Addressing the issues of weak generalization capability of equipment models, poor algorithm compatibility, and insufficient adaptability to multi-device coupling and dynamically changing topologies in the simulation modeling of multi-chiller HVAC systems, this paper proposes a modular, configurable steady-state universal modeling framework. It adopts a three-level decoupled architecture: (1) At the equipment level, a high-precision hybrid model library is built by integrating physical principles and data-driven techniques, supporting the plug-and-play of heterogeneous devices. (2) At the topology level, adaptive representation of topological structures and flow distribution are achieved through basic connection configurations, no longer relying on complex pipeline network parameters. (3) At the solution level, an iterative algorithm is employed to solve the coupled cooling water inlet and outlet temperatures. Experimental validation based on a real hospital system in Taizhou demonstrates the framework’s excellence: The hybrid equipment models achieve a coefficient of variation of the root mean square error (CV-RMSE) of 4.38
rapid prediction of wind fields in complex urban built environments is essential for ventilation design and pollutant control. However, computational fluid dynamics (CFD) methods are computationally expensive, while machine learning approaches rely on mesh-based representations, limiting their applicability to large-scale urban geometries. To overcome these challenges, this paper proposes a geometry-driven neural operator model that establishes a direct mapping from building surface geometry to 3D flow fields, enabling mesh-free and efficient prediction. The model extracts geometric features through multi-scale spherical queries and leverages continuous operator learning to represent flow structures governed by urban morphology. Evaluations on both idealized building arrays and realistic urban configurations demonstrate that high-density sampling around buildings improves prediction accuracy in complex regions, reducing the median errors of velocity magnitude (Vmag) and turbulence kinetic energy (k) by 17.1
City-scale indoor overheating assessment is limited by the lack of methods that are simultaneously scalable, equitable, and applicable across diverse residential building stocks. This review evaluates existing indoor temperature prediction methods for this purpose, particularly given that outdoor temperature proxies explain less than 40
Identifying the main influencing factors of surface urban heat island (SUHI) and their marginal effects is crucial for developing effective mitigation countermeasures. However, existing studies mainly focus on a single period and lack a multi-period perspective. This study took Tianjin as the study area, analyzing SUHI intensity (SUHII) distribution characteristics and differences among local climate zone (LCZ) types across five different periods (morning, before noon, noon, afternoon, and evening) based on multi-temporal ECOSTRESS data acquired under similar background meteorological conditions. Furthermore, XGBoost-SHAP model was used to explore nonlinear relationships and marginal effects between SUHII and influencing factors. The key findings are: (1) the number of high SUHII areas first increased and then decreased, with peak values at noon. The distribution characteristics of SUHII and LCZ were most consistent before noon and in the evening; (2) LCZ1, LCZ2, LCZ3, LCZ5, and LCZ8 generally exhibited highest SUHII during daytime, whereas LCZ6 and LCZ9 exhibited lowest SUHII across all periods; (3) during daytime periods, urban morphological and land cover factors were main influencing factors, while human activities played the main role in the evening. Some factors exhibited nonlinear relationships with SUHII, with distinct thresholds and marginal effects. Notably, high-rise LCZ types showed turning marginal effect with average building height (ABH) at noon: SUHII was negatively correlated with ABH below 70 m, while positively correlated above this threshold. Based on these findings, specific countermeasures were developed across LCZ built types, offering practical contributions to the mitigation and management of SUHII.
Thermal sensation vote (TSV) is a key indicator in thermal comfort research due to its integration of subjective perception with environmental conditions. Existing methods of estimating TSV rely on linear estimation, overlooking the intrinsically nonlinear influence of human thermal perception. To address such limitations, this study aims to systematically investigate how indoor meteorological variables shape studentsu2019 thermal sensation by coupling measurements, questionnaires, and explainable machine learning (ML). Environmental parameters, including air temperature (Ta), globe temperature (Tg), relative humidity (RH), and air velocity (Va), together with studentsu2019 thermal sensation votes (TSV), were collected across transitional, summer, and winter conditions. Based on the coupled subjectiveu2013objective dataset, a random forest model was developed to quantify TSV under realistic classroom conditions. The model demonstrates good performance across seasonal scenarios (R2 = 0.992, RMSE = 0.094), indicating its effectiveness in capturing nonlinear relationships between subjective thermal sensation and environmental variables. Model interpretation was supported through SHapley Additive exPlanations analysis, enabling identification of the most dominant driver as Tg, followed by Ta, RH and Va. Scenario-based analysis further shows that a 1 u2103 increase in Tg is associated with an increase in TSV of approximately 0.35 to 0.50, compared with only 0.10 to 0.18 for an equivalent increase in Ta, highlighting the substantially stronger influence of radiant environment on perceived comfort. Overall, this study provides a robust, data-driven framework for analyzing thermal sensation in secondary school classrooms and offers practical support for performance-oriented indoor environmental control and design in hotu2013humid climates.
