.csv", "my_demand.csv", n_clusters=8, random_state=0)
print(days)          # the 8 representative days
print(sequence[:7])  # which representative day stands in for each day of the first week

Similarly, one scenario of the study can be rerun with all four formulations, returning a table:

from src.reproduce import run_case

df = run_case("WND", k=8, h=1)
print(df[["Model", "Total cost", "Storage Capacity"]])

3. Use it as a reference for other models. For work in another framework, the modules are a compact recipe: how many representative days to use, how to weight them, and how each formulation links storage across days. Reading src.update_yearsplit and the update_conversion* modules next to About Storage Modelling shows exactly what each formulation adds to a model, which can then be ported to another tool.