Research On Complexity Of Carbon Market And Chaotic Forecasting Of Carbon Prices In EU ETS

In recent years, the problem of global warming has gradually aroused wide attention of many countries. The international community has made much efforts. The carbon emissions trading mechanism is one of the effective ways to deal with climate change. Hence the carbon market and its financial feature has gradually become the research focus in the field of energy economy. As effected by international political environment, economic crisis and other factors, carbon prices present typical features of complex systems, such as dynamic, nonlinearity and uncertainty. For the research of carbon market and its prices, it not only can provide a certain reference value for the long-term development of the carbon market, market traders and participants, but also give some important reference for the construction of Chinese carbon futures market.Based on the research background of complex system in EU ETS, this paper explores the complex mechanism of carbon market, the nonlinear dynamic behavior and neural network prediction model of carbon prices. By adopting the mathematical technique of finance, econometrics and statistics et al., we analyze qualitatively the impact factors of the complexity mechanism of the carbon market, describe the nonlinear dynamic behavior of carbon prices and construct a multi-layer neural network model. This study reveals complex behavior of carbon market and the chaotic characteristics of carbon prices. Besides, the significant test and generalization ability are carried out for the neural network model.The study mainly includes three aspects as follows:(1) Analysis of complex mechanism of carbon market. Based on the multiscale entropy method, we research the complexity behaviors of carbon market from international political environment, financial crisis and economic development. The results show that the entropy is higher in 1 to 10 time scales and lower in larger than one month time scales. The dependence of entropy on to time scale reveals that the daily price returns of carbon prices follow the mean reversion in the long run.(2) Chaotic characteristics identification of carbon prices. Based on phase space reconstruction technology, we adopt the largest Lyapunov exponent、the correlation dimension and the Kolmogorov entropy to study the chaotic characteristics of carbon prices from the point view of nonlinear dynamic. The results show that the carbon price time series has a positive maximum Lyapunov exponent, a non-zero Kolmogorov entropy and the correlation dimension increases with the embedded dimension。Hence we judge that carbon price time series is chaotic.(3) Construction of multi-layer neural network model. By selecting single hidden layer artificial neural network and changing the hidden layer number and adopting four statistical measurement monitoring the advantages and disadvantages of different models, we choose the optimal 3-7-3 artificial neural network model and carry out the significance test and generalization ability for the constructed model. Results show that the selected optimal 3-7-3 neural network model has higher prediction accuracy for carbon price.