Reproduce “Foundation-Style Pretraining for Scalable and Transferable Distributed PV Forecasting” 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
Dayin Chen, Xiaodan Shi, Xin Xu, Zhongyue Su, Shibo Zhu, Ruizhe Deng, Dongxiao Zhang, Yuntian Chen and 1 more
The authors present a shared PV forecasting model pretrained across stations and horizons, treating future PV as a missing part of a sequence containing historical power, weather, time information, and available weather forecasts. In zero-shot tests on 139 held-out Ningbo stations, its mixed-reconstruction model achieved seven-day MAE 0.127 and RMSE 0.168, outperforming a 330M-parameter baseline; evaluation was limited to Ningbo.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.