Carbon Market Price Forecast Analysis And Risk Management Research

Carbon market is a cost-effective way to deal with climate change. The Twelfth Five-Year Guideline of China proposed "gradually establishing carbon market" and launched regional carbon market plot project in Beijing, Shanghai, Guangdong and other regions, and will continue to popularize to build national carbon market progressively, which will play an important role in energy conservation strategy. It is, therefore, an essential task that how to design a strong and stable carbon market in our country. In recent years, as the representative of European Union Emission Trading Scheme (EU ETS), global carbon market has influences on reduction performance significantly with its dramatic price volatility. Aiming at the carbon price of EU ETS, this dissertation explores prediction analysis, risk measure and dynamic behavior, it’s contributions are as follows:1. Aiming at carbon price multiscale feature, we propose carbon price multiscale decomposition using an improved Hilbert-Huang transform (HHT) algorithm. Firstly, the ensemble empirical mode decomposition (EEMD) algorithm is built by introducing the Gaussian white noises into the EMD, which is applied to resolve the mode-mixing phenomena during EMD. Next, for the end effect phenomena, the extension method which is appropriate to deal with carbon price is obtained by comparing five different kinds of extension methods for EEMD. Finally, taking two carbon future prices with different maturities called DEC 12 and DEC 14 under EU ETS as samples, empirical results show that improved HHT algorithm can significantly enhance decomposition accuracy, and the application scope of HHT is extended.2. The improved EEMD, statistical analysis, Hilbert spectrum analysis and Bai-Perron structural breakpoints test are adopted to identify the different scales effects of an extreme event on carbon price and the interaction influences between different scales. Empirical results show that both global financial crisis in 2008 and European debt crisis in 2011 made significant effects on carbon price, and caused structural breakpoints in the carbon price.3. Making full use of advantages of self-adaptive data decomposition and artificial intelligence models, we present an integrated model based on improved EEMD and Support Vector Machine (SVM) for carbon price turning point prediction. Firstly, the improved EEMD is applied to decompose carbon price series which have been removed seasonality. Meanwhile, the sequence reconstruction method is adopted to identify circulating factors, which is then used to distinguish turning point of carbon price by the Bry-Boschan (BB) method. Then the sequence reconstruction method is used to make the selection of input-layer units easily.The grid search method based on K-fold Cross Validation (K-CV) theory is used to train SVM model with the training samples and obtain the optional parameters, which is then used to forecast turning point of testing samples. Empirical results show that the proposed model is an effective way to forecast turning point.4. As the premise of assumption that the heterogeneity of the carbon market, we put forward carbon price multiscale risk measure based on improved EEMD and conditional value at risk (CVaR). Firstly, the improved EEMD algorithm is applied to decompose carbon price return series. Secondly, both the different scales of intrinsic mode function (IMF) and residuals are modeled respectively by GARCH models. Then, the iterated cumulative sums of squares (ICSS) method is applied to recognize the extreme event window,in which each scale component are modeled respectively by exponentially weighted moving average (EWMA). Lastly, we measured the time-varying conditional variance of carbon price return series based on multiscale GARCH and EWMA, which improved variance estimation using traditional CVaR. Empirical results show that the proposed multiscale CVaR measure is obviously better than the traditional variance covariance method’s.5. In view of the Zipf model, we map the τ-returns of carbon price into relative frequencies and absolute frequencies, and analyses different τ (speculator’s time scale of investment) and ε (speculator’s expectation of returns) affecting on fluctuation behavior and formation mechanism of carbon price. The results show that:first, carbon price behavior is asymmetric, and the long-term bearish probability is greater than the long-term bullish probability. Second, time-scales of investment and speculators’ expectations of returns have dual effects on carbon price behavior. Lastly, speculators’expectations of returns have critical points. Once the critical points are reached, they will no longer be able to distort carbon price behavior.