Building intelligence remains limited by fragmented data, labor-intensive semantic modeling, and AI interfaces that lack grounding in verifiable building facts. This study presents an Agentic AI framework that combines a simplified semantic modeling approach, termed KebGraph, with large language model (LLM)-based AI agents. The framework enables end-to-end workflows from natural language instructions to engineering services. KebGraph organizes heterogeneous building context into a queryable semantic layer, while a multi-agent architecture maps user requests to concrete engineering actions through tool execution. Evaluated on a real office building, the framework is tested across four large language models on a 213-question benchmark. Grounding on KebGraph attains the highest factual question-answering accuracy of the compared representations, ahead of a conventional Brick model and a raw-document baseline, while using the fewest input tokens among the grounded representations. In an agentic function-calling evaluation, it also achieves the highest average argument-formation score across the evaluated models. On invalid or adversarial requests, the capable models largely decline or ask for clarification rather than retrieving fabricated data. An indoor thermal modeling workflow further demonstrates agent-driven construction, calibration, and evaluation of thermal models end to end over a one-month winter period. These results indicate that grounding LLM agents in a lightweight building semantic layer improves the accuracy, efficiency, and reliability of information retrieval and tool use on this building, while broader generalization across buildings and seasons remains future work.
Large-scale HVAC demand response can provide grid flexibility, but market participation requires fast aggregate allocation, device-level feasibility and protection of occupant comfort. Although recent research has advanced building-flexibility assessment, learning-based control and distributed service provision, evidence is still limited for allocating a post-clearing dispatch request across many lower-level HVAC aggregators while accounting for accumulated thermal comfort burden. Here we develop a hierarchical aggregation framework in which a deep reinforcement learning (DRL) allocator assigns market-requested flexibility among resource aggregators using both current conditions and accumulated comfort impact. Local linear programming (LP) controllers then schedule HVAC operation under comfort and power constraints. The framework is evaluated for supply-demand adjustment requests defined by flexibility magnitude, direction, response-time tolerance and continuous delivery period. Compared with a centralised LP benchmark and a rule-based Greedy+LP baseline, the proposed DRL+LP framework reduces computation from more than 1,000 to a few seconds at large scale. At the 3,600-room scale, it assigns requests up to 0.792 MW for three hours and keeps aggregate assigned requests within a ± 10% tolerance in nearly all reported extraction intervals. Its comfort-violation metrics remain close to the centralised LP reference, with absolute differences no larger than 0.052 ∘C in mean absolute error (MAE) and 0.045 ∘C2 in mean squared error (MSE) in the reported cases. It also avoids the uneven comfort burden observed in greedy allocation under low-flexibility requests. In an unseen two-request daily scenario, the learned policy maintains request allocation within tolerance and median comfort deviations of approximately 0.04–0.05 ∘C. These results establish a scalable and comfort-aware allocation approach within the simulated market-dispatch and continuously modulatable HVAC setting considered here. Evidence remains limited to a single calibrated room model, cooling operation and one climatic region; performance under heating, heterogeneous buildings, diverse occupancy and field communication or model uncertainty has not yet been established.
Building-integrated greenery (BIG) is increasingly used to support environmental performance, human experience, and urban ecological functions, yet research priorities vary across geographical contexts. This study compares China-context and rest-of-corpus research on biophilic sky gardens and related green systems. Publications from 2018 to 2026 were retrieved from Scopus and Web of Science, screened by study or case location, deduplicated, and divided into 203 China-context and 2095 rest-of-corpus records. The analysis combined VOSviewer keyword co-occurrence networks with standardized occurrence frequencies, relative prominence ratios, temporal patterns, thematic clusters, and a structured review of traceable case evidence. Both corpora shared a green-roof-centred core encompassing green infrastructure, environmental performance, and ecosystem services. Rest-of-corpus research showed broader associations with green walls and façades, nature-based solutions, biodiversity, ecosystem services, and urban green systems. China-context research placed comparatively greater emphasis on roof-based thermal and microclimatic performance, runoff and water management, sponge-city and SWMM modelling, and optimization. Occupant health appeared in both corpora but had less longitudinal or building-level evidence than technical and environmental performance. Case comparisons showed that climate, building use, construction, accessibility, operation and maintenance, and urban morphology and scale condition the performance of BIG interventions beyond broad regional distinctions. These findings inform a nested context-adaptive framework in which urban morphology and scale provide the overarching boundary condition for three intersecting pillars: integrated greenery performance, human–environment interaction, and ecosystem resilience, while long-term operation and assessment form the framework's evidential base. The comparison treats the two corpora as research-emphasis lenses rather than competing design paradigms and provides a basis for cross-context verification and refinement in future comparative research, performance evaluation, and iterative design across diverse contexts.
In the context of increasing energy demand and thermal discomfort in the building sector, the assessment of thermal comfort and cooling energy consumption has become a key factor of research. This paper presents the passive cooling strategies performing in the tropical school building, including Phase Change Materials based on BioPCM, straw bale insulation covering fibers parallel and perpendicular to the heat flow, and building orientation. The study was conducted in two public elementary schools of Antsiranana, Madagascar, and focused on in-situ measurement using arduino recorder to highlight the outdoor and indoor environmental parameters and numerical approaches employing DesignBuilder as thermal simulation and building energy modeling software. As all the models are validated following ASHRAE Guideline 14 specification, the evaluation reveals that in the original configurations, both schools failed the comfort range during the rainy and the dry season spanning across the teaching period. Results demonstrate that straw bale insulation 40 cm with fibers oriented perpendicular to the heat flow exhibits the highest thermal performance and efficient payback period under 2 years for residential and business pricing scheme, followed by straw parallel configuration, Roof embedded PCM, and wall integrated PCM, and the north or south-facing. Since the electricity access remains very limited, this research focusing on the passive cooling principles provides relevant approaches in order to improve indoor thermal comfort, academic performance, and well-being of pupils.
Accurate online identification of building thermal states, model parameters, and unknown thermal disturbances is essential for reliable building thermal modeling and advanced energy management. However, conventional Kalman filter-based approaches are highly dependent on manually adjusted noise covariance matrices, making them sensitive to non-Gaussian and time-varying uncertainties commonly encountered in building operations. To address this challenge, this study proposes a neural network-assisted unscented Kalman filter under unknown input (UKF-UI-Net) for the joint estimation of thermal states, RC model parameters, and unknown heat inputs. The proposed framework combines a recurrent neural network with an unscented Kalman filter under unknown input (UKF-UI), enabling adaptive learning of the evolution patterns of process and measurement uncertainties from filtering residuals. RC parameters are augmented into the state vector for simultaneous online identification, while unknown heat inputs are reconstructed through weighted least-squares estimation. This hybrid physics-data framework eliminates the need for manual covariance tuning and enhances adaptability to complex noise environments. The framework is evaluated against fixed-covariance UKF-UI and Sage–Husa-UKF-UI under Gaussian, t-distributed, and nonstationary noise conditions using 35 Monte Carlo trials, and is further validated through experiments conducted in a south-facing test room. Compared with the baseline methods, the proposed method generally improves the overall accuracy and stability of parameter and unknown-input identification while maintaining reliable temperature estimation under complex noise conditions. These results demonstrate that UKF-UI-Net improves the robustness and accuracy of building thermal model identification under uncertainty conditions, providing a practical solution for online model calibration, digital twins, and model predictive control applications.
