Cyber-Physical Systems (CPSs) are combinations of physical processes and network computation. Modern CPSs such as smart buildings, power plants, transportation networks, and power-grids have shown tremendous potential for increased efficiency, robustness, and resilience. However, such modern CPSs encounter a large variety of physical faults and cyber anomalies, and in many cases are vulnerable to catastrophic fault propagation scenarios due to strong connectivity among their sub-systems. To address these issues, this study proposes a graphical modeling framework to monitor and predict the performance of CPSs in a scalable and robust way.

This thesis investigates on two critical CPS applications to evaluate the effectiveness of this proposed framework, namely (i) health monitoring of highway traffic sensors and (ii) building energy consumption prediction. In highway traffic sensor networks, accurate traffic sensor data is essential for traffic operation management systems and acquisition of real-time traffic surveillance data depends heavily on the reliability of the physical systems. Therefore, detecting the health status of the sensors in a traffic sensor network is critical for the departments of transportation as well as other public and private entities, especially in the circumstances where real-time decision making is required. With the purpose of efficiently determining the traffic network status and identifying failed sensor(s), this study proposes a cost-effective spatiotemporal graphical modeling approach called spatiotemporal pattern network (STPN). Traffic speed and volume measurement sensors are used in this work to formulate and analyze the proposed sensor health monitoring system. The historical time-series data from the networked traffic sensors on the Interstate 35 (I-35) within the state of Iowa is used for validation. Based on the validation results, this study demonstrates that the proposed graphical modeling approach can: (i) extract spatiotemporal dependencies among the different sensors which lead to an efficient graphical representation of the sensor network in the information space, and (ii) distinguish and quantify a sensor issue by leveraging the extracted spatiotemporal relationship of the candidate sensor(s) to the other sensors in the network.

In the building energy consumption prediction case, we consider the fact that energy performance of buildings is primarily affected by the heat exchange with the building outer skin and the surrounding environment. In addition, it is a common practice in building energy simulation (BES) to predict energy usage with a variable degree of accuracy. Therefore, to account for accurate building energy consumption, especially in urban environments with a lot of anthropogenic heat sources, it is necessary to consider the microclimate conditions around the building. These conditions are influenced by the immediate environment, such as surrounding buildings, hard surfaces, and trees. Moreover, deployment of sensors to monitor the microclimate information of a building can be quite challenging and therefore, not scalable. Instead of applying local weather data directly on building energy simulation (BES) tools, this work proposes a spatiotemporal pattern network (STPN) based machine learning framework to predict the microclimate information based on the local weather station, which leads to better energy consumption prediction in buildings.