Research On Fault Diagnosis Of Pumped Storage Unit Based On Machine Learning

Along with the reform of the Chinese energy structure,new energy sources such as water,wind and energy with foam accelerate the connection of the grid,but these new energy sources have been affected by uncertain factors such as climate and light,causing the energy grid to fluctuate,Influence the new power system and make the pumped power plant perform more top-level shaving tasks.Currently,the pumping plant industry started late.Due to unstable operating conditions and complex causes of pump-pumping plant failures,research on the diagnosis of defects in pumped storage is still very weak.The key problem to be solved in this document,is to ensure safe operation of pumped power net.Based on the theory of machine learning,certain technical problems and diagnostic failures of pumped storage units are examined.Firstly,on the basis of an analysis of vibration signal processing technology,with the aim of identifying the problem that small signals are difficult to process,a method of measurement based on empirical distribution and wave transformation threshold is proposed.Separation of background noise at an early stage of pump failure In order to obtain relatively clean vibration signals,the method proposed in this document has proved to be an effective method of measurement.according to the characteristics of the vibration signals of pumped warehouses with different stationary sources of vibration,The temporal characteristics and the characteristics of the vibration signals of the pumped warehouses are analysed on the basis of the mathematical characteristic parameter analysis technology.In accordance with the engineering requirements for intelligent diagnostics of storage unit failures,an automated model of diagnostic unit failures is proposed based on the machine learning algorithm.according to the characteristics of many types of sources of activation and frequent switching of operating conditions in conventional water plants,the state of flux under different operating conditions shall be taken into account,and the accuracy of the classification is improved by an analysis of the state of performance.By a comprehensive modelling of the parameters of working conditions and variable vibrations simulation results verify the results of the diagnosis.Analysing the vibration signals collected in the daily work of the pumped power plant,in conjunction with data on the state of work and the history of defects Pull out the time-frequency characterisation of the error signal.It can efficiently diagnose defects in the operation of pumped warehouses and ensure the efficient functioning of the power plant..Figure[17]Table[16]Reference[93]...