The Study of Log Anomaly Detection Strategy for Electric Equipment of Space Environment Simulation and Research Infrastructure

Monitoring the abnormal states of power equipment in the Space Environment Simulation and Research Infrastructure (SESRI) is crucial to ensure the efficient and safe operation of these devices. This paper introduces a power equipment abnormal state monitoring strategy based on log analysis technology. The log parser, Drain, is used to analyze the log data of power equipment, transforming it into structured data. Subsequently, the Template2Vec algorithm is employed to convert this data into word vectors and segment it into log sequences. Next, an attention-based Bi-LSTM (Bi-directional Long Short-Term Memory) model is constructed to extract feature vectors from the log sequences, and these vectors are input into an SVM classifier to detect whether the log sequences are abnormal. Finally, the feasibility of the proposed strategy is validated using the HDFS log dataset from Loghub. Experimental results demonstrate that this strategy efficiently detects and identifies abnormal behavior in power equipment, thereby reducing equipment failures and downtime.