Profiler

This profiler loads an OpenDSS model and its connected loadshape/pvshape profiles to generate new synthetic data.

Setup

This profile utilizes TSGM.TimeGAN to generate synthetic data. If you wish to speed up results, we recommend setting up your GPU to speed up model training.

git clone https://github.com/pnnl/oedisi_dopf.git
cd oedisi_dopf/profiler
poetry update

Run

Extract Features

The first step is to extract features to train TimeGAN with. Run the extract_features.py command with the path to the model folder and the name of the model that will be used for storing the new synthetic information.

poetry run python src/feature_extraction.py <model_path> <model_name>

Example:

poetry run python src/feature_extraction.py ../builds/lindistflow/ieee123/feeder/opendss ieee123

Train Model

Once the features have been extracted, the main.py function will load the features from the specified model_name and generate new synthetic profiles for each feature.

poetry run python src/main.py <model_name>

Example:

poetry run python src/main.py ieee123

Evaluate

If you are interested in evaluating the new profiles, you can generate the maximum mean discrepancy (MMD) for random samples and visualize the t-SNE plot for model inspection. run the evaluate.py

poetry run python src/evaluate.py <model_name>

Example:

poetry run python src/evaluate.py ieee123