نشریه هوش محاسباتی در مهندسی برق
سال پانزدهم شماره 1 (بهار 1403)

The development of the application of artificial intelligence and machine learning methods and the expansion of the use of phasor measurement units (PMU) have made it possible to analyze the power system as online using measurement data. In this article, a deep learning method based on a Bidirectional Gated Recurrent unit (Bi-GRU) with convolution layers is presented for online prediction of short-term voltage stability (STVS) using PMU data. In order to investigate the dynamic behavior of the power system in STVS, the database includes the time series of voltage magnitude and phase angle. A three-class classification problem (stable, alert and unstable) is defined with the help of a dynamic index including Lyapunov function and voltage deviation. The ability of bidirectional neural network in the simultaneous analysis of past and future data and the ability of the convolution layer in extracting the temporal characteristics of the data have led to an increase in the accuracy of the online STVS assessment. The simulations on IEEE 39 and 118 bus show that the proposed method can predict the voltage stability based on the measurement data pre and post-fault with good accuracy and timely, so that There will be more time for corrective actions in the network. According to the obtained results, the proposed algorithm is resistant to changing the topology of the power grid as well as changing the operating points.

In this paper, using a deep learning technique, the sensing characteristic of a refractive index (RI) sensor based on a multimode optical fiber speckle-gram has been extracted. The investigated sensor structure consists of a tapered optical fiber or microfiber (MF) induced on a standard multimode optical fiber (MMF). Due to the presence of evanescent waves around the MF, the guiding status of the end MMF is dependent on the external RI around the taper. Thereby, the propagating modes and resulting interference pattern or speckle-gram become sensitive to the changes in the taper environment. Herein, by analyzing and quantifying the speckle-gram modulation, the RI changes around the taper are characterized. To extract the sensing characteristic and improve its measurement accuracy, deep learning algorithms and specifically the ResNet deep neural network are used in the processing of the output speckle-grams. Using this algorithm, while the tapered fiber is immersed in different refractive index solutions in the range of 1.33 to 1.368, a set of output speckle images is captured for learning and then a number of them are used for measurement, and finally the accuracy of detecting the refractive index of the solutions RI has been achieved up to 95%.

Reconfiguration of the distribution network as well as the optimal use of distributed generation resources in the distribution system are very effective methods to reduce losses and improve the voltage profile or in other words power quality in the electricity distribution system. In recent years, researchers have paid attention to the use of distributed production resources. The use of these resources has several advantages, the most important of which are the reduction of network losses and the increase of voltage stability. In this study, a differential evolutionary algorithm is presented to solve the desired optimization problem to reduce losses and bus voltage deviation. On the other hand, since the system load is always changing and is not constant, therefore, for the simulation results to be close to the real conditions of the distribution network, it is suggested in this study that the uncertainty of the consumption load should also be modeled and applied to the optimization problem. The mentioned problem has different discrete and continuous variables, which necessitates the use of algorithms that can search in discrete and continuous spaces. Therefore, to overcome this issue and apply different constraints to the problem, the differential evolutionary algorithm has been used. The mentioned method has been tested on a standard 33-bus test network and the results have been compared in three different scenarios. In the second and third scenarios, reconfiguration of the distribution network has been resolved in the absence/presence of scattered production units (wind turbines) respectively. The results of the proposed method have been compared with other references. The results show that the proposed differential evolutionary algorithm performs better than the other two algorithms and has achieved better results.

Neuromarketing uses cognitive science in marketing-related research to study consumer sensorimotor effects, such as cognitive and emotional responses to marketing stimuli by using new technologies. This research is a predictive framework for participants’ decision-making processes in terms of liking and disliking when viewing and selecting the products of an online store. For this purpose, dispersion entropy (DE) and hierarchical dispersion entropy (HDE) features are extracted from the participant’s brain signal (EEG) when displaying different products. Dispersion pattern evaluation is used for the first time to distinguish between two categories of likes and dislikes using EEG signals. In this study, the ensemble learning approach with a random forest classifier is investigated to create a model for predicting neuromarketing to distinguish between two categories of likes and dislikes. From the analysis of the model results, 73.67% accuracy has been obtained. Studies in this area can change and improve marketing strategies to improve production and consumer processes, and ultimately lead to mutual benefit.

Surveillance cameras can provide more accurate traffic information compared to GPS and infrared radiation sensors. With the intelligent processing of image information provided by them, the analysis of traffic problems is done more precisely. Meanwhile, the speed of car positioning is of particular importance because, after positioning the car, other processes are needed to make decisions that require time management. The purpose of the present study is to propose an algorithm based on deep networks that solves this challenge. In the presented algorithm, a fast and accurate detector network is designed based on multiple receptive fields, segmentation, and differential images, which increases the accuracy of the detector by producing semantic feature maps and filtering them. Also, it increases detector speed by reducing the parameters. Each vehicle's maneuver is decided based on the time and location information of that vehicle and the vehicles nearby. Then, according to the type of maneuver, the predictive network is selected in one of the modes of lane keeping, left turn or right turn. The networks designed in the proposed algorithm are complementary to each other. The performance of the proposed algorithm is demonstrated by experiments on the Highway and UA-DETRAC datasets.

Nowadays, facial plastic surgeries are performed for the correction of congenital anomalies, correction of injuries caused by accidents, treatment of burns, and the creation of a beautiful structure in the face. Therefore, surgeons need to analyze and measure quantitative facial metrics. Facial landmark detection plays an important role in the measurement of the facial golden ratio. In this study, 33 landmarks from the frontal view were detected using the Cascade Regression Model (CRM). Face detection is the first step in facial landmark detection to put the initial shape on the face. Therefore, the fuzzy c-means algorithm optimized by Harris Hawk Optimization (HHO) has been used for facial segmentation. In CRM, the regression function is used for mapping input face images to the desired output. For feature extraction, the Histogram Orientation Gradient (HOG) method was applied around the defined landmarks on three channels of RGB color space. Experiment results showed that feature extraction in three channels of RGB color space is more accurate than feature extraction in the gray scale. Also, results confirmed that the proposed method has a high accuracy in facial landmark detection and can be used as an effective method in the Cephalometry analysis of facial soft tissue.

Simultaneous use of an efficient routing protocol and mobile sink not only prevents rapid sensor energy depletion but also effectively improves the energy balance in wireless sensor networks. In this paper, a new data aggregation method is proposed called “Ant colony-based Routing and Petal-shaped Data Aggregation (ARPDA)”, which includes clustering, Cluster Head (CH) selection, intra-cluster routing, determining polling points, and designing the sink node trajectory. In the proposed method, the network is divided by several virtual concentric circles that are equidistant from each other. The intersection point of these circles with the virtual lines passing through the origin of the network determines the probable polling points for the sink node. First, by efficiently clustering the sensor nodes and forming an intra-cluster routing tree based on the ant colony optimization algorithm, the sensor data is aggregated in CHs; then, by selecting the appropriate polling points and designing a petal-shaped path, the sink node periodically collects the aggregated data in CHs based on two linear and circular movements. Considering the sink movement in the CH selection process is another advantage of the proposed algorithm. Numerical results confirm the better performance of the proposed algorithm compared to the EDT algorithm.