Energy Consumption Optimization in Cloud Data Centers Using Deep Reinforcement Learning and Workload Forecasting

Energy Consumption Optimization in Cloud Data Centers Using Deep Reinforcement Learning and Workload Forecasting

Zahra rohani1

1) Instructor, Computer Science Department, Shirvan Branch, Islamic Azad University, Shirvan, Iran

IOI Code : AAHD-ABKGHF
Publication : National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
    Cloud data centers, as the backbone of the modern digital economy, are facing unprecedented growth in energy consumption. The increasing use of cloud computing, big data, and AI-based applications has put significant pressure on energy infrastructure. This paper examines methods for optimizing energy consumption in cloud data centers using a combination of deep reinforcement learning and workload forecasting. First, the main challenges of energy consumption in data centers, including IT equipment, cooling systems, and workload distribution, are reviewed. Then, a deep reinforcement learning architecture is introduced for dynamic control of cooling systems and resource allocation. In the following, workload prediction methods based on deep neural networks, such as long short-term memory and attention networks, are presented for proactive energy management. The results of experimental studies show that the proposed hybrid approach can reduce energy consumption by up to 15% in the IT subsystem and up to 11% in the cooling system, while improving the power efficiency index to the level of 1.33. This paper also discusses the challenges of implementing these methods in real environments and future research directions.

Keywords : energy optimization cloud data centers deep reinforcement learning workload prediction power efficiency index