Multi-agent based Models for the Distributed Coordination of Energy Micro-grids

Abstract

The recognized harmful environmental impact of emissions from fossil fuels (such as their contributions to global warming), as well as the diminishing long-term availability of fossil fuels, will necessitate a shift towards renewable energy sources to supply vital electrical energy needs in the future. Two abundant renewable energy sources, sun and wind, are increasingly cost-competitive and also offer the potential of decentralized, and hence offer a more robust sourcing. In this connection, the concept of decentralized electric energy micro-grids (MGs), in which communities share their locally-generated power, is receiving greater attention day by day. Intrinsic to this idea of energy micro-grids is that the energy is generated and consumed locally from renewable energy sources. Because of the variability of power generated by these renewable resources though, sometimes there are energy surpluses (supply is more than demand) and sometimes deficits (supply is less than demand). Researchers in multi-agent systems investigate (through modeling and simulation) how to overcome the effect of the sporadic nature of renewable sources. One of the ways to reduce the ill effects of intermittency in energy supply is through trading energy with a connected electricity grid or with other MGs. One of the standard ways of facilitating this trading is to employ a market mechanism. Most of the existing research on power trading among MGs use a multi-agent framework that focuses on economics (improve profitability or reduce the cost). To the best of our knowledge, no prior work has investigated power trading strategies for MGs from the lens of all three pillars of sustainability taken together (i.e. economic, environmental, and social aspects). Therefore, there is a need for the study of power trading mechanisms among MGs that focus on the three pillars of sustainability. To this end, the first part of the thesis aims to study the market-based trading strategies for communities having micro-grids that are in-line with economic and environmental aspects of sustainability (the first two pillars of sustainability). We assume that communities having micro-grids are connected to the main grid and have the choice to trade power either in the market or with the utility grid. The thesis also describes an agent-based architecture that can be used by MGs to trade power with other MGs or with the utility grid. We also developed power-trading mechanisms for a community having MGs that employ Markov-Decision-Process(MDP)- based reasoning and reinforcement learning (Q-learning) to improve their goals.

MGs make an electric system resilient and robust. When the main grid is down due to a natural disaster (such as an earthquake or a tsunami) or a technical fault, MGs have the ability to go into island mode and try to satisfy the demand with their local generation. Due to the intermittent nature of renewable sources, sometimes it is not possible for MGs to meet their local demand. To that end, the second part of the thesis discusses mechanisms to reduce the power deficit of communities having MGs by forming a group called a coalition (i.e., addressing social aspects of sustainability). The mechanisms discussed in the second part of the thesis address the social aspect of sustainability (the third pillar of sustainability). In these considerations, this thesis presents a notion of “discomfort” (hardship faced by the community due to power deficit) which has a nonlinear relationship with energy deficits. The thesis also describes mechanisms for dynamic coalition formation among MGs to reduce the overall discomfort level of the coalition. In this connection, the thesis investigates various aspects of coalition formation and demonstrates how: a) singleton communities form a coalition; b) a community from one coalition moves to another coalition; c) coalitions can merge into one coalition; d) a coalition can split into many coalitions; and e) a new coalition can be formed from existing coalitions.

The experimental work presented in the thesis employs agent-based simulations to investigate various mechanisms discussed in the thesis. The results reveal insights for practical use of these mechanisms in real-life energy trading domains. The mechanisms presented in this thesis could also be applied to other real-life scenarios involving communal sharing of resources such as water and food. The power trading strategies presented in this thesis can be used by the communities to improve its resilience (in the case of power failure), economic, and environmental outcomes.