Abstract: Carbon emission projections are pivotal in addressing global climate change and advancing green, low-carbon development. Accurate forecasts of China's carbon emissions provide critical insights for policymakers to understand future emission trends and potential peak levels, thereby enabling the formulation of scientifically grounded and practical mitigation strategies. Discretization has proven to be an effective approach for enhancing the accuracy of grey prediction models. To further refine the performance of discrete grey prediction models, this study integrates integer-order polynomials with time fractional power terms to develop an optimized discrete grey prediction model, DGMPT(1,1,N, ), the particle swarm optimization (PSO) algorithm is utilized to optimize hyperparameters. To validate the model's superiority and predictive accuracy, this study applies it to the carbon emission data of Xinjiang, Shaanxi, and Gansu provinces in China. Empirical results demonstrate that, based on the mean absolute percentage error (MAPE) criterion, the proposed model achieves the lowest MAPE values in both the training and prediction datasets across all three case studies. Finally, the proposed model is employed to forecast China's carbon emissions over the next decade. The results indicate that under current conditions, China is unlikely to achieve its peak carbon emissions by 2030. This underscores the urgency of implementing effective and comprehensive policy measures to improve carbon emission systems and foster a sustainable emission environment. 

Key words: Carbon emissions, Discrete grey model, Hyperparameter, The PSO algorithm