Structured review examines the integration of machine learning and blockchain across peer-to-peer energy trading, highlighting operational, consensus, and privacy trade-offs.
Key Points
To evaluate how machine learning and blockchain operate across peer-to-peer energy trading layers, distinguishing market clearing and ledger mechanisms from physical power delivery.
Applied a structured review methodology to 52 peer-reviewed journal articles and 10 supplemental foundational sources using defined search families and qualitative synthesis.
Categorized literature across architecture, market operations, consensus protocols (including PBFT, IBFT, and PoA), and privacy mechanisms (including federated learning and zero-knowledge proofs).
Conducted two illustrative, deterministic MATLAB simulations modeling a five-prosumer forecasting scenario and a 10-peer trading workflow.
In the five-prosumer forecasting example, regression reduced mean absolute error from 0.4240 to 0.2219 kWh and decreased hourly grid-import mismatch from 30.3529 to 7.9029 kWh.
In the 10-peer workflow simulation, five trades settled 7.7587 kWh, accounting for 59.35% of the horizon-level surplus–deficit denominator.
The review identified key operational divisions between digital consensus and physical electricity flow, noting that existing simulations do not confirm grid feeder feasibility, consensus latency, or full deployment readiness.