Research And Application On Early Fault Warning Of Power Plant Fans Based On Multivariate State Estimation Technique

Modern large scale generator units have complex systems and numerous equipments.With the increasing competition of the power generation industry,it is more important to reduce the unplanned shutdown of the generator units.And the base of the units safety and economic operation is the stable and reliable operation of the equipment.Fan system is one of the important auxiliary system of power plant,which is directly related to the production safety of power plant.Therefore,in the early stage of equipment failure,the prompt message of abnormal equipment state should be given to the equipment manager clearly by the fault of the early signs and intelligent real-time monitoring of the running state of the equipment should be realized.Thus,the passive fault processing mode is changed into a kind of more active matter,early warning prevention.In this paper,the multivariate state estimation technique(MSET)based on the similarity principle of equipment history operating state is studied and applied to the field of fault diagnosis for power plant fan.Firstly,the structural features and the common faults of plant fan system are analyzed.Secondly,the data acquisition and pretreatment process of fault warning system method are studied,in which mainly discusses the simplification of modeling parameters and the selection of important parameters based on the principal component analysis,and the screening of historically normal operation time based on wavelet analysis.A kind of multivariate state estimation technique method was proposed to solve fan condition monitoring and fault prognostic in this article,based on the algorithm of clustering by fast search and find of density peaks(DPC).Firstly,the DPC algorithm was used to study the historical data of the fan under normal working conditions,extract data that contains characteristic information of the equipment under normal working conditions,and construct the memory matrix.Secondly,the level of similarity between the observation vectors and the memory matrix were analyzed by the correlation principle,and the observation vectors were estimated by the MSET.The statistically residual errors and similarity degrees between the estimated and measured values were calculated,and the work conditions of the fan were ascertained.Then,the fault simulation analysis is carried out by using Matlab,which verifies the validity of the principle.At last,the dynamic fault early warning model is established to analyze the application effect of this method in practical engineering.The results show that the method can real-timely and accurately predict fans working state,discover the fault symptom in advance,and guide the operation and maintenance of equipments.