Energy Consumption Forecasting in Cloud Computing Using CNN with Bidirectional Gated Cycle Units
Keywords:
Energy Consumption Forecasting, Cloud Computing, Convolutional Neural Network (CNN), Bidirectional Gated Cycle Units (BGCU),, Spatiotemporal Modeling, Data Center Efficiency, Predictive Analytics, Energy Optimization.Abstract
The surge in cloud computing services has led to increasing energy consumption in data centers, driving the need for intelligent energy forecasting and optimization mechanisms. This paper proposes a novel deep learning model integrating Convolutional Neural Networks (CNN) with Bidirectional Gated Cycle Units (BGCUs) to accurately forecast energy consumption patterns in cloud environments. The model leverages CNN layers to extract spatial features from heterogeneous cloud resource metrics and employs BGCUs to capture bidirectional temporal dependencies. Experimental results demonstrate that the proposed CNN-BGCU architecture outperforms traditional methods such as LSTM and GRU in terms of prediction accuracy, achieving an average reduction of 15% in mean absolute error (MAE) and 18% in root mean square error (RMSE). These findings highlight its potential for enabling proactive resource scheduling and reducing energy wastage in modern cloud infrastructures.
