The invention discloses an abnormal data processing method for wind turbines with power limiting operation, and belongs to the technical field of wind farms. The method comprises the steps that by introducing the reasonable hypothesis of the wind turbines under the power limiting operation, a data processing problem of the wind turbines under the power limiting operation is transformed into a mixed probability distribution model, K-Means clustering algorithm is used for initializing the model parameters, then according to the expectation maximization (EM) algorithm, the parameter update expression is deduced, and the optimal model parameters are obtained by iteratively maximizing the logarithmic likelihood function to process the abnormal operation data of the wind turbines under power limiting operation. According to the abnormal data processing method for wind turbines with the power limiting operation, the abnormal operation data with power limiting can be found, the operation datawith different power limiting levels is distinguished, and outlier abnormal data points are eliminated to get a fast and stable data processing effect.