The invention provides a distribution network multi-objective reactive power optimization method based on two-stage particle swarm optimization, which comprises the steps of building a distribution network reactive power optimization model, and acquiring preference information of a decision maker; performing a first stage of particle swarm reactive power optimization on the built distribution network reactive power optimization model, acquiring a rough Pareto front, solving by combining the preference information of the decision maker to acquire a preference Pareto front, and constructing a preference search area; and performing a second stage of particle swarm reactive power optimization on the distribution network reactive power optimization model in the preference search area, and acquiring a preference solution set of the decision maker. According to the invention, the preference search area of a multi-objective optimization algorithm is calculated by using the preference information of the decision maker, so that limited optimization resources are centralized, the quality, the solving speed and the pertinence of an optimization solution are improved, more meaningful optimization scheme selection is provided for the decision maker, and the method is provided for multi-objective optimization engineering applications.