In allusion to the scene that multiple photovoltaic distributed generation（PV-DG） will be connected in the distribution power system

particle swarm optimization（PSO） based on K-means cluster for day-ahead allocation plan of multiple PV-DG is proposed. In this algorithm

the hourly network loss of nodes connected with PV-DG is alternately combined with the designed formulas to initially allocate the penetration level of multiple PV-DG connected with nodes

this penetration is used to initialize the particle in the PSO algorithm. Moreover

the ARMA is compared with various predicting parameters the traditional ARMA. The simulation results indicate that the proposed ARMA can improve the predicting accuracy. The comparison is carried on among the PSO based K-means cluster

classical PSO

and fuzzy PSO

the results show that PSO-based K-means cluster can reduce more network loss.