Research On Key Technologies Of Operation Maintenance And Management In Wind Farm

In decades,the exhaustion of fossil energy resources and environmental pollution caused by power generation have become severe factors that hinder the sustained and rapid development of our national economy,highlighting the importance of an alternative green and renewable ener-gy development.Because wind energy is much abundant in reserves,wind power generation has the most mature conditions for industrialization.Therefore,in the process of continuous reform in national power system,wind power industry has been regarded as one of the national strate-gic emerging industries and achieved great progress.Nowadays,our national energy system has changed dramatically and wind power generation becomes the third largest energy source after thermal power generation and hydropower generation.However,frequent faults in wind turbines and the lack of sophisticated management of spare parts result in the high operation and mainte-nance cost,which further restricts the competitiveness of wind energy generation in electric energy market.Based on the challenges mentioned above,aiming at enhancing the market competitiveness of wind power energy generation industry,this thesis focuses on methods for fault symptoms ex-traction/recognition,troubleshooting and spare parts inventory management.The main contents of this paper are organized as follows:Firstly,a fault pattern recognition method for identifying/recognizing early signs of failure is presented and applied for fault prognosis of the thermostatic valve damage in gearbox.In this part,a dual-stage attention-based encoder-decoder neural network is employed to characterize the nonlinear relationship between the target sequence,i.e.the gearbox oil temperature,and other exogenous variables in various normal conditions.Based on the residual between the estimated value of the normal behavior model and the actual measured value of the oil temperature,failure symptoms of thermostatic valve can be extractedNext,an approach for conducting the failure mode and effect analysis based on the combi-nation of fault tree analysis and binary decision diagram is introduced in this thesis.With the qualitative and quantitative results in failure mode and effect analysis,importance measures of different basic events in failure modes can be obtained and served to figure out the most crucial components of wind turbine.Moreover,with the qualitative and quantitative results,Bayesian inference is employed to guide the troubleshootingFinally,this thesis deals with two kinds of inventory systems:single-item system with unlim-ited storage resource and multi-item system with limited and sharable-common storage resource.Correspondingly,two different model are established with non-stationary stochastic demands un-der continuous review(r,Q)policy considering the sum of set-up cost,holding cost and stock-out penalty cost.Especially,the qualitative and quantitative results of failure mode and effect analysis are employed to quantify the penalty cost of each individual item.