With the challenges posed by high penetration of distributed photovoltaic integration into low-voltage distribution networks, such as substation reverse overload, voltage violations, and three-phase imbalances, a multi-stage evaluation method for assessing the maximum available capacity of distribution substations considering high proportions of distributed resources is proposed. Firstly, the inverse peak load time-series characteristics of distributed PV and load are analyzed, considering constraints related to reverse overload and voltage violations caused by PV over-generation, as well as the impact of three-phase imbalance on the substation transformer's load capacity. A robust optimization model for substation demand response to enhance distribution transformer load capacity is proposed. Secondly, considering phase selection configuration of distributed photovoltaics, coordinated charging of electric vehicles, as well as the demand response of flexible loads, a multi-stage collaborative evaluation model is proposed for determining the maximum available capacity of distribution substations under high proportions of distributed photovoltaic and electric vehicle integration. This model incorporates phase selection configuration optimization, virtual capacity expansion, and verification and correction stages. Finally, on the basis of nested column and constraint generation algorithm（N-C&CG）, the model is transformed into three sub-problems: photovoltaic phase selection configuration optimization, demand response of flexible resources, and capacity verification at the upstream substation, which are solved interactively. Through comparative analysis based on simulations of typical distribution substations, the proposed method demonstrates a 9.84% improvement in the available capacity of the distribution substations, providing technical support for optimized connection schemes and reliable operation in the expansion of distribution networks.