Research On Wind Power Prediction Approach Based On CFD Pre-calculated Flow Fields In Complex Terrain

As the stochastic and intermittent nature of wind energy, large-scale wind power has been bringing about huge negative impact on the secure and economic operation of power system. While accurate wind power prediction is one of effective way to mitigate the above problems. As for the wind power prediction based on Computational Fluid Dynamics (CFD). For one thing, the computational model can be refined pointly, and the prediction accuracy could be improved by complex flow fields calculation. For another, it operates flow fields calculation before the wind power prediction, and the computational burden from flow fields calculation is dramatically reduced. Moreover, a CFD-based method has no need for historical data to train the model which is suitable for those wind farms with insufficient operational data or with complex terrain and variable meteorological pattern.The contributions in this paper are summarized as follow:(1) CFD based power prediction model and pre-computational database for wind farms with complex terrainThe error characteristic of CFD-based power prediction method used in flat terrain is studied to better understand the influential law of different weather factors on prediction errors. In this way, an optimization scheme based on wind condition discrete is proposed to improve the structure of pre-computation database. The CFD model and pre-computation database are improved by local refining the model related with sensitive wind condition. Therefore, the wind field could be accurately simulated, meanwhile save redundancy of insensitive computation efforts.(2) Pre-computation wind power prediction modelPre-computation method is to build a database according to wind field characteristic and to pre-store the results from time-consuming CFD computation. In real time operation, the corresponding prediction could be searched and extracted from the pre-established database. A wind farm with complex terrain in northwest China is taken as example to validate the model performance. The results show that the proposed prediction model perfectly tracks the output fluctuation in three conditions when wind power exerts most negative impacts on power system. These three conditions contain wind power generated from full capacity rapidly down to zero, zero rapidly up to full output, and dramatic fluctuation in a very short period of time.(3) Three optimization models for numerical weather prediction dataThree optimization model are proposed based on least squares, support vector machine (SVM) and radical basis function(RBF) respectively to improve the accuracy of NWP wind speed and then the wind power prediction accuracy. The performance of each optimization models is compared and the contribution of model input error on power prediction is examined by put observed wind speed, raw NWP, improved NWP into the proposed prediction model respectively.