MRI and PET (or SPECT) image fusion enable the simultaneous visualization of anatomical details provided by the MRI image and functional details provided by the PET (or SPECT) image. This paper introduces a new image fusion framework for MRI and PET (or SPECT) medical images, which utilizes the features obtained through cartoon-texture decomposition in a pulse-coupled neural network. Firstly, the medical images are decomposed into cartoon and texture components such that the cartoon components represent piecewise smooth parts of the images with clear boundaries, and the texture components represent details of the images, such as edges and textured areas. Secondly, the cartoon components are fused by comparing the regional energy of coefficients, while the texture components are fused by feeding EOG (energy of image gradient) of coefficients into the pulse-coupled neural network to motivate the neurons. Experiments using several pairs of brain images show a state-of-the-art quality of the fused image, in terms of both visual appearance and objective criteria. Thanks to the efficient decomposition, the proposed method produces clear edges and details and does not suffer from color distortion.
The use of electric vehicles (EVs) as an emerging sector in transportation is effective in reducing global warming. In this paper, the optimal replacement of a fast charging station (FCS) considering the impact of EV users’ behavior is investigated. The behavior of users is investigated to reduce the charging demand in EVs due to the long driving distance to reach the station. In this article, the nature of the charging behavior of EV users is considered as a fixed point equation, which is calculated by establishing a relationship between the distance entry rate, and its spatial and temporal penalty. The optimal deployment of fast charging stations is modeled by a non-linear exact optimization problem that will determine the optimal locations for FCS construction. A proposed method based on a genetic algorithm is presented to solve this problem. The simulation results prove the effectiveness of the proposed algorithm in maximizing the profit for the FCS construction company, considering the power grid constraints.
The most common method for detecting internal faults in power transformers is the dissolved gas analysis (DGA) technique. The application of deep learning approaches in analyzing dissolved gases in oil faces challenges related to the scarcity of reliable labeled datasets, leading to overfitting and limited generalizability of these methods within the DGA domain. To address this issue, this study proposes the use of a large volume of synthetic DGA data for training deep models. A synthetic DGA database is generated by applying a Gaussian copula-based probabilistic function to the limited standard IEC TC 10 dataset, resulting in a semi-standard, extensive, labeled synthetic DGA database. Subsequently, various statistical and technical methods are employed to verify the similarity between the generated synthetic database and the original dataset. The deep convolutional neural network (CNN) model is then trained using this synthetic data. The outcome of this research introduces a hybrid model combining a deep convolutional neural network with a long short-term memory (LSTM) network for internal fault detection in transformers, trained on the semi-standard IEC TC 10 dataset. The model achieved a training accuracy of 95% and a testing accuracy of 93.3%.
Accurate localization is of significant importance in various applications, such as search and rescue operations. UAVs are regarded as a suitable solution for this purpose due to their agility and greater line-of-sight capabilities. In this research, a UAV is responsible for two-dimensional localizing the ground signal source by collecting received signal strength measurements. As the UAV moves through the environment, it creates virtual reference nodes and estimates the source location in the continuous space using the least squares method. The UAV’s trajectory is optimized using the Q-learning algorithm, a reinforcement learning method. This algorithm navigates the UAV within a search space divided into discrete cells. The main innovation of this research is the presentation of a hybrid algorithm that simultaneously integrates the least squares, Kalman filter, and Q-learning algorithms. In this framework, following the estimation of the current position, the Kalman filter predicts the future position of the moving source. Subsequently, the predicted position is discretized, and Q-learning determines the UAV’s optimal trajectory. The results indicate that this triple combination provides accurate source localization and the design of an optimal trajectory for the UAV, simultaneously improving localization accuracy and navigation efficiency. Additionally, the integration of the Kalman filter into the proposed framework enhances the tracking accuracy of the moving signal source and leads to a 42% reduction in the mean squared error.
