Bi-level coordinated expansion planning of high-PV-penetration active distribution networks incorporating multiple flexible resources considering WGAN-GP-based uncertainty generation

Integrating high-penetration random photovoltaics (PVs) into the active distribution network (ADN) often causes overvoltage issues, which can even trigger ADN failures. To accommodate high-penetration PVs integration while maintaining economic efficiency and flexibility of ADN, a bi-level collaborative expansion planning model considering uncertainty generation and reduction is proposed in this paper. Based on described uncertainty scenarios of PV and load obtained by proposed Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) and K-means method, the upper level firstly configures soft open points to adjust the ADN topology, then optimizes the siting and sizing of PVs, electric vehicle charging stations, energy storage systems, and active management resources (on-load tap changers, and static var compensators) with minimum investment cost. Lower level uses multi-objective mixed-integer second-order cone programming (MISOCP) model with network loss, PV curtailment and voltage deviation of different attributes to optimize ADN scheduling. To solve this planning model, a hybrid algorithm that integrates an adaptive particle swarm optimization algorithm with SOCP is proposed, employing the technique for order preference by similarity to an ideal solution (TOPSIS) to address the multi-objective issue. Finally, the presented model was tested in the actual rural 41-bus and urban 40-bus ADN of Gansu, China, demonstrating that total costs decreased by 14.4 % and 12.5 % respectively compared to plans without active management resources.