The modern power grid is undergoing unprecedented levels of transformation. As such, two major trends can be observed. First is the confluence of information and communication technologies (ICTs) with the physical power grid components (e.g. advanced metering infrastructure (AMI), phasor measurement units (PMUs), data concentrators, etc.), which is rendering real-time monitoring and control operations a practical reality. Second is the rising prevalence of diverse power components such as renewable generation sources, storage systems, flexible consumer appliances and smart protection devices that have the potential to increase sustainability and resilience in the grid. These also introduce interesting open challenges, specifically in the areas of power system security and efficiency. The RISE lab strives to overcome these issues by leveraging on optimization, statistical analysis and machine learning constructs which are fundamental in the broad field of artificial intelligence (AI) to address open research problems in three specific areas: grid security assessment, adversarial attacks on the smart grid and preventive/reactive mitigation. These research efforts strive to empower individual cyber-enabled power entities with the ability to effectively infer, predict and respond to adversarial/natural perturbations in the system.

Generic placeholder image

Grid Security Assessment

The modern electric grid is increasingly prone to cascading failures and adversarial attacks mainly due to the uncertainties introduced by highly-fluctuating grid components and security flaws present in ICT- enabled power devices. Thus, traditional contingency analysis that assesses the impact on system stability when one or more critical grid components fail (e.g. synchronous generation/transmission lines) no longer suffices. Vulnerability analysis and anomaly detection techniques that utilize advanced prediction and data analytic techniques are now necessary to account for increasing uncertainties and the timely detection of cyber attacks in the system. As such, our research is focussed on three main areas: 1) Online contingency analysis, 2) Vulnerability assessment; and 3) Anomaly detection.

Adversarial Attacks on the Smart Grid

Due to the cyber-enabled nature of the modern grid, stealthy adversarial attacks can now exploit security flaws in ICTs to gain remote access into power system devices. Thereafter, the adversaries either passively glean information for espionage purposes (e.g. Havex malware) or actively alter data or control signals generated in the grid to affect the physical states of the system (e.g. Stuxnet malware). A thorough analysis of various modes and forms of these attacks will allow system operators to construct proactive measures to minimize or completely mitigate the impact of these in the system. As such, our research work has explored adversarial attacks under two different contexts: 1) Attacks that are contained within the ICT domain, and 2) Attacks that directly actuate cyber-enabled power nodes.
Generic placeholder image
Generic placeholder image

Preventive/Reactive Mitigation

Both preventive and reactive mechanisms are vital for mitigating the undesirable effects of system uncertainties and adversarial attacks. Preventive mechanisms infer adverse grid trends and compute actuations for mitigating potential failures that may occur in the near future. Reactive techniques, on the other hand, are activated in the immediate aftermath of failures so that the faults can be contained in an efficient manner while maximizing service up-time. In the area of preventive mitigation, we have investigated how flexible power components, reconfigurable physical networks and data protection can be utilized to mitigate grid congestions, increase stability margins and prevent information leakage in the power grid. Reactive mitigation necessitates immediate actions to contain faults within the system. Hence, pro- tection devices (e.g. relays, circuit breakers) which typically act to contain faults in the grid are not delay tolerant and thus computations are based on local state measurements. As global states are not accounted for in decision-making, these actions will be myopic and possibly sub-optimal. Thus, effective mitigation techniques that enable self-healing and resilient grid operations are crucial for maintaining the integrity of the power grid.