Machine Learning Carbon Emission Allowance Price Predictions
Bingzi Jin, Xiaojie Xu
Abstract
Accurately predicting the value of carbon credits stands as an indispensable task for policymakers aiming to craft potent strategies against climate change and implement robust market-driven regulatory schemes. Precise forecasts of these allowance prices empower crucial choices concerning carbon taxation policies, the design and adjustment of cap-and-trade mechanisms, and directing capital towards sustainable infrastructure projects, ultimately fostering a smoother transition to low-carbon economies. This research specifically tackles a notable void in current academic work by concentrating on the Shenzhen carbon emission allowance (SZCEA) exchange, a pioneering regional testbed within China’s broader emissions trading system initiatives. Utilizing the powerful pattern recognition and adaptive learning strengths inherent in artificial neural networks (ANNs), we construct and deploy an innovative predictive model targeting the daily settlement prices for SZCEA contracts. Our analysis spans a significant timeframe– from January 4, 2016 to August 19, 2021. This era witnessed considerable flux, characterized by evolving government regulations, progressive maturation of the market structure itself, and shifting behaviors among participants responding to these dynamics. To the best of our understanding, this investigation marks the inaugural comprehensive utilization of advanced neural network techniques applied to this distinct regional market. Consequently, it yields valuable new perspectives on the SZCEAs particular price discovery processes and the complex interplay of factors driving its unique market behavior, insights previously unexplored in depth.