View in EDS

A new deep learning model for short term electricity load forecasting using fuzzy granulation and attention mechanism to support clean energy systems.

Saved in:
Bibliographic Details
Title: A new deep learning model for short term electricity load forecasting using fuzzy granulation and attention mechanism to support clean energy systems.
Authors: Kaptan, İlayda1 (AUTHOR) ikaptan@atu.edu.tr, Aksu, İnayet Özge2 (AUTHOR)
Source: Advances in Mechanical Engineering (Sage Publications Inc.). Mar2026, Vol. 18 Issue 3, p1-22. 22p.
Subjects: Load forecasting (Electric power systems), Granular computing, Deep learning, Machine learning, Global optimization, Forecasting, Clean energy
Geographic Terms: Turkey
Abstract: Providing accurate electricity load forecasts is crucial for ensuring a cleaner and more sustainable environment, as these forecasts enable the balancing of energy supply and demand, reduce unnecessary reserve capacity requirements and lower dependence on fossil fuels. Objective of this study is to improve the limited performance of existing electricity load forecast models using time series data and to make more accurate forecasts. In this study, the Fuzzy Information Granulation method was applied to Türkiye's actual hourly electricity load data to enhance the interpretability of time series data and better manage uncertainties. Then, the structure based on the combination of Long-Short-Term Memory encoder and Gated Recurrent Unit decoder structure is hybridized by implementing the Attention Mechanism to increase the ability to focus on the features of time-dependent sequential data. In addition, the Bayesian Optimization method was applied to maximize the performance of the obtained hybrid model and analyze the effects of the model's hyperparameters on prediction accuracy performance. It is observed that the proposed model outperforms four different benchmark models and provides an improvement in electricity load forecasting by applying the Diebold-Mariano statistical test and 10 different evaluation metrics used to compare the performance of time series forecasting models. [ABSTRACT FROM AUTHOR]
Copyright of Advances in Mechanical Engineering (Sage Publications Inc.) is the property of Sage Publications Inc. and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:Providing accurate electricity load forecasts is crucial for ensuring a cleaner and more sustainable environment, as these forecasts enable the balancing of energy supply and demand, reduce unnecessary reserve capacity requirements and lower dependence on fossil fuels. Objective of this study is to improve the limited performance of existing electricity load forecast models using time series data and to make more accurate forecasts. In this study, the Fuzzy Information Granulation method was applied to Türkiye's actual hourly electricity load data to enhance the interpretability of time series data and better manage uncertainties. Then, the structure based on the combination of Long-Short-Term Memory encoder and Gated Recurrent Unit decoder structure is hybridized by implementing the Attention Mechanism to increase the ability to focus on the features of time-dependent sequential data. In addition, the Bayesian Optimization method was applied to maximize the performance of the obtained hybrid model and analyze the effects of the model's hyperparameters on prediction accuracy performance. It is observed that the proposed model outperforms four different benchmark models and provides an improvement in electricity load forecasting by applying the Diebold-Mariano statistical test and 10 different evaluation metrics used to compare the performance of time series forecasting models. [ABSTRACT FROM AUTHOR]
ISSN:16878132
DOI:10.1177/16878132261429041