Overview

ProLoaF is a probabilistic load forecasting project.

Key Capabilities

ProLoaF comes with the following scripts as entry points

  • preprocess.py: Preprocess your input data for use with ProLoaF.

  • train.py: Train an RNN model for load forecasting based on provided data.

  • evaluate.py: Evaluate a previously trained model.

  • baselines.py: Train and compare performance with typical statsmodels.

Dependency diagrams of basic entry points to ProLoaFDependency Diagrams

Example Workflow

  • Add the data a model should be trained on to the data/ folder

  • Preprocess the data using prep.py or custom scripts if needed.

  • Specify all required and desired settings and values in config.json in the targets/ folder

  • Train a model using python3 ./src/train.py -s <path_to_config_in_targets_folder>

  • Evaluate the model using evaluate.py in the same fashion.

Test Data

This repository contains the following submodules:

  • proloaf-data: Example gefcom2017 and open power system data for testing purposes.

Where should I go next?


Last modified April 2, 2022 : update links in docs (0f90685)