Research on adaptive online anomaly detection for satellite sensor data under concept drift and missing values
Zhizhuang Li, Jian Quan Liang, Lei Song, Lili Guo · IET conference proceedings. · 2025
To address the core challenges of missing value interference and concept drift in aerospace satellite online anomaly detection tasks, this paper proposes a deep anomaly detection framework based on an adaptive model pool strategy. Relying on the adaptive model pool mechanism, the framework enables multi - model collaborative work and dynamic updates, adapts to changes in data streams through flexible dynamic management, and effectively improves the anomaly detection performance for unknown concept drift. In experiments, the proposed model was extensively evaluated on public datasets and validated on space station payload datasets. The results show that the model exhibits excellent performance in adapting to concept drift and handling scenarios with missing data, successfully solves the problems of missing value interference and concept drift in data streams, which fully verifies the effectiveness and practicality of the proposed model.