Search

To solve the problem that maintenance tools are difficult to retrieve once being lost in maintenance of an electric section of the railway system, this study proposed a new method for estimating the object location optimized by a fuzzy neural network based on received signal strength indicator. Through a radio frequency identification reader and antennae, the received signal strength indicator values feed back by the dropped tools were collected. The environmental attenuation factor was fitted with field data. In the pre-processing stage, K-means algorithm optimized by the genetic algorithm, thus improving accuracy of data. The processed data were trained and tested by the fuzzy neural network, so that they could more accurately estimate the actual location of dropped tools. Data were collected on the railway lines of National Local Joint Engineering Research Center of Safety Guarantee Technology for Operation and Maintenance of Rail Transport Infrastructure. The collected data were used for subsequent processing and locating tools. The experimental results show that the proposed location method offers advantages in improving the location accuracy of dropped tools, which can improve the retrieval efficiency of tools in the field.

This article addresses the problem that the remaining useful life (RUL) prediction accuracy for a high-speed rail catenary is not accurate enough, leading to costly and time-consuming periodic planned and reactive maintenance costs. A new method for predicting the RUL of a catenary is proposed based on the Bayesian optimization stacking ensemble learning method. Taking the uplink and downlink catenary data of a high-speed railway line as an example, the preprocessed historical maintenance and maintenance data are input into the integrated prediction model of Bayesian hyperparameter optimization for training, and the root mean square error (RMSE) of the final optimized RUL prediction result is 0.068, with an R-square (R2) of 0.957, and a mean absolute error (MAE) of 0.053. The calculation example results show that the improved stacking ensemble algorithm improves the RMSE by 28.42%, 30.61% and 32.67% when compared with the extreme gradient boosting (XGBoost), support vector machine (SVM) and random forests (RF) algorithms, respectively. The improved accuracy prediction lays the foundation for targeted equipment maintenance and system maintenance performed before the catenary system fails, thus potentially saving both planned and reactive maintenance costs and time.

This paper presents a novel differential evolution (DE) algorithm, with its improved version (IDE) for the benchmark functions and the optimal reactive power dispatch (ORPD) problem. Minimization of the total active power loss is usually considered as the objective function of the ORPD problem. The constraints involved are generators, transformers tapings, shunt reactors, and other reactive power sources. The aim of this study is to discover the best vector of control variables to minimize power loss, under the premise of considering the constraints system. In the proposed IDE, a new initialization strategy is developed to construct the initial population for guaranteeing its quality and simultaneously maintaining its diversity. In addition, to enhance the convergence characteristic of the original DE, two kinds of self-adaptive adjustment strategies are employed to update the scaling factor and the crossover factor, respectively, in which the detailed information about the two factors can be exchanged for each generation dynamically. Numerical applications of different cases are carried out on several benchmark functions and two standard IEEE systems, i.e., 14-bus and 30-bus test systems. The results achieved by using the proposed IDE, compared with other optimization algorithms, are discussed and analyzed in detail. The obtained results demonstrated that the proposed IDE can successfully be used to deal with the ORPD problem.

To improve Asian food image classification accuracy, a method that combined Convolutional Block Attention Module (CBAM) with the Mobile NetV2, VGG16, and ResNet50 was proposed for Asian food image classification. Additionally, we proposed to use a mixed data enhancement algorithm (Mixup) to have a smoother discrimination ability. The effects of introducing the attention mechanism (CBAM) and using the mixed data enhancement algorithm (Mixup) were shown respectively through experimental comparison. The combination of these two and the final test set Top-1 accuracy rate reached 87.33%. Moreover, the information emphasized by CBAM was reflected through the visualization of the heat map. The results confirmed the classification method’s effectiveness and provided new ideas that improved Asian food image classification accuracy.

As a fine-grained classification problem, food image classification faces many difficulties in the specific implementation. Different countries and regions have different eating habits. In particular, Asian food images have a complicated structure, and the related classification methods are still very scarce. There is an urgent need to develop a feature extraction and fusion scheme based on the characteristics of Asian food images. To solve the above problems, we proposed an image classification model SLGC (SURF-Local and Global Color) that combines image segmentation and feature fusion. By studying the unique structure of Asian foods, the color features of the images are merged into the representation vectors in the local and global dimensions, respectively, thereby further enhancing the effect of feature extraction. The experimental results show that the SLGC model can express the intrinsic characteristics of Asian food images more comprehensively and improve classification accuracy.

A vehicle logo occupies a small proportion of a car and has different shapes. These characteristics bring difficulties to machine-vision-based vehicle logo detection. To improve the accuracy of vehicle logo detection in complex backgrounds, an improved YOLOv4 model was presented. Firstly, the CSPDenseNet was introduced to improve the backbone feature extraction network, and a shallow output layer was added to replenish the shallow information of small target. Then, the deformable convolution residual block was employed to reconstruct the neck structure to capture the various and irregular shape features. Finally, a new detection head based on a convolutional transformer block was proposed to reduce the influence of complex backgrounds on vehicle logo detection. Experimental results showed that the average accuracy of all categories in the VLD-45 dataset was 62.94%, which was 5.72% higher than the original model. It indicated that the improved model could perform well in vehicle logo detection.

