Comprehensive transmission sequence anomaly detection based on RFE-PAA and BI-LSTM-CNN

With the rapid advancement of industrial technology, the structure of transmission systems has become increasingly complex, posing more severe challenges to their real-time monitoring and anomaly detection systems. To address these challenges, this paper proposes a novel comprehensive transmission sequence anomaly detection method. Firstly, the method employs Recursive Feature Elimination (RFE) to select the most relevant temporal features for precise identification of the anomaly detection target variables. Subsequently, Piecewise Aggregate Approximation (PAA) is utilized for effective dimensionality reduction of the time series data to simplify the subsequent processing steps. Finally, by integrating Bidirectional Long Short-Term Memory (Bi-LSTM) and Convolutional Neural Network (CNN), a deep anomaly detection model is constructed to identify abnormal states efficiently. An example analysis of time series anomaly detection in transmission system oil leakage demonstrates the significant advantages of this method in improving the accuracy and efficiency of anomaly detection. The results indicate that this method can provide a new, reliable, and efficient technical solution for anomaly detection in complex transmission systems.