Due to various factors such as communication interference, operational faults, maintenance shutdowns, protective shutdowns, and grid dispatching instructions, wind farms often accumulate a significant amount of invalid data in their collected operational data. This low-quality data presents challenges for tasks such as wind power prediction and wind turbines condition monitoring. To address this issue, this paper presents a novel method for identifying invalid data in wind turbine operations, which combines segmented quartiles and curve morphological feature detection. The proposed method first employs the segmented quartiles approach to identify sparse invalid data points. Subsequently, leveraging the curve morphological characteristics associated with downrating invalid data, a peak detection algorithm integrated with curve fitting techniques is utilized to identify invalid data points specifically related to downrating operations. To validate the effectiveness and applicability of the proposed method, both real-world wind farm data and artificially generated simulation data are employed. The results demonstrate the efficacy and universality of the proposed approach in identifying and filtering out invalid data, thereby enhancing the reliability of wind power prediction and wind turbines condition monitoring.