Energy-Efficient Cloud Usage Forecasting Using SEGRU for Power and Carbon Emission Reduction

Raj, Rishabh (2025) Energy-Efficient Cloud Usage Forecasting Using SEGRU for Power and Carbon Emission Reduction. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

Cloud data centers face growing energy consumption and carbon emissions due to quickly fluctuating CPU workloads by making accurate forecasting which is important for sustainable and cost-efficient cloud operations. There are some existing deep learning approaches like LSTM, BiLSTM and GRU provide reasonable predictions but struggle with sudden workload spikes, higher latency and insufficient linkage which is in between CPU forecasting and subsequent power and carbon estimation. This study addresses these limitations by proposing a Squeeze-and-Excitation GRU (SE-GRU) model to improve workload sensitivity and forecasting accuracy. Also the trained SE-GRU model was deployed on an AWS EC2 instance to validate real-time inference capability. Synthetic workload data has generated every five minutes to simulate live CPU behavior and the system produced forecasts. When high utilization was predicted then an automated SNS notification was triggered which shows its potential for future integration with AWS CloudWatch metrics. There are some experimental evaluations which shows that SE-GRU outperforms all baseline models by achieving the lowest MSE (9.48e-06), highest R² (0.980), reduced latency (61.03 ms) and improved throughput. It also gives the lowest estimated power and carbon outputs by showing stronger environmental performance. The findings advance current state-of-the-art CPU forecasting by using attention-enhanced lightweight modeling with sustainability analysis.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Mijumbi, Rashid UNSPECIFIED
Uncontrolled Keywords:	CPU forecasting; SE-GRU model; cloud energy; power; carbon prediction
Subjects:	T Technology > T Technology (General) > Information Technology > Cloud computing H Social Sciences > HC Economic History and Conditions > Natural resources > Power resources > Energy consumption
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	01 Sep 2026 10:34
Last Modified:	01 Sep 2026 10:34
URI:	https://norma.ncirl.ie/id/eprint/9731

Actions (login required)

View Item