Short-Term Wind Power Prediction Based On Principal Component Analysis And Fruit Fly-Neural Network For Jiuquan Base

As Wind-generatedelectricity.has the characteristics of randomness,intermittence and so on,which leads to the large fluctuation of the output power of the wind farm,Direct access to the grid will seriously threaten the stability,continuity and adjustability of the power grid.In particular,Jiuquan and other large scale wind power base using centralized grid,to further enlarge the impact of wind power fluctuations caused by the impact of the grid,resulting in a huge security risks.Therefore,in view of the fluctuation of wind power output,this paper puts forward an accurate method of wind power prediction,which is of great p ractical value to realize the high precision prediction of wind power generation and the safety and economic dispatch.Study on the optimization of the dynamic Elman neural network based on a new adaptive fruit fly algorithm for short term wind power prediction in the future 24 h and the measured data of a wind farm and the surrounding wind tower in Jiuquan wind power base in order to meet the actual demand of wind power prediction in Jiuquan wind power base.Specific work as follows:First of all,according to the original input variables of wind power formula used in this paper,acquisition,transmission and storage process leads to data loss,error and other issues for the raw data,soto detect integrality and rationality of the data,and the abnormal data by deleting and filling method for processing.Based on the investigation of the error distribution data of each wind farm in Gansu,we choose the appropriate error evaluation index to evaluate the prediction results.Secondly,according to the Elman neural network's gradient descent learning algorithm of slow convergence and easy to fall into local optimum,put forward the adaptive fruit fly optimization algorithm(FOA)with a good global optimization performance and computing performance,which improve the parameters ofElman neural network mode and an improved FOA-Elman neural network model is established.Finally,in order to improve the accuracy of power prediction,considering the factors that affect the prediction accuracy,in addition to the model selection and learning algorithm,the effectiveness of input data is also crucial.Therefore,the principal component analysis(PCA)is used to deal with the input characteristics of the short-term wind power prediction,and four kinds of uncorrelated principal components are obtained after analysis.The four principal components are used as input variables of the improved FOA-Elman neural network and the PCA-FOA-Elman neural network model is established.Through the experimental simulation,the PCA-FOA-Elman model and the FOA-Elman neural network model based on the adaptive fruit fly optimization algorithmand the improved Elman neural network model are compared and analyzed.The results show that: in the short-term wind power prediction,the prediction results of the new PCA-FOA-Elman model are 39.44% lower than the Elman neural network model,and the root mean square error is reduced by about 36.45%,The above model provides a new idea for improving the accuracy of short-term wind power predictionand is of great significance to the realization of measurable,controllable and adjustable wind power.