Research Article
Deep Reinforcement Learning-Based Energy Management of Battery-Supercapacitor Hybrid Storage Systems in Renewable Microgrids
Daniel Kumi Owusu*
Issue:
Volume 12, Issue 2, December 2026
Pages:
40-56
Received:
3 July 2026
Accepted:
20 July 2026
Published:
17 August 2026
Abstract: The increasing integration of renewable energy sources into microgrids has intensified the need for intelligent energy management strategies capable of addressing the intermittency of solar and wind generation while ensuring reliable and cost-effective operation. Although Rule-Based Control (RBC) methods are straightforward to implement, their limited adaptability often leads to suboptimal utilisation of Hybrid Energy Storage Systems (HESS). This study develops and evaluates a Deep Reinforcement Learning (DRL)-based energy management system employing a Deep Q-Network (DQN) to coordinate battery–supercapacitor operation within a renewable microgrid. A Gymnasium-compatible simulation environment was constructed using a publicly available time-series dataset comprising renewable generation, load demand, electricity prices, battery state of charge (SoC), and supercapacitor SoC. Feature engineering, incorporating sinusoidal temporal representations and Min-Max normalisation, was applied to enhance learning stability and capture cyclical demand and generation patterns. The DQN agent was trained over 50 episodes and benchmarked against a conventional RBC strategy under identical operating conditions. Training performance demonstrated progressive policy improvement, with cumulative rewards increasing from approximately -1200 to -400, indicating enhanced decision-making capability during learning. The learned controller exhibited adaptive energy scheduling through dynamic utilisation of the supercapacitor and selective grid interaction in response to varying operating conditions, whereas the RBC followed a deterministic control strategy with limited flexibility. However, comparative evaluation revealed that the DQN did not consistently outperform the RBC in cumulative economic performance, suggesting the need for further refinement of the reward function, training process, and hyperparameter configuration. Nevertheless, the proposed framework demonstrates the feasibility of applying deep reinforcement learning to coordinated battery–supercapacitor energy management and highlights its potential to enhance operational flexibility and intelligent resource utilisation in renewable microgrids. The study contributes a dataset-driven reinforcement learning framework that provides a foundation for future research on advanced AI-based energy management systems and the integration of more sophisticated reinforcement learning algorithms for resilient and sustainable microgrid operation.
Abstract: The increasing integration of renewable energy sources into microgrids has intensified the need for intelligent energy management strategies capable of addressing the intermittency of solar and wind generation while ensuring reliable and cost-effective operation. Although Rule-Based Control (RBC) methods are straightforward to implement, their limi...
Show More
Research Article
LIMD: LBP-InceptionV3 Hybrid Approach for Mpox Detection Using Skin Lesion Images
Issue:
Volume 12, Issue 2, December 2026
Pages:
57-75
Received:
5 September 2026
Accepted:
15 September 2026
Published:
28 September 2026
Abstract: Mpox skin lesions may exhibit visual similarities with other infectious skin conditions, particularly chickenpox and measles, making reliable image-based differentiation challenging. This study presents an LBP InceptionV3 hybrid approach for Mpox detection using skin lesion images. The proposed framework combines high level visual representations extracted using pretrained InceptionV3 with complementary local texture information generated using Local Binary Pattern (LBP). The secondary dataset comprised 228 original images. 102 images belong to Mpox class, while 126 images belong to the others class (chickenpox and measles). The images were expanded through augmentation to 3,192 images comprising 1,428 Mpox and 1,764 Other images. Images were standardized to 224 × 224 pixels and divided into training, validation, and testing sets using a 70:15:15 ratio. InceptionV3 generated a 2,048-dimensional deep feature representation, while LBP texture maps were processed through a shallow convolutional network to obtain 512 texture features. The resulting 2,560-dimensional representation was fused for classification. The proposed model achieved 94.15% accuracy, 94.16% precision, 94.15% recall, 94.16% F1 score, and an AUC of 0.99. On the 479 image test set, 451 images were correctly classified, including 202 Mpox and 249 Other images. The ablation study demonstrated that incorporating LBP improved accuracy from 88.72% for InceptionV3 alone to 94.15%, representing a 5.43 percentage point improvement, while AUC increased from 0.95 to 0.99. In addition, clinically sourced images from Owerri General Hospital were used to provide an independent assessment of the model on previously unseen skin lesion samples. The findings demonstrate that combining deep visual and local texture representations provides an effective approach for Mpox skin lesion classification.
Abstract: Mpox skin lesions may exhibit visual similarities with other infectious skin conditions, particularly chickenpox and measles, making reliable image-based differentiation challenging. This study presents an LBP InceptionV3 hybrid approach for Mpox detection using skin lesion images. The proposed framework combines high level visual representations e...
Show More
Research Article
Personalized Calories Burned Prediction for Human Support Fitness Using Neural Network Models
Issue:
Volume 12, Issue 2, December 2026
Pages:
76-86
Received:
8 May 2026
Accepted:
25 May 2026
Published:
26 October 2026
Abstract: Accurate prediction of calorie expenditure is essential for personalized fitness management yet conventional estimation methods based on Metabolic Equivalent of Task (MET) values fail to account for individual physiological differences. This study developed and compared three neural network models: Artificial Neural Network (ANN), Deep Neural Network (DNN), and Recurrent Neural Network (RNN) for predicting calories burned during aerobic exercise using Calories.csv and exercise.csv datasets sourced from Kaggle. The dataset was preprocessed through categorical encoding and feature engineering to expand the feature set. Exploratory data analysis was performed ant it identified exercise duration, heart rate, and body temperature as the strongest predictors of calorie expenditure. All three models were trained using back propagation and the Adam optimizer, and evaluated using R2, RMSE, MAE, and MAPE among other metrics. All models achieved over 99% explained variance, with the DNN emerging as the best performer (R2 = 0.9998), followed by the ANN (R2 = 0.9997) and the RNN (R2 = 0.9975). The RNN's comparatively weaker performance could be attributed to the non-sequential nature of the dataset. The DNN was subsequently deployed as a Flask-based web application providing real-time calorie predictions alongside health metrics such as BMI and exercise intensity. The study demonstrates that deep learning, particularly DNN architectures, offers a highly accurate and practical solution for personalized calorie prediction in fitness and health applications.
Abstract: Accurate prediction of calorie expenditure is essential for personalized fitness management yet conventional estimation methods based on Metabolic Equivalent of Task (MET) values fail to account for individual physiological differences. This study developed and compared three neural network models: Artificial Neural Network (ANN), Deep Neural Netwo...
Show More
