Research On Key Technologies Of Intelligent Fault Diagnosis For Wind Turbine Gearbox

Wind energy is a kind of green and clean renewable energy with rapid development.Its newly installed capacity and total installed capacity rank first in the world for five years in a row.With rapid growth of installed capacity,especially the increase in offshore installed capacity,the overall operation and maintenance costs of wind farms and wind turbines are increasing.At the same time,the continued decline in the price of wind turbines has further compressed the profitability of equipment manufacturers.Therefore,under the condition of combining big data and modern artificial intelligence technology,how to break through the key technology of intelligent fault diagnosis of wind turbines is the key to achieving "cost reduction and efficiency increase" for wind power equipment companies,and it is also an important challenge facing the current wind power industry.According to the actual demand of wind power enterprises,this paper studies the key technologies of wind power gearbox,such as wind turbine gearbox operation status warning,fault diagnosis,analysis platform integration,etc.The specific research work is as follows:To realize the early warning of wind turbine gearbox operating status with complex operating conditions and time sequence characteristic information,this thesis proposes a wind turbine gearbox status early warning method based on Gated Recurrent Unit(GRU)fusion multi-source data.Firstly,the inlet oil pressure of the wind turbine gearbox is used as the target variable of the GRU state warning model,and the wind turbine control logic,the physical mechanism of the fault are combined with the correlation coefficient method to determine the input variables of the GRU warning model;secondly,the time series characteristics of the model input variables are fused by GRU function to establish an early warning model for the operating state of the wind turbine gearbox;finally,the residual of the target variable is calculated through the calculated value of the model and the actual value of the operation,and the operating state of the gearbox is evaluated according to its change,thereby the early warning of the operation status of the wind turbine gearbox is realized.Validation and comparative analysis of instance data with other models showed that the GRU state-based early warning algorithm model is superior to other algorithm models,the gearbox fault state warning is issued 46 days ahead of the on-site real-time early warning system,and can detect the abnormal status of wind turbines in a timely and effective manner,and issue early warning information about fault status.This thesis proposes a fault diagnosis approach of wind turbine gearbox bearing based on multi-core learning support vector machine and firework parameter optimization algorithm.The main contribution of this study is to detect the gearbox bearing failure with small training samples.Starting from the gearbox CMS vibration data,this paper uses the classification advantages of the multi-core learning support vector machine method under the condition of small samples,combining the wavelet packet energy algorithm,the firework parameter algorithm and the multi-core learning support vector machine,and proposes a firework parameter optimization algorithm Multi-core learning support vector machine gearbox bearing fault diagnosis method(LOTF-MKSVM);building model result push and actual information feedback mechanism,realizing model parameter self-optimization,fault knowledge sub-refinement and push,and self-construction of knowledge base.On the basis of combining the online vibration(CMS)data collection mechanism of wind turbines,a multi-core learning SVM classification algorithm model and implementation mechanism is established based on the firework parameter optimization algorithm of wind turbine data.Then we apply the wind turbine Validation analysis using the actual operating data,apply verification analysis and make result explanations using the example data.Finally we use data analysis results to prove the effectiveness and superiority of the method.Aiming at the multi information integration state early warning problem of wind turbine early warning analysis platform,based on the currently used big data platform of wind turbine manufacturers,we develop a wind power gearbox early warning analysis platform on the basis of service architecture system-the key technology development,platform design and prototype system architecture design method.According to the actual needs of wind power operators,combined with the currently used data information platform,and the practical condition of wind turbine operation,this paper studies the key technologies of the wind turbine early warning analysis platform based on the micro-service architecture system,On this basis,the design method of service granularity division of wind turbine gearbox service entity object and the design method of early warning information push mechanism based on multi-source information joint diagnosis are proposed,and the fault map is constructed according to the structure of wind turbine gearbox.Finally,the application system integration of multi-source data state monitoring,multi model / multi information joint early warning and fault maintenance strategy is realized based on the prototype system of wind turbine gearbox early warning analysis platform.The case study is carried out for the theme expansion and service deployment based on service topic and application topic.