Extreme weather events, technical and human failure, or sabotage (e.g., with weapons of mass destruction), can disrupt infrastructure and impact the electrical grid, causing cascading failures or blackouts that can affect millions of people. In power grid simulations, the role of human failure has been mostly ignored as a potential factor. In a recently published paper, researchers from Fraunhofer CSE and MIT focus on what human intervention can accomplish to influence the recovery and limit the extension of blackouts.
What policies should be enacted to ensure the stability of the grid in a crisis, such as a storm, or a malicious attack? What kind of human intervention can help stabilize the grid in such an event? In a recent study, the U.S. Department of Defense (Defense Thread Reduction Agency), the University of New Mexico (UNM), and scientists at Fraunhofer CSE researched what factors affect the stability of electric grids. UNM simulated the physical electric grid, while Fraunhofer CSE provided data and an assessment of how grid operator stress coupled with incomplete and inaccurate information can affect grid stability. Ultimately, the results of the human factors assessment were integrated into the grid model developed by UNM.
A team led by Dr. Joana Abreu, Behavioral Scientist at Fraunhofer CSE, applied a methodology originally used to calculate the probability of human error in nuclear power plants. The Standardized Plant Analysis Risk Model Human reliability analysis (SPAR-H) methodology quantifies error probability by considering factors that influence the perception, processing and response to events. Factors considered in the context of a grid operations center are the amount of time that the operator has available to respond, fatigue, stress, stressors, level of complexity of the situation, level of experience and training, procedures, alarms, and work processes. The methodology incorporates variables that describe the cognitive status of the operator and variables describing the situation.
To assess the circumstances of actual incidents in electric grid operation, the team interviewed operators and asked them to illustrate a timeline, identify key decisions made, and the consequences of those decisions. To document the effects of weather zones in outage reports, the team interviewed operators from two distinct climate zones: New England and New Mexico. Each operator described one or more incidents that took place during a shift, such as power loss on an island that is difficult to serve. In a next step, multipliers for each performance shaping condition, such as time available, level of complexity, and level of experience were assigned to the actions taken and/or decisions made by the operator. These multipliers reflect the error probability in performing an action or making a decision.
The researchers classified each narrated scenario according to a phase in a blackout – precursor, escalation, and cascading. Then they calculated the frequency of occurrence that is associated with factors like fatigue (fitness for duty), complexity, experience, along with their level of severity (e.g., unfit, moderate complexity, high experience). Finally, they created a probability distribution for the likelihood that different levels of severity would occur as a function of each factor. The resulting probability distribution was integrated into UNM’s dynamic model to simulate the progression of failures.
Previous research indicates that calculated human error probabilities significantly correlated 72% of the time with actual values. Yet, as the researchers point out, more data are needed to more accurately represent error probability. Research also points out that methods to assess human reliability rely on the subjective interpretation of experts about situations researchers? did not directly observe, and depend on operators’ ability to narrate a past event objectively. In conclusion, the SPAR-H methodology could be further improved by considering the dependency between performance shaping factors, such as time available and stress.
The project is funded by a grant from the Department of Defense’s Defense Threat Reduction Agency (DTRA).