<p>Accurately identifying key factors influencing carbon emissions and forecasting future emission trajectories is crucial for policymaking and achieving sustainable development. While many existing studies focus on large-scale predictions at the national or regional level, there is a lack of attention to the city-industry scale. In this research, on the basis of the objective identification of critical influencing factors, a carbon emission prediction framework is presented. As a major energy and resource supplier, Shaanxi Province in China is adopted as a case study, and this research identifies the primary factors affecting carbon emissions in the nonferrous metals industry using socio-economic and sectoral energy consumption data via the Lasso regression model. Thus, a backpropagation neural network optimized by the particle swarm optimization algorithm (PSO-BPNN) is constructed to conduct scenario-based carbon emission forecasts for the period 2022–2035 under three development scenarios. The results indicate that six factors significantly influence carbon emissions. The PSO-BPNN model exceeds the traditional BPNN model in the area of prediction accuracy, achieving a coefficient of determination (<i>R</i>²) of 0.996, a mean absolute percentage error (<i>MAPE</i>) of 0.010%, a root mean square error (<i>RMSE</i>) of 2.397%, and a mean absolute error (<i>MAE</i>) of 2.022%. Scenario-based forecasts show that there will be a continuous rise of emissions under the baseline and high-carbon scenarios, while under the low-carbon scenario, it is projected that carbon emissions will peak in 2028 at 6.34&#xa0;million tons. These findings offer theoretical support for emission reduction policymaking in regions with huge energy consumption and carbon emissions.</p>