With the increasing of distributed generators（DGs）， the effect of reactive power and voltage optimization can be problematic due to the uncertainty of DGs outputs. Using fuzzy characteristics of photovoltaic generation and wind power generation， the model of reactive power and voltage optimization in a district grid is proposed. The output of DGs is expressed by trapezoidal fuzzy number in this model. The fuzzy fitness function is formed based on the membership function of the object function and constraints. The most sensitive control devices are selected to carry on genetic operators by sensitivity calculation. Single point crossover operator is used in the algorithms in order to adapt to the mixed encoding of floating point numbers and integers. Actual district grid data is used to simulate and the results show that the model and algorithm are feasible.