To achieve accurate prediction of carbon emissions and trends, a GRA-PCA-Transformer prediction method is proposed. Twelve potential factors, such as demographic, economic, energy, and technological factors, are considered comprehensively, and the key influencing factors are screened using grey relation analysis (GRA). Four principal components are extracted by principal component analysis (PCA) to simplify the features, while a high-precision time series prediction model based on Transformer is constructed. The optimal sliding window size is determined by experimental analysis and compared with LSTM, GM(1,1), and ARIMA methods. The experimental results show that the method significantly outperforms other methods in prediction accuracy and trend-capturing ability. It provides a new idea for regional carbon emission prediction.