AI-Driven P2P Blockchain Energy Trading System for Optimized Microgrid Balance

The intermittent character of distributed renewable energy resources (DERs), especially the sun, is a major obstacle to stabilizing localized smart microgrids. In this paper, a hybrid AI-based P2P energy trading platform is introduced that employs machine learning (ML) forecasting to actively balance the supply-demand of the microgrid. The system incorporates an RNN-based predictor executed on a device at the edge (e.g., NVIDIA Jetson Xavier NX) to predict near-term energy surpluses (prosumers) and deficits (consumers). Such predictions are the input for smart contracts implemented on a consortium blockchain, which trigger P2P trading instructions automatically before physical imbalances arise. The system's unique feature is its predictive auto-execution of trades, which significantly reduces transaction latency and contingency reserves. Based on past weather and consumption trends, simulation results demonstrate that the AIbased method improves grid stability and maximizes local use of renewable energy by reducing the Grid Imbalance Factor (GIF) by over 35% and transaction decision time to less than 100 ms.