Data and code for nonlinear effects and multi-objective optimisation of traditional courtyard morphology in Lhasa
Description
This dataset accompanies the manuscript entitled “Nonlinear effects and multi-objective optimisation of traditional courtyard morphology on building energy use and winter indoor–outdoor thermal environments in Lhasa”. The study hypothesised that courtyard morphological parameters exert nonlinear and differentiated effects on building energy use, indoor cold discomfort, and courtyard outdoor cold stress, resulting in both synergies and trade-offs among these performance objectives. The ranges of the morphological variables were determined from field surveys of 33 courtyards in the historic Barkhor district, relevant planning constraints, and building-design standards. A field-validated parametric building-performance simulation workflow was then used to generate the dataset. Latin hypercube sampling produced 250 continuous morphological configurations, each of which was combined with four courtyard opening-orientation variants, resulting in 1,000 simulation cases. The CSV file contains seven input variables: courtyard length-to-width ratio (L/W), courtyard depth (D), perimeter-to-height ratio (P/H), south-facing window-to-wall ratio [WWR(S)], courtyard orientation (θ), enclosure degree (ED), and opening-orientation code (OP_code). The three outputs are annual building energy use intensity (EUI_annual, kWh/(m²·year)), winter indoor operative-temperature cold-discomfort degree-hours (OT_CDH_winter, °C·h), and winter courtyard UTCI cold-discomfort degree-hours (UTCI_CDH_winter, °C·h). Lower output values indicate better performance. The data reveal nonlinear responses and objective-specific effects. In particular, increasing enclosure degree generally reduced annual energy use and indoor cold discomfort but could increase outdoor cold discomfort, indicating a trade-off between indoor and outdoor performance. Courtyard depth and window-to-wall ratio also exhibited performance-dependent effects, while opening orientation influenced the distribution of the optimised solutions. The multi-objective results indicate that no single courtyard configuration simultaneously minimises all three outputs; therefore, the Pareto-optimal solutions should be interpreted as alternative design options reflecting different performance priorities rather than as one universally optimal form. The accompanying Jupyter Notebook contains the code used for BO-XGBoost modelling, sample-size adequacy analysis, SHAP-based interpretation, and NSGA-II multi-objective optimisation. The files can be used to examine nonlinear parameter–performance relationships, reproduce the reported analyses, compare courtyard configurations, and support performance-oriented courtyard renewal under comparable climatic and modelling assumptions.