Research On Short-term Wind Power Prediction Based On Information Fusion

Due to the limitation of traditional energy utilization and increasing environmental problems,wind energy is widely used as a kind of green renewable energy.Wind energy is mainly converted into electricity by the rotation of the wind blower blades.However,due to the complexity of the earth’s atmosphere,the instability of wind energy leads to the fluctuation and intermittencies of output power,which brings challenges to the planning and construction of large wind power projects and power grids.Accurate wind power forecasting is one of the most effective responses to these challenges.In order to improve the accuracy of short-term wind power prediction,this paper proposes a short-term wind power prediction method based on information fusion.Firstly,the wind measurement data and wind power data were preprocessed.In terms of wind measurement data processing,the rationality of the wind measurement data was tested according to the national standards for wind farm data auditing.On this basis,the wind speed data were processed by using the smoothing z-score method for abnormal data,and finally the wind measurement data were processed by interpolation.In wind power data processing,according to the wind speed-power scatter diagram analysis of wind power distribution characteristics of abnormal data and appears the reason.In view of the different types of abnormal data identification,different methods for the numerical characteristics of the abnormal data is significant according to its numerical characteristics,which can identify as the numerical characteristics is not obvious abnormal data using local anomalies factor algorithm processing,Finally the paper adopt the interpolation method to repair the wind power data.Then,the outlier robust extreme learning machine model based on the rich poor optimization algorithm is used to predict the wind power.The outlier robust extreme learning machine uses the sparsity principle of outliers to replace the loss function of ?2 norm in the classical limit learning machine with the loss function of ?1 norm,which reduces the interference effect of outliers in the data set and improves the robustness and generalization performance of the model.The rich and poor optimization algorithm is simple in structure,easy to control and stable in performance,which can be used to find the optimal model parameters of outlier robust extreme learning machine.The results of an example show that the outlier robust extreme learning machine based on the rich and poor optimization algorithm can improve the prediction accuracy of wind power.Finally,in order to illustrate the importance of information fusion of high-dimensional meteorological data for wind power prediction,a wind power prediction model based on local linear embedding dimensionality reduction algorithm is established.Through the comparison of calculation examples,it is found that the prediction model after information fusion has better prediction results.At the same time,due to the nonlinear characteristics of meteorological data,compared with the linear dimensionality reduction algorithm principal component analysis,the local linear embedding of the nonlinear dimensionality reduction algorithm can better restore the essential structure of the data.