A multi-timescale reactive power optimization strategy for active distribution networks based on spectral clustering

Abstract

With a significant increase in the integration of distributed photovoltaic （PV） systems into distribution networks， traditional optimization approaches struggle to effectively mitigate voltage fluctuations， and the reactive power control capability of distributed PV inverters remains underutilized. In response， this paper proposes a reactive power optimization strategy for active distribution networks based on spectral clustering across multiple timescales. The approach consists of two stages： day-ahead optimization and real-time optimization. Firstly， the temporal coupling of discrete equipment actions is decoupled. Using distribution network power loss， average voltage deviation， and voltage fluctuation severity as objective functions， a day-ahead reactive power optimization model is formulated based on a social network search algorithm. This model determines the static optimal operating sequences for discrete equipment. Secondly， employing spectral clustering for coupling， the dynamic optimal operating sequences for discrete equipment are determined. The strategy incorporates an improved local control strategy for distributed PV inverters and establishes a real-time optimization model， thereby mitigating voltage fluctuations caused by discrepancies in day-ahead forecast data. Finally， the proposed strategy is validated through simulations on an improved IEEE 33-node system. Simulation results demonstrate that the proposed strategy effectively reduces computational complexity， enhances solution efficiency， and verifies the effectiveness and superiority of the approach.

Keywords

• active distribution network
• multi-timescale
• dynamic reactive power optimization
• decoupling method using spectral clustering
• social network search algorithm