Accurate short-term building energy consumption forecasting is crucial for the intelligent operation of Building Energy Management Systems (BEMS), but it is challenging due to strong nonlinearity, weather uncertainty, and occupancy variability, as well as the sensitivity of Long Short-Term Memory (LSTM) models to hyperparameter settings. This study proposes an ICPO-LSTM framework, in which an Improved Crested Porcupine Optimizer (ICPO) is developed to efficiently optimize multiple LSTM hyperparameters. ICPO incorporates Cat chaotic mapping initialization, nonlinear dynamic regulation, and adaptive T-distribution perturbation to enhance convergence stability and global search ability. The proposed method is evaluated using two heterogeneous datasets, including an Internet of Things (IoT)-based high-rise office building in Xi’an, China, and a public library building from the Building Data Genome 2 (BDG2) dataset. Experimental results demonstrate that ICPO-LSTM consistently outperforms other optimization-based LSTM models and advanced deep learning methods, while achieving a favourable trade-off between prediction accuracy and computational efficiency.

The integration of ceiling fans with radiant systems remains underexplored despite their potential to address cooling capacity limitations. This study adopts a two-step approach to quantify the impact of elevated air movement on thermally activated building systems (TABS). First, we used OpenFOAM to calculate convective heat transfer coefficients under varying airflows, air-to-surface temperature differences, and zone sizes. These coefficients also apply to ceiling fans in buildings without radiant systems. Second, we implemented these coefficients in EnergyPlus to evaluate key radiant design parameters. Scenario 1 results show median steady-state cooling heat transfer rates increase of up to 47% relative to the no-fan cases when operative temperature is held constant. Scenario 2 demonstrates a median cooling effect of up to 4.8 K under fixed capacity, reflecting both lower zone temperatures and direct air movement on occupants. Overall, TABS-fans systems offer a scalable strategy to enhance comfort, increase capacity, and reduce energy demand. Highlights We use CFD to develop simplified surface convective heat transfer models that account for air movement with ceiling fans on room surfaces. We use simplified surface convective models in EnergyPlus to evaluate radiant system performance with ceiling fans. Ceiling fans increase the zone’s steady-state cooling heat transfer rate by a median of 47% at the highest fan speeds. Ceiling fans have a median cooling effect of up to 4.8 K at the highest fan speeds. A radiant and ceiling fan couple system has the potential to decrease HVAC energy use while maintaining occupant thermal comfort.

Model Predictive Control (MPC) in building energy management requires transient thermal models balancing thermodynamic accuracy with computational efficiency. Standard spatial discretization triggers state-space inflation, paralyzing real-time solvers, while Transfer Matrix Methods (TMM) suffer from high-frequency numerical overflow and assume material homogeneity. This paper introduces a novel frequency-domain framework based on the continuous spatial Riccati equation. A recursive admittance mapping strictly bounds exponential growth, preventing numerical instability. The primary contribution of this work is extending this linear propagator via regular perturbation theory to analytically resolve spatial property gradients ( λ ( x ) ) and non-linear T 4 radiative boundaries as equivalent harmonic source terms. This meshless approach eliminates spatial truncation errors. It corrects peak heating load deviations of 21.9% in wetted media and mitigates artificial nocturnal cooling fluxes of 12.0 W/m 2 . Preserving O ( N ) spatial complexity, the framework avoids state-space inflation, ensuring high-speed execution for multi-week MPC optimization.

Traditional building energy modelling workflows remain labor-intensive and error-prone, requiring specialized expertise that limits broader adoption. This paper introduces a novel Model Context Protocol (MCP)-enabled framework that connects AI assistants to EnergyPlus through MCP, a standardized interface for tool invocation and context management. Two complementary integration paradigms are presented and compared: conversational integration, where users interact through natural language while an AI assistant orchestrates MCP tools on demand, and agentic workflow integration, where specialized agents coordinate autonomously to complete multi-step tasks. Using an experimental testbed for residential buildings, the end-to-end workflows are demonstrated. The conversational approach reduced typical inspection and modification tasks from 1-2 h to under 15 min, while maintaining full transparency through visible tool invocations. The agentic approach automated parametric analysis. These demonstrations establish MCP as a foundational layer for AI-assisted building energy modelling, enabling natural language interactions with simulation tools while preserving professional oversight and decision-making authority. Highlights An MCP-enabled framework provides standardized integration between AI assistants and EnergyPlus for building energy modelling workflows. Two complementary paradigms are demonstrated: conversational integration for exploratory tasks and agentic workflow integration for systematic automation. End-to-end demonstrations on a residential energy model show an 80% to 90% time reduction for model inspection, modification, and analysis tasks. The framework augments rather than replaces professional judgment, with AI handling tool orchestration while practitioners retain decision-making authority. The approach establishes a foundational infrastructure for advanced capabilities including parametric simulations, BIM-to-BEM translation, and automated model calibration.

Centralized heating systems are a promising heating solution for multi-apartment buildings. Dynamic simulations are crucial for optimizing building and hydronic system performance, but physical models are computationally expensive. Simplified models using lumped components can reduce simulation effort and time. This study investigates the appropriate level of model simplification in terms of computational performance and accuracy for a central 2-pipe hydronic system with decentralized heat transfer stations (HTSs), providing space heating and DHW (excluding the generation system). Various lumping strategies for building thermal zoning, piping, radiators and HTSs are applied to determine the recommended modelling approach depending on specific goals. For assessing overall energy trends, a lumped model is sufficient; however, more detailed approaches are necessary to accurately calculate the return temperature. Lumped models are sensitive to lumping parameters, posing risks when using such simplified approaches. Results are supported by a sensitivity analysis across different buildings, hydronic systems and user settings.

