Doctoral Theses and Dissertations

Permanent URI for this collectionhttps://hdl.handle.net/10294/2900

Browse

Browsing Doctoral Theses and Dissertations by Author "Bais, Abdul"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    ItemOpen Access
    Machine learning-based models for failure prediction and propagation in smart grid systems
    (Faculty of Graduate Studies and Research, University of Regina, 2024-09) Salehpour, Ali; Al-Anbagi, Irfan; Bais, Abdul; Wang, Zhanle (Gerald); Yow, Kin-Choong; Louafi, Habib; Ameli, Amir
    The smart grid connects components of power systems and communication networks in an interdependent two-way system that supplies or receives electricity to or from prosumers and collects data that enables it to react to usage levels and interference from threats, such as cyber-attacks. Cascading failures resulting from cyberattacks are one of the main concerns in smart grid systems. The use of artificial intelligence (AI)-based algorithms has become more relevant in identifying and forecasting such cascading failures. However, existing models that study the propagation of cascading failures either omit the impact of the communication network or power characteristics on the propagation process. To address this gap, in this thesis, we propose a set of novel cyber-attack failure propagation models in smart grids. First, our realistic failure propagation (RFProp) model addresses the system’s heterogeneity by assigning different roles to its components. We define rules and interdependencies for failure propagation and propose a new model for studying cascading failures. In addition, the RFProp graph-based model identifies the most vulnerable nodes and implements power flow analysis to guarantee that all transmission lines work below capacity and remove lines exceeding capacity. Our results establish that by considering both power and communication characteristics and interdependencies, cascading failures are modeled more accurately. In the second step, we propose a novel earlystage failure prediction (ESFP) model based on supervised machine learning (ML) algorithms. We use the RFProp model to generate a dataset for training these algorithms and predicting the state of a system’s components after a failure propagates in that system. Using the ESFP model, we predict failures of all of a system’s elements in the early stages of failure propagation. We use the XGBoost algorithm and consider the features of both the power and communication networks that provide high accuracy in the prediction process for failures. We also identify the location of the initial failures, as this allows for further protection plans and decisions. In the third step, we use the real-time digital simulator (RTDS) to develop a real-time early-stage failure prediction (RESP) model that simulates the power system in real time and makes it more realistic. We evaluate the RESP model’s effectiveness using the IEEE 14-bus system, which results in the XGBoost algorithm achieving a high accuracy in predicting attacks and with a lower testing time. Finally, we introduce a real-time attack prediction (RTAP) model based on a real-time testbed designed to examine the impact of cyber-attacks on smart grid systems. We utilize real-time simulators, including RTDS and network simulator 3 (NS3) to emulate the behavior of power and communication networks. Using this model, we employ various ML algorithms to detect cyber-attacks. We evaluate the effectiveness of the proposed model using an IEEE 14-bus test case, demonstrating high accuracy and efficient testing time.