Attention-Aware Deep Learning Framework for Smart Home Energy Forecasting

Elgamoni, Harishwar Yadav (2025) Attention-Aware Deep Learning Framework for Smart Home Energy Forecasting. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (616kB) | Preview

Abstract

Accurate energy consumption forecasting is essential for smart home automation and sustainable energy management. This study introduces an attention-aware transformer-based deep learning framework capable of modeling both multivariate time-series data and structured tabular data for appliance-level energy prediction. Leveraging the self-attention mechanism, the model captures complex temporal dependencies and contextual variations—such as temperature, humidity, seasonality, and appliance usage patterns—without relying on recurrent structures. Evaluations were conducted on two benchmark datasets: the SmartHome Environmental Time-Series Energy Dataset and the Appliance-Level Household Energy Usage Dataset. The model achieved strong results across both, with an MSE of 2.19 (R² = 0.37) for time-series data and an MSE of 0.36 (R² = 0.74) for tabular data. Compared to traditional machine learning models and dense neural networks, the transformer-based model outperformed across key metrics, demonstrating superior accuracy, generalization, and training stability. This work highlights the transformer model's versatility and effectiveness in energy forecasting tasks and underscores its potential as a scalable and intelligent solution for future smart energy systems.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Kumar, Teerath UNSPECIFIED
Subjects:	H Social Sciences > HC Economic History and Conditions > Natural resources > Power resources > Energy consumption Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	25 Aug 2026 13:45
Last Modified:	25 Aug 2026 13:45
URI:	https://norma.ncirl.ie/id/eprint/9628

Actions (login required)

View Item