Identification Of Operating Condition Of Wind Turbine And Fault Diagnosis Of Pitch System

With the rapid development of wind power technology,wind turbine unit capacity is growing.However,with the construction of large-scale wind farms,wind turbine generator system(WTGS)failure rate is high,the problem of high operating and maintenance costs are also highlighted.How to improve the reliability and utilization of WTGS operation becomes an urgent problem to be solved in wind power generation.Pitch system is an important part of WTGS,but because of its poor operating environment,many components,frequent start and stop,causing it to frequent failures.Based on the analysis of the working principle of WTGS and its pitch system,this paper studies the fault condition of WTGS operating condition identification and operation condition of pitch system based on SCADA data using WTGS historical data and pitch system fault information.The main contents of the study include:1)The structure and working principle of the direct drive WTGS and electric servo pitch system and the SCADA system are briefly introduced.The operating parameters of the WTGS are analyzed and the data of the historical data collected by the SCADA monitoring system is preprocessed.Analysis of operating parameters of pitch system based on correlation analysis of feature parameters based on information entropy.2)The energy control mode and the power limit symbol of the WTGS operating parameters are taken as the categorical characteristic parameters,we have proposed a mixed attribute data clustering algorithm based on self-organizing neural network and used it to classify the operating conditions of WTGS.The proposed algorithm employs the framework of self-organizing neural network,and adopts heterogeneous difference metric to measure dissimilarity on mixed attributes of WTGS operating parameters.The frequency of each category occurring in the Voronoi sets is utilized as on basis of the rule on categorical reference vectors.The mixed update rule of neurons is adopted for both numeric and categorical data simultaneously.On the basis of this,an additional output layer is added to the competitive layer of self-organizing neural network,which becomes a supervised classification network,and a supervised mixed attribute data self-organizing map classification model is proposed to realize the operation of WTGS condition identification.3)Aiming at the fault diagnosis of wind turbine electric servo pitch system,the characteristic attribute selection of the pitch system is carried out,and the abnormal identification model of the pitch system is established under different working conditions,the model is based on principal component analysis,the electric servo pitch system operation data is projected onto the principal subspace and the residual subspace of the principal component model,and it is determined by determining whether the corresponding T2 and SPE exceed the corresponding control limit;also,we use the contribution map method to find abnormal properties,the fault reconstruction based on SPE is carried out by the theory of fault subspace,and finally the fault diagnosis of WTGS pitch system is realized.