Optimising P2P Energy Trading using Internet of Things and Agentic AI Cluster Zooming

Artificial intelligence (AI) has become the game changer in smart grids-an enabler of network autonomy, self-healing, and reconfiguration. This study integrates AI and Internet of Things (IoT) to organise peer-to-peer (P2P) energy prosumers into virtual clusters without altering the physical topology of the power network. The aim is to enable an autonomous, scalable and dynamic virtual microgrids (VMG) by leveraging federated learning, agentic AI, AI agents, IoT, and cluster zooming to optimise P2P energy trading costs for prosumers and operational expenditure (OPEX) for network operators, depending on the number of prosumers available. The study employs a central controller AI to coordinate multiple local AI agents. Each AI agent resides in the network server and monitors energy trading traffic for each long-range wide-area network (LoRaWAN) gateway to optimise trading and OPEX costs via cluster zooming achieved by the spreading factor (SF) via adaptive data rate (ADR) mechanism of LoRaWAN. The agentic AI module in the cloud autonomously selects and adapts the network coverage based on SF, via the AI energy trading agent configured in the LoRaWAN access network server, to zoom the clusters (i.e., VMGs) in grid-connected and island modes. The study formulates an energy trading model connecting the physical (electrical) and virtual (telecom) distances and OPEX in the VMG. With agentic AI-assisted cluster zooming, over 70% of the energy is traded at lower SF. At the same time, the energy costs decrease by 40% in proportion to the network size and the number of prosumers. For the network operator, OPEX reduces by 21% and 38% in base-station power consumption. Ultimately, grid-connected prosumers pay higher charges than their off-grid counterparts. The agentic AI model in this study exemplifies a use case of the 3GPP model of the future 6G network.