Multi-Source Deep Learning for TSO-Level Electricity Demand Forecasting

Thomas, Jismol (2025) Multi-Source Deep Learning for TSO-Level Electricity Demand Forecasting. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (2MB) | Preview

Abstract

Proper short-term electricity demand forecasting is critical for the maintenance of secure and cost-effective grid operation at the Transmission System Operator (TSO) level. This paper compares the performance of three deep learning models, namely, Long Short-Term Memory (LSTM), the Temporal Fusion Transformer (TFT), and a hybrid TFT-LSTM residual model, on a multi-source dataset, which combines SMARD consumption and generation data, Open-Meteo weather variables, Eurostat socioeconomic indicators, and calendar features. The models are evaluated based on the point-forecast, probabilistic, and operational measures. The findings indicate that the TFT that has been optimized with Optuna scores the best in total point-forecast accuracy, whereas the hybrid TFT-LSTM Residual model has better peak-load accuracy by fixing systematic underestimation of the TFT results. Even though the transformer-based models tend to under-calibrate uncertainty intervals, the tuned TFT and hybrid architectures are highly effective in terms of early warning. These results indicate the power of transformer-based and hybrid deep learning techniques to forecast the peaks and multi-source TSO-level using various sources.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Nagahamulla, Harshani UNSPECIFIED
Subjects:	T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Electricity Supply 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:	09 Sep 2026 10:35
Last Modified:	09 Sep 2026 10:35
URI:	https://norma.ncirl.ie/id/eprint/9922

Actions (login required)

View Item