Miniaturization has been widely applied in biosensing and immunosensing applications nowadays. Due to low Reynolds number flow, microchips usually suffer from mass transport and low detection yield. It should be noted that the antibody is immobilized on the surface of the biosensor, and the antigen is in the mobile phase. In this research, we propose an electrode structure to generate rotational flow near the biosensor and enhance the antibody-antigen binding process on the surface of the immune-sensor. Electrodes are excited at a proper voltage and frequency to generate an AC Electrothermal flow effect near the biosensor and concentrate the antigen to the immobilized antibodies. To study the AC electrothermal effect for an immunsensor, the physical equations, including electrostatic, fluid mechanics, temperature field, species concentration, and antibody-antigen binding reaction, is solved by the finite element method. By generating the electrothermal flow, the best location for the immunosensor is determined. The binding reaction is investigated both with and without applying the AC electrothermal effect. Based on the results with application of electrothermal flow (12-volt, 500 kHz), a rotational flow is induced and leads to a 9-fold improvement in antibody-antigen binding. The proposed system is studied for different Damköhler and Peclet numbers, and high throughput binding enhancement is observed for miniaturized and on-chip immunosensors.
Voltage stability assessment is one of the critical topics concerning the stability of power systems. With the increasing loading of transmission networks and its impact on downstream networks, the importance of studying this issue has gained attention even more. By presenting a new index, this paper evaluates the voltage stability of transmission systems. Employing the proposed index, network lines can be assessed in terms of voltage stability, and this means identifying weak network lines. Besides decreasing the calculation burden, the proposed index can boost accuracy. The index has been extended to consider the impact of the presence of distributed generations (DGs) and on-load tap changers (OLTC). To analyze and calculate the efficiency of the proposed index, two standard 9-bus and 14-bus transmission networks are utilized and the results are compared with other indices, showing that the results related to this index are reasonable and acceptable in comparison with those of other indices and impose a lower computational burden.
Because of the inherent limitations of traditional AC-DC-AC converters, especially in electric motor drives, Matrix Converters (MCs), have attracted increasing attention as an alternative. Unlike traditional converters, MCs eliminate the necessity of a DC-link capacitor, support bidirectional power transfer, and enable precise reactive power management. These unique capabilities make MCs as a promising solution where advanced control is required. This study focuses on enhancing the performance of MCs by integrating Model Predictive Torque Control (MPTC) and uses a model predictive control (MPC) framework with a finite control set. The key aim is to control an induction motor (IM) drive fed through MC, exploiting the predictive nature of MPC to improve dynamic response and control flexibility over traditional approaches. The proposed controller’s cost function includes terms for accurate torque and flux tracking, as well as grid-side reactive power regulation. To evaluate the performance, the MC-driven IM drive with the MPTC scheme was compared against a standard AC-DC-AC converter under identical operating scenarios. The results reveal superior torque tracking and adequate reactive power compensation by the MC. Notably, the MC-driven IM drive achieved approximately a 40% reduction in current Total Harmonic Distortion (THD) compared to the AC-DC-AC converter, and the grid-side reactive power was significantly mitigated compared to the conventional system.
