Financial and carbon-aware cashback mechanisms applied in blockchain-enabled VPP ensuring user data privacy through federated learning

The implementation of Virtual Power Plants (VPPs) with sophisticated energy management can significantly reduce consumer energy costs and promote the use of distributed renewable energy, thereby lessening the environmental impact of the energy sector. This study presents a novel methodology that integrates AI-driven meteorological and demand forecasts, using Federated Learning for privacy preservation. The NSGA-III multi-objective optimization algorithm is applied to the Battery Energy Storage System (BESS) to minimize both the VPP's total energy costs and its carbon emissions. A smart contract, written in Solidity, implements a VEST settlement vault in which a two-of-three oracle committee authorizes Merkle-root commitments, while high-resolution measurements and metric calculations remain off-chain. A deterministic weighted-proportional rule is employed to allocate the value generated by the Demand Side Management (DSM) operation among all participants. This cashback mechanism utilizes a multi-objective approach based on Load Factor (LF), Peak Shaving Efficiency (PSE), and Carbon Avoidance Intensity (CAI) to distribute financial and carbon-aware rewards. The results of the case study demonstrate substantial improvements in the VPP, including an 11.07% reduction in total energy costs and 3.51% reduction in CO2e emissions when applying the BESS approach. CU savings reflect system-shared profits based on each unit's operational performance relative to the baseline. This profit-sharing incentivizes improved demand profiles, boosting DSM efficiency and financial returns. The blockchain proof of concept was evaluated over 30 rounds on a four-validator Hyperledger Besu QBFT network, while off-chain allocation was benchmarked for 18–1000 participants. An analysis of deployment and operating costs across Ethereum-based networks complements the experimental evaluation.