Abstract: The accurate forecasting of provincial carbon emissions is pivotal for China as it strives to meet its carbon neutrality goals. To this end, an improved grey multivariable convolution model has been developed, employing a unified new-information-based method for the preliminary accumulation of data. The particle swarm optimization (PSO) algorithm is then applied to determine the optimal parameters within this sophisticated model. Moreover, to identify the relevant factors for provincial carbon emissions, a comprehensive determination of these factors was conducted from two aspects: literature research and grey relational analysis. For validation, carbon emission data from two provinces are analyzed, and the model’s efficacy is thoroughly compared with five competitors across three different predictive horizons. The empirical results indicate that the proposed model has distinct advantages over the competing models. Additionally, the model’s robustness and comprehensive forecasting abilities for provincial carbon emissions are confirmed through detailed Monte Carlo simulations and parameter sensitivity analyses across various forecasting horizons.  

Key words: Grey multivariable convolution model, Accumulating generation operators, Particle swarm optimization, Provincial carbon emissions forecasting