Attention-Based BiLSTM with Mutual Information Feature Selection for Improved Solar Radiation Prediction

DOI

10.2991/978-94-6239-791-0_5How to use a DOI?

Keywords

Solar radiation prediction; Mutual information; Feature selection; BiLSTM; Attention mechanism; Deep learning

Abstract

Solar radiation prediction is one of the important factors to be considered for the efficient usage of solar energy. Solar radiation is highly unpredictable because of the complex weather conditions and it is also highly variable. Therefore, the prediction of solar radiation is considered to be a challenging task. In this paper, an effective framework is presented for prediction of solar radiation. It is based on the concept of temporal feature engineering, feature selection using mutual information, and the usage of the Attention-based Bidirectional Long Short-Term Memory model combined with Mutual Information-based feature selection and lag-sequence temporal modelling. Lag based temporal features are used to efficiently predict the time varying data. Moreover, the mutual information feature selection method is used to select the most pertinent meteorological attributes. Unlike conventional LSTM-based forecasting methods, the proposed framework integrates nonlinear feature relevance analysis with temporal attention learning to improve forecasting robustness. In the proposed model, Performance is assessed using the publicly available solar radiation dataset and evaluated using the standard performance metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination (R²). From the experimental results, it is evident that the proposed model is able to achieve high accuracy with an RMSE of 0.0477 and R² of 0.9171. The results demonstrate that the proposed model is effective in modelling complex relationships in solar radiation data and shows capability for real world solar energy applications.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Gautam Kumar
AU  - Yash Hooda
AU  - Ajay Kumar
AU  - Aayush Kumar
AU  - Sayan Khan
AU  - Shovon Dey
PY  - 2026
DA  - 2026/10/09
TI  - Attention-Based BiLSTM with Mutual Information Feature Selection for Improved Solar Radiation Prediction
BT  - Proceedings of the International Conference on Artificial Intelligence and Machine Learning for Agro-Ecosystems (ICAIML-AE 2026)
PB  - Atlantis Press
SP  - 35
EP  - 49
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-791-0_5
DO  - 10.2991/978-94-6239-791-0_5
ID  - Kumar2026
ER  -