Decision Support Independence in a Smart Grid Kendall E. Nygard, Steve Bou Ghosn, Md. Minhaz Chowdhury, Ryan McCulloch, Davin Loegering, Anand Pandey, Md. M. Khan Department of Computer Science North Dakota State University Fargo, ND, USA {Kendall.Nygard, Steve.Boughosn, Md.Chowdhury, Ryan.Mcculloch, Davin.Loegering, Anand.Pandey, Mahbuburrahman.Khan}@ndsu.edu Prakash Ranganathan Department of Electrical Engineering University of North Dakota Grand Forks, ND prakashranganathan@mail.und.nodak.edu Abstract—We describe a novel framework for designing and implementing agent based simulations of the smart electrical grid. The framework is based on two primary concepts. First, the eletrical grid system is separated into semi-autonomous microgrids, each with their own set of hierarchically organized agents. Second, models for automating decision-making in the grid during crisis situations are independently supported. Advantages of this framework are scalablity, modularity, coordinated local and global decision making, and the ability to easily implement and test a large variety of decision models. We believe that simulators based on this kind of framework will be valuable for evaluating the effectiveness and reliability of alternative methodologies for configuring automated self-healing in the grid with little human intervetion. The primary achievement of work is the software design for directly supporting decision model independence. Keywords-Multi-agent Multi-agent System; SmartGrid; Distributed Computing; Intelligent Systems; Self-healing. I. INTRODUCTION A primary objective of smart grid software architectures is to provide intelligence and communications technology to support powerful and efficient automation. A power system is exposed to faults created by natural calamity, terrorism and equipment or operator failures. Once a fault occurs in a power system, it is necessary to quickly isolate the malfunctioning components from the rest of the network to minimize outages. A power failure can range in magnitude and impact from a relatively modest curtailment to a catastrophic regional blackout. Because most power failures cannot be prevented [6], it is desirable for the Smart Grid to have self-healing capabilities that respond appropriately to disruptions when they occur, restore the power system to a healthy state, minimize consumer outages, and involve little or no manual intervention. Due to the large scale and complexity of the Smart Grid, anticipating all possible scenarios that lead to performance lapses is difficult [7]. There is a high degree of uncertainty in accurately estimating the impact of disruptions on the reliability, availability and efficiency of the power delivery system. These uncertainties result in hesitation on the part of decision makers in committing to smart systems for grid management. We report here on research that is focused on the use of simulation models to promote trust in Smart Grid solutions in safe and cost effective ways. Recently developed Smart Grid simulators and analysis tools include GridLAB-D and the Graphical Contingency Analysis (GCA). Both are projects developed at the U.S. Department of Energy's (DOE) Pacific Northwest National Laboratory (PNNL). GridLAB-D is a sophisticated simulator that provides detailed information of the power grid’s state, including power flow, end-use loads, and market functions and interactions. The GCA is a visual analytic software tool that aids power grid operators in making complex decisions. By using human friendly visualizations and classifications of critical areas and by allowing the operators to simulate possible actions and their consequences, the tool helps human operators to analyze large amounts of data and make decisions in a reasonable amount of time. Due to the large amounts of data representing the grid status at any given time, even when aided by simulation and analysis tools, there are still limitations on how quickly human operators can make efficient decisions in near real time. Because of the limitations of human operators in comparison with automated control, there is considerable research being done on how to fully automate control of the electrical grid by using software agents. A software agent is an encapsulated software system situated in an environment where it can conduct flexible and autonomous actions to meet its design objectives [2]. A Multi Agent System (MAS) is composed of multiple interacting intelligent agents that can sense, act, communicate and collaborate with each other. In our previous work [1], we presented guidelines for an agent- oriented smart grid simulation design. The agents in our MAS exhibit autonomy or partial autonomy, are decentralized, and have local views and knowledge. This design is related to other agent-based simulators that have been developed [4][5][16][17]. A fully automated grid relying on a multi-agent system will also presents some challenges and disadvantages along with its many advantages however. For example, developing agents able to function on par with human experts for the various scenarios that can happen in the smart grid, this will require a significant amount of research and experimentation. Relying on autonomous agents will also introduce a number of security issues. An agent could be hacked and controlled by an attacker who could manipulate the decisions and communications of the agent to perform malicious behavior. The trustworthiness of any particular agent or even the system as a whole could be called into question because of both the security risks and the general difficulty in replicating human expertise. Much must be done in order to overcome these inherent disadvantages of autonomous agent based systems. 69 Copyright (c) IARIA, 2012. ISBN: 978-1-61208-189-2 ENERGY 2012 : The Second International Conference on Smart Grids, Green Communications and IT Energy-aware Technologies