Analysis and Forecasting of CO₂ Emissions in the Power Sector Based on GRA and LSTM

The continuous increase in global carbon dioxide (CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>) emissions has triggered severe climate change in recent years. To effectively curb the rise in carbon dioxide emissions, it is essential to forecast the emissions from the power sector, the largest emitter, as well as examine the influencing factors of emissions. This paper proposes a two-stage forecasting model, where the first stage employs the Grey Relational Analysis (GRA) to identify the strongly correlated factors among the pre-selected 12 potential influencing factors. Subsequently, a Long Short-Term Memory (LSTM) model is developed to forecast carbon emissions in the power sector in Hong Kong. The results indicate that the accuracy of carbon emission predictions using LSTM surpasses that of the Backpropagation Neural Network (BPNN) and Polynomial Regression (PR), demonstrating the effectiveness of the proposed method in forecasting carbon emissions.