Resilience characteristics in power systems and distribution networks refer to the system's ability to withstand severe disturbances with a low probability of occurrence. This paper proposes a two-level planning framework designed to increase resilience and manage outages in the distribution system by incorporating operational situation awareness. At the first level of the model, decisions are made regarding the allocation of repair crews and emergency generators within the distribution system, and the second stage evaluates the effectiveness of these decisions across various operational scenarios. The actions at the second level include preventive and emergency actions that aim to minimize power outages and the expected costs associated with operating the distribution system. Numerical analyses have been performed using the IEEE 69-bus test system. Simulation results not only validate the model's effectiveness but also demonstrate that integrating operational situation awareness into the resilience planning process significantly mitigates additional investment costs. Keywords:
In recent years, dynamic emotion recognition has become a crucial component for enhancing natural human-computer interaction. This paper proposes an advanced method for recognizing emotions in video files by combining facial and speech features, with a novel focus on improving the segmentation process to enhance recognition accuracy. The approach utilizes facial images and speech signals as the two primary inputs, which are integrated through an improved deep architecture. Feature fusion occurs at the extraction stage, employing an enhanced spectrogram for speech signals and AlexNet for both facial expression analysis and speech emotion recognition. To further enhance the accuracy of emotion recognition, the segmentation of extracted facial images is optimized using the Grey Wolf Optimization (GWO) algorithm, a powerful evolutionary technique inspired by the hunting behavior of grey wolves. This algorithm ensures better segmentation by adaptively dividing facial regions into meaningful parts, providing more informative features for dynamic recognition tasks. Feature selection is performed using Correlation-Based Feature Selection (CFS), while information fusion is guided by the SOAR cognitive model. The dynamic aspects of emotion recognition are addressed by employing dynamic deep learning techniques to implement the SOAR model. The proposed model is tested on the RAVDESS dataset, achieving an impressive accuracy of 91.23%, which demonstrates the effectiveness of the GWO algorithm in segmentation alongside multimodal fusion for dynamic emotion recognition.
This paper addresses the design and implementation of correct distance relay operation for transmission lines during single-phase asymmetric faults in the presence of SVC. The method used is based on utilizing negative, positive, and zero sequence currents to detect the effect of the Static Var Compensator during single-phase fault occurrence, without relying on costly communication systems. The importance of this method lies in its low cost, independence from remote devices and relays at the distance relay installation site, the elimination of noise effects, and the prevention of cyberattacks that are typically probable in communication systems and can disrupt the correct operation of distance relays. This method has been practically implemented with internal circuits in the Siemens 7SA522 distance relay configuration. Simulation results and practical relay testing using relay test equipment demonstrate the effectiveness of the proposed method for correct distance relay operation in power systems with SVC, without the use of remote equipment and communication systems.
Nowadays, communication networks play a crucial role in the integrated management of power systems, improving operational conditions, and implementing protective algorithms. However, the extensive utilization of these networks introduces challenges, including increased vulnerability of power systems to cyberattacks. Given the critical role of power networks in a country’s infrastructure, ensuring the security of these systems is both essential and unavoidable. One common type of attack in this domain is false data injection, which can alter measured values, disrupt the state estimation process, lead to incorrect decision-making, and ultimately compromise system stability. In this paper, a differential-distance-based protection algorithm is proposed to secure double-circuit transmission lines, one of the most important components of power networks, against false data injection cyberattacks, and its performance is evaluated on the standard IEEE test system. Case study results indicate that the proposed algorithm achieves a detection rate (DR) ranging from 91% to 100% across six scenarios, a false positive rate (FPR) below 3.7%, and a response time (DD) varying between 8 and 31 milliseconds, demonstrating its high speed and reliability.
In this article, a new hybrid control strategy for full-bridge DC-DC converters based on the phase shift control method is presented. The proposed control method has two loops of current control and voltage control, which is designed by using two operational amplifiers and creating a type 3 compensator structure. In fact, by comparing the current of the primary side of the transformer in the full-bridge DC-DC converter and the value of the error amplifier supplied from the output of the type 3 compensator, control parameters such as set and reset are extracted for use in the phase shift controller. In this regard, to control the power switches in the converter, a peak current control method is used, and a cycle-to-cycle current control method and pulse width modulation comparator are considered. The main purpose of the proposed control method is to be able to maintain control stability to a very desirable level in continuous and high load changes in the output of a full-bridge DC-DC converter. To provide a controller strategy, all the required theoretical relationships have been extracted and presented. To verify the proposed control structure, a laboratory sample with 12kW power has been built in the application of electric vehicles, and practical results have been presented.
