Based on the control characters of distributed generators (DG) integrated into a local power system, DG is divided into voltage-controlled type and reactive-compensated type. DG participates in reactive power optimization control. A dispatch model for fuzzy dynamic reactive power optimization control of a local power system is presented with a compositive objective of reducing the network loss and restricting node voltage variations of a local system. In this model, a fuzzy evaluation function is established to transfer the objective function into the satisfaction degree on optimization results. A method based on self-adaptive weight is proposed to normalize the compositive objective. Improved genetic algorithm (IGA) is used to solve dynamic reactive power optimization dispatch in a local power system with distributed generators. The numerical example analysis shows the rationality of the model as well as its reference value.