Supervisor
Dr Maqsood Shah
Programme
MSc in Data Analytics
Subject
Computer Science
Abstract
Accurate forecasting of electricity demand is critical for reliable energy planning, resource allocation, and policy design. Traditional statistical models, such as ARIMA, SARIMA, and ARIMAX, have been widely applied but remain constrained by linear assumptions, limited temporal memory, and inflexibility in handling multiple exogenous drivers. In this study, we systematically compare these approaches with multivariate Long Short-Term Memory (LSTM) networks, which are capable of capturing nonlinear dependencies, long-term temporal dynamics, and multivariate interactions. Historical electricity consumption data were combined with weather variables, including temperature, wind speed, and rainfall, and pre-processed through cleaning, scaling, and temporal alignment. Statistical baselines and deep learning models were developed under consistent training–testing configurations, with post-hoc bias adjustment applied to all models to improve calibration. Model evaluation, based on RMSE, MAE, and R², demonstrated the superior performance of LSTM models: the best-performing variant (multivariate LSTM without rainfall and with bias adjustment) achieved R² ≈ 0.79, compared to ≈0.64 for ARIMAX and < 0.20 for SARIMA. Results further revealed that not all exogenous variables contribute equally; while temperature and wind provided strong predictive signals, rainfall added little value at the monthly scale. These findings highlight the limitations of classical models, the potential of bias adjustment as a calibration tool, and the robustness of LSTMs as a step-change in electricity demand forecasting. Beyond empirical performance, the study underscores the importance of feature selection, methodological rigor, and the balance between accuracy, interpretability, and computational cost in advancing forecasting frameworks.
Date of Award
2025
Full Publication Date
2025
Access Rights
open access
Document Type
Capstone Project
Resource Type
thesis
Recommended Citation
Sant’Ana, L.
(2025) Traditional vs Deep Learning Approaches for Efficient Electricity Consumption Prediction. CCT College Dublin.
DOI: https://doi.org/10.63227/652.299.119