Simulated Dataset for Thermal Resilience Assessment of Affordable Housing in Surakarta, Indonesia under Current and Future Climate Scenarios

This dataset contains simulation-based results used to develop and evaluate a probabilistic surrogate model for thermal resilience assessment of affordable housing under progressive climate change. The data were generated using a calibrated EnergyPlus model of a naturally ventilated two-story dwelling representative of low-income housing in tropical regions.

The dataset comprises 3,500 parametric simulation cases generated using Sobol’ sequence sampling across ten passive design parameters, combined with seven climatic scenarios representing current conditions and future projections (2050, 2070, and 2090) under Representative Concentration Pathways (RCP4.5 and RCP8.5). Climate projections were derived using Meteonorm v8.2 based on CMIP5 ensemble data.

The weather files (WF) are encoded into a single climatic indicator (Cooling Degree Days; CDD), and the resulting annual percentages of thermal discomfort were computed using a regionally adjusted adaptive thermal comfort model using the 80% acceptability limits (±3.5°C around the comfortable temperature). The file consists of:

• Input features:
  • Orientation (Or) [Note that this parameter is encoded into sine and cosine fesatures during training]
  • WWR for the front window (WWR-Fr)
  • WWR for the side window (WWR-Si)
  • Overhang shading projection (Shd_Proj)
  • XPS insulation thickness for roof (R_XPS_t)
  • XPS insulation thickness for walls (W_XPS_t)
  • Solar absorptance for roof (R_SA)
  • Solar absorptance for wall (W_SA)
  • Windows thermal transmittance (Win_U)
  • Windows solar heat gain coefficient (SHGC)
  • WF [excluded from training]
  • CDD [Climate scenario indicator]
• Outputs for both the bedroom (BR_) and the living room (LR_):
  • Annual percentage of hours exceeding the upper acceptability limit (pct_hot)
  • Annual percentage of hours below the lower acceptability limit (pct_cold)
  • Total discomfort hours (TDH)

Although the dataset contains multiple comfort-related outputs for both rooms, only the bedroom's upper-limit exceedance metric (BR_pct_hot) was used as the target variable for surrogate model training and subsequent design applications. The remaining outputs are provided for transparency and potential reuse. Potential applications include cimate-resilient building design studies, uncertainty-aware building performance optimization, and educational research in building performance simulation.

The numerical results reported in this study are specific to the modeled dwelling, construction system, and climatic context. The relatively higher overheating percentages observed in this dataset, compared to generic reference-building simulations (e.g., standard U.S. DOE residential archetypes), arise from the calibrated representation of ventilation behavior, envelope characteristics, and spatial constraints informed by field observations of the target dwelling. Consequently, direct generalization of absolute discomfort values to other building typologies, construction systems, or regions should be undertaken with caution.