Research On Fault Detection And Visualization Of Wind Turbine Generators Integrating Machine Learning And Digital Twin

Wind power generation,as a clean and renewable energy source,is of great significance for addressing climate change and environmental protection.However,the high equipment maintenance costs of wind turbines seriously affect the profits of wind power enterprises.Realizing accurate real-time visual monitoring of status and fault diagnosis can effectively improve the power generation efficiency of wind turbines and reduce equipment maintenance costs.Digital twin and machine learning,as enabling technologies that drive traditional industries towards intelligent manufacturing,provide new solutions for intelligent maintenance and fault diagnosis technology of wind turbines.By studying the structure and operating characteristics of wind turbines,discovering their fault mechanisms,designing a fault diagnosis method for wind turbines that integrates digital twins and machine learning technology,and constructing a prototype system for fault diagnosis of wind turbines based on digital twins,achieving real-time perception of the state of wind turbine groups,it has significant theoretical and practical significance.This article specifically completes the following research work on fault diagnosis and visualization of wind turbines:(1)In view of the problems of low visibility of current wind turbine status detection and insufficient real-time data interaction,inspired by the concept of digital twin five-dimensional model,the digital twin five-dimensional model framework of wind power plant is constructed by analyzing the five dimensions of physical entity of wind power plant,virtual entity of wind power plant,data center of wind power plant,application service of wind power plant,and data connection of wind power plant;Based on the proposed five dimensional model framework of wind power plant digital twin,a visualization platform for wind power plant digital twin is built using software such as Unity3 D.The connection between the wind turbine database and the twin platform is completed through C# scripts.Driven by real-time data,real-time monitoring and precise control of the wind turbine operation status are achieved.(2)Aiming at the problem that the accuracy and recall rate of traditional fault detection algorithms are not high enough,and there is imbalance between positive and negative samples in the data set of fan fault detection,the loss function of CatBoost algorithm is improved,and the cross-entropy loss function is replaced by an asymmetric loss function.ASL CatBoost algorithm is proposed to realize real-time detection of fan fault,capture the information of fan fault status,and learn the algorithms GBDT,XGBoost,Light GBM CatBoost conducted an experimental comparison on a fan icing fault dataset provided by a certain wind power plant,and the experimental results proved the effectiveness of the improved algorithm.At the same time,the two improved points of the asymmetric loss function proposed in this paper are ablated,and the results are better than the original algorithm,which verifies the effectiveness of the improved method.(3)Aiming at the difficulty of setting appropriate hyperparameter for ASL CatBoost algorithm,a reptile search algorithm based on Tent chaotic map and t-distribution mutation strategy is proposed.The hyperparameter of ASL CatBoost algorithm,such as learning rate,iteration times,tree depth,are optimized,and the hyperparameter obtained after optimization are brought into ASL CatBoost algorithm for model training.In order to verify the optimization ability of the improved reptile search algorithm,TtRSA algorithm is compared with classic swarm intelligence optimization algorithms such as PSO,WOA,SSA on 11 benchmark test functions.The experimental results show that the improved reptile search algorithm has better convergence speed and accuracy.The TtRSA-ASL-CatBoost algorithm,which has been optimized by the TtRSA algorithm,has higher detection accuracy and efficiency compared to the original ASL-CatBoost algorithm.In summary,this paper builds a digital twin wind power visualization platform based on the digital twin five dimensional model,and proposes an improved ASL CatBoost algorithm to detect icing faults for the fault detection service of wind turbines in the digital twin five dimensional model,and optimizes hyperparameter through the improved crawler search algorithm to improve the detection accuracy of the algorithm,Finally,based on Unity3 D design,an integrated platform for digital twin wind power plant operation,maintenance,and control is implemented to verify the feasibility of the proposed architecture and algorithm.