Ultra-short-term Wind Speed Prediction Of Wind Turbines Considering Wake Effect

Wind power has both environmental,economic and social benefits,and has developed rapidly in recent years.The ultra-short-term prediction of wind speed can guide the wind turbine to move in advance,which is of great significance for improving the efficiency and safety of wind energy conversion.Most of the existing wind speed prediction studies are based on the wind speed data measured by the mechanical anemometer on the wind measuring tower or the tail of the wind turbine nacelle.However,due to the large number of wind turbines in modern wind farms and the large power,the wake effect is obvious.Therefore,the wind speed measured by the mechanical anemometer cannot truly and comprehensively reflect the actual input wind speed near the hub of the wind turbine.Using it to predict the wind speed may have problems with low accuracy and hysteresis.Lidar can know the input wind speed of the wind turbine in advance.However,considering the cost,it is unrealistic to install lidar on all wind turbines in the wind farm.Based on this,this paper considers the influence of upstream wind turbine wake on downstream wind speed and the spatial correlation of upstream and downstream wind speeds,and uses deep learning to establish the relationship between the few wind turbines with lidar installed in the wind farm and the wind speed of upstream wind turbines,and then realize Use the upstream output wind speed to predict the ultra-short-term downstream input wind speed.This article mainly carried out the following work:(1)A Jensen wake model was constructed,and based on the Jensen wake model,the influence of upstream initial wind speed,wind direction and other factors on the downstream wind speed and the dynamic change of the downstream wind speed with the upstream wind speed were analyzed based on the simulation of the Jensen wake model.The simulation results show that the downstream wind speed is obviously affected by the upstream wake and there is a strong correlation between the upstream and downstream wind speeds.(2)Based on the idea of deep learning,a CNN-GRU model that can simultaneously mine the temporal and spatial characteristics of wind speed series is constructed.The input wind speed measured by the lidar on the target wind turbine and the output wind speed measured by the upstream wind turbine mechanical anemometer are input into the model training,and the nonlinear mapping relationship between the two is established,and the upstream wind speed is used to predict the downstream wind speed.,With good accuracy.Based on the idea of integrated learning,all wind turbines installed with lidar and the corresponding upstream wind turbine data sets in the entire wind farm except for the first row of the upstream are used to form a new sample set,and Bagging is constructed by combining the Bagging algorithm with the CNN-GRU model.-CNNGRU model.The experimental results show that the prediction accuracy and stability of this model are further improved compared with the single CNN-GRU model.(3)Using the simulation experiment model constructed in the previous article,comprehensively considering functions,architecture,and interface,a wind speed prediction platform based on deep learning was designed and developed.The visual operation of wind speed prediction is realized,and the efficiency of wind speed prediction is improved.