A dual-layer planning method based on improved MOPSO for distribution networks considering source–load temporal uncertainty

Abstract

The integration of distributed generations (DGs) and time-varying loads introduces significant uncertainties in distribution network planning. Existing methods often rely on simplified scenarios (e.g., typical days), which fail to capture the full temporal volatility of wind, solar, and load profiles. To address this challenge, this paper proposes a dual-layer planning framework integrating scenario reduction and multi-objective optimization. First, the AP-DTW-K-medoids method is used to reduce 500 wind–solar–load scenarios to six representative clusters, enhancing the Davies–Bouldin index (DBI) by 25.5% compared to traditional clustering. Second, a dual-layer model decouples investment decisions (upper layer) and operational dynamics (lower layer), enabling cost-effective DG and energy storage (ES) allocation. Third, an improved multi-objective particle swarm optimization (MOPSO) algorithm with adaptive inertia weights accelerates the convergence by 25%. Case studies on the IEEE-33-bus system demonstrate a 1.41% reduction in total costs and 7.87% lower voltage deviations compared to conventional methods. The proposed framework provides a scalable solution for uncertainty-aware distribution network planning.