Electric power systems are undergoing significant transformations that have brought fundamental challenges to the system's situational awareness. This research aims to develop reliable strategies to manage such challenges for two typical anomaly-detection problems in modern power systems: cyber-attacks and faults. The growing complexity of power grids due to the penetration of intelligent monitoring devices and recent technological advances in energy management systems have rendered cyber-physical systems like electric grids increasingly vulnerable to cyber threats. These anomalies have been estimated to have a considerable economic impact during a severe cyber-attack. This work is essential for protecting modern and clean power systems because it implements effective methods for detecting and isolating faults in emerging technologies for renewable electric energy generation. Consequently, this research not only reduces the frequency of disruptions but also reduces the risk of damage, lowers maintenance costs, and extends the lifespan of equipment. This work is vital in preventing economic and power-related disruptions caused by cyber-attacks and faults.

Cyber-physical power systems (CPPS) rely on wide-area monitoring, protection, and control (WAMPAC) technologies, allowing operators to drive modern systems more efficiently and reliably. WAMPAC technologies use modern devices such as intelligent relays, digital recorders, and phasor measurement units (PMUs) to monitor the CPPS state. However, such devices produce continuous and unbounded data streams, posing significant data handling, storage, and processing challenges. Moreover, WAMPAC devices and their communication links are vulnerable to cybersecurity risks. To cope with these challenges, this study introduces two novel event detection methods for cyber and non-cyber contingencies based on the Hoeffding Adaptive Tree classifier. This robust machine learning classifier can handle high-velocity and high-volume data streams with limited memory and low computational burden. We consider several cyber and non-cyber events affecting the physics and monitoring infrastructure of CPPS, such as short-circuit faults, line maintenance, remote tripping command injection, relay setting change, and false data injection. The data are generated based on a modified IEEE 9-bus system. Simulation results show that our proposed approach outperforms the state-of-the-art method in the literature.

Fault detection is vital in ensuring the reliable and resilient operation of modern power systems. Its importance lies in swiftly identifying and isolating faults, preventing cascading failures, and enabling rapid power restoration. This work proposes a strategy based on observers and residuals for detecting internal faults in grid-forming inverters with power-sharing coordination for clean and small-scale AC microgrids. Grid-forming inverters will be the future drivers of modern and clean power systems to accelerate the adoption of renewable energy sources such as solar and wind. The dynamics of the inverters are captured through a nonlinear state space model. The design of our observers and residuals considers $H_{-}/H_{\infty}$ conditions to ensure robustness against disturbances and responsiveness to faults. The proposed design is less restrictive than existing observer-based fault detection schemes by leveraging the properties of quadratic inner-boundedness and one-sided Lipschitz conditions. Also, this work introduces a residual-based adaptive threshold for fault detection. An inequality for the upper bound on the $\ell_2$ norm of the residual is derived and used for designing the adaptive threshold. The upper bound is obtained via semidefinite programming with two linear matrix inequality constraints. The internal faults considered in this paper include actuator faults, busbar faults, and inverter bridge faults, which are modeled using vector-matrix representations that modify the state space model of the inverters. One significant advantage of the proposed methods is its cost-effectiveness, as it does not require additional sensors. Experiments are conducted on an islanded AC microgrid with three inductive lines, four inductive loads, and four grid-forming inverters to validate the merits of the proposed fault detection strategy. The results demonstrate that our designs outperform existing methods in the field.