Green AI: Enhancing Sustainability and Energy Efficiency in AI-Integrated Enterprise Systems

Abstract

The rise of Artificial Intelligence (AI) in automating tasks and driving decision-making within enterprise systems has led to growing concerns over the significant energy consumption involved in model training and inference processes. This paper introduces an innovative framework focused on optimizing energy efficiency in AI models, all while preserving high performance. The system employs advanced optimization algorithms aimed at minimizing energy usage during both AI training and inference, ensuring minimal impact on model accuracy. A dynamic, multi-objective optimization approach is used to achieve an optimal balance between energy reduction and performance, identifying Pareto-optimal solutions tailored to various operational needs. Validated within large-scale enterprise settings, the system delivers a 30.6% decrease in overall energy consumption, with only a slight 0.7% reduction in model accuracy. Furthermore, scalability is demonstrated through a 5.0% improvement in task execution time and a 4.8% increase in system throughput. The findings highlight the practicality of this framework for promoting sustainable AI deployment, aiding both cost efficiency and environmental responsibility. The paper concludes by discussing limitations and outlining potential avenues for future research to further enhance scalability and broaden the framework’s application.

Keywords

• Energy-efficient AI
• multi-objective optimization
• enterprise systems
• AI model optimization
• sustainability
• scalability