In contemporary electrical networks, reliability stands as a paramount attribute. Since the introduction of distributed generation (DG) and microgrid (MG) concepts has considerably changed the structure of distribution networks, it is necessary to consider reliability aspects while going toward modern smart grids. This study introduces a novel methodology focusing on the partitioning of radial distribution systems into distinct MGs, strategically aimed at minimizing energy not supplied (ENS) to fortify overall system reliability. Two types of faults are considered including DG and busbar failures, and the impact of each kind of fault is investigated. Besides, a new Markov model is proposed to evaluate the reliability of DGs and MGs. Given the stochastic nature of the majority of DGs, the incorporation of energy storage systems (ESSs) can significantly enhance system performance. Therefore, this paper dedicates attention to exploring optimal ESS placement. To address the complex challenge of optimizing system partitioning, a modified version of the group search optimization (GSO) algorithm, named dominated GSO (DGSO) is proposed. The PG&E 69-bus distribution system is studied with two different structures to accurately investigate the effectiveness of the proposed partitioning approach on the system reliability. Moreover, the optimal number of MGs to be constructed within the intended system is determined.