Buildings are interconnected cyber-physical systems (CPS) that couple with occupancy, heating, ventilation, and air-conditioning (HVAC), distributed energy resources (DERs), and power grids. This increasing complexity poses significant challenges for scalable modeling, control, and optimization. Existing approaches either rely on physics-based models with limited scalability, data-driven methods with weak physical consistency, or simplified representations that cannot fully capture system dynamics, restricting their applicability in real-world scenarios. To address these limitations, this study presents BESTOpt, a modular physics-informed machine learning (PIML)-based runtime environment for unified modeling, control, and optimization of interconnected occupancy–building–HVAC–DER–grid systems. The framework introduces a hierarchical structure (cluster–domain–system/building–component) and a standardized state–action–disturbance–observation data typology, enabling scalable coordination and integration of heterogeneous subsystems. Three case studies demonstrate the capabilities of the proposed framework. First, BESTOpt is compared with EnergyPlus, long short-term memory (LSTM), and a reduced-order 3R2C model to evaluate computational performance, prediction accuracy, physical consistency, and scalability from 1 to 200 buildings. Second, a 30-building residential cluster simulation evaluates 12 demand-response and retrofit strategies, showing that PV–battery integration can reduce peak-hour grid import by up to 97.8% and operating cost by up to 26.6%. Third, three deep reinforcement learning (DRL) algorithms are evaluated. Among them, Soft Actor-Critic (SAC) achieves the most balanced performance, reducing operating cost by 24.8%, grid import by 33.0%, peak demand by 41.9%, and PV curtailment by 87.3% compared to a rule-based baseline. Overall, BESTOpt provides a flexible and extensible platform for grid-interactive building analysis and advanced control evaluation.

Across the U.S., the increasing demand for electric vehicle (EV) charging infrastructure is placing new demands on the power grid, challenging its stability and efficiency. To address these challenges, this study proposes a Vehicle-to-Building-to-Grid (V2B2G) framework that incorporates urban-scale human mobility modeling to optimize EV charging and discharging. Using anonymized GPS traces from the study community, we extracted individual mobility patterns and applied kernel-density estimation to predict departure and arrival times for each user. The framework was tested in a mixed-use community in Phoenix, Arizona that includes both residential and commercial buildings. A comprehensive decentralized model predictive control (MPC) framework is implemented to minimize energy costs and enhance grid flexibility through demand-side management while maintaining occupant comfort. Four different control strategies were designed and evaluated, the strategy which balances both user and grid benefits demonstrated the best performance, achieving: (1) a flattened grid net load curve, with a 56% reduction in on-peak demand and a 56% decrease in peak load rebound; (2) a 37.96% reduction in grid net load compared to the baseline control; and (3) a 68.05% performance improvement when considering six flexibility factors: cost savings, total-energy reduction, on-peak demand reduction, off-peak demand reduction, load shifting from peak to non-peak hours, and peak-load-rebound reduction. These findings enhance our understanding of the impacts of urban mobility and EV charging optimization on grid management.

Buildings use a large amount of energy in the United States. It is important to optimally manage and coordinate the resources across building and power distribution networks to improve overall efficiency. Optimizing the power grid with discrete variables was very challenging for traditional computers and algorithms, as it is an NP-hard problem. In this study, we developed a new optimization solution based on quantum computing for BTG integration. We first used MPC for building loads connected with a commercial distribution grid for cost reduction. Then we used discretization and Benders Decomposition methods to reformulate the problem and decompose the continuous and discrete variables, respectively. We used D-Wave quantum computer to solve dual problems and used conventional algorithm for primal problems. We applied the proposed method to an IEEE 9-bus network with 3 commercial buildings and over 300 residential buildings to evaluate the feasibility and effectiveness. Compared with traditional optimization methods, we obtained similar solutions with some fluctuations and improved computational speed from hours to seconds. The time of quantum computing was greatly reduced to less than 1% of traditional optimization algorithm and software such as MATLAB. Quantum computing has proved the potential to solve large-scale discrete optimization problems for urban energy systems.

In the United States, the buildings sector consumes about 76% of electricity use and 40% of all primary energy use and associated greenhouse gas emissions. Occupant behavior has drawn increasing research interests due to its impacts on the building energy consumption. However, occupant behavior study at urban scale remains a challenge, and very limited studies have been conducted. As an effort to couple big data analysis with human mobility modeling, this study has explored urban scale human mobility utilizing three months Global Positioning System (GPS) data of 93,000 users at Phoenix Metropolitan Area. This research extracted stay points from raw data, and identified users’ home, work, and other locations by Density-Based Spatial Clustering algorithm. Then, daily mobility patterns were constructed using different types of locations. We propose a novel approach to predict urban scale daily human mobility patterns with 12-hour prediction horizon, using Long Short-Term Memory (LSTM) neural network model. Results shows the developed models achieved around 85% average accuracy and about 86% mean precision. The developed models can be further applied to analyze urban scale occupant behavior, building energy demand and flexibility, and contributed to urban planning.

Weather forecasting has been a critical component to predict and control building energy consumption for better building energy management. Without accessibility to other data sources, the onsite observed temperatures or the airport temperatures are used in forecast models. In this paper, we present a novel approach by utilizing the crowdsourcing weather data from neighboring personal weather stations (PWS) to improve the weather forecast accuracy around buildings using a general spatial-temporal modeling framework. The final forecast is based on the ensemble of local forecasts for the target location using neighboring PWSs. Our approach is distinguished from existing literature in various aspects. First, we leverage the crowdsourcing weather data from PWS in addition to public data sources. In this way, the data is at much finer time resolution (e.g., at 5-minute frequency) and spatial resolution (e.g., arbitrary location vs grid). Second, our proposed model incorporates spatial-temporal correlation information of weather variables between the target building and a set of neighboring PWSs so that underlying correlations can be effectively captured to improve forecasting performance. We demonstrate the performance of the proposed framework by comparing to the benchmark models on temperature forecasting for a building located at an arbitrary location at San Antonio, Texas, USA. In general, the proposed model framework equipped with machine learning technique such as Random Forest can improve forecasting by 50% compares with persistent model and has 90% chance to outperform airport forecast in short-term forecasting. In a real-time setting, the proposed model framework can provide more accurate temperature forecasting results compared with using airport temperature forecast for most forecast horizon. Moreover, we analyze the sensitivity of model parameters to gain insights on how crowdsourcing data from the neighboring personal weather stations impacts forecasting performance. Finally, we implement our model in other cities such as Syracuse and Chicago to test the model’s performance in different landforms and climate types.

Since the coronavirus disease 2019, the extended time indoors makes people more concerned about indoor air quality, while the increased ventilation in seeks of reducing infection probability has increased the energy usage from heating, ventilation, and air-conditioning systems. In this study, to represent the dynamics of indoor temperature and air quality, a coupled grey-box model is developed. The model is identified and validated using a data-driven approach and real-time measured data of a campus office. To manage building energy usage and indoor air quality, a model predictive control strategy is proposed and developed. The simulation study demonstrated 18.92% energy saving while maintaining good indoor air quality at the testing site. Two nationwide simulation studies assessed the overall energy saving potential and the impact on the infection probability of the proposed strategy in different climate zones. The results showed 20%–40% energy saving in general while maintaining a predetermined indoor air quality setpoint. Although the infection risk is increased due to the reduced ventilation rate, it is still less than the suggested threshold (2%) in general.