Reliable day-ahead electricity forecasts are needed for building energy management, demand-side operation, and risk-aware scheduling. In common data-driven pipelines, targets are normalised for training, whereas forecast errors and uncertainty intervals are interpreted after inverse transformation to kWh. This creates a cross-space mismatch for heterogeneous building loads. We propose DSRC-TFT, a lightweight calibration extension of the Temporal Fusion Transformer that combines original-scale interval regularisation, linear residual refinement, and horizon-aware quantile calibration while retaining the TFT backbone. On BDG2, DSRC-TFT reduces MAE, RMSE, WAPE, pinball score, Winkler score, and prediction-interval width relative to TFT and quantile-regression baselines, with empirical coverage remaining close to the nominal 90% level. Sensitivity analysis reveals an accuracy–sharpness–coverage trade-off rather than a universally optimal interval-loss weight. Five independent random initialisations show limited variation in the main test metrics. Computational profiling further shows that the added modules increase the parameter count by less than 0.4% and profiler-reported FLOPs and inference latency by about 0.1%. The results support original-scale calibration as a practical way to improve probabilistic building electricity forecasts without materially increasing model complexity.
The residential building sector accounts for 17–20% of global CO 2 emissions, making it a critical decarbonisation target. This study presents an automated framework for carbon risk assessment that integrates scalable energy and emissions data into City Information Models (CIMs). It leverages TABULA building archetypes and power-law scaling (optimal exponent b = 0.85) to extrapolate energy demand and CO 2 emissions from minimal inputs (building geometry, heating system, and renovation status), and is designed for portfolio-scale screening rather than asset-level prediction. Validated against measured data from 10 Bavarian residential buildings, it achieved low aggregate bias, with normalised mean bias errors of 4.77% for energy and 4.17% for CO 2 , within a conservative ±5% benchmark. Individual-building deviations are larger (approximately −29% to +32%; MAPE 14–18%), as expected for archetype-based stock modelling and consistent with the documented energy performance gap; single-building estimates are therefore indicative. Strong correlations were observed (Pearson’s r = 0.825 for energy, 0.774 for CO 2 ). Stranding year predictions under CRREM 1.5°C pathways showed deviations of up to 4 years, and excess carbon costs were quantified using EU ETS price trajectories, providing policymakers with a scalable tool for retrofit prioritisation. Practical application With minimal user input and low computational requirements, this automated framework enables building services engineers, housing asset managers and municipal planners to estimate residential energy demand, emissions and carbon transition risk at portfolio and regional scales. Integrating building archetypes with City Information Models enables scalable assessment of extensive building stocks, where measured energy data are scarce. Spatial risk maps, indicative stranding years and excess carbon cost estimates help practitioners compare areas, prioritise retrofit programmes and target detailed assessments. These outputs support strategic decarbonisation planning and resource allocation across residential portfolios and regions.
Accurate prediction of building energy demand is essential for efficient HVAC system design and energy-aware building operation. Although machine learning models can achieve high predictive accuracy, their limited interpretability restricts their practical use in engineering workflows. This study proposes an interpretable analytical surrogate modeling framework that integrates data-driven learning with explicit analytical formulations for building energy prediction. A systematic feature selection strategy combining correlation analysis, mutual information, and physically informed reasoning identifies a compact subset of meaningful design variables. Machine learning benchmarking demonstrates that this reduced representation preserves predictive performance comparable to the complete feature set while substantially reducing model complexity. A Generalized Additive Model (GAM) is then employed to reveal nonlinear relationships between building design variables and annual energy demand. Based on the learned GAM response functions, explicit analytical surrogate equations are derived using piecewise quadratic and low-order polynomial approximations, enabling rapid benchmark-supported design screening. The transferability of the proposed workflow is further evaluated using an independent EnergyPlus-based ASHRAE prototype building. The results demonstrate that the same methodology can be successfully applied beyond the original UCI Energy Efficiency benchmark, providing an effective balance between predictive accuracy, engineering interpretability, and practical applicability for early-stage building design. Practical application The proposed analytical surrogate framework provides building services engineers with an efficient decision-support tool during the conceptual design stage. By combining physics-informed feature reduction with explicit analytical equations, the framework enables rapid estimation of annual heating and cooling energy demand, visualization of design trade-offs, and benchmark-supported screening of alternative building configurations without repeatedly performing computationally expensive building energy simulations. Rather than replacing detailed simulation-based analyses, the proposed workflow serves as a computationally efficient preliminary screening approach that helps designers identify promising design regions before undertaking comprehensive building performance assessment.
