Research On Price Dynamic Prediction Of Carbon Emission Trading Market In China Based On LSTM Neutral NETwork

To effectively address global warming,countries around the world,including China,actively explore solutions and establish trading to reduce carbon emission.In2020,the commitment of "reaching the peak by 2030 and achieving carbon neutrality by 2060" brought new opportunities for Chinese carbon emission trading market.At present,Chinese carbon market is still at an early stage,and the carbon price has sharp fluctuations,which affects the smooth operation of the market.Because the prediction method of carbon price is still relatively traditional,the guidance on the dynamic prediction of carbon price at the operational level is obviously insufficient,and further research is needed.This research adopted the method of combining theory with empirical analysis.At First,this paper found that it is difficult for scholars to comprehensively cover the influencing factors of carbon price through experience and data statistics,therefore selected the sequential data of carbon price as the research object,and used LSTM neural network to predict carbon price to overcome this problem.These effective carbon price prediction results would help carbon market players avoid extreme investment behavior and maintain market stability.In terms of theoretical analysis,this paper summarized the existing research results.This paper systematically introduced the carbon emission trading market and carbon emission trading price at home and abroad,including the research situation of carbon market mechanism,efficiency,risk and influencing factors of carbon price,and summarizes the management and research status of carbon trading at home and abroad.This paper also summarized the relevant theories of carbon market,systematically reviews the construction process of carbon market in China,focusing on the direction and basic way of five factors influencing carbon trading price:macroeconomic,energy price,carbon price in the European Union,meteorology and climate,and COVID-19 epidemic.In terms of empirical analysis,this paper used literature review method to make theoretical analysis and technical discussion on time series prediction,and proposed a prediction method based on LSTM neural network.Then,this paper used computer software for deep learning simulation experiments,including model construction,parameter selection and model training.This study selected carbon emissions trading price time-series data as the research object——taking trading price of Guangdong Carbon Emission Allowance(GDEA)as an example—— built the LSTM neural network time series prediction model and conducted simulation prediction experiments.This simulation discussed the technical feasibility of different price forecasting methods,such as forecast-based or observation-based price forecasting methods.On the basis of studying the technical feasibility and comparing the prediction performance of different methods,three kinds of wavelet transform noise reduction technology,hard threshold,soft threshold and fixed threshold,were used to improve the prediction effect of LSTM neural network operation.