Reproduce “Towards Continuous Power Forecasting: Practical Continual Learning for Real-World Energy Systems in Nonstationary Time Series” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yujiang He, Frederic Uhrweiller, Bernhard Sick
The authors frame power forecasting as continual learning: a model monitors incoming data and updates when novelty buffers fill. CLeaR uses reconstruction and prediction errors to detect change, with replay- and EWC-based update methods. Across 95 entities, Familiarity-based Replay had lower test error than the full-data baseline, while Generative Replay had the lowest fitting error and forgetting; rankings differed for reconstruction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.