To address the complexity of individually controlling high-penetration distributed photovoltaics (PV) and the impact of correlation between random variables on distributed PV planning, a dual-layer planning method for high-penetration distributed PV considering operational partitioning and correlation is proposed. Firstly, load nodes in the distribution network are clustered based on modularity and active power balance indicators. Secondly, to analyze the correlation between PV and loads, a 3PEM stochastic power flow model based on the Nataf inverse transform is established. Then, a dual-layer planning method for high-penetration distributed PV is developed. The outer layer minimizes the annual comprehensive cost by optimizing the allocation capacities of distributed PV and energy storage for each cluster, utilizing an improved genetic algorithm. The inner layer, based on chance constraints, minimizes the deviation rate of comprehensive node voltage expectation and the operational risk value as multi-objective functions, optimizing the output of distributed PV and the placement of energy storage at each node, employing non-dominated sorting genetic algorithm 2(NSGA2) and entropy-based technique for order preference by similarity to ideal solution(TOPSIS) to find the optimal configuration. Finally, the feasibility of the proposed model is demonstrated through a case study of IEEE 33-node system.