Similarly, the London blackout in August of 2003 has been attributed to an incorrect relay setting.

Improved procedures and system designs can help avoid errors. With good surveillance and training, many errors can be detected and corrected before they lead to problems. But errors do happen. If they were to occur in the face of an unfolding terrorist attack, they could considerably complicate an already serious situation. This prospect further strengthens the importance of contingency planning, training, and simulated exercises.

The explosion in available information has made attention time an extremely valuable commodity for all workers. Most automated networks require some human intervention not only for routine control, but also especially when exhibiting anomalous behavior that may suggest actual or incipient failure. Progress continues to be needed in the design of interfaces that help users retain good situational awareness while allowing them to focus on the most important factors in a complex and rapidly evolving dynamic situation. Improved displays of the state of the electric power grid are being installed in control centers (Christie and Mahadev, 1994, Overbye and Weber, 2001), but there is room for a great deal of imaginative innovation in this area.

Humans have cognitive limitations that can cause them to make serious mistakes when they are interrupted. While actual or imminent local failures can be detected automatically, operators can easily be distracted by other tasks” including responding to multiple systems warnings. In the worst case, a detected failure can set off a multitude of almost simultaneous alarms as it begins to cascade through the system. Under this scenario, system operators may be unable to accurately determine the real source of the problem, which in turn could lead to the whole network shutting down automatically.

In recent years, systems have been designed that allow users to delegate tasks to intelligent software assistants (“softbots”) that operate in the background, handling routine tasks and informing the operators in accordance with some protocol that establishes the level of their delegated authority to act independently. In this arrangement, the operator becomes a supervisor, who must either cede almost all authority to subordinates or be subject to interruption by them. At present, there is very limited understanding of how to design user interfaces to accommodate interruption.

Two products developed by EPRI for substation operations and maintenance (O&M) could lead to tools for analysis of human performance. The first is the Maintenance Management Workstation (MMW), a data integration, analysis, and display tool that is used to guide decisions on equipment maintenance and replacement. Since it can connect to any database and data source, it could be adapted to analyze operational decision making. The other tool is the Planning and Resource Optimizer (PRO), which is a planning tool to assist in task scheduling and resource allocation (including labor). It allows for consistent and efficient work planning, optimized schedule and resource allocation, and facilitation of unexpected changes, and it can be used for backlog management. It also integrates with the MMW.

The degree of field information available to operators is also an area of concern. In many cases, there is little feedback from the maintenance crews to operations engineering and design engineering personnel with regard to the actual work done during a maintenance task and the as-found condition of the asset being maintained. Insufficient coordination and communication among these various personnel can result in a lack of information that can lead to less than optimal configurational control of the system and to incorrect decision making in responding to a system alarm or failure. As one example of attempts to address this issue, ConEd is evaluating a hand-held reporting system that requires specific feedback that can be uploaded to the work order management system. Such a system could enable an operator to quickly assess field work performed in evaluating the implications of an alarm. Despite the progress made to date in addressing the shortcomings of automation and human performance, the following challenges remain:

•   Application of statistical methods to extract information and trending on human performance. These analytical techniques can be combined with enhanced visualization and techniques to improve situational awareness of the state of the system (perhaps using multimedia user interfaces and virtual reality) to assist the human operator.

•   Network visualization and situation awareness. The exact nature of the information needed by operators, managers, users, and the general public may vary, but all need to understand what is going on in the infrastructure network. Adequate visualization of the state of the system is required for situation awareness. The proliferating new technology for multimedia user interfaces, and for virtual reality in particular, needs to be evaluated and fitted into this context of human performance. Such technology also should be incorporated into existing training simulators having adequate modeling and database capabilities at a regional transmission operator or an independent system operator level so that any entity in the region could use the same setup for its training facilities.

•   Interface design. Little use has been made of esthetic considerations in the design of interfaces, yet it is clear that humans are attracted to, and seek to use more frequently, that which is esthetically pleasing. Such considerations may also be important if means are provided (e.g., on cable or broadcast television) to pass disaster mitigation information to the general public.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement