Power, Control, and Data Processing Systems

Abstract Accurately estimating the state of charge (SOC) and state of health (SOH) of lithium-ion batteries remains a significant challenge due to complex electrochemical dynamics and varying operational conditions. This research aims to overcome these challenges by proposing a novel hybrid neural network that combines octave convolution (OctConv) and long short-term memory (LSTM). The methodology begins with preprocessing battery voltage, current, temperature, internal impedance, and capacity data through outlier removal and min-max normalization, followed by partitioning into training, validation, and test sets. The OctConv module then decomposes the input into high- and low-frequency feature maps, enhancing extraction efficiency by reducing spatial redundancy. These multi-scale features are subsequently processed by LSTM layers to model both short-term fluctuations and long-term degradation trends across battery cycles. The model was trained using the Adam optimizer with MSE loss and ReLU activations and evaluated against established deep learning approaches, including OCT, LSTM, and CNN-LSTM, on the NASA dataset. Experimental results demonstrate the superior performance of the proposed OCT-LSTM model. It achieved significant improvements in SOC and SOH estimation, reducing the mean squared error (MSE) by 84.53% and 82.44%, respectively, compared to the CNN-LSTM benchmark. Furthermore, the model attained a 6.21% reduction in simulation parameters, highlighting greater parameter efficiency. The proposed method outperforms state-of-the-art approaches across multiple metrics, including the number of parameters, MSE, RMSE, MAE, prediction accuracy, and computational speed, demonstrating its potential as a significant advancement in this domain.

Abstract With the rapid advancement of sensing technologies, cloud computing, and communication systems, interaction between physical sensors and cloud-based applications has become more seamless and intelligent. This technological convergence has given rise to the concept of the Sensor-Cloud, a hybrid architecture that tightly integrates wireless sensor networks with scalable cloud infrastructures. One of the core challenges in Sensor-Cloud systems lies in efficiently managing resource allocation and task scheduling between sensors and cloud-hosted applications. Ensuring that sensors are assigned to appropriate tasks while maintaining network efficiency, fairness in resource distribution, and minimizing communication overhead remains a complex optimization problem. Traditional methods for resource management, such as heuristic algorithms and mathematical optimization techniques, often face significant trade-offs: they may offer good solutions but at the cost of high computational complexity, or achieve faster execution while compromising accuracy and fairness. To address these limitations, this paper introduces a novel game theory-based approach for resource allocation within Sensor-Cloud infrastructures. In the proposed model, each sensor acts as an autonomous rational agent that strategically selects an application to serve, aiming to maximize its own utility while contributing to overall system efficiency. By formulating this interaction as a non-cooperative game, the system dynamics naturally evolve toward a Nash equilibrium, where no sensor has an incentive to unilaterally change its allocation decision. This equilibrium represents an optimal and stable allocation of resources that balances efficiency, fairness, and computational cost. The simulation results demonstrate that the proposed method is near optimal.

Abstract Reliable fault diagnosis in photovoltaic (PV) arrays is essential for reducing energy losses, preventing safety hazards, and maintaining system availability; however, conventional supervised learning methods require large labeled datasets that are costly and risky to obtain from operating plants. This paper proposes a data-efficient diagnostic framework that combines convolutional neural networks, semi-supervised generative adversarial learning, and few-shot meta-learning. A 5.3-kW, 6 × 3 PV array is modeled in MATLAB/Simulink to generate healthy and faulty operating data under variations in irradiance, temperature, fault resistance, and partial shading. Four conditions are considered: normal operation, line-to-line faults, hot-spot or high-series-resistance faults, and open-circuit faults. Current, voltage, irradiance, and temperature signals are normalized and transformed using the short-time Fourier transform; their spectrograms are then concatenated into composite images for automatic feature extraction and classification. A three-layer CNN provides a high-accuracy supervised benchmark, reaching approximately 99.5% accuracy on noise-free data. To reduce dependence on labeled samples, a semi-supervised GAN exploits unlabeled and synthetically generated data and achieves 81.57% test accuracy using only 100 labeled samples on the one-dimensional noisy dataset. To further address the computational cost of GAN training and extreme label scarcity, a few-shot meta-learning model is introduced. In the 5-shot, 4-way setting, it achieves 94.03% accuracy on previously unseen samples, substantially outperforming a conventional CNN trained with limited labeled data. The main contribution is therefore a unified comparison of three complementary learning strategies and a practical few-shot solution that preserves strong diagnostic performance with minimal labeled PV fault data under realistic operating variability.

