Decentralized Voltage Prediction in Multi-Area Distribution Systems: A Privacy-Preserving Collaborative Framework

With the high penetration of renewable energy and the deregulation of electricity markets, future active distribution networks are anticipated to comprise areas with diverse ownership. Frequent power exchange between these areas leads to strong interarea influences, making it essential to account for voltage coupling in decentralized voltage prediction (DVP). To ensure the accuracy of DVP, collaborative modeling using power information from neighboring areas is imperative. However, potential privacy concerns pose significant challenges to data sharing and interarea collaboration. To address these challenges, this paper proposes a privacy-preserving collaborative prediction framework. First, a privacy-preserving multi-objective XGBoost algorithm is developed to mitigate curious behavior through homomorphic encryption and minimal information sharing. Then, an equitable incentive mechanism is introduced to allocate rewards based on the relative marginal contributions, which discourages potential adversarial behavior. The proposed framework further modularizes subproblems, enabling parallelization to improve computational efficiency. Finally, case studies demonstrate that the proposed framework can achieve prediction accuracy comparable to centralized methods while ensuring privacy preservation and collaboration reliability, highlighting its potential for broad adoption in multi-area systems.