Distributed Multi-Agent System Approaches for Microgrid Power Management
By Kaveh Dehghanpour and Hashem Nehrir
A microgrid (MG) is considered the backbone of the smart grid. We are proposing an intelligent distributed multi-agent-system-based MG power management to find a unique and fair trade-off optimal solution for its operation point, using the game-theoretic concept of Nash bargaining solution.
Quality of a standard
Integrating distributed renewable and non-renewable power generation systems and energy storage units into electrical distribution energy networks has led to the introduction of the concept of microgrids. Microgrids are clusters of different types of small-scale micro-sources, including electric vehicles, on-site generators, battery storage systems, and smart loads. Multiple microgrids, along with wider use of market-based control mechanisms in electrical power distribution systems result in formation of hierarchical energy networks with multiple levels of autonomous decision-making agents and bi-directional flow of energy and information.
In this vision, power distribution networks are composed of multiple Microgrids that are able to manage their own private assets, participate in retail electricity markets, interact with each other to share power, and provide ancillary services at the behest of distribution system operator. Given the highly interactive nature of this interconnected network of decision-makers, solving control and power management tasks to satisfy system-wide operational constraints and achieving desirable global behavior, while respecting the data-privacy and autonomy of smaller entities in the system will become a challenging task.
This task becomes even more difficult when different parties and elements in the system have different sets of objectives and goals. In general, the control and power management tasks in microgrid-based energy systems are interactive multi-time-scale, multi-level, and multi-objective optimization problems. Another aspect of the challenging nature of these computational tasks is the huge amount of information and data that is required to reach a desirable global performance.
Hence, we believe conventional control and power management techniques are too burdensome and, in most cases, unable and unfit to address the challenges in control and power management of Microgrids and power systems. Instead we need to investigate and look into solution concepts that take advantage of distributed computational techniques. In other words, control and management of a distributed power system demands a distributed control approach. Apart from the fact that distributed computational techniques introduce modularity and plug-and-play capabilities into automation systems, they also provide efficient frameworks to rapidly and efficiently process and deal with the huge amount of data and information throughout the system without violating the data privacy of different parties.
The backbone of distributed and decentralized computational techniques is the Multi-Agent System (MAS) theory, which has been receiving intense scientific interest in recent years in different fields of research. MAS theory is an intersection of various research arenas, including artificial intelligence, network science, data science, complexity science, and game theory. A MAS is a community of intelligent “agents” that interact with each other over a communication medium (e.g., the Internet) to collectively solve a problem of system-wide proportion, subject to local/global constraints. From a power engineering point of view, the goal is to design an agent-based control system to solve system-wide multi-objective optimization tasks, efficiently, using local controller and decision-making agents that have access to local information only. Hence, the big challenge is to build a bridge to remove the gap between the locality of decision functions and the global requirements of system operation.
In general, MAS-based decision-making tracks the optimal trade-off solutions to the power management problem of Microgrids in the decision space, known as the Pareto-front. The Pareto-front represents a “multitude” of optimal non-dominated solutions of a multi-objective optimization problem. While this approach is quite effective in systems with a low number of objective functions and decision variables, as the number of objective functions grow along with the number of decision-making agents, tracking the whole Pareto-front in real-time becomes computationally expensive. Hence, we need to investigate other distributed computational techniques to tackle the problem.
An alternative solution to this challenge is the use of the concept of Nash Bargaining Solution (NBS) from the field of cooperative game theory. In this context, the multi-objective power management problem is modeled as a bargaining process among the different agents with different sets of objective functions. Then, NBS is applied as a solution concept for the bargaining game. NBS is guaranteed to lie on the Pareto-front of the bargaining problem.
Moreover, NBS introduces a “fair” balance into the problem, in the sense that it does not discriminate among the objectives of different agents that participate in the bargaining process. As a result, NBS is employed as the unique and fair solution to solve the multi-objective power management problem of a microgrid, without the need to track the whole Pareto-front in real-time, which alleviates the computational burden of the problem considerably.
Another advantage of employing NBS is that it can be formulated as a distributed optimization task, which is solved using an agent-based distributed bargaining framework. Within this framework, control agents only share their estimated best solutions with each other, without disclosing their private cost and constraint data to their peers. Hence, agents are able to collectively reach a system-wide desirable behavior (i.e., the NBS of the multi-objective power management problem) with only having access to local information and performing local computational functions. This approach ensures that as the number of agents grow with the size of the system, the computational tasks remain efficiently trackable and scalable.
Effective use of distributed MAS-based computational techniques (e.g., distributed optimization algorithms) in future power systems depends greatly on the breakthroughs in multiple areas of research in applied mathematics and computer science. This implies that the power engineers need to adopt a more multi-disciplinary attitude towards the problems they face in order to find relevant theoretically sound and practically viable solutions to engineering challenges.
Kaveh Dehghanpour, an IEEE Student Member, received his B.Sc. and M.S. degrees from the University of Tehran, Iran. Currently, he is working towards his Ph.D. degree in electrical engineering at Montana State University. His research interests include agent-based modeling and control of power systems, smart grids, and market-based control and power management.
Hashem Nehrir, an IEEE Life Fellow, has over 40 years of university teaching and research experience. He is a professor in the Department of Electrical and Computer Engineering at Montana State University. His active research includes modeling, control, and energy management of renewable energy–based distributed power generation, applications of intelligent control to power management of microgrid and smart grid, and load control (demand response). Dr. Nehrir is the current Vice Chair of the Renewable Technologies Subcommittee of the Energy Development and Power Generation Committee of the IEEE Power & Energy Society. He is the recipient of the 2016 IEEE-PES Ramakumar Family Renewable Energy Excellence Award and 2010 recipient of the Wiley Faculty Award for Meritorious Research at Montana State University.