Using Artificial Intelligence to Improve Energy Efficiency in Libyan Data Centers: An XGBoost-Based Approach

Abstract

The expanded proliferation of virtual services in Libya has brought about a marked escalation in statistics middle electricity consumption, compounded with the aid of pervasive power outages and inadequate cooling infrastructure. The proposed method makes use of an XGBoost-based totally method to decorate energy performance, leveraging 18 months of operational records from 3 statistics facilities in Tripoli, Benghazi, and Misrata. The version attained ninety-four.7% forecasting accuracy for short-term electricity call for (MAE = 3.2 kW, RMSE = four.8 kW), thereby facilitating proactive cooling manage and workload control. The proposed machine verified a 32.6% improvement in the common PUE, reducing it from 2.18 to at least one.47, and it achieved approximately 28% power price financial savings. Cross-validation and impartial trying out confirmed the machine's sturdy overall performance beneath various conditions, providing a practical framework for sustainable statistics middle operations in environments with restrained sources.