Cluster-driven multi-objective PV curtailment for fair and efficient voltage regulation

The rapid integration of distributed photovoltaic (PV) systems necessitates fair and efficient curtailment strategies to mitigate overvoltage. This study proposes a cluster-driven multi-objective framework that integrates spectral clustering and ε-constraint optimization. The network is first partitioned into coherent zones using a data-driven approach that integrates electrical topology, geographic proximity, and load-profile similarity. A multi-objective optimization then jointly minimizes total curtailed energy and maximizes equity, generating the full Pareto frontier of fairness–efficiency trade-offs. The framework is validated on a 2,532-bus low-voltage distribution network using realistic load and photovoltaic generation profiles and is compared against four benchmark strategies. Results demonstrate fairness improvements: the two-stage benchmark reduced maximum node-level disparity by 89.1% without increasing total curtailment, while the best-fairness Pareto solution achieved a 40% reduction at a 57.3% curtailment cost. This framework provides system operators with a transparent tool to select policies balancing technical and social objectives.