Agent Technologies for Control Applications in the Power Grid Angel A. Aquino-Lugo, Thomas Overbye University of Illinois Urbana-Champaign aaquino3@illinois.edu, overbye@illinois.edu Abstract Remotely controlled network devices will transform the way power systems are operated and studied. One possible application for this is the use of agent based technologies to implement decentralized control algorithms. The agents would perform local control actions instead of centralized control actions. Two cases are presented. The second case would be examined in detail. In the first case, the power losses were minimized using a decentralized algorithm and the results were similar to those found using a centralized algorithm. In the second case, transmission line overloads are relieved by controlling the load in the system. For this case, a detailed algorithm to control loads was presented to show the integration that would be required between the transmission and distribution network. This work showed that to implement decentralized control a reliable communication network within the power system will be necessary. 1. Introduction Today, much conversation is being made about how the electric power grid would look in the future. The common consensus is that it would incorporate new technologies that would let us control the grid in a “smart” way. The problem is that many have different definitions about what “smart” grid means. A “smart” grid can be defined as the utilization of new digital and intelligent devices to replace the old analog devices in the power network. In this paper, “smart” grid relates to using those new intelligent devices to allow for remote control providing a new opportunity for decentralized control. The challenges, whatever the definition, are enormous. The stimulus law of 2009 provides billions of dollars for smart grid funded projects and studies. Certainly the transformation of the grid would change the way it is operated and analyzed. In this paper, some new ideas on how to control the power grid in a decentralized but intelligent scheme are studied. Some examples are presented, as well as the challenges they bring to the electric power gird. 2. Agent Technologies Previous Work Typically, the power grid system is operated and coordinated in a centralized way. Every time the power network fails the central control center determines which system elements and control actions should be implemented to either save the system from collapse or to reconfigure the system after an outage [1]. In recent years, with the introduction of remotely controlled power network devices, new possibilities for control strategies are starting to emerge. A number rely on a more local control but with decentralized algorithms. Thus the concepts of agents were studied to get more local and faster responses for applications that formerly were coordinated by a centralized control. One of the first concepts was the self-healing of power distribution networks in combat ships. During battles, the ships can suffer severe damage to the electrical system, and in a combat situation it is important to maintain the availability of energy to the loads to keep the ship operational [2-3]. People quickly realized that this concept could be applied to power system distribution networks. In [4], the authors present a multi-agent system (MAS) approach for a decentralized solution for the power system reconfiguration problem using Matlab Simulink S-functions as agents. Following the same approach as in [1], a restoration algorithm applying an expert system type of solution was presented using Matlab Simulink and the Stateflow toolbox. In [3] an intelligent power routers (IPR) scheme was proposed where control can be detached from the central control sites, and delegated to IPRs that would be distributed over the entire electric network to initialize and coordinate control actions. More recently, research has focused on studying centralized problems, other than power system restoration, in a decentralized scheme. For example, the optimal power flow (OPF) can be solved by using decentralized algorithms to parallelize the solution to get faster results. In [5] a parallelized OPF suitable for a coarse-grained distributed implementation is 1 Proceedings of the 43rd Hawaii International Conference on System Sciences - 2010 978-0-7695-3869-3/10 $26.00 © 2010 IEEE