LAUSR

Multiple factors influence carbon emission factors, resulting in high-dimensional data characteristics. Simultaneously, real-world data often contain noise that complicates data processing. To address these challenges, a multi-constraint control algorithm incorporating nonlinear causal inference and knowledge graph technologies is developed. The methodology involves generating synthetic data through randomized sampling, calculating dynamic node-specific carbon emission factors, and incorporating elasticity coefficients to construct a carbon emission sensitivity analysis model. A comprehensive carbon emission assessment framework is established using Granger causality graphs, integrating multiple control constraints, including total emissions, reduction targets, energy consumption metrics, and stakeholder parameters. The experimental results demonstrate the designed method’s effectiveness, achieving carbon emissions below 60 000 tons with a margin of error between 2.69% and 2.98%, maintaining a coefficient of determination exceeding 0.9, and completing computations within 2.0 s. This provides reliable decision support for industrial enterprises to achieve total carbon emission control and emission reduction targets, helping high energy consuming industries promote green and low-carbon development while ensuring economic benefits.