Vibration Modeling And Analysis Of Wind Turbine Towers Based On Edge Computing

Aiming at the instability problem caused by mechanical fatigue and vibration of a single wind turbine flexible tower,an improved regression method is used to establish a tower vibration prediction model.Under different operating conditions of the wind turbine,the multi-source heterogeneous data is optimized by correlation analysis,and the related variables affecting the vibration of the flexible tower are obtained.Based on the grey wolf optimization algorithm,the optimal parameters of the support vector regression method are obtained,and the tower vibration prediction model is established.Under the flexible tower of multiple wind turbines,due to the limited computing resources of the wind turbine intelligent terminal equipment,some terminal equipment frequently has task congestion or abnormal operation when performing tower vibration modeling tasks,resulting in particularly large execution delay and energy consumption.Therefore,a server with computing power is introduced on the edge side as an edge node to share the computing pressure of the terminal equipment to improve the task processing rate and reduce energy consumption.In order to improve the computational performance of edge nodes,it is necessary to design a reasonable computational offloading strategy.However,traditional computational offloading algorithms usually require complex operations to optimize the target.In order to solve the problem of computing offloading decision-making of wind turbine intelligent terminal equipment in the process of tower vibration modeling,while improving the performance of tower vibration modeling,the time delay and energy consumption in the modeling process are reduced as much as possible.A tower vibration modeling task offloading algorithm based on deep reinforcement learning to achieve multiple optimization objectives is proposed.Firstly,the tower vibration modeling task is divided into several sub-tasks,and the edge nodes are set as agents.According to the sub-task dependency and the state of computing resources,the agents continuously adjust the strategy through the reward function according to multiple optimization objectives of time delay,energy efficiency and cache hit rate,and generate the optimal offloading decision for each sub-task to achieve parallel computing.The simulation analysis is carried out with the 120 m flexible tower data of a 2 MW wind turbine in a wind farm.The results show that under the flexible tower of a single wind turbine,the SVR model based on GWO optimization has the minimum error between the predicted value and the actual value under the two working conditions below and above the rated wind speed,which effectively improves the vibration prediction accuracy of the flexible tower.Under the flexible tower of multiple wind turbines,the influence of the number of wind turbine terminal equipment and the computing power of edge nodes on the cost of energy consumption is compared.The results show that the proposed algorithm reduces the energy consumption and improves the operation speed,which effectively solves the optimization problem of tower vibration modeling of multiple wind turbines.