Marine wind energy resources are an important part of the new power system with new energy as the main body. However, offshore wind power shows a trend of large-scale and centralized development in coastal areas, and has the characteristics of anti-peak regulation and volatility, which is easy to produce a large number of wind curtailment. In order to maximize the dispatching capacity of offshore wind power systems, a "source-network-load-storage" optimization scheduling model considering energy storage capacity configuration is proposed. At the same time, Kullback–Leibler divergence is used to characterize the uncertainty of wind power output, and a two-stage optimal scheduling model of "source-net-load-storage" is constructed. The results show that the probability of fluctuation of offshore wind farm is less than 5% under 5 min time scale is 98.24%. On a 15-min time scale, the probability drops to 90.51%. In the optimal scheduling of different scenarios, the model reduces the load cutting operation, protects the users with interruptible load, and thus reduces the operating cost and fossil fuel consumption of thermal power units. It is proved that the optimization cost of this scheduling model is lower, the robustness is stronger, the operating cost and fossil fuel consumption are reduced effectively, and the robustness and economy of the system scheduling operation are taken into account.