Research On CFD Simulation Method And Fast Calculation Model For Wind Farm Wake Flowfield

The wake effect has a significant impact on the power generation capacity of the wind farm,resulting in a loss of wind power generation to be about 10%to 20%.The development of wind farm wake flowfield simulation methods with high precision and fast calculation is beneficial to improving the optimal design and safe operation of wind farms.As an essential ingredient of wind farm wake flowfield,the structure of wind turbine wake flowfield and its modeling method are crucial to understand the mechanism of multiple turbine wake interaction and accurately simulate the wind farm wake flowfield distribution.Currently,among the mainstream wake simulation methods,the computational precision method based on the actuator disc is controllable,but the accuracy of wake simulation needs to be improved,and the computational resources are expensive;the traditional analytical model has fast calculation speed but low accuracy.For the pros and cons of the two models mentioned above,how to accurately and efficiently simulate the wake flowfield of wind farms has become a critical issue to improve the power generation in the design stage of wind farms.Hence,taking wind turbines and wind farms as the research objects,from the physical change laws and data of the wake,we explore the simulation method of actuator disc and analytical model with higher computational accuracy by combining the law of wake evolution on the one hand;on the other hand,to balance the computational accuracy and efficiency of the model,we develop the data-driven wake simulation technology based on the data from the simulation method of actuator disc,combined with reduced order model and the machine learning method.The following work has been accomplished around this topic:(1)A simulation method of an improved actuator disc for wind turbine wake flowfield is proposed.In view of the issues that the standard actuator disc cannot accurately reflect the turbulence characteristics,an improved actuator disc model considering the structure of nacelle and tower is developed,according to the canopy blockage effect and the principle of actuator disc.To address the issue that the standard k-ε turbulence model tends to overpredict the rapid wake recovery,the modified k-ε turbulence model is used to control the dissipation rate of turbulence in the near wake region,which aims to further improve the wake distribution.The proposed model of the actuator disc is developed by combining the improved actuator disc model with the modified k-ε turbulence model.The proposed model is validated by GH wind tunnel experiments in UK and TNO wind tunnel experiments in Netherlands.The results show that the proposed numerical model improves the simulation accuracy of wind turbine near wake,while ensuring the accuracy of far wake simulation.However,for the situation that the interwind turbine spacing in large wind farms has exceeded the near wake area,it is recommended that structures such as nacelle and tower be ignored when conducting wind farm wake simulations.(2)The effects of CFD engineering turbulence model parameter on wind farm wake flowfield are investigated.To address the issues that the parameter C4ε in the modified k-ε turbulence model is difficult to be calibrated in wind farm wake simulation and how the change of the parameter C4ε affects the wind farm wake recovery.Based on the wind farm actuator disc model with the thrust coefficient modification,the effects of the parameter C4ε on the wind farm wake simulation have been studied.The Horns Rev,Nysted offshore wind farm and Wieringermeer onshore wind farm are utilized as cases for analysis.The results conclude that reducing the parameter C4ε in the modified k-ε turbulence model can effectively promote the turbulent kinetic energy,inflow wind speed and power outputs of each rotor along the downwind direction.When the parameter C4ε is 0.15,the numerical simulation results are in the highest agreement with the measured power values.(3)A two-dimensional entrainment model for wind turbine wake flowfield is developed.For the high demand for computational resources and time consuming problems of the simulation method of the actuator disc,an analytical wake flowfield modeling research of wind turbines is developed to rapidly evaluate the wake distribution of wind turbines.However,traditional one-dimensional analytical models are based on a single conservation theorem and the uniformly distributed assumption,which cannot accurately capture the non-uniform wake distribution along the radial direction of the wind turbine.Contrarily,the onedimensional entrainment model can satisfy the conservation of mass and momentum concurrently within the control volume.On this basis,the Gaussian entrainment model is initially established with the assumption of Gaussian distribution,and the wake simulation results are closer to the real value.In view of the issue that the entrainment coefficient E=0.15 of the entrainment model is determined by fitting finite wind tunnel experiments and field measurement data,which lack certain mathematical and physical principles to derive the entrainment coefficient,resulting in the calculation accuracy of the wake model to be affected.A two-dimensional entrainment model with entrainment parameterization is developed,which considers the effect of variable entrainment coefficient on the wake recovery of wind turbines.The results indicate that the proposed model has a better simulation accuracy of wind turbine wake compared to the other four Gaussian wake models in the three scenarios.(4)An entrainment model for wind farm wake flowfield is proposed.An entrainment model for wind farm flowfield is proposed to address the shortcomings of several wind farm flow models that are widely applied in wind power engineering to overestimate the wake deficits.The proposed model consists of the modified linear entrainment wake(MLEW)model and the Root Sum Square superposition model.The MLEW model considers the effect of surface roughness on the wake recovery of downstream wind turbine,based on the basis of a one-dimensional LEW model.An entrainment wind farm flow model,combining with RSS superposition model widely used in engineering,for wind farm power generation simulation is established.The TNO wind tunnel experinments,Horns Rev offshore wind farm,and Lillgrund offshore wind farm are used to verify the MLEW model and proposed wind farm flow model.The results show that the MLEW model can accurately predict the normalized streamwise velocity at the wake centerline.The proposed wind farm flow model has a better accuracy in power compared with other analytical wind farm models.(5)A data-driven wake simulation method for wind turbines and wind farms is developed.To address the problems of high cost and long time-consuming of obtaining wake by high-precision CFD simulation,A data-driven wake simulation method for wind turbines and wind farms is established.The proper orthogonal decomposition(POD)method is used to extract the representative modal basis of wake flow.Four Machine Learning(ML)models such as Back Propogation(BP)neural network,Radial Basis Function(RBF)neural network,Generalized Additive Model(GAM),and Gaussian Process Regression(GPR)have been utilized to predict the modal basis coefficient from the target operating conditions.Based on the modal basis coefficient prediction results,the wake distribution of wind turbine and the power output of wind farm are quickly evaluated by linear multiplication with representative flow modal basis.TNO wind tunnel test and Horns Rev offshore wind farm are deployed to validate the proposed data-driven model.The results demonstrate that the four data-driven models based on PODML greatly improve the computational efficiency on the premise of ensuring the CFD simulation accuracy.