An Ensemble Prediction System Based on Artificial Neural Networks and Deep Learning Methods for Deterministic and Probabilistic Carbon Price Forecasting

Carbon price prediction is an important work to decrease greenhouse gas emissions and cope with climate change. At present, many models are widely used to predict irregular, nonlinear and nonstationary carbon price series. However, they ignore the importance of the feature extraction and the inherent defects of the single model, so that the accurate and stable prediction of carbon prices by the relevant industry practitioners and the government is still a huge challenge. Therefore, this research proposes an ensemble prediction system (EPS), which includes improved data feature extraction technology, three prediction sub models (GBiLSTM, CNN and ELM), and multi-objective optimization algorithm weighting strategy. At the same time, based on the best fitting distribution of prediction error of the EPS, the carbon price prediction interval is constructed to explore its uncertainty. More specifically, EPS integrates the advantages of sub models and provides more accurate point prediction results; the distribution function based on point prediction error is used to establish the prediction interval of carbon price, and to mine and analyze the volatility characteristics of carbon price. In addition, numerical simulation experiments are conducted on the historical data of three carbon price markets. The experimental results present that the ensemble prediction system can provide more effective and stable carbon price forecasting information, and can provide more valuable suggestions for enterprise managers and governments to improve the carbon price market.