Skip to content

EnergiCast v1.0.0 — Initial Public Release

Latest
Latest

Choose a tag to compare

View all tags
@TyMill TyMill released this 25 Sep 14:12
· 3 commits to main since this release
v1.0.0
a4682ee
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

EnergiCast v1.0.0 — Release Notes (Zenodo)

Release date: 2025-09-25
Repository: https://github.com/TyMill/EnergiCast
Documentation: https://tymill.github.io/EnergiCast/
License: MIT

Authors

  • Dr Tymoteusz Miller — University of Szczecin
  • Dr inż. Ewelina Kostecka — Maritime University of Szczecin

Summary

EnergiCast is a domain‑specific Python library for energy forecasting (load, PV, wind, prices).
It combines domain‑aware AutoML, physics–ML hybrid features (pvlib‑based solar/wind inputs), synthetic gap filling, hierarchical reconciliation, probabilistic metrics, and a standardized backtesting workflow with a simple CLI. See the project README and docs for details.

What’s included in v1.0.0

  • AutoML: tailored search spaces for energy time series; rolling‑origin validation and backtesting.
  • Pipeline: ForecastPipeline for data loading → imputation → feature generation → training → prediction → export.
  • Models (initial set): ARIMA/ETS wrapper and XGBoost forecaster (TFT on the roadmap).
  • Physics features: pvlib‑based solar features and energy‑specific feature generators.
  • Synthetic imputation: seasonal heuristics and stubs for generative variants (diffusion/GAN roadmap).
  • Hierarchical reconciliation: MinT‑style coherent forecasts (roadmap extensions).
  • Metrics: pinball loss, empirical CRPS, energy‑weighted errors.
  • CLI: train, backtest, export, report commands.
  • Docs site: guides for pipeline, models, metrics, backtest, CLI.

Installation

pip install energicast

Quickstart

python -m energicast.cli train --config examples/pv_config.yaml --out runs/demo_model
python -m energicast.cli backtest --config examples/pv_config.yaml --out runs/demo_backtest
python -m energicast.cli export --model-dir runs/demo_model --fmt pickle
python -m energicast.cli report --backtest-dir runs/demo_backtest

System requirements

  • Python: 3.9+
  • Core deps: pandas, numpy, scikit‑learn, xgboost, pvlib, optuna, matplotlib.
  • Optional: prophet, torch, pytorch‑lightning (some packages may be limited on ARM64).

Module overview

  • energicast.pipeline — end‑to‑end training/prediction/export pipeline.
  • energicast.models — ARIMA/ETS, XGB (TFT roadmap).
  • energicast.automl — Optuna‑based search + rolling‑origin validation helpers.
  • energicast.features — calendar & energy features (ramp‑rate, weekend/holiday, lags); pvlib solar features in data/pv.
  • energicast.impute — gap filling utilities.
  • energicast.metrics — probabilistic & energy‑weighted metrics.
  • energicast.backtest — standardized backtesting with plots/metrics.
  • energicast.scenarios, energicast.hier, energicast.bench, energicast.deploy — scenario generation, reconciliation, benchmarks, and export helpers.

How to cite

Please cite the Zenodo record once it is created. Example format:

Miller, T., & Kostecka, E. (2025). EnergiCast (v1.0.0) [Software].

Related identifiers

Keywords

energy forecasting; automl; time series; pvlib; xgboost; arima; ets; probabilistic forecasting; hierarchical reconciliation; backtesting; feature engineering

Support

Issues and feature requests: https://github.com/TyMill/EnergiCast/issues

This release note describes the initial public release of EnergiCast, providing a stable CLI, pipeline, metrics, and documentation suitable for research and prototyping.

Assets 2
Loading