Chaotic Time Series Prediction Based On Variational Mode Decomposition And Neural Network

With the intensive study of nature,more and more time series are analyzed to have chaotic characteristics,such as precipitation,temperature and sunspots time series.Analysis of chaos and prediction of time series can reveal the essence of things,discover its internal laws,and the value of these researches has a profound impact on the progress of human society.Variational mode decomposition(VMD)is a new adaptive signal decomposition method,which is widely used in chaotic signal processing.In this paper,VMD,BP neural network,extreme learning machine and intelligent optimization algorithms are used to make hybrid prediction models.Taking precipitation,temperature and monthly mean sunspots as the research object,this paper explores the advantages of these hybrid models in chaotic time series prediction.The main work and innovations are as follows:(1)Chaotic theory is used to analyze the chaotic characteristics of precipitation,temperature and sunspots time series.The phase space reconstruction parameters,maximum Lyapunov exponent and Kolmogorov entropy are calculated.The simulation results show that these three types of meteorological time series have chaotic characteristics and can be used for short-term chaotic time series prediction.(2)A hybrid prediction model based on VMD and extreme learning machine is employed to forecast the monthly precipitation chaotic time series.VMD can effectively suppress the modal aliasing phenomenon.Extreme learning machine has a fast training speed and simple setting parameters.Compared with various models,the experimental results show that the model can predict the precipitation trend and improve the prediction accuracy.(3)A hybrid prediction model based on improved extreme learning machine and VMD is proposed to forecast temperature chaotic time series.Particle swarm optimization algorithm is used to select the optimal weights of extreme learning machine to improve the prediction performance of network.The experimental results show that the improved model can predict the temperature effectively and further improve the prediction accuracy.(4)A novel hybrid prediction model based on BP neural network optimized by firefly algorithm and VMD is proposed,and applied to the prediction of monthly mean sunspots.Using the firefly algorithm,the optimal weight of BP neural network can be found quickly,and the accuracy of the prediction model is effectively improved.The experimental results show that compared with the unoptimized neural network,the proposed model has a certain improvement in prediction accuracy,and the prediction effect is better.