Keywords: human supervisory control, decentralized systems, human-computer interface, smart grid, decision support, power plantsIn the push to develop smart energy systems, there is increasing focus on how to design systems that measure and predict user behavior in order to effect optimal energy consumption. While such focus is clearly an important component in the success of these future systems, substantially less attention is paid to the human on the other side of the energy system loop -the supervisors of power generation processes, the proverbial men (or women) behind the curtain. Out of sight and sadly in terms of technological advancements, out of mind, today these operators perform high risk jobs in often datarich, but information-impoverished settings. For these operators, pervasive computing of the future will likely add to an already complex array of data streams, and introduce a new layer of supervisory complexity in response to the goal of dynamically adapting energy management.The Three Mile Island nuclear power plant accident in 1979 was caused primarily by operator misunderstanding of sensor data from an overwhelmingly complex control panel [1]. More recently in 2003 in the Northeast, operators were not able to both see and understand critical system states for nearby power grids, ultimately leading to the largest blackout in North American history which contributed to at least 11 deaths and cost an estimated $6 billion [2]. In these high profile cases, and in countless other more minor electric and nuclear power plant incidents, a significant problem was and continues to be the lack of explicit design to support rapid data aggregation and information visualization to support supervisors' time-pressured decision making.The development of smart energy systems that leverage pervasive computing could further add to the workload of these supervisory control operators who will have to predict possible power plant load and production changes due to environmental and plant events, as well as dynamic system adaptation in response to customer behaviors. Contrary to many assumptions, the insertion of more automation, both in terms of distributed sensors and algorithms to post-process data for operators, will not necessary reduce workload, nor necessarily improve system performance. These concepts are explored in more detail in the following sections.
Supervisory Control and Workload in Power GenerationCurrent power generation operations are highly automated. In normal, day-to-day operations, automation controls the adjustment of system parameters, while human operators generally take the role of system supervisors, monitoring system states and typically intervening in only non-monitoring operations, such as responding to an alarm, managing a plant start-up, or overseeing other off-nominal operations. However, in present-day power generation operations, while the system itself is highly automated, little automation is used to support and augment supervisor decision-making and performance, especially in time...