AI-Driven Energy Prediction Models for Sustainable Cloud Infrastructure: A Comprehensive Survey and Unified Framework

Abstract

The Rapid rise of AI and cloud services are increasing in the modern data centers, which consume a large amount of energy consumption. There are certain key concerns like Sustainability, operational cost and environmental impact. However, many AI-based approach had been implemented to maintain workload prediction, Resource optimization and renewable energy integration in cloud infrastructure. In existing studies, there is a lack of unified sustainability-oriented framework. This research does a Comprehensive survey which includes AI-based energy prediction model for unified sustainability, that includes techniques like machine learning, Deep learning and reinforcement learning that does predictions such as workload forecasting, intelligent scheduling, predictive maintenance and energy-aware resource management. while doing this study we found some problems in current cloud systems, which includes limited cross-layer coordination, lack of carbon-aware scheduling, data heterogeneity, explainability issues and minimal real-world validation. To overcome these issues we can use a unified multi-layer sustainable cloud framework, this is nothing but building one complete system that connects all layers of the cloud together that is data acquisition, AI-driven prediction, optimization and sustainability monitoring. A Modified and simpler mathematical model is introduced to reduce the maximum energy consumption along with satisfying Quality of Service (QoS), carbon intensity, and renewable usage constraints. This proposed models approach is to provide a better point of view for building intelligent, energy-efficient and environmentally applicable cloud infrastructures that ensures the sustainability as well as supporting the future research in green computing and sustainable digital ecosystems.