After acquiring the power grid dispatching signals, traditional deep confidence identification systems are mostly used for anomaly data extraction, which can only obtain the anomaly information parameters contained in low-dimensional data, resulting in a lower Area Under the Curve (AUC) value of the final data extraction result. Therefore, in order to improve the AUC value of the anomaly data extraction results of the power grid dispatching signals, an anomaly data extraction method for power grid dispatching signals based on data mining algorithms is proposed. The power griddispatching signals are processed using the Independent Component Analysis(ICA) algorithm to remove noise from the signals. The denoised signals are then subjected to wavelet decomposition to obtain multiple sub-signal datasets. Clustering algorithms in data mining algorithms are employed to analyze the sub-signal datasets to obtain the characteristics of the data samples, and data feature classification is completed considering the attribute feature density index to obtain the anomaly data characteristics.Finally, with the assistance of the Support Vector Data Description(SVDD), the abnormal data in the power grid dispatching signals are detected, and summarizing this part of the data can complete the anomaly data extraction. The experimental results show that the AUC value of the anomaly data extraction results obtained after applying the proposed method is always greater than 0.85, proving its superior application effect.