MANAGING CARBON FOOTPRINT IN GLOBAL SUPPLY CHAINS: A DATA-DRIVEN STRATEGY EMPLOYING AI ALGORITHMS

Abstract

The global supply chain is a characteristic of the new economy and has also greatly influenced the production patterns, consumption trends as well as global trade. Despite the fact that globalization has assisted in cost reduction, optimization of resource usage as well as diversification of sourcing, it has resulted in massive increase in carbon emission. The manufacturing business associated with logistics, transportation, and supply chain involves a significant percentage of global CO 2 emissions; therefore, the carbon footprint management is an urgent business issue, but the traditional methods of carbon accounting, based on historical averages, constant emission factors, manual calculations, and linear models are not sufficient to reflect the dynamic and uncertain nature of the modern supply chain. All of this has been exacerbated by such issues as different fuel efficiency, multimodal routes, real time traffic congestion, different energy sources, climate change, and geopolitical events, which have led to artificial intelligence as a strong technology to manage carbon with the data. Artificial intelligence can more effectively predict emissions and make decisions because it can examine large volumes of various data and detect nonlinear relationships between them. Transformer model, in this regard, has proven to be quite impressive in terms of learning long-term dependencies and high-dimensional time-series data. This renders the Transformer model to be applicable in the context of emission prediction in the global supply chain where the information is obtained through such sources as IoT sensors, GPS, enterprise, transport logs, and energy utilization. Nonetheless, accurate forecasting is not enough to achieve the 6 sustainability goal. Such is the nature of the supply chain problem, where there exists a trade-off between cost and environmental aspects. Conventional optimization problems, like the single objective linear programming, are not useful in the resolution of such conflicting goals. In that regard, this research work integrates multi-objective optimization methods, i.e. Non-Dominated Sorting Genetic Algorithm II (NSGA-II) and Multi-Objective Particle Swarm Optimization (MOPSO), to come up with Pareto-optimal solutions that can, at the same time, take into consideration cost effectiveness and carbon emissions, and meet service level restrictions. The paper has a hybrid AI solution consisting of Transformer-based carbon emission forecasting and multi objective optimization of sustainable supply chain management. The core findings of the research work are the enhanced accuracy of carbon emission forecast, the flexibility of solutions of cost-carbon trade-offs and the integrated decision support system which enhances the sustainability and transparency of long-term planning. To conclude, the suggested solution is a practical way forward in building smarter, more resilient and sustainable global supply chain in the light of the increasing regulatory and sustainability demands.