With the increasing penetration of distributed generation and diversified reliability requirements of end-users, existing methods for evaluating power supply capability based on the N-1 criterion are no longer applicable. For fully exploiting the impact of differentiated reliability requirements and distributed generation integration on power supply capability, this paper proposes an innovative methodology to evaluate the power supply capability of distribution systems. First, the power supply capability is set as objective function, and the various reliability requirements of users are modelled as the main constraints. Then, an integrated reliability evaluation method is proposed based on the pseudo-sequential Monte Carlo simulation. A tailor-made genetic algorithm is employed to solve the black-box optimization problem. In the case study, a real distribution system is used to validate the proposed methodology. The simulation illustrates that the proposed method can improve the economic efficiency of distribution systems while enhancing the power supply capacity. It is found that the relaxation of reliability constraints can significantly improve power supply capability. In addition, the effect of increasing distributed generation capacity on reliability improvement is more obvious when system load is higher.