基于暂态录波与多元状态估计的阀冷系统故障预警及识别OA

Valve cooling system is one of the important equipment in DC transmission engineering,and improving the level of its fault warning and identification has important practical value.The dual frequency transient recording data triggered by the main pump cycle switching is used to extract the system's characteristic quantities under pump switching disturbance,thereby improving operational efficiency.A dynamic process memory matrix construction method based on a combination of clustering algorithm and switching conditions is proposed.An estimation vector based on multivariate state estimation is obtained.The cosine similarity between the estimation vector and the ob-servation vector are calculated.3σ outlier detection is used as the warning threshold,and then fault identification is achieved based on the coupling degree between the warning observation vector and the switching condition,as well as the contribution of the three fault types to the error.Finally,the effectiveness of the proposed method is verified through case analysis.