Research on Reactive Power Source Planning for Active Distribution Networks with Large-scale Photovoltaic Integration

To address the bottleneck of non-convex nonlinear reactive power voltage operation domain caused by strong bilateral source-load uncertainties in active distribution networks under large-scale photovoltaic integration, and the limitations of traditional planning methods in providing accurate support, this study proposes a reactive power source planning approach integrating probabilistic modeling, situational awareness, and source-grid coordination. A non-parametric kernel density estimation is employed to fit the dynamic stochastic characteristics of photovoltaic output, while a semi-invariant method-based probabilistic power flow model characterizes the multimodal impact of PV fluctuations on node voltage probability distributions. A deep confidence network-gated recurrent neural network fusion model enables online identification and prospective assessment of reactive power deficit spatial-temporal distribution. A two-layer collaborative planning model for reactive power sources considering source-grid coordination is established with dual objectives of optimizing lifecycle cost and minimizing voltage security risks, employing a multi-objective chaotic whale optimization algorithm to solve Pareto front optimization. Case studies demonstrate that voltage compliance rate increased from 96.8% to 99.5%, annual grid losses decreased by 20.8%, and voltage risk indicators declined by 65.7%, significantly enhancing system safety resilience and planning economic adaptability.