Abstract: Aiming at the problems of excessive voltage generation in the distribution network under the high penetration of distributed power generation (DG), the conflict of multi-objective optimization and the easy to fall into local optimization, an adaptive weight multi-objective gray wolf optimizer-taboo hybrid algorithm (AW-MOGWO-TS) is proposed, and a multi-objective reactive voltage optimization model of the station distribution network is constructed. Firstly, based on the principle of voltage safety first and economic operation second, a four-objective function is established, including the smallest network loss, the smallest voltage deviation, the lowest operating cost of discrete equipment and the lowest DG abandonment rate. Secondly, the AW-MOGWO-TS algorithm is set up: the Kent chaos-Sobol sequence is used to initialize the population to ensure diversity, the adaptive weight mechanism is introduced to dynamically balance the multi-target priority, and the taboo search (TS) is triggered by the variance of population fitness to jump out of the local optimum, and the efficient decision-making of the Pareto solution is realized through fuzzy interception theory. Finally, the IEEE33 node station model is verified by examples. The results show that compared with the MOPSO and MOGWO algorithms, the proposed algorithm reduces the network loss by 62.8%, the number of voltage exceeding the limit to 0, the light curtailment rate to 2.76%, and the solution time by 15.6%~22.4% in the DG penetration scenario of 100% permeability.