Run EPM from Python#

Using the Python interface enables advanced features such as scenario creation, sensitivity analysis, and Monte Carlo simulations.

You don’t need to know Python—just follow the steps below.

You must have Python installed. See the prerequisites for setup instructions.

1. Create a Python Environment#

A Python environment ensures that all required libraries for EPM are available and isolated from other projects.

Important:

  • Before creating the environment, you should have GAMS installed on your computer, with a recent version (ideally >= 48).

On Windows#

On Windows, you first need to install Microsoft C++ Build Tools to compile some of the packages (specifically, chaospy):

  1. Go to VSCode build tools

  2. Download and install Build Tools for Visual Studio.

  3. During installation, check the option: C++ build tools

Then you can follow these steps:

  1. Open a terminal or command prompt (Anaconda Prompt).

  2. Navigate to the folder where you cloned EPM:

    cd EPM

  3. Create a new environment named epm_env:

    conda create -n epm_env python=3.10

  4. Activate the environment:

    conda activate epm_env

  5. Install all required libraries:

    pip install -r requirements_mac_and_windows.txt

On Mac#

You can follow the 5 to 9 steps above (the prior installation of Microsoft C++ Build Tools is not needed).

On Linux (remote server)#

  1. Open a terminal or command prompt.

  2. Navigate to the folder where you cloned EPM:

    cd EPM

  3. Create a new environment named epm_env, specifying some Linux compilers.

conda create -n epm_env python=3.10 numpy cython gcc_linux-64 gxx_linux-64 -c conda-forge

  1. Activate the environment:

conda activate epm_env
export TMPDIR=~/pip_tmp

  1. Install with conda the chaospy package, necessary for Monte-Carlo simulations

conda install -c conda-forge chaospy

  1. Install additional packages from the requirements.txt file specific to Linux

pip install -r requirements.txt

2. Run the Model (Basic Test)#

Once the environment is set up, you can test the model:

  1. Navigate to the epm code directory:

    cd epm

  2. Run the model:

    python epm.py

This runs the model using the default input folder and configuration.

3. Input Data#

All input data need to be within the epm/input folder. The path of the input file are defined based on this reference.

Input data are defined in a folder folder_input and controlled via a config.csv file, which specifies what CSV files to use for each parameter in the model.

Example input structure is provided in the GitHub main branch under the input folder. See the input documentation for full details.

By default, the model uses the data_test input folder and input/config.csv file, but command line arguments allow you to specify different folders and configurations.

Input data for EPM is specified through two key components:

  1. Input Folder (FOLDER_INPUT)
    This folder contains all the necessary .csv input files for the model, organized by type (e.g., data_test_region). It holds the raw data the model reads.

--folder_input#

  • Type: string

  • Default: data_test

  • Purpose: Folder containing the CSV input data for the simulation.

  • Notes: Should contain the input files referenced in the config.

  1. Configuration File (config.csv)
    This CSV file defines which input files correspond to which parameters in the model for the baseline scenario. It includes the parameter name and the relative file path inside the input folder. An example of a configuration file is available here: Example config.csv

--config#

  • Type: string

  • Default: input/config.csv

  • Purpose: Specifies the configuration file that defines baseline input CSV files and parameters.

  • Notes: The config file is required for proper data loading and scenario configuration.

Important Configuration Options in config.csv

Option

Description

solvemode

How the model is solved:
2 = normal (default)
1 = write savepoint PA_pd.gdx
0 = generate model only (no solve)

trace

Logging verbosity:
0 = minimal (default)
1 = detailed debugging output

reportshort

Report size:
0 = full report (default)
1 = compact report for multiple runs (e.g., Monte Carlo)

modeltype

Solver type:
MIP = default
RMIP = force LP relaxation

4. Specifying Optional Arguments for your Run#

Check Run EPM Advanced Features for additional explanations on advanced features like sensitivity analysis, Monte Carlo simulations, and scenario management.

--simple#

  • Type: list of strings

  • Default: [‘DiscreteCap’, ‘y’]

  • Purpose: Enables simplified model modes by toggling specified parameters.

--cpu#

  • Type: integer

  • Default: 1

  • Purpose: Number of CPU cores to use for running scenarios in parallel.

--engine#

  • Type: string

  • Default: None

  • Purpose: Path to a GAMS Engine file for running simulations remotely.

--postprocess#

  • Type: string

  • Default: None

  • Purpose: Runs only postprocessing on an existing output folder, skipping simulation.

--plot_selected_scenarios#

  • Type: list of strings

  • Default: "all"

  • Purpose: Specifies scenarios to plot in postprocessing.

--no_plot_dispatch#

  • Type: flag (boolean)

  • Default: False (dispatch plotting enabled by default)

  • Purpose: Disables automatic creation of dispatch plots to save time and memory.

--graphs_folder#

  • Type: string

  • Default: img

  • Purpose: Directory where postprocessing graphs and plots will be saved.

--scenarios#

  • Type: string

  • Default: None

  • Purpose: Path to a scenario definition file specifying multiple scenarios and their overridden inputs.

  • Notes: If omitted, only the baseline scenario will run.

--selected_scenarios#

  • Type: list of strings

  • Default: None

  • Purpose: Specifies a subset of scenarios from the scenarios file to run.

  • Example: --selected_scenarios baseline ScenarioA

--sensitivity#

  • Type: flag (boolean)

  • Default: False

  • Purpose: Enables sensitivity analysis which automatically modifies select parameters to evaluate model sensitivity.

  • Usage: Include --sensitivity to enable.

--project_assessment#

  • Type: list of strings

  • Default: None

  • Purpose: Specifies projects to exclude from the simulation to assess counterfactual scenarios.

  • Example: --project_assessment SolarProject

--montecarlo#

  • Type: flag (boolean)

  • Default: False

  • Purpose: Enables Monte Carlo simulation by sampling uncertain parameters multiple times.

  • Usage: Include --montecarlo to enable.

--montecarlo_samples#

  • Type: integer

  • Default: 10

  • Purpose: Number of random samples/scenarios to generate during Monte Carlo analysis.

  • Notes: Larger numbers increase accuracy but require more computation.

--uncertainties#

  • Type: string

  • Default: None

  • Purpose: CSV file that specifies uncertain parameters and their distributions for Monte Carlo sampling.

  • Notes: Required if --montecarlo is enabled.

--reduced_output#

  • Type: flag (boolean)

  • Default: False

  • Purpose: Generates reduced-size outputs to save memory, useful for large Monte Carlo runs.

Example Usage#

Run EPM with a custom input folder, sensitivity analysis enabled, and multiple CPU cores:

It is important to specify the config.csv path when running python. Otherwise, a default value will be used, calibrated on the data_test_region folder.

python epm.py --folder_input data_test_region --config input/data_test_region/config.csv --sensitivity --cpu 4

Run postprocessing only on a previous simulation folder:

python epm.py --postprocess simulations_run_2025-05-01_12-00-00

Run Monte Carlo analysis with 20 samples and a specified uncertainties file:

python epm.py --folder_input data_test_region --config input/data_test_region/config.csv --montecarlo --montecarlo_samples 20 --uncertainties input/data_test_region/uncertainties.csv