The Research On Wind Power Prediction Algorithm

When the fossil energy reserves are reduced,wanton harms to human health,the energy industry is gradually changing,the new energy that owns the nature of clean,renewable,rich reserves has been changing the human Energy awareness in a strong momentum.Under the constraints of technical conditions,though China's new energy industry started late,China's wind energy,solar energy and other clean energy are rich and widely distributed.With the rapid growth of economy,the energy demand growing,and the construction of ultra-high voltage lines in China,wind capacity integrated into grid and utilization rate of wind have also increased several times over.However,the problems of voltage fluctuations,frequency deviation and power quality have been caused by the quality of wind intermittent,when wind power was integrated into gird.The regional interconnection of power gird and energy globalization strategy will also have been Obstructed by the shortcomings of wind.In order to eliminate the social controversy about wind power,to avoid the phenomenon of abandonment wind,to build a grid-friendly wind farm,governments require wind farms to use wind power forecasting technology to achieve a reasonable allocation of rotation reserve capacity of wind turbines,significantly reduce the power of energy storage facilities of wind farm.Reasonable wind power dispatching plan will also be an important support toensure that wind power becomes controllable and smooth.Therefore,when formulating the power dispatching plan,the grid has been rendering the more stringent control on the accuracy of uploading rolling forecast information of wind power.This paper will start a the rolling forecast study based on the nature of wind from shanxi wind farm.This paper uses the combination of artificial intelligence,pattern recognition,signal decomposition and group intelligence to improve the accuracy of wind power prediction based on the data of Wind Tower.Firstly,introduce the wind power characteristics and analyze the factors in affecting the output of the Wind turbines quantificationally in this paper.Using Fruit Fly Algorithm FOA and support vector machine SVM to establish FOA-SVM wind power prediction model.Secondly,In ordering to resolve the FOA weakness of being susceptible to local extreme interference and strengthening the optimizing ability of FOA,an intelligent optimization algorithm that names Simulated Annealing Fruit Fly Optimization Algorithm based on Gaussian Disturbance GDSAFOA is proposed,so GDSAFOA-SVM wind power prediction model is established.By the optimizing SVM parameters,a reasonable SVM regression Hyperplane is set up to carried out to realize the wind power prediction and raise forecast accuracy.Then,for the sake of improving the similarity between the training sample and the predicted sample in the learning model and reducing the interference of irrelevant data,the paper bring GDSAFOA to the fuzzy C-means algorithmFCM and establish GDSAFOA-FCM cluster model.By this way,realize to cluster historical days and forecasting days with similar characteristics,and raise forecast accuracy.From the GDSAFOA-FCM results we can see that the similarity day's trend is close to the forecast day's,nevertheless,some similarity day's wind power fluctuation is very different from the forecast day's.The error of forecasting is also increased by the wind power fluctuation.It is proposed to introduce the turbulence value IT into the FCM algorithm.Controlling the similar volatility by the turbulence value IT as the characteristic information of similar days so as to increase the similarity between the training samples and the predicted samples in the fluctuation degree.By the means of establishing GDSAFOA-FCM-IT cluster model,the samples that possess large fluctuation deviations with predicted days are removed,and the data of similar fluctuation trend with predicted day is kept.GDSAFOA-FCM-IT-SVM wind power prediction model is established based on GDSAFOA-FCM-SVM model,and raise accuracy of GDSAFOA-FCM-SVM model.Finally,the EEMD and EMD algorithm are used to smooth the wind power signals and are compared in the ability of decomposition.The EEMD-GDSAFOA-FCM-IT-SVM wind power prediction model is used so that suppressed wind power volatility.All the above models and algorithms are verified by Matlab platform and wind farm data.From the simulated results we can see that EEMD-GDSAFOA-FCM-IT-SVM combined prediction model possess thequality of Less training samples and high prediction accuracy,which provide valuable analysis and research results for engineering realization,compared with other predicted models.