To flexibly address the issues of unreasonable integration of distributed photovoltaic （PV） systems and the temporal fluctuations of PV and loads, while improving voltage quality and operational efficiency in distribution networks, a bi-level planning method is proposed under typical source-load temporal scenarios considering demand response. Firstly, the SOM-KFCM clustering algorithm is introduced. This method combines the topological preservation characteristics of self-organizing maps （SOM） and the nonlinear clustering capability of kernel fuzzy C-means （KFCM） to enhance clustering accuracy, and typical scenarios are derived using the probability-weighted Pearson correlation coefficient. Then, by incorporating demand response and reactive power optimization, the active and reactive power of the distribution network are jointly regulated, forming a bi-level planning model. The upper model optimizes the PV planning scheme with the goal of minimizing total annual costs and maximizing voltage stability indices, while the lower model aims to minimize daily operational costs for each scenario. To effectively solve this model, an improved whale optimization algorithm with multiple strategies is adopted to enhance global search capability and escape from local optima, and it is combined with second-order cone programming for solving, achieving a globally optimal and locally coordinated planning solution. Finally, simulation analysis through case studies demonstrates that the proposed model and planning scheme enhance both the economic benefits and stability of the distribution network.

photovoltaic power  /  demand response  /  reactive power  /  distribution networks  /  bi-level planning