Carbon Emission Forecasting in Electric Power Sector in Shanxi Province Based on ARDL Model and Machine Learning Algorithm

With the nation's introduction of the "double carbon" objective, given that the power industry's total carbon emissions constitute a major part of these emissions, precise forecasting of these carbon emissions becomes crucial. Consequently, this document develops a combined model utilizing the autoregressive distributed lag model (ARDL) and machine learning algorithm (Transformer) for forecasting Shanxi's power sector's carbon emissions. Shanxi, a significant coal resource province, boasts a precise carbon emission measurement technique, making it an ideal location for gathering the dataset. Initially, data pertaining to Shanxi's electric power sector over the last decade undergoes preprocessing, followed by normalization of yearly electricity production figures. Subsequently, these figures are fed into the time series ARDL model in a suitable manner to examine the long-term and short-term correlations between electricity production and carbon emissions. Ultimately, the residuals are processed by entering the data into the ARDL and Transformer models, leading to the creation of the fusion model. model of fusion. The study demonstrates the hybrid model's extensive benefits in merging predictive precision, capturing long-term trends, and theoretical clarity, making it ideal for policy simulations and intricate energy structure analyses. Consequently, this paper's methodology introduces an innovative approach for forecasting carbon emissions in other coal-dependent provinces, serving as a benchmark for future prediction techniques.