Coordinating distributed energy sources in modernized power markets is another field where MAS-based cooperation can be employed. PowerMatcher, a concept developed by ECN in the Netherlands [51, 68, 69], is an example. The objective is
to coordinate agents so as to balance supply and demand in an economically efficient way. To solve this problem, agents have the capability to competitively trade energy on a common market. These market-based negotiations provide a decision-making framework based on microeconomics.
The negotiation process relies on dynamic pricing schemes, in which prices can vary throughout the day depending on the balance between supply and demand: the higher the demand, the higher the prices, and vice versa. Each agent buys or sells energy depending on its type (load, source, etc.), and commits to this bought or sold amount. In this competitive general equilibrium market, all agents have access to the same information on price, obtained by searching for the equilibrium between supply and demand.
The PowerMatcher architecture is based on a logical tree structure, as shown in Fig. 15.26. Each group of agents is coordinated by a concentrator agent, which is in turn coordinated by another concentrator at a higher level. Four types of agents are used:
• Device agents control the distributed sources, loads, and storage units of the system, and are able to converse with concentrators. They have their own goals and properties, and issue bids based on what they are willing to pay (resp. be paid) for a given amount of energy to consume (resp. produce).
• Concentrator agents (Fig. 15.25) locally concentrate and aggregate the bids they receive, and back propagate the price chosen by the auctioneer to lower level agents (device agents and other concentrators).
• The auctioneer agent, at the top of the tree, centralizes the bids of the agents connected to it and finds the equilibrium price based on bids it has received to clear the market. The obtained price is then sent back to the agents if it has significantly changed from its earlier value.
• The objective agent is connected to the auctioneer agent and defines the objective of the system.
Fig. 15.26 The MAS proposed in PowerMatcher uses a hierarchical architecture, based on four types of agents: an auctioneer (A), an objective agent (O), concentrator agents (C) and device agents (D). The above diagram shows an example with a limited number of agents
Each agent, whether it is a device or a concentrator, only knows its direct neighbors.
A negotiation round is initiated with a given frequency (of a few minutes) by the auctioneer agent. The agent starts by asking other agents to submit their bids, and computes the equilibrium price based on information it has gathered. Other events can trigger a round, such as a sudden change in demand. DF and AMS agents are used, so that each agent can find other agents to negotiate with.
The main interests of this concept are its ability to operate in environments with a large share of renewable energy sources, where the energy sector has been fully liberalized, and consumers can also be producers. The selected bottom-up approach, combined with the decentralized decision-making process, enable a very flexible architecture, where a change in the structure, by adding or removing functionalities or agents, does not affect the overall operation of the system. This structure is plug-and-play and scalable: contrary to other architectures, this solution is not based on a central optimization algorithm that would hinder the ability to scale the system. By aggregating local bids, the need for intensive data communication is avoided. Similar concepts can also be applied to trade other commodities, such as heat in a microgrid where cogeneration is available. The system is being tested in a demonstration project called PowerMatching city, located in the Netherlands.
Comparable and interesting MAS – and market-based approaches were proposed by other researchers, such as DEZENT , MASCEM , work by Dimeas and Hatziargyriou [72, 73], Pipattanasomporn et al. , and Funabashi et al. .