Abstract: The large-scale cluster development of offshore wind power has led to an increase in power loss in wind farms. In order to solve the problems of large loss and high time-consuming reactive power dispatching in offshore wind farms,a data-driven strategy of optimization reactive power distribution in offshore wind power collector network is proposed. Firstly,considering the reactive power ability of doubly-fed wind turbines,an optimization model of reactive power allocation with the goal of comprehensive minimization of loss reduction rate and voltage deviation is constructed,and the optimal reactive power distribution strategy of DFIG is designed.Secondly,based on the electrical topology and optimal allocation strategy of offshore wind power collector network,the graph convolutional network is used to train the reactive power distribution relationship network offline. Finally,the particle swarm optimization algorithm based on prior knowledge is applied to refine the model’s output for practical use. Simulations demonstrate the effectiveness of the proposed method in reducing losses and computation time while meeting reactive power and voltage control needs for offshore wind energy.