Publications

Advanced data analytics modeling for evidence-based data center energy management

Published in Journal - Physica A: Statistical Mechanics and its Applications, 2023

This paper presents AI based modeling strategies for effective energy management with a particular emphasis on DC’s two most energy intensive systems (i.e., cooling and IT systems). This study addresses the issues of IT equipment performance degradation, inappropriate IT room thermal conditions, inefficient workload placement, and excessive energy waste. This research entails the application of machine learning for DC thermal classification, and deployment of deep learning models to predict resource utilization and energy consumption in DC.

Download Paper

A Comparative Analysis of Machine Learning Models for a Scalable Electrical Model-Free Voltage Calculation Methodology

Published in : Ready for submission in a conference, 2025

This paper presents a methodology to capture the underlying relationships among historical smart meter data (P, Q, and V) and the corresponding low-voltage (LV) network. The research include a comprehensive analysis and comparison of three machine learning (ML) models: Random Forest Regression, XGBoost, and Artificial Neural Networks (ANN).

OPTAAS for Computational Efficiency in Energy Applications

Published in : Ready for submission in a conference, 2025

The proliferating energy data generation poses significant challenges across various energy applications. This study focuses on one such energy application: cost-optimization of single households. Households with limited computing capacity need an online decision-making process to efficiently utilize their renewable resources and effectively manage their energy data. This paper presents OPTAAS (Optimization as a Service), a framework designed to implement online decision-making for households to resolve optimization problems efficiently and in a timely manner.

Exploratory data analysis for data center energy management

Published in Proceedings of the Thirteenth ACM International Conference on Future Energy Systems, 2022

Generated heat in the data center is categorized into different granularity levels namely: server level, rack level, room level, and data center level. Several datasets are collected at ENEA Portici Data Center from CRESCO 6 cluster - a High-Performance Computing Cluster. This research aims to conduct a rigorous exploratory data analysis on each dataset separately and collectively followed in various stages. This work presents descriptive and inferential analyses for feature selection and extraction process. Furthermore, a supervised Machine learning modelling and correlation estimation is performed on all the datasets to abstract relevant features, that would have an impact on energy efficiency in data centers.

Download Paper