@article{M9C4F73D8,
title = "Solar Power Generation Forecasting using a Hybrid LSTM-Linear Model with Multi-Head Attention",
journal = "The Transactions of the Korea Information Processing Society",
year = "2025",
issn = "null",
doi = "https://doi.org/10.3745/TKIPS.2025.14.2.123",
author = "Hyeonseok Jin, David J. Richter, MD Ilias Bappi, Kyungbaek Kim",
keywords = "Solar Power Prediction, Power Generation, Prediction, Time series, Deep Learning",
abstract = "Due to the negative impact on the environment, the demand for solar energy, which can effectively replace fossil fuels, is increasing. 
In order to operate solar energy efficiently, deep learning has been used recently to predict future power generation, but it is still a 
challenge to provide accurate predictions because the power generation greatly depends on external factors such as weather and time 
of day. In this paper, we propose a hybrid model that combines LSTM and Linear models using Multi-Head Attention to provide more 
accurate predictions. The proposed model can improve prediction accuracy and learn richer time series features because each Linear 
model can capture trends and irregularities in time series features. We conducted extensive experiments, and the results showed that 
the proposed model outperformed other prediction models by about 10%, and the ablation study confirmed that combining three models 
based on Multi-Head Attention is the most effective way to consider trends and irregularities."
}