Abstract: Accurate prediction of offshore wind power output is the basis for guaranteeing the dispatch operation of offshore wind power grid-connected system. Aiming at the highly complex marine environment of offshore wind power and the strong coupling of stochastic space-time, this paper proposes a new type of interval prediction method for offshore wind power output based on Vague soft set. First, the concept of Vague soft set is introduced, and the interleaved offshore wind power interval division method integrating the true affiliation and pseudo-affiliation functions of Vague set is proposed to realize the Vague soft intervalization of wind power data. Then, an offshore wind power combination prediction model based on Vague-convolutional neural network (CNN)-long short-term memory neural network (LSTM) is established. The double affiliation interval probability vector is transformed into the interval prediction result under the complex uncertain information of offshore wind power by the Vague-like soft interval transformation method. Next, Vague soft interval prediction evaluation indexes such as prediction interval coverage accuracy, prediction interval width and prediction synthesis level are established from the perspective of prediction accuracy, clarity and compatibility. Finally, the actual data of an offshore wind turbine in the eastern part of China are used as an example for validation, and the results show that the proposed prediction model can take into account the coverage accuracy and clarity of the prediction intervals, which can provide support for the operational requirements of offshore wind power under different working conditions.