Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Alzheimer's disease detection based on MR images using the quad convolutional layers CNN approach
Get PDFoai:ojs.beei.org:article/10304Alzheimer’s disease is a progressive neurodegenerative disorder requiring early and accurate detection for effective intervention. Deep learning (DL) techniques, particularly convolutional neural networks (CNNs), have shown promise in medical image classification. However, conventional CNN models often suffer from high computational complexity and inefficiency in handling imbalanced datasets. This study proposes a quad convolutional layers-CNN (QCL-CNN) for Alzheimer’s disease detection using magnetic resonance images (MRI) scans from the open access series of imaging studies (OASIS) dataset, which includes four dementia stages, non-dementia, very mild dementia, mild dementia, and moderate dementia. The QCL-CNN model employs four sequential convolutional layers for enhanced multi-level feature extraction, ensuring efficient classification while minimizing computational overhead. The experimental results demonstrate that QCL-CNN outperforms traditional CNN architectures, achieving an accuracy of 99.90%, recall of 99.89%, specificity of 99.93%, and an F1-score of 99.52%. The model surpasses VGG19, Xception, ResNet50, and DenseNet201 while maintaining a significantly lower parameter count (4.2M), making it computationally efficient. These findings confirm that network optimization is more crucial than model depth, ensuring robust performance even with fewer layers. Future research should explore multi-modal imaging, class balancing techniques, and real-world clinical validation to further improve the model’s diagnostic capabilities. The QCL-CNN model offers a promising artificial intelligence (AI)-powered approach for early Alzheimer’s detection, enabling precise, and efficient medical diagnosis
Bilateral transactions impact voltage stability and nodal pricing in power networks
Get PDFThis study investigates the impact of bilateral transactions on voltage stability and nodal pricing in the Indian power grid using a modified IEEE 30-bus system. A high voltage direct current (HVDC) link is integrated into the network to enhance control and system flexibility. Two advanced transmission pricing mechanisms— megawatt (MW)-Mile and megavolt-ampere (MVA)-Mile—are employed to allocate costs based on power flow magnitude and distance. The analysis incorporates hybrid AC-DC optimal power flow (OPF) modeling under various transaction levels. Simulation results show that a 100 MW bilateral transaction reduces the voltage at the receiving bus (bus 28) by 2% (from 1.05 to 1.03 p.u.) and increases the nodal price by 6.25% (from ₹4.80 to ₹5.10/kWh). The use of HVDC technology reduces total generation cost by approximately 8.2% (from ₹85 lakhs to ₹78 lakhs) and decreases real power loss from 70 MW to 50 MW. These findings confirm that bilateral transactions influence voltage profiles and market pricing. Moreover, MW-Mile and MVA-Mile methods demonstrate effective cost allocation capabilities. The proposed approach offers a practical framework for improving grid reliability and economic transparency in evolving power markets
Digital twins and IIoT: comparison of Prometheus and InfluxDB
Get PDFThis article presents a comparative analysis of data monitoring and visualization tools—Prometheus and InfluxDB—in the context of digital twins (DTs) applied to industrial settings. DTs optimize production processes using industrial internet of things (IIoT) technologies. Mathematical models assessed the tools based on response time, resource consumption, throughput, and reliability. Prometheus is better suited for high-frequency monitoring, achieving a response time of 0.01 seconds and processing up to 10,000 metrics per second—10–15% better than InfluxDB. It consumes 1.5 times less memory (100 MB versus 150 MB), making it faster and more resource-efficient. Conversely, InfluxDB excels in long-term storage and analytics, handling up to 8,000 metrics per second with a response time of 0.09 seconds. However, it requires more resources, including higher CPU usage (20% versus 15%). Both tools integrate seamlessly with Grafana for visualization, offering flexibility for real-time monitoring and decision-making. The study provides actionable insights for selecting monitoring systems based on project-specific requirements, highlighting Prometheus’s efficiency in dynamic scenarios and InfluxDB’s strength in analytics-focused tasks
Electrical active power optimization of the SEIG-WTS based on perturb and observe method