How to predict heating load more accurately plays an important role in energy conservation and promoting scientific and green development of society. This study analyzes heating load data from two consecutive seasons (October 2021–April 2023) in Harbin, integrated with meteorological data, and proposes a novel hybrid model combining a Long Short-Term Memory Network (LSTM) with a Particle Swarm Optimization (PSO) algorithm and an Attention Mechanism. This model leverages LSTM's strength in extracting temporal features, the Attention Mechanism's capability for optimal feature weight allocation, and PSO for hyperparameter optimization. Comparative analysis against K-Nearest Neighbors (KNN), Support Vector Regression (SVR), Recurrent Neural Network (RNN), Attention-LSTM, PSO-LSTM, GA-LSTM, and DE-LSTM demonstrates that the proposed PSO-Attention-LSTM achieves superior performance, yielding the lowest Mean Squared Error (MSE) and Mean Absolute Error (MAE) while maintaining a comparable Coefficient of Determination (R²). Furthermore, an integrated model integrating the optimal KNN and PSO-Attention-LSTM was developed. This integrated model further enhances accuracy, reducing MAE and MSE by approximately 14.40% and 18.93% compared to KNN, and by 7.82% and 11.71% compared to the standalone PSO-Attention-LSTM, respectively.

Rising temperatures and urban heat islands (UHI) affect human health and well-being. Green infrastructures (GI) can mitigate UHI and reduce indoor building temperatures through shading (lower solar gains) and evapotranspiration. However, quantifying these effects is challenging, as most building performance simulation (BPS) tools accurately model conventional cooling but lack greenery effects. Therefore, we developed facade greening parameters in IDA ICE using ENVI-met outdoor microclimate simulations that include ecosystem services from greenery. A case study in Vienna, Austria, assessed renovated and unrenovated buildings under future climate scenarios (RCP4.5 and RCP8.5). Cooling effects in IDA ICE were negligible in renovated buildings and up to 1.1°C in unrenovated ones. The study demonstrates that modelling facade greening in IDA ICE is feasible, though fully capturing GI’s benefits requires further BPS tool development.

This study investigates how uncertainty in occupancy sensor performance propagates through residential smart-thermostat control and shapes energy savings, thermal comfort, and peak demand. A four-step framework was used. First, stochastic models of occupancy schedules and sensor accuracy were calibrated using long-term field data from a single-family home in Texas. Second, these models were embedded in an EnergyPlus simulation platform via PyEMS to dynamically couple occupant behaviour, sensing errors, and building physics. Third, over ten thousand Monte Carlo simulations were conducted on an HPC cluster across seven IECC climate zones, varying setpoints, occupancy patterns, and sensor sensitivity. Fourth, standardized metrics were used to quantify energy, comfort, and demand impacts. Results show that sensor uncertainty substantially widens the performance range of smart thermostats. High sensor accuracy does not guarantee optimal outcomes, as false-positive and false-negative errors propagate nonlinearly through HVAC operation. Moderate sensitivity levels (0.4–0.7) best balance energy efficiency and comfort.

Conventional data-driven modeling often fails under unseen conditions. This study introduces bidirectional knowledge sharing via federated learning (FL) across multiple systems without data exchange. The method was tested on two dedicated outdoor air system (DOAS) units with imbalanced datasets. FL improved sensible heat prediction accuracy, reducing CVRMSE from 39.1 to 17.7% for DOAS 1 and 13.9 to 9.8% for DOAS 2 compared to a conventional ANN. Under unseen control conditions, FL further reduced CVRMSE from 33.1 to 15.1% for DOAS 1 and 21.3 to 14.6% for DOAS 2, demonstrating enhanced extrapolation. Intervention-based analyses also showed that FL promotes more physically consistent relationships between control variables and system responses. Although limited to a short-term case study of two similar systems, the results highlight FL's potential to improve the robustness of data-driven HVAC models.

Buildings consume a large share of global energy, making accurate building energy models (BEMs) critical for improving efficiency. Existing calibration methods often face high computational demands, overlook parameter identifiability, and lack independent validation, limiting their practical reliability. This study proposes a comprehensive framework for the efficient calibration of multi-zone BEMs, integrating graph-based modelling (GBM), global sensitivity analysis (GSA), parameter identifiability assessment, and optimization. GBM captures complex thermal interactions across zones with interpretability and accuracy. Sobol'-based GSA identifies the most influential parameters, improving calibration efficiency, while parameter identifiability assessment ensures uniqueness of parameter estimates before optimization. Calibration is conducted using the golden search optimization algorithm, balancing exploration and exploitation. Validation through both statistical metrics and independent validation data not used during model calibration demonstrates the robustness and generalizability of the proposed approach. Results demonstrate that explicitly addressing sensitivity, identifiability yields more accurate and reliable BEMs, supporting energy-efficient operation and control strategies. Highlights Developing the graph-based model of multi-zone building dynamics with a reduced-complexity structure. Optimized calibration integrating GSA, identifiability, and golden search method. Validation robustness and generalizability under static metrics and independent data not used during model calibration. Providing a foundation method for energy-efficient operation and control strategies.