The invention relates to the field of wind
turbine generator fault early warning, and provides a wind
turbine generator fault early warning method based on an SVR
algorithm and kurtosis. The method comprises: carrying out data collection of historical data of a wind
turbine generator; performing data cleaning to remove abnormal data; establishing an
early warning model by using a support vector regression
algorithm; carrying out residual analysis and early warning; and based on skewness and kurtosis in statistics, calculating the residual error of the output value of the
early warning model; and calculating the kurtosis and skewness of the residual error day by day through a sliding window
algorithm, taking a mean value of a maximum value of the skewness and a maximum value of the kurtosisas a maximum value of an
early warning model threshold, taking a mean value of a minimum value of the skewness and a minimum value of the kurtosis as a minimum value of the early warning model threshold, and carrying out online monitoring and early warning on real-
time data of the wind turbine generator. According to the method, pre-fault pre-judgment can be provided in time before the wind turbine generator breaks down,
fault analysis and control are achieved in the first time, and huge economic losses and safety accidents are prevented from being brought.