Mobile radio communications have been used for a long time, and in recent years, new generations, including 5G and 6G, have been studied by researchers and made available to the public. In addition to these high-speed communications, very high-speed wireless optical communications in the core network section have also been considered in new standards. In this paper, the scenario of mixed radio frequepncy (RF)-free space optical (FSO) uplink transmission has been investigated using reconfigurable intelligent surfaces (RISs). By assuming the Nakagami-m fading channel for the RF links and the double generalized gamma (DGG) turbulence channels, the outage probability, the error probability, and the ergodic capacity have been obtained in closed form. Subsequently, due to the importance of security of wireless communication systems, the issue of security in the mentioned scenario has been examined, and in the presence of an eavesdropper, security parameters such as the secure outage probability and secrecy rate have been calculated in the form of mathematical relations. The results show the important role of RISs in increasing the security of RF-FSO uplink communication systems, so that by doubling the RIS elements, the performance is improved by at least 3 dB.
In this paper, a multi-objective performance optimization of an energy hub, including electrical, heating, and cooling power, has been conducted to simultaneously minimize fuel and environmental costs. To account for the uncertainties in the modeling parameters, such as the amount of power generated from renewable sources and the electrical, heating, and cooling loads during a day, the Monte Carlo method was employed to generate 1000 random scenarios. To manage calculation costs, 10 scenarios were selected from the generated scenarios by reducing the number of similar scenarios and those with low probabilities. Also, a demand response program has been implemented to encourage consumers of electrical, heating, and cooling loads to shift their usage. In this research, the multi-objective optimization process was conducted using GAMS software to evaluate the performance of the energy hub to identify a set of non-dominant solutions (the Pareto front) for both summer and winter seasons in the presence of energy storage devices. The simulation results for the determination of the dimensionless Pareto front support long-term planning for selecting optimal energy policies. This approach considers uncertainties and various operating scenarios while considering the interests of decision-makers in various conditions, such as fluctuating fuel prices and changing environmental penalties.
Stroke is the third leading cause of death and the largest cause of acquired disability worldwide. Classification of stroke lesions is vital in recovery, diagnosis, outcome assessment, and treatment planning. The current standard approach for segmenting ischemic stroke lesions is based on thresholding of computed tomography perfusion (CTP) maps. However, this detection approach is inaccurate (the dice similarity coefficient is around 68%). Accordingly, several machine learning-based techniques have recently been proposed to improve the segmentation accuracy of ischemic stroke lesions. Although these studies have achieved significant results, they still need to be improved before being used in real practice. This research presents a new technique based on deep learning for the segmentation of ischemic stroke lesions in CTP maps. The proposed network architecture includes the 7 Graph Convolutional layer, which can automatically perform feature selection/extraction and classify the resulting feature vector. In this study, the ISLES 2018 database was used to train the proposed network. The indices of the Dice Similarity coefficient and Jaccard Index based on the proposed model are 75.41% and 74/52%, respectively, which is a significant improvement compared to recent studies. In addition, the performance of the proposed model in noisy environments is very promising; so, at SNR=-4, the accuracy of networks is still above 60%.
The diagnosis of Parkinson's disease (PD) is usually done clinically by a doctor. This diagnosis is based on the initial symptoms, motor symptoms, and meditation of the doctor's experience. Since the diagnosis is made with the help of a doctor and based on the clinical description and received information, there is always an error in the diagnosis. Also, early clinical diagnosis is very difficult and almost impossible. Using methods based on machine learning is very useful for early diagnosis of Parkinson's disease. Brain signals and brain function can be a suitable solution for early diagnosis of this disease. Conventional methods are not effective due to the dynamics and complexity of the brain signal. Machine learning methods are a suitable solution with their high capabilities in the process of disease diagnosis. In this article, an efficient method based on machine learning is presented. In this method, after brain signals are pre-processed, time and frequency domain features are extracted from each signal and the best features are selected with the help of the improved intelligent gray wolf algorithm. The selected features are classified using a support vector machine classifier, K nearest neighbor, and random forest. Accuracy higher than 97% shows the superiority of the method in predicting Parkinson's disease.
