Abstract: Icing conditions generally lead to a decrease in blade aerodynamic performance and an increase in turbine load, which poses a threat to the operational efficiency and operational safety of wind farms. In this regard, in order to achieve effective prediction of wind turbine blade icing, a wind turbine blade icing prediction algorithm based on time series data analysis is proposed. On the basis of clarifying the data source, the existing data are filled and corrected based on the linear interpolation method and the 3σ principle, and the corrected and processed multi-dimensional time series data of wind turbine blades are extracted and combined with the long and short-term memory network to achieve the prediction of wind turbine blade icing. The experimental results show that the design method is capable of capturing the actual icing situation of fan blades, and can accurately and effectively predict the icing situation of fan blades, which provides a new research programme reference for the optimization of the operation and maintenance of fan safety.