Study on Dynamic Threshold Multivariate State Estimation of Condenser Based on Marine Environment

The static threshold analysis of condenser running states only based on running parameters can lead to missed detections and false alarms of abnormal states caused by variations under the marine environment. This study proposes a condenser running state analytical method based on the dynamic threshold multivariate state estimation approach related to the marine environment. Firstly, dynamic threshold for running parameters is formed based on the thresholds obtained from different marine environmental samples under time-series data. These thresholds are derived by applying an improved Gaussian Mixture Model to conduct multi-sample clustering of running parameters, in combination with evaluation criteria functions. Secondly, the condenser’s running state is determined by comparing the running parameter values with the dynamic threshold. An improved Multivariate State Estimation Technique is applied to estimate the condenser’s running parameters values using a distinguished state matrix related to marine environments. Additionally, the condenser’s running state are identified through the sliding window similarity assessment. Finally, through validation with time-series data of the condenser, it is demonstrated that the proposed method can accurately distinguish the running environments of the condenser and correctly evaluate its running states.