A Hybrid Deep Learning Model for Energy Consumption Forecasting in Cloud Computing Environments

Abstract

The exponential growth of cloud computing has led to a substantial increase in energy consumption across data centers, posing challenges in terms of operational cost, sustainability, and environmental impact. Accurate energy consumption forecasting is therefore crucial for optimizing resource allocation and improving the energy efficiency of cloud infrastructures. This study presents a hybrid deep learning framework that integrates a Convolutional Neural Network (CNN) with Bidirectional Gated Cycle Units (Bi-GCUs) to forecast energy consumption in cloud computing environments. The CNN component efficiently captures spatial dependencies and workload-related patterns, while the Bi-GCU layer models temporal correlations and bidirectional dependencies in time-series energy data. Experimental evaluations conducted on real-world cloud datasets demonstrate that the proposed CNN–BiGCU model outperforms traditional forecasting techniques and standard deep learning architectures in terms of prediction accuracy, convergence speed, and stability. The results confirm that the hybrid model effectively reduces prediction errors and enhances adaptive energy management strategies. This work contributes to the advancement of intelligent, sustainable, and energy-aware cloud systems through the integration of explainable and efficient deep learning techniques.