Abstract: It is of significant importance to scientifically predict the peak carbon emissions and the timing of the peak under different pathways and propose tailored strategies for the planning and deployment of carbon emission reduction policies in large cities. This study employed an expanded Stochastic Impacts by Regression on Population, Affluence, and Technology (STIRPAT) model, integrating the ridge regression and scenario analysis methods, to systematically investigate the evolution trends of carbon emissions in Guangzhou City from 1997 to 2035. By integrating core driving factors such as population, per capita GDP, and carbon emission intensity, it constructed three scenarios (Baseline Scenario, Low-Carbon Scenario, and Enhanced Low-Carbon Scenario) to quantify the carbon emission trajectories under varying intensities of policy intervention. The findings revealed that under the Baseline Scenario, Guangzhou's carbon emissions would not peak by 2035. However, under the Low-Carbon and Enhanced Low-Carbon Scenarios, the peak was reached in 2030 and 2025, respectively. Building on the results of the dynamic scenario simulations, this study investigated the unique "carbon lock-in" effect in Guangzhou. Based on the dynamic simulation results and the city's regional characteristics, the study formulated differentiated, multi-dimensional emission reduction strategies including policy, industrial technology, population, and infrastructure.