ניהול הקונפליקטים בין חקלאות, חקלאות סולרית ותשתיות הולכה לשם השגת יעדי אנרגיות מתחדשות

The first level focuses on maintaining the grid’s electrical quality and stability while considering a large number of electricity consumers and producers. The second level deals with the management of the electrical market, in which the different producers and consumers trade electricity, and with the regulation of prices within the market.

While these overgrowing challenges are currently mainly handled using centralized methods, these will soon become inadequate due to their inability to support the dynamic and constantly evolving nature of the electrical grid. This raises the need to develop decentralized and adaptable methods for grid management. With a hierarchical decentralized approach decision making can be propagated to local entities while maintaining electrical quality and reducing the overload of communication.

To support the decentralizing of the grid and to mitigate the challenges that are associated with this process, we suggest a multi-agent reinforcement learning (MARL) framework comprised of multiple autonomous and semi-autonomous artificial intelligence (AI) agents. Such agents are modeled with an internal utility measure, and can designed to collaborate and compete on a general set of resources in a given environment. In the context of a smart grid, agents are used to model the grid at the component level with each agent designed to pursue its own objectives while complying with the constraints of the grid. The MARL approach is used to simulate the effect of different outcomes on the system, to minimize information flow needed for optimal control, and for optimizing the interaction between the agents and between other components of the network.