Research On Wind Farm Micro-siting And Short-term Wind Speed Prediction Over Mountainous Terrain

The wind field within the atmospheric boundary layer(ABL)in the mountainous terrain will become complicated due to the influence of terrain conditions,etc.,making it difficult for wind farm construction from wind turbines micro,siting to wind power generation prediction.Accurately predicting the wind speed distribution over mountainous terrain becomes the premise of wind farm construction,and Computational Fluid Dynamics(CFD)is one of the most effective means.The following works have been completed about wind farm construction in mountainous terrain:1.A turbulent inlet generation method for Large Eddy Simulation(LES)based on time history cross-correlation correction is proposed,and it is combined with high-order Spectral Element Method(SEM)to simulate wind flow over mountainous terrain.Second-order Finite Volume Method(FVM)which introduces large dispersion and dissipation,is always adopted in existing CFD softwares,and it is difficult to give correct simulation results for the multi-scale flow phenomena such as separation and re-attachment accompanying in the mountainous terrain.Therefore,the high-order SEM with LES model is used to simulate ABL wind flow over mountainous terrain.Nevertheless,the turbulence characteristics of the inlet in LES should be satisfied,and the directly generation of fluctuating wind speed of all the inlet grids will:increase the storage and cost.In this study,the fluctuating wind speed of the coarse grids are generated,which satisfy the target turbulence intensity,power spectrum,turbulent integral scale and time history correlation,and the wind data.of the inlet grids are obtained by linear interpolation and cross-correlation correction.The turbulent inlet generation method is elaborated,and the validations are accomplished by two simple hilly terrain with different slopes.The open source platform Nek5000 based on SEM and the commercial software Fluent based on FVM are introduced for simulating wind flow over mountainous terrain,and the validation is accomplished by the Askervein case.The method is also used to simulate wind flow over a complex terrain in Hunan Province.The results are compared to the wind tunnel test,and the wind filed characteristics over mountainous terrain are studied.2.A coupled model combining refined CFD model and meteorological data is proposed for wind resource assessment and wind farm micro-siting over mountainous terrain.Existing commercial softwares for wind resource assessment are mostly based on Reynold Averaged Navier-Stokes(RANS),and some simplifications are treated to the governing equations,which will definitely lead to some limitations.For example,the nonlinear convection terms are linearly treated in the software WAsP,the one-equation turbulence model is used in the software WT,and the standard k-s turbulence scheme is applied in the software Wind Sim.The Realizable k-ε turbulence scheme is adopted based on software Fluent,to simulate wind flow over a complex terrain in Hunan Province,and the long-term meteorological measured data are transformed to the Generalized Wind Climate(GWC)by removing the local terrain effects.Combining the CFD results and GWC data,the wind energy resource map is obtained,and the wind farm micro-siting is determined preliminarily.A semi-analytical approach is proposed for wind power generation prediction at the wind turbine site by fitting the Gaussian function to the wind power curve and the Weibull function to the probability density distribution function of the mean wind velocity at hub height.The wake effects of the wind farm are considered by introducing the Jensen model and Gauss Distribution Wake Model.3.A combined statistical model of short-term wind speed/power prediction for wind farm is proposed.The commonly used statistical short-term wind speed prediction models,such as Autoregressive(AR)model,Autoregressive Moving Average(ARMA)model,Grey Prediction,Kalman Filtering,Artificial Neural Networks(ANNs),Support Vector Machin,Wavelet Transform(WT)and Fuzzy Logic,etc.,have a poor short-term prediction accuracy on wind speed data with non-stationary characteristics.In this study,the Brown method is introduced to remove the non-stationary characteristics of wind speed,and the Persistence model,AR model,ARMA model,ANNs and WT are used to predict the short-term wind speed in future.Moreover,the combined model is obtained by weighted averaging of several single models according to the error optimization theory.The short-term wind speed prediction models are tested by measured data from a wind tower near Ansai Wind Farm in Shanxi Province.The results show that the accuracy of shor-term wind speed prediction improves a lot after adopting the Brown method,and the combined model always provides the best prediction result.