Wind Speed Time Series Prediction Based On Variational Mode Decomposition And Echo State Network

Economic development will constantly promote social development,and any resource use should be moderate and appropriate,excessive demand will cause a serious problem of energy shortage and further damage to the environment.So it needs to explore the use of reproducible energy and mining to maximize economic benefits.wind power is a clean energy which has the superiorities of reproducible wind energy.at the same time the advent of the industry has become increasingly mature wind power technology,wind speed is influenced by many elements,however,wind power integration is often accompanied by randomness and stability,which brings considerable difficulties to wind power grid connection and has a huge impact on the safety and maintenance of power systems,so the exact forecasting of wind speed change on wind power has great research significance.In this thesis,the wind speed forecasting model is researched as follows:(1)Comparison of wind speed time series decomposition models.There are components of different frequencies in the time series,so it is necessary to decompose the time series by some method.This thesis introduced a set of compound containing cycle oscillation signal and uses the empirical mode decomposition method with the variational mode decomposition method to decompose at the same time,through the composite signal contrast can be seen that using the variational mode decomposition(VMD)can effectively solve the empirical mode decomposition(EMD)endpoint effect,reduce the error of the prediction model,this thesis finally adopted VMD algorithm for time series decomposition method.(2)The original wind speed time series was pretreated by variational mode decomposition method.Given by wind speed time series itself is affected by many factors lead to strong randomness make direct prediction accuracy is not high,this thesis will through the VMD algorithm do decomposition with historical data of wind speed,all the signals of different frequency component is got,thus to set up prediction model of the components in the below the stage.(3)According to the component characteristics of different frequencies after VMD decomposition,a suitable model is selected to predict the wind speed.echo state network(ESN),because of its unique structure,reserve pool increases the network’s memory ability,and improves the efficiency of network training and fitting precision,a whale ESN parameters in the model and algorithm optimization and comparing echo state network model can be seen after the model had the very big enhancement in the prediction precision.(4)A combined prediction model based on VMD-ESN-WOA was proposed by combining the VMD-ESN-WOA model with the whale algorithm.After the wind speed sequence was decomposed into several components by VMD algorithm to reduce the coupling of wind speed internal signals,the ESN prediction model was established by using the wind speed sequence as input.Then,the four parameters in the ESN model were optimized and predicted by the whale algorithm for each component after decomposition.The final results were got by the summations of the estimates of every component and compared to other models,thus verifying the precision of the forecasting model proposed in this thesis.