Abstract This paper proposes a collaborative energy management in the water-energy nexus multi-microgrid systems. In this model, several microgrids collaborate together to form a large coalition and perform a centralized energy scheduling in order to minimize the total costs. In the water section, the microgrids utilize the water desalination unit and underground water sources to supply the freshwater. Also, they operate the water storage tank to enable the water flexibility for the microgrids. In the electrical section, the renewable energy sources, capstone microturbine, and battery storage systems are considered to keep the balance between the generation and demands. In this section, a dynamic demand response program is introduced that calculate the up-ward and down-ward potential of different microgrids based on their load consumption profiles. The numerical results shows that performing the dynamic demand response programs reduces the operating cost and energy not supplied of microgrids by 8.35 % and 91.43 %, respectively.

Abstract This study aimed to improve breast tumor segmentation in ultrasound images using a modified U‑Net architecture that reduces trainable parameters while enhancing segmentation accuracy and diagnostic performance. A lightweight modified U‑Net with fewer trainable parameters was developed for breast ultrasound tumor segmentation using the BUSI and UDAIT datasets, and to compensate for the simplified structure and reduce overfitting risk, the input channels were diversified using complementary tumor and non‑tumor region enhancement information, which improved both direct and indirect segmentation metrics while maintaining lower computational complexity than recent U‑Net‑based models. The model performance was evaluated using accuracy, precision, sensitivity, specificity, F1‑score, Dice coefficient, and Intersection over Union (IoU). On the BUSI dataset, the proposed method achieved an accuracy of 0.99, precision of 0.95, sensitivity of 0.96, specificity of 0.99, F1‑score of 0.95, Dice coefficient of 0.95, and IoU of 0.92; on the UDAIT dataset, it obtained an accuracy of 0.98, precision of 0.88, sensitivity of 0.96, specificity of 0.98, F1‑score of 0.92, Dice coefficient of 0.92, and IoU of 0.85. Compared with previous approaches, the proposed method achieved superior segmentation performance, particularly in Dice coefficient and F1‑score, while maintaining lower computational complexity due to fewer trainable parameters. In conclusion, the proposed modified U‑Net model improved breast tumor segmentation in ultrasound imaging while maintaining lower computational complexity than newer U‑Net variants, and its high accuracy and Dice scores indicate its potential as an efficient decision‑support tool for breast cancer diagnosis.

Abstract Fault detection and classification in photovoltaic arrays are important for

improving the reliability, safety, and energy production of photovoltaic systems.

This study presents a two-stage evaluation of machine learning-based PV fault

diagnosis using both simulation-generated and measured experimental data. First,

a 64-module PV array with an 8×8 configuration is modeled in

MATLAB/Simulink under ten normal and faulty operating conditions. Support

Vector Machine and Naive Bayes classifiers are evaluated before and after

Principal Component Analysis preprocessing. PCA-SVM achieves the highest

accuracy of 97.93% in the ten-class simulation study, compared with 86.50% and

43.125% for SVM and Naive Bayes, respectively. Second, external validation is

performed using twelve measured current-voltage curves obtained from six

physical PV panels under clean and partial-shading conditions. The measured

curves are processed into 3,600 observations and evaluated using leave-one-panel

out validation. Six classifier families, including SVM, Naive Bayes, Random

Forest, XGBoost, LightGBM, and MLP are investigated both with and without

PCA. Random Forest achieves the highest experimental-data accuracy of 97.53%.

PCA increases SVM accuracy from 68.14% to 79.69% and XGBoost accuracy

from 83.33% to 90.03%, while its effect is unfavorable for several other classifiers.

The results demonstrate that PCA can improve feature separability, but its

effectiveness is classifier- and dataset-dependent. The external validation confirms

the usefulness of electrical and environmental features for measured partial

shading diagnosis while also highlighting the need for broader field validation.