Hybrid Deep Learning-Based Renewable Energy Classification for Smart Grid Optimization

Abstract

The classification of renewable energy sources is crucial for optimizing energy management and advancing sustainable practices. This study proposes a robust classification framework using a publicly available renewable energy dataset comprising multivariate time-series data from solar, wind, and hydro sources. Standard preprocessing techniques, including normalization and segmentation, were applied to prepare the data for modeling. We evaluate several machine learning and deep learning models Logistic Regression, Support Vector Machine (SVM), XGBoost, Artificial Neural Networks (ANN), and 1D Convolutional Neural Networks (1D-CNN). To further enhance performance, we introduce a hybrid 1D-CNN model integrated with an attention mechanism to improve feature extraction and model focus on relevant temporal patterns. Experimental results show that the attention-enhanced hybrid model achieves superior performance with an accuracy of 97.8%, precision of 97.5%, recall of 97.7%, and F1-score of 97.6%, outperforming all baseline models. Compared to the best traditional model (XGBoost, 93.2% accuracy), our approach shows a 4.6% improvement. This demonstrates the effectiveness of attention-based deep learning for renewable energy classification and lays a foundation for future intelligent and sustainable energy management systems.