Monitoring the operation status of wind turbines is important for improving wind turbine operation efficiency and lowering wind farm operating costs. Thus， a new method of wind turbine condition monitoring is proposed， which is based on supervisory control and data acquisition （SCADA） and a firefly improved sparrow search algorithm optimized deep belief network （FISSA-DBN）. The SCADA data is first preprocessed and analyzed， and the input and output parameters are chosen using an expert system and Pearson's correlation analysis. Second， using the preprocessed data set， a new wind turbine operation monitoring model based on FISSA-DBN is established. The reconstructed value error between the predicted value and the actual output value of the model and the exponential weighted moving average （EWMA） are used to determine if there is an abnormality. Finally， actual data from a wind farm in Central China are used to validate the results. Verification results show that the early warning time of the proposed method can be more than four days earlier than the actual recording time. When compared to other methods， the proposed method provides earlier warning sand has a lower prediction error.