Research On Improvement Algorithm For Wind Turbine Blade Fault Prediction Based On LightGBM

For the last few years,With the made in China 2025"," Industry 4.0" and other concepts put forward and implemented,China speeds up its industrial strategy with intelligent manufacturing as its core.The installation capacity of wind turbines is developing rapidly,becoming the fastest-growing clean energy source.In the current wind farm,a large amount of equipment operation data has been stored,showing many characteristics of big data,which provides a valuable opportunity for data mining of wind power generator operation.A fault detection algorithm for wind turbine blades based on improved LightGBM is designed and implemented in order to ensure the normal and stable operation of the equipment,timely and fast detection of equipment faults and improve the accuracy and speed of fault prediction.On this basis,the fault prediction software is developed.The main work including.1.Aiming at the problem that the prediction accuracy of the algorithm is not high,and combining the data characteristics of the fault data,a fault detection model based on the improved LightGBM is established First of all,the historical fault information of generator blade is excavated,and the parameters of GRU algorithm are optimized,and extract the fault factors of the historical time to improve the prediction accuracy of the algorithm.Then compare the improved algorithm with the simulation results of the original LightGBM algorithm,GRU algorithm and other models.Finally analysis shows that the F1 value is increased by 3% and 3.6% respectively.2.Aiming at the problem of predicting running time,Increased operational speed through data compression.Firstly,the feature of the data set is extracted,such as blade angle,speed of overspeed sensor,blade variable propeller temperature and other characteristic physical quantities are extracted,then the data samples with mutex features or small gradient are compressed to improve the running speed under the condition of ensuring the prediction accuracy.Finally the improved algorithm is compared with other models in the same environment.It is proved that the fault detection algorithm has good performance.3.On the basis of the improved prediction model,design the wind turbine fault prediction software to perform real-time fault prediction.First,a detailed demand analysis is made,and the various functional modules that the software needs to be implemented are determined,and then the software’s overall framework and Detailed functional design,clarify the specific functions realized by each module,design the database structure,and provide interface functions such as login,device management,user management,historical data management,and fault warning function management from different users.This paper combines fault detection methods with the construction of fault prediction software to achieve the supervision of equipment status and the rapid location of faulty equipment,sending timely warning information to the management side.