Artificial intelligence (AI) technology has rapidly advanced and transformed the nature of scientific inquiry. The recent release of the large language model Chat Generative Pre-Trained Transformer (ChatGPT) has attracted significant attention from the public and various industries. This study applied ChatGPT to autonomous building system operations, specifically coupling it with an EnergyPlus reference office building simulation model. The operational objective was to minimize the energy use of the building systems, including four air-handling units, two chillers, a cooling tower, and two pumps, while ensuring that indoor CO₂ concentrations remain below 1000 ppm. The performance of ChatGPT in an autonomous operation was compared with control results based on a deep Q-network (DQN), which is a reinforcement learning method. The ChatGPT and DQN lowered the total energy use by 16.8% and 24.1%, respectively, compared with the baseline operation, while maintaining an indoor CO₂ concentration below 1000 ppm. Notably, compared with the DQN, ChatGPT-based control does not require a learning process to develop intelligence for building control. In real-world applications, the high generalization capabilities of the ChatGPT-based control, resulting from its extensive training on vast and diverse data, could potentially make it more effective.

End-use consumption provides more detailed information than total consumption and reveals the mechanism of energy flow through a given building. Specifically, for weather-sensitive energy end-uses, it enables the prioritization and selection of heating and cooling areas requiring investigation and actions. One of the major barriers to acquiring such heating and cooling information for small- and medium-sized buildings or low-income households is the high cost related to submetering and maintenance. The end-use data, especially for heating and cooling end-uses, of such-sized buildings are a national blind spot. In this study, to alleviate this measurement cost problem, two weather-sensitive energy disaggregation methods were examined: the simplified weather-related energy disaggregation (SED) and change-point regression (CPR) methods. The first is a nonparametric approach based on heuristics, whereas the second is a parametric approach. A comparative analysis (one-way ANOVA, correlation analysis, and individual comparison) was performed to explore the disaggregation results regarding heating and cooling energy perspectives using a measurement dataset (MEA) from eleven office buildings. The ANOVA results revealed that there was no significant difference between the three groups (SED, CPR, and MEA); rather strong correlation was observed (r > 0.95). Furthermore, an analysis of the building-level comparison showed that the more distinct the seasonal usage in the monthly consumption pattern, the lower the estimation error. Thus, the two approaches appropriately estimated the amount of heating and cooling used compared with the measurement dataset and demonstrated the possibility of mutual complements.

This study introduces a hybrid model that combines physics and machine learning (ML) models to describe the behaviour of variable refrigerant flow (VRF) systems. The standalone ML model was developed with identical data and conditions for comparison between the hybrid and ML models. A Bayesian neural network (BNN) was used for both the models, and the predictive abilities and uncertainties were investigated. For the experimental dataset, the predictive performances of both models were similar. For example, the predictive performance of the hybrid and ML models showed mean absolute error of 0.73 and 0.78 kW, respectively. However, the epistemic uncertainty of the hybrid model quantified using the BNN was 36.4% lower than that of the ML model. A parametric study showed that the hybrid model combined with the physics model can achieve better generalization performance than the ML model, yielding results that are more reliable and physically explainable.

Indoor CO₂ concentration is considered a metric of indoor air quality that affects the health of occupants. In this study, a hybrid model was developed for forecasting the varying indoor CO₂ concentration levels in a residential apartment unit in the presence of occupants by controlling the ventilation rates of a heat recovery ventilator. In this model, the mass balance equation for a single zone as a white-box model was combined with a Bayesian neural network (BNN) as a black box model. During the learning process of the hybrid model, the BNN estimated an aggregated unknown ventilation rate and transferred the estimation to the mass-balance equation. A parametric study was conducted by changing the prediction horizons of the hybrid model from 5 to 15 min, and the forecasting performance of the hybrid model was compared with the stand-alone mass balance equation. The hybrid model showed better forecasting performance than that of the mass balance equation on the experimental dataset for a living room and bedroom. The average MBE and CVRMSE of the hybrid model for the prediction horizon of 15 min were 0.65% and 5.23%, respectively, whereas those of the mass balance equation were 0.99% and 9.30%, respectively.

The building sector is responsible for more than one-third of total global energy consumption; hence, increasingly efficient energy use in this sector will contribute to achieving carbon neutrality. Most existing building-energy-benchmarking methods evaluate building energy performance based on total energy use intensity; however, energy usage in buildings varies with the seasons, and as such, this approach renders the evaluation of cooling and heating energy difficult. In this study, an information gain-based temporal segmentation (IGTS) method was used to identify the seasonal transition times based on patterns of hourly weather and corresponding building energy use. Twelve commercial buildings were considered for the study and four seasons were identified using IGTS; base-load, cooling energy, and heating energy data were gathered. For the 12 buildings, the estimated and measured heating and cooling energy during the summer and winter periods showed a linear relationship (R² = 0.976), and the average of those differences was 9.07 kWh/m². In addition, differences in the benchmarking results based on these energies were marginal. The results indicated that the IGTS approach can be effectively used for determining the actual heating and cooling energy consumption in buildings, as well as for energy benchmarking. This can, in turn, improve building energy use, with positive implications for achieving carbon neutrality.