Research On Wind Power Ultra-Short-Term Power Prediction Based On Dragonfly Optimization Algorithm

With the gradual expansion of energy demand for social development,the exhaustion of fossil energy is inevitable,and the problem of environmental pollution is becoming more and more serious.As an important part of my country’s energy strategic action plan,wind energy has received extensive attention.However,due to the strong nonlinearity and volatility of wind energy,it brings severe challenges to the safe and stable operation of the power system.Providing accurate and reliable wind power forecasts for the power system dispatching department can effectively alleviate the "wind abandonment" problem,and can also provide data guarantee for the safe hydrogen production of the microgrid hydrogen production system,which can improve the comprehensive utilization of energy.Based on this,this paper takes "data preprocessing-parameter optimization-basic model" as the main line to conduct wind power ultra-short-term power prediction research on historical power data.The main researches in this article are as follows:(1)The paper summarizes the topic background,development status and classification methods of wind power forecasting.A general prediction process of "data preprocessing-basic prediction model-parameter optimization algorithm-power prediction and data conversion" is proposed.Then take the general forecasting process as the main line.The common algorithms in each part and the characteristics and applicable scenarios of each algorithm are respectively summarized.It also summarizes the common prediction performance evaluation indicators for wind power ultra-short-term power prediction.(2)Aiming at the problem of inaccurate prediction models caused by the randomness of the original wind power data.This research proposes three signal analysis methods:Wavelet Packet Decomposition(WPD),Empirical Mode Decomposition(EMD),and Variational Modal Decomposition(VMD)for data preprocessing.And experiment with the actual historical power of a wind farm as the research object.The data preprocessing effect is judged according to the thoroughness of the processed data decomposition,the number of sub-modalities,and the degree of modal overlap.Provide reliable data samples for the establishment of combined forecasting models.(3)Aiming at the problem of insufficient accuracy of wind power forecasting models.Establish combined prediction models based on support vector machine(SVM)and least square support vector machine(LSSVM)respectively.Aiming at the problems of the penalty factors and the difficulty in selecting the core parameters of the two models,the Dragonfly Algorithm(DA),which has fast convergence and strong ability to jump out of the local optimal solution,is used to optimize its parameters.Two optimized forecasting models,DA-SVM and DA-LSSVM,were constructed respectively to provide a method basis for establishing a combined forecasting model.(4)This article takes a 50 MW wind power station in Hebei Province as an example.Select the winter sample group in January 2018 and the summer sample group in July2018 for simulation experiments respectively.First,a combined prediction model based on WPD-DA-SVM is constructed.The model uses WPD to decompose and process the original data to reduce the volatility of the original data.Then DA-SVM is used to model its sub-sequences and calculate the predicted values.Then superimpose the sub-predicted values to obtain the final predicted value.The results show that the accuracy of the summer data sample of this method is 96.12%,and the accuracy of winter data sample is97.33%.Later,in order to improve the prediction accuracy,LSSVM is used to replace SVM to construct a combined prediction model based on WPD-DA-LSSVM.Since LSSVM uses the least squares linear system as the loss function,it effectively reduces the prediction error and optimizes the prediction result.The results show that the accuracy of the summer data sample of this method is 96.61%,and the accuracy of the winter data sample is 97.79%,which has a good prediction effect;because the VMD decomposition has the advantages of thorough decomposition,fewer sub-modes,and fast decomposition,data can be extracted quickly feature.The constructed VMD-DA-LSSVM combined prediction model has the highest prediction accuracy.The results show that the accuracy of the summer data sample of this method is 98.03%,and the accuracy of winter data sample is 98.60%,which can meet the requirements of wind power day-ahead dispatch and wind power hydrogen production system,and has certain practical engineering significance.