Sample Data and Inference Code for<b>“Data-driven solar forecasting enables near-optimal economic decisions”</b>

Sample Data and Inference Code for<b>“Data-driven solar forecasting enables near-optimal economic decisions”</b><b>Authors:</b><br>Zhixiang Dai¹⁺<i>, Minghao Yin²˒³⁺, Xuanhong Chen⁴⁺</i>, Boris Bonev¹, Alberto Carpentieri¹, Chengzhe Zhong²˒³, Jussi Leinonen¹, Thorsten Kurth¹, Jingan Sun¹, Ram Cherukuri¹, Yuzhou Zhang¹, Ruihua Zhang¹, Farah Hariri¹, Xiaodong Ding³, Chuanxiang Zhu⁴, Dake Zhang⁵, Yaodan Cui², Yuxi Lu², Yue Song², Bin He², Jie Chen², Yixin Zhu⁶˒⁷, Chenheng Xu⁷, Maofeng Liu⁸, Zeyi Niu⁹˒¹⁰, Wanpeng Qi¹¹, Xu Shan¹², Siyuan Xian¹³, Ning Lin¹³˒¹⁴, Michael Oppenheimer¹⁵˒¹⁶˒¹⁷, and Kairui Feng²˒³*<b>Affiliations:</b>NVIDIA Corporation, Santa Clara, CA, USAState Key Laboratory of Autonomous Intelligent Unmanned Systems, Tongji University, Shanghai, ChinaShanghai Innovation Institution, Shanghai, ChinaSchool of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, ChinaAntai College of Economics and Management, Shanghai Jiao Tong UniversitySchool of Psychological and Cognitive Sciences, Peking UniversityInstitute for Artificial Intelligence, Peking UniversityDepartment of Atmospheric and Oceanic Sciences, School of Physics, Peking UniversityShanghai Typhoon Institute, Shanghai, ChinaKey Laboratory of Numerical Modeling for Tropical Cyclone of the China Meteorological Administration, Shanghai, ChinaQinghai Meteorological Bureau, Qinghai, ChinaMax Planck Institute for Biogeochemistry, Jena, GermanyDepartment of Civil and Environmental Engineering, Princeton University, Princeton, USAAndlinger Center for Energy and the Environment, Princeton University, Princeton, USADepartment of Geosciences, Princeton University, Princeton, NJ, USAPrinceton School of Public and International Affairs, Princeton University, Princeton, NJ, USAHigh Meadows Environmental Institute, Princeton University, Princeton, NJ, USA<b>Data Description</b><b>/Solar_Economy-main</b><br>Contains the Python notebook for downloading input data and running inference with <b>SunCastNet</b>, demonstrating the complete workflow for data preparation and forecasting.<b>/Sample_Power_Industrial_Usage</b><br>Provides typical power consumption profiles for representative industrial sectors, used as benchmark demand scenarios in the economic backtesting experiments.<b>/Background_field</b><br>Includes ERA5 reanalysis data fields, which serve as boundary conditions for generating hindcasts and as the meteorological foundation for model training and validation.<b>/Annual_variability</b><br>Contains annual variability data for 2020–2025, used to evaluate the robustness of strategies under interannual fluctuations in solar resource availability.<b>/Daily_Prediction</b><br>Provides sample daily prediction outputs for the year 2020, illustrating the format and structure of SunCastNet forecasts at 10-minute temporal resolution and 0.05° spatial resolution.