This study conducted a field survey of open-plan offices in industrial enterprises and divided them into two groups based on their distance from the production workshop: near (N-IOPO) and far (F-IOPO). Objective acoustic parameters (, and ) were measured in 7 N-IOPOs and 6 N-IOPOs, and occupants' acoustic satisfaction, noise sources disturbance and its interference on work productivity were obtained through a questionnaire, as well as perceived restoration from the Perceived Restorativeness Scale, which includes being away, extent, fascination, and compatibility. The results showed that speech noise (e.g., nearby and distant colleague chatting and telephone conversations) was the main source affecting acoustic satisfaction and work-related outcomes. In terms of perceived restoration, both N-IOPOs and F-IOPOs had low overall scores and showed no significant difference between them. Among the four dimensions, the sense of being away was the lowest, indicating that the office's inherent characteristics limited the upper limit of restoration. In addition, the acoustic environment can affect perceived restoration, and its effects on restoration differed between F-IOPOs and N-IOPOs. In F-IOPOs, objective acoustic parameters were positively correlated with restoration, while in N-IOPOs, restoration was correlated only with subjective acoustic satisfaction and not with objective acoustic parameters.Practical Application This study provides a practical basis for the acoustic environment design and assessment of open-plan offices in industrial enterprises. Firstly, noise levels meeting the requirements of the acoustic standard do not necessarily guarantee acoustic satisfaction or good restoration in IOPOs. Assessments should not focus solely on sound levels but should prioritise the control of speech noise. Secondly, steady-state industrial noise from neighbouring workshops could mask speech noise, potentially increasing acoustic satisfaction. Therefore, acoustic improvement should make good use of the steady-state background noise rather than simply focusing on noise reduction.
Accurate point-wise classification of structural objects on building surfaces is essential for high-fidelity 3D modeling and digital twin applications. However, maintaining spatial coherence in complex fa & ccedil;ade environments remains a significant challenge due to geometric heterogeneity. This study proposes an explainable machine learning framework for classifying fa & ccedil;ade-mounted structural elements from terrestrial laser scanning (TLS) point clouds using multiscale geometric features. The dataset, comprising over 10 million labeled points across nine classes, represents dense infrastructural components attached to building surfaces. To capture both fine- and coarse-scale structural details, nine geometric descriptors were evaluated across multiple neighborhood radii ranging from 0.02 m to 1.0 m. Feature relevance was assessed using three filter-based selection methods-Fisher score, Gini importance, and Chi-square-under an 80% cumulative contribution threshold. The resulting subsets were used to train Random Forest, XGBoost, and LightGBM classifiers, with macro-F1-based hyperparameter optimization to mitigate class imbalance. XGBoost combined with Fisher score-selected features yielded the best performance, achieving a 0.920 overall accuracy and 0.811 macro-F1. Class-wise SHAP analysis revealed that vertical positioning (Z) and scale-dependent surface descriptors are the primary drivers of class discrimination. These findings enable diagnosing systematic misclassifications, proving explainable frameworks offer transparent alternatives for enriching complex building fa & ccedil;ades.Practical Application This explainable framework provides AEC and digital twin professionals with a transparent, efficient tool to process massive TLS data. By identifying scale-dependent geometric features, engineering practitioners can accurately automate the semantic enrichment of complex fa & ccedil;ade elements (e.g., pipelines and structural components). Ultimately, these diagnostic insights minimize manual annotation costs, mitigate systematic misclassifications, and directly enhance Scan-to-BIM modeling and structural asset monitoring.