There are many methods to diagnose transformer faults, including dissolved gas analysis (DGA), done in two conventional and smart ways. This paper presents a new method based on fuzzy logic and 5 DGA methods (key gas method, Durenberg's ratio method, Rogerˊs ratio method, IEC method, and Duval triangle method) to evaluate the condition of power transformers. At first, it is determined whether the transformer is healthy or defective. This step is performed based on key gas methods and Durenberg's ratio method with a fuzzy logic approach. Then, if it is detected that the transformer is defective, the type of error is detected using Rogers and IEC methods. If the error detection by these two methods reaches the same result, the error type detection is terminated and if the result is not the same, the Duval triangle method with fuzzy logic approach is used to detect the error type. The presented algorithm has been tested on 30 transformer devices and the results confirm its high accuracy (96.7%) in fault detection. In this article, in addition to detecting the type of fault, the location of the fault has been investigated and diagnosed using the CO2/CO ratio and the key gas method with a fuzzy logic approach.
The possibility of transferring the sense and force of the user on the driver side to the follower robot has always been discussed in master/slave systems. It has gained special importance in recent research. In the present study, the control torque of the slave side is adjusted by transferring the user's force (with the haptic handle) from the master side to the slave robot. The proposed system is a four-channel approach where the position signal and force signal are transmitted to the slave on the master's side, and reciprocally, the position signal and the contact force of the environment on the slave reach the master's side. The input-to-state stability for the master-slave system is investigated with the proposed approach. The approach of this research, unlike the approach of many previous studies, does not require the acceleration of the joints to parameterize the robot's dynamics or control it. In this method, the amount of force/torque of the user's hand of the master robot will be directly involved in the control command of the slave robot. In this article, in addition to the four-channel system, a three-channel system is proposed. Both systems, in addition to following the path, can follow the user's will on the master's side when encountering unexpected obstacles and stopping joints. In addition to software simulation, experimental implementation has been done on a leader-follower system with haptic robots, which confirms the correctness of the proposed approach.
To improve the resiliency of the distribution system during normal and emergency conditions, approaches for planning and operating are required so that the network can supply more loads during normal and emergency times. Thus, with the cellular approach in network planning, according to the production and consumption, the cellularization of the network to supply more loads can be considered. Consequently, the identification of energy cells and the boundary of energy cells is a challenge in the planning issue. This article achieves and identifies the maximum number of energy cells and the border among energy cells of distribution systems. The proposed objective functions for solving the resilience problem include the number of supplied loads and the number of unsupplied loads in normal or abnormal conditions. Furthermore, the eigenvalues and eigenvectors of the Graph Laplacian matrix of the distribution network are used to identify the maximum number of cells in the distribution network and the boundary among the generated cells, considering the prediction of distributed generation and consumption load based on fuzzy modeling. To show the efficiency of the proposed method, the test systems of 69 and 118 buses have been simulated. The results prove the proper performance of the proposed method.
In recent years, the use of permanent magnet synchronous motors has become popular. These engines are cheaper, lighter, and more durable. However, to control these motors, accurate information about the position of the rotor is necessary. For this reason, various methods have been used to improve the motor position estimation performance. One of these methods is estimation by injecting harmonic current into the stator winding. In this paper, using high-frequency currents injected into the stator, the rotor position in IPMSM motors is estimated. However, the response obtained has a DC offset due to the inaccuracy of the current measurement sensors, which causes an error in the estimation of the rotor position. The improved ePLL method was used to solve this problem. This method causes the convergence of the estimated position of the rotor to its real position by removing the noises of the current measurement circuit and the DC offset effect. The efficiency of this method has been proven first by simulation in MATLAB software, then in a Psim environment, and then implemented on the drive of a permanent magnet motor using a TMS320F28034 processor.
