Transformer-based AI predicts grid constraints for P2P energy trading
New framework lets microgrids self-assess trade feasibility using transformer regression...
A new paper from researchers Devangi, Ankit Singhal, and Yashasvi Bansal introduces a learning-augmented framework for grid-aware peer-to-peer (P2P) energy trading. As distribution networks increasingly integrate distributed energy resources (DERs), P2P trading among aggregated microgrids (MGs) has emerged as a viable local energy exchange mechanism. However, existing frameworks lack predictive capabilities that allow prosumers to anticipate network feasibility or Distribution System Operator (DSO) responses during trade formulation. The proposed solution trains a supervised transformer-based regression model that enables MGs to locally predict the DSO's response to proposed trades without sharing their trade details, thereby preserving information privacy and reducing computational burden on the DSO.
The framework is validated on a modified IEEE 33-bus distribution power system with interconnected microgrids. Case studies demonstrate improvements in market efficiency, trade acceptance rates, and reduced transaction overhead. By allowing prosumers to self-assess and refine trading decisions based on predicted DSO responses, this work addresses a key gap in ensuring P2P transactions remain physically feasible and grid-compliant. The approach represents a step toward more autonomous, privacy-preserving, and scalable decentralized energy markets.
- Supervised transformer regression predicts DSO response without prosumers sharing trade data, preserving privacy
- Validated on modified IEEE 33-bus distribution system with interconnected microgrids
- Improves market efficiency and trade acceptance while reducing DSO computational burden
Why It Matters
Enables scalable, privacy-preserving P2P energy markets by letting prosumers self-assess grid feasibility in real time.