Get PDFThis paper proposes an electrical active power optimization for self-excited induction generator-wind turbine system (SEIG-WTS) using a perturb and observe (PO)-based maximum power point tracking. The main advantage is the optimization of the SEIG active power of SEIG-WTS and the simply practical implementation with rotor speed sensor, current sensor and three phase inverter. The active power optimization of SEIG-WTS is achieved by perturbing angular magnetic field speed with stator reference voltage. To test the effectiveness of the proposal, numeric simulations were carried out under very challenging conditions with wind speed profile in steps and actual wind speed profile. The proposal reaches the maximum power in 7 seconds for hardest condition when the system works at high rotor speed. The proposal is useful for the development of maximum power point tracking control (MPPT) controllers due to its simplicity in implementation.This paper proposes an electrical active power optimization for self-excited induction generator-wind turbine system (SEIG-WTS) using a perturb and observe (PO)-based maximum power point tracking. The main advantage is the optimization of the SEIG active power of SEIG-WTS and the simply practical implementation with rotor speed sensor, current sensor and three phase inverter. The active power optimization of SEIG-WTS is achieved by perturbing angular magnetic field speed with stator reference voltage. To test the effectiveness of the proposal, numeric simulations were carried out under very challenging conditions with wind speed profile in steps and actual wind speed profile. The proposal reaches the maximum power in 7 seconds for hardest condition when the system works at high rotor speed. The proposal is useful for the development of maximum power point tracking control (MPPT) controllers due to its simplicity in implementation
Real-time IoT security framework for detecting a person with a weapon using Raspberry Pi, Google Vertex AI, and AWS
Get PDFRealtime crime scene detection is a vital issue for ensuring security in various environments. Building on recent advancements in machine learning algorithms, this paper presents an IoT framework for real-time weapon and face detection. By deploying a convolutional neural network (CNN) architecture in Vertex AI and utilizing the portable camera module of a Raspberry Pi, to detect whether a person is carrying a weapon. This is achieved by pre-processing, which we resize and annotate the images. Then, train and validate the CNN model with the annotated label dataset. The trained model is saved in Google Cloud’s Vertex AI portal. Then we tested the model by uploading live images from a camera as well as a few video clips, to a Django application in amazon web hosting services (AWS) to Vertex AI. The model exhibited an accuracy of 97.2% along with a F1 score of 0.97. In addition, the model outperforms the other state-of-the-art models by less trainable parameters and higher accuracy
Analysis of human emotions through speech using deep learning fusion technique for Industry 5.0
Get PDFEmotions are important for human well-being and social connections. This work focuses on the issue of effectively understanding emotions in human speech, specifically in the context of Industry 5.0. Traditional approaches and machine learning (ML) techniques for identifying emotions in speech are limited, such as the requirement for complicated feature extraction. Traditional methods yield recognition accuracies of no more than 90% because to the restricted extraction of temporal/sequence information. This paper suggests a ground-breaking fusion-based deep learning (DL) method to overcome these limitations. Specifically, one-dimensional (1D) and two-dimensional (2D) convolution neural network (CNN) can automatically extract significant characteristics and handle enormous datasets in real time. Furthermore, a fusion-based DL network, speech emotion recognition deep learning fusion network (SER_DLFNet), has been proposed, which combines CNN with long short-term memory (LSTM) to collect sequence information and increase recognition accuracy. The proposed model shows impressive results, with a test accuracy of 95.52% on the ryerson audio-visual database of emotional speech and song (RAVDESS) dataset. This research contributes to the advancement of more precise and efficient emotion identification algorithms for voice analysis, especially within the framework of Industry 5.0
An interior penalty function method for solving fuzzy nonlinear programming problems
