| In September 2020,China formally put forward the emission reduction goal of"achieve carbon peak by 2030 and carbon neutrality by 2060".Such a formidable task has pointed out the direction and drew a specific road map for our country’s energy transformation towards clean and low-carbon,and demonstrated China’s firm determination to tackle climate change and its commitment to responsibility.As one of the main utilization methods of clean energy at present,wind power generation will inevitably develop at a faster speed,larger scale,and higher quality in the future,which puts forward higher requirements for the design,manufacturing,and operation and maintenance of wind power equipment.In the early stage of large-scale wind power construction in our country,there are outstanding problems of high wind turbine equipment failure rate and high operation and maintenance costs,especially the high failure rate of key large components(gearbox,generator and main bearing,etc.),which seriously affects power generation.Increasing the cost of later operation,maintenance and overhaul of wind farms will bring greater economic losses to the operation of wind farms.As the wind power industry is gradually moving towards large-scale and offshore development,this has put forward higher requirements for the safety,reliability and economy of wind turbine operation.Therefore,continuous improvement of the state monitoring and fault diagnosis technology level of key components of wind turbines,wider component monitoring scope,more timely fault warning,and more accurate fault diagnosis will surely become the consensus of the entire wind power industry.The research results of this paper can provide a scientific and reasonable basis for the timing of maintenance of key components of wind farms,and can also provide important technical support for the safe and economic operation of wind farms.This paper conducts in-depth research on the current status of fault diagnosis of key components of wind turbines and key technical issues in life assessment of key components,and the following contents are studied separately:(1)Fault diagnosis method based on composite decomposition of vibration signal combined with feature dimensionality reduction.The wind turbine transmission chain has complex structure and many failure modes,and the vibration monitoring signal has the characteristics of multi-source coupling modulation,which makes it difficult to quickly and comprehensively extract weak fault characteristic information.In response to this key problem,this paper studies and proposes a vibration signal composite decomposition combined with principal component analysis dimensionality reduction processing method.First,the advanced ensemble empirical mode analysis method(EEMD)and the local mean value decomposition method(LMD)are used to compound and decompose the vibration monitoring signal,so as to fully reveal the weak fault information that causes the state change in the vibration monitoring signal.Secondly,select the decomposed high-frequency components for feature extraction,and fuse the eigenvalues of each component obtained by the two decomposition methods to form a high-dimensional feature vector.Finally,Principal Component Analysis(PCA)is used for dimensionality reduction,and the square prediction error(SPE)index is used as a quantitative evaluation index reflecting state changes.Through an example of actual fault signal analysis of wind turbines,the analysis effect of this method is compared with several other feature extraction and dimensionality reduction methods based on different signal analysis methods.The results show that the proposed "composite decomposition combined with principal component analysis" method Not only meet the requirement of feature extraction accuracy but also improve the efficiency of feature recognition.(2)A weak fault feature extraction method based on blind deconvolution combined with sparse decomposition.In view of the strong interference and noise of wind turbine vibration monitoring signals,wide frequency range of fault information distribution,and strong non-stationarity,this paper studies and proposes a vibration signal analysis method that combines blind deconvolution and sparse decomposition.First,the blind deconvolution method(mainly using the minimum entropy deconvolution and the maximum kurtosis deconvolution method)is used to preprocess the vibration monitoring signal,so as to suppress the random noise component in the monitoring signal and highlight the fault characteristic signal,enhance the fault impact component in the signal.Secondly,the pre-processed vibration signal is sparsely decomposed,and the vibration monitoring signal is processed in sections according to the minimum interval of the periodic impact component of the wind turbine transmission chain fault.The matching tracking algorithm is used to decompose each segment of the signal to extract the fault component.Then reconstruct the signal in order of the fault impact components obtained.Through the analysis and comparison of the actual operation monitoring signals of wind turbines,the results show that the method can achieve the purpose of weak fault information enhancement and extraction at the same time.By selecting appropriate deconvolution filter parameters,significant vibration signal extraction effects can be obtained.(3)Fault early warning method based on classification and recognition of operation monitoring data.There is a huge amount of big data in the wind turbine operation monitoring system(SCADA),among which various logical relationships between the data are complicated,and there are many interference data.This brings a lot to the rapid and accurate construction of the target model to efficiently diagnose and predict various typical faults.Aiming at this difficulty,this article uses the attributes of various data in the SCADA system to construct a modeling method for classification and recognition.First,based on the data of similar monitoring indicators in the SCADA system(gearbox oil temperature),a fault warning method with detailed working conditions is proposed.The method is divided into bins for different impeller speeds in the grid-connected section of wind turbines,and the normal behavior of gearbox oil temperature is established.The model also delineates the oil temperature distribution and the abnormal limit of the rate of change,and adopts the methods of timing execution,quantitative analysis,and fixed window evaluation to realize the functions of gearbox oil temperature abnormality detection and fault warning;In the end,the correlation between the SCADA data is more direct and can quickly identify the relevant components(blade icing fault),and the BP_Adaboost fault detection strong classifier is used to build the model,and its accuracy is better than single a wind turbine blade icing fault detection model constructed by BP neural network.(4)Wind Turbine Life Evaluation System Based on the Theory of Fatigue Cumulative Damage.Due to the complex and changeable characteristics of the structure,operating conditions and environmental conditions of wind turbines,domestic wind turbines in the middle and late stages of the design life have a relatively weak research foundation for the life evaluation technology of the whole machine and special key structural components.This is a bottleneck.problem,this paper has developed a set of fatigue load and life evaluation system for key components of wind turbines.Firstly,a wind turbine load database(including air density,turbulence intensity and calculated wind speed conditions)was constructed;secondly,based on the Miner fatigue cumulative damage theory,a life evaluation system for key components of the wind turbine was developed.The remaining life predictions of the key components of the two wind farms are verified as calculation examples,and the results show that the system can realize the functions of optimizing the configuration of the curtailed wind farms and predicting the parts where fatigue damage will occur. |
PDF Full Text Request