Research On Optimization Maintenance Strategy Of Wind Turbine Based On Fault Data And Monitoring Data

In recent years,the development of renewable energy has been increasingly valued by many countries.As a relatively mature power generation technology,wind power has been widely used in recent years.The number of installed wind turbine is increasing,and the installed capacity is also increasing.However,due to the complex and volatile working environment of the wind turbine,it takes a lot of manpower and material resources to maintain them.This greatly increases the operating cost of the wind farm and reduces the profitability of the enterprise.Therefore,state monitoring and maintenance decisions for wind turbines are becoming more and more important.If the fault of the wind turbine can be judged accurately,the operation and maintenance cost of wind farm can be effectively reduced.At present,the research on the maintenance of wind turbines is widely based on the its failure data recorded on the operation ticket from wind farm,and the Weibull distribution method is used to model and analyze components.This method only considers the fault data of wind turbine and is not comprehensive enough.The status of wind turbine will change over time and its actual state cannot be considered.At present,wind turbines are basically equipped with SCADA system.The monitoring data of the SCADA system reflects the state change information of each component of the unit,but the massive SCADA data stored in the control center is not fully utilized.In view of the above problems,this paper first counts the fault information of each component in the fault database,determines the key components.For the irreparable components,using its historical monitoring data and fault data,a proportional hazards model based on state information is established.And its parameters are identificated by using maximum likelihood estimation combined with genetic algorithms.The minimum cost method was used to determine the optimal preventive maintenance interval and calculate the component threshold in combination with model parameters.Upper and lower limits are added to the state maintenance curve to distinguish the normal working area,the monitoring enhanced area and the maintenance area.We substitute the status information of the parts that need maintenance into the graph to achieve condition based maintenance.Finally,an example analysis of the generator bearing and carbon brush is carried out.For the repairable parts,the age-return factor is introduced and a proportional intensity model is established.Based on the existing data,polynomial theory is used to fit more realistic simulation data and solve the problem of insufficient data required for modeling.The SPSS software is used to perform principal component analysis on multiple monitoring data items to perform data dimensionality reduction.Finally,the proportional intensity model of the repairable component is established,and the method is applied to the maintenance decision of the generator to realize the condition based maintenance of the generator.