Get PDFIn this article, we investigate fuzzy interior penalty function method for solving fuzzy nonlinear programming problems (FNLPP) based on a new fuzzy arith-metic, unconstrained optimization, and fuzzy ranking on the parametric form of triangular fuzzy numbers (TFN). The main objective of this paper is to solve constrained fuzzy nonlinear programming problems using interior penalty func-tions (IPF) by converting it into unconstrained optimization problems. We prove an important lemma and a convergence theorem for the interior penalty functions method. Interior penalty function techniques favor sites near the boundary of the feasible region in the interior. We present a numerical example of the suggested method and compare the results to those produced by existing methods
Deep learning based photovoltaic generation on time series load forecasting
Get PDFIn recent years, solar irradiance forecasting has become essential to managing, developing, and effectively integrating photovoltaic (PV) systems properly into the smart grid. The foundation of a conventional variational autoencoder (VAE) is an entirely coupled layer that includes both decoder and encoder components. In this study, a novel deep attention-driven model for forecasting named bidirectional long short-term memory (BiLSTM) which is combined with the VAE model is introduced as an enhanced version of the VAE. BiLSTM is integrated at the encoder side of VAE to effectively extract and learn temporal dependencies that are embedded in the panel irradiance data. Additionally, a self-attention mechanism (SAM) is added to bilateral variational autoencoder (BiVAE) which is known as BiVAE-SAM that highlights the important characteristics. The proposed BiVAE-SAM permits the VAE’s capacity to design the temporal dependency. The examined models are assessed using sun irradiance measurements from New York City, Turkey, Canopy, Los Angeles, California, and Florida. The outcomes exhibit that the proposed BiVAE-SAM model performs better mean absolute percentage error (MAPE) with values of 1.7935, 0.7828, 1.3491 and 2.8346 respectively for California, Los Angeles, New York City, and Florida, over existing stacked denoising auto-encoders (SDA) model at projecting solar irradiance
Low insertion loss open-loop resonator–based microstrip diplexer with high selective for wireless applications
Get PDFThis paper presents a low-insertion-loss open-loop resonator (OLR)-based microstrip diplexer with high-selective for wireless applications. We used two series capacitive gaps in the microstrip transmission line, loaded with rectangular-shaped half-wavelength OLRs, to create a high-selectivity bandpass filter (BPF). The planned BPFs are linked through a T-junction combiner, precisely tuned to align with both filters and the antenna port in order to produce the proposed diplexer. The system is implemented on a rogers TMM4 substrate with a loss tangent of 0.002, a dielectric constant of 4.7, and a thickness of 1.52 mm. The suggested diplexer has dimensions of (90×70) mm². It achieves a modest frequency space ratio of R=0.1646 in both transmit and receive modes by having two resonance frequencies of ft=2.191 GHz and fr=2.584 GHz, respectively. The simulated structure displays good insertion losses of approximately 1.2 dB and 1.79 dB for the two channels, respectively, at fractional bandwidths of 1.24% at 2.191 GHz and 0.636% at 2.584 GHz. The simulated isolation values for 2.191 GHz and 2.584 GHz are 53.3 dB and 66.5 dB, respectively
Comparative study of teachable machine for forest fire and smoke detection by drone
Get PDFForests play a vital role in maintaining ecological equilibrium and serving as vital habitats for wildlife. They regulate global climate, safeguard soil and water resources, and provide crucial ecosystem services such as air and water purification, essential for human well-being and sustainable development. Forest fires wreak havoc on ecosystems and wildlife, emitting harmful pollutants, disrupting communities, and increasing the risk of erosion and landslides. Detecting forest fires through satellite imaging, aerial reconnaissance, and ground-based sensors is pivotal for early detection and containment, safeguarding human lives, wildlife, and preserving natural resources for future generations. Utilizing drones and deep learning (DL) algorithms can significantly enhance early fire detection and minimize their devastating impact. In this paper, we examine teachable machine, a Google tool for creating DL models. We compare the top model generated by teachable machine for fire and smoke detection to models obtained through transfer learning from established DL models in image recognition and computer vision (CV), such as VGG16, VGG19, MobileNet, MobileNetv2, and MobileNetv3. The results underscore the significance of employing the teachable machine model in specific fire and smoke detection scenarios