A Hybrid Benchmark Dataset for Multi-Home Residential Energy Management under Heterogeneous Building and Energy Conditions

Each household represents heterogeneous residential conditions, including occupancy patterns, building characteristics, HVAC behavior, photovoltaic generation, battery storage, base-load demand, and shiftable appliance operation. Indoor thermal dynamics are generated using reduced-order RC models with home-specific parameters to provide physically consistent comfort modeling. The dataset is organized into chronological training, validation, and testing subsets without random shuffling, enabling evaluation under future unseen conditions.

Additional files include household metadata and real-time price forecasting data used for predictive energy management studies. This dataset is intended for research on multi-home energy management, constrained reinforcement learning, demand response, HVAC control, battery scheduling, photovoltaic integration, and scalable smart-grid optimization. The complete dataset documentation, including data sources, construction methodology, variable descriptions, and reference basis, is provided with this release.