Carbon Price Forecasting Based On Combination Of VMD And ELM

The carbon price fluctuations will affect the production and carbon trading market steady development,the key problems of risk management and the carbon price volatility rely on the carbon market participants,it can provide scientific decision-making tools for carbon emissions trading investors,guide investors to make better use of the carbon trading market for investment,promote the rational development of the carbon market.The international carbon price presents non-stationary,multi-frequency,non-linear and other irregular characteristics.The traditional single model is difficult to describe the characteristics of the volatility of carbon price.The multi-frequency prediction model can dig into the various inherent laws of carbon price at different frequencies,so as to better grasp the law of the fluctuation of carbon price.Therefore,this paper adopts the variational mode decomposition method(VMD)to decompose the European Union's carbon emission quota(EUA)spot price into different frequency components which have their own characteristic components.In view of the existing research on the use of BP neural network and SVR model to predict each frequency sequence of carbon price,it will lead to prediction accuracy problems and slow running speed problems.In this paper,we use the algorithm of extreme learning machine(ELM),which has good fitting effect and high computational efficiency,to predict each frequency sequence of carbon price,so as to get the final prediction result of carbon price.At the same time,it is considered that the number of modes obtained by the decomposition of VMD algorithm and the number of hidden layer nodes of ELM have a great influence on the final prediction accuracy.It is first put forward that the optimal combination of modal number and ELM hidden layer node number is determined by establishing the three-dimensional relationship among modal number,hidden layer number and average absolute error for the first time,so as to get more accurate prediction results of carbon price.In order to verify the prediction effect of the hybrid prediction model proposed in this paper,Empirical mode decomposition(EMD),as a widely used decomposition model,is also applied in this paper to compare with VMD.The popular prediction models such as GM(1,1),ARIMA,BP,SVR and ELM are compared with ELM model.Research shows:(1)The mixed VMD-ELM method is used to predict the carbon price,and the best prediction result is obtained,which solves the problem that single model is difficult to fully characterize the volatility of carbon price.Meanwhile,the calculation speed is greatly improved.(2)Introducing the corresponding ELM to predict each frequency sequence of the carbon price,and it solves the prediction performance problems of BP neutral network and SVR model.(3)Using VMD to decompose the carbon price effectively solves the modal aliasing problem of the EMD model,which makes the change law and the hierarchical characteristics of the carbon price clearer.(4)Compared with the statistical models such as GM(1,1)and ARIMA,BP,SVR,ELM and other intelligent algorithms can better grasp the trend of carbon price and solve the problem of non-linear prediction more effectively.The research contribution of this paper is that it is of great practical significance and application value to expand the theoretical research on the price prediction method of international carbon market.