This study first calculates China's CO₂ emissions from 2000 to 2022 using an emission source classification approach. Second, it constructs an indicator system for CO₂ emission influencing factors, covering five dimensions: economy, society, environment, energy, and technology. Lasso regression is then applied to identify key drivers of carbon emission changes. Based on this, the research represents pioneering work to apply the difference creative search (DCS) algorithm and kernel density estimation (KDE) to the CNN-BiLSTM-Attention model, and thereby construct the DCS-CNN-BiLSTMAttention- KDE interval prediction model. The prediction results are then compared with those of other point and interval prediction models. Finally, carbon emissions in China from 2023 to 2035 are predicted using the scenario analysis method. The research results indicate that: (1) MAE, RMSE, MAPE, PICP, and PINAW of the DCS-CNN-BiLSTM-Attention-KDE are 10063.0250, 13100.8491, 0.0092, 0.9021, and 0.2817, respectively. In comparison to other models, the proposed model exhibits higher prediction accuracy. (2) The carbon peaking times for the baseline scenario, the green and low-CO₂ scenario, and the strong low-CO₂ scenario are projected to be 2033, 2030, and 2027, respectively. Under the green and low-CO₂ scenarios, as well as the strong low-carbon scenario, China is projected to achieve its carbon peaking target on schedule.