The Phasor Measurement Unit provides data support for dynamic analysis and power fluctuation monitoring in renewable energy stations. Due to harsh operating environments

deployed PMUs often exhibit measurement anomalies

and continuous faulty data can impact the detection of real disturbances. Based on measurement data from a renewable energy demonstration base

this paper proposes a disturbance detection method for renewable energy stations considering PMU data characteristics. The method analyzes four types of PMU frequency anomalies in renewable energy stations and reveals the differences in spatiotemporal similarity between faulty data and disturbances across different PMUs. Considering both point and sequence anomalies’ local and global features

this paper proposes an improved Local Outlier Factor algorithm for point anomaly detection

which effectively reduces false alarm rate by incorporating clustering algorithms to compare the differences in point density. An improved denoising variational autoencoder model is introduced for sequence anomaly detection

enhancing feature extraction and setting dynamic thresholds

which improves detection accuracy. To further distinguish faulty data from disturbances

this paper develops an improved dynamic kernel correlation disturbance detection method

which calculates a spatiotemporal similarity matrix to measure feature differences

ensuring accurate disturbance event detection. Simulation and real data confirm the effectiveness of the proposed method.