SUMMARYThis paper proposes an improved reinitialized social structures particle swarm optimization (IRS-PSO) for solving optimal multiple distributed generations (DG) placement in a microgrid (MG) system. The movement of each particle in IRS-PSO is pulled by an inertia term, a cognitive term (personal best), and three social learning terms including global best, local best, and near neighbor best. The objective is to minimize the real power loss within real and reactive power generation limits and voltage limits. Five types in a MG system are considered including MG with DG supplying real power only, MG with DG supplying reactive power only, MG with DG supplying real power and consuming reactive power, MG with DG supplying real power and reactive power, and MG with four different types of DG regulating the bus voltage. For a given number of DG units in each type, IRS-PSO can find better sizes and locations of multiple DGs than repetitive load flow, basic particle swarm optimization (BPSO), adaptive weight particle swarm optimization (APSO), and global best, local and near neighbor best particle swarm optimization (GLN-PSO) on the 69-bus radial MG distribution system.