Aiming at the problem that the RSSI value of the received signal strength indication (RSSI) ranging location is vulnerable to environmental interference, and the collected value has fluctuation, which leads to the reduction of localization accuracy, a localization algorithm based on genetic algorithm optimization clustering was proposed. Firstly, the field data collection was completed, and then the K-means clustering parameters were optimized by using genetic algorithm. The data set extracted by clustering was combined with the deep neural network to build a localization model for training and testing, so as to realize the accurate localization of the objects to be tested in different areas. Experimental results show that the proposed localization method has great advantages in improving the localization accuracy and reduce tedious environmental layouts.

The location selection of logistics distribution centers is a crucial issue in the modern urban logistics system. In order to achieve a more reasonable solution, an effective optimization algorithm is indispensable. In this paper, a new hybrid optimization algorithm named cuckoo search-differential evolution (CSDE) is proposed for logistics distribution center location problem. Differential evolution (DE) is incorporated into cuckoo search (CS) to improve the local searching ability of the algorithm. The CSDE evolves with a coevolutionary mechanism, which combines the Lévy flight of CS with the mutation operation of DE to generate solutions. In addition, the mutation operation of DE is modified dynamically. The mutation operation of DE varies under different searching stages. The proposed CSDE algorithm is tested on 10 benchmarking functions and applied in solving a logistics distribution center location problem. The performance of the CSDE is compared with several metaheuristic algorithms via the best solution, mean solution, and convergence speed. Experimental results show that CSDE performs better than or equal to CS, ICS, and some other metaheuristic algorithms, which reveals that the proposed CSDE is an effective and competitive algorithm for solving the logistics distribution center location problem.

In order to optimize the network coding resources in a multicast network, an improved adaptive quantum genetic algorithm (AM-QEA) was proposed. Firstly, the optimization problem was translated into a graph decomposition problem. Then the graph decomposition problem was represented by the binary coding, which can be processed by quantum genetic algorithm. At last, a multiple-operators based adaptive quantum genetic algorithm was proposed to optimize the network coding resources. In the algorithm, the individual fitness evaluation operator and population mutation adjustment operator were employed to solve the shortcomings of common quantum genetic algorithm, such as high convergence rate, easy to fall into local optimal solution and low diversity of the population in later stage. The experimental results under various topologies show that the proposed algorithm has the advantages of high multicast success rate, fast convergence speed and strong global search ability in resolving the network coding resource optimization problems.

In foggy weather, the occlusion of smoke dust and the variation of imaging scale of foreign objects will mislead the existing algorithm to learn the wrong target information, causing the drift of the tracking box. A kernel correlation filtering target tracking dimensionality reduction algorithm combining dark channel prior and scale estimation is proposed, effectively improve target tracking accuracy in foggy weather. Firstly, in the process of detecting intruding foreign objects along the railway with visual background extractor ViBe, the atmospheric scattering model based on dark channel prior was used to defog video sequence. After that, FHOG features of the initial tracking box were extracted by dense cyclic sampling and scale pyramid technology, and then a Kcf position filter, as well as a scaling filter with PCA dimensionality reduction was trained to realize the scale-adaptive rapid tracking of railway foreign objects in foggy weather. The experimental results show that in terms of tracking accuracy, the proposed Defog-PSA-Kcf algorithm is superior to the non-scale estimation link the generation algorithm Mean Shift devoid of scale estimation and the native Kernel Correlation Filter algorithm (Kcf), and while linear kernel algorithm Dual Correlation Filter algorithm (Dcf), which is higher than the scale-adaptive SA-Kcf and the SAMF algorithm; in terms of tracking speed, it's faster than the Mean Shift, SA-Kcf, and SAMF algorithms, and in tracking speed. The algorithm can achieve fast tracking results comparable to as fast as the Kcf algorithm.

The status of the Pantograph and Catenary is the guarantee for the safe operation of the railway. However, the traditional Pantograph and Catenary status judgment efficiency is not satisfactory, and it is impossible to timely repair the catenary, which may lead to the greater economic loss. In this paper, a new GA-ADNN-based (genetic algorithm-Adadelta deep neural network-based) optimization method for the prediction model for catenary comprehensive pantograph and catenary monitor (CPCM) status is proposed. According to the status values of the CPCM parameters such as height, stagger, hard point, contact force, and height difference within span, the status value of the pillars in the catenary has been calculated by the analytic hierarchy process, and then the prediction model for predicting catenary CPCM status has been established and then optimized by genetic algorithm to avoid prediction model falling into local optimum. Finally, the CPCM test parameters of each pillar of the catenary in the actual example are input and the CPCM status value of the corresponding pillar is predicted. With the smallest prediction error found, the genetic algorithm is used for optimization, the optimal learning rate of the prediction model is 0.0559, and the optimal number of the hidden layer of the CPCM status prediction model is determined to be 14. The experimental results show the feasibility of GA-ADNN-based prediction model for predicting the catenary CPCM status, and that compared with the support vector machine and traditional artificial neural network prediction methods, the GA-ADNN-based prediction model has higher prediction precision and better generalization ability.

The purpose of the study is to solve problems, i.e., increasingly significant processing delay of massive monitoring data and imbalanced tasks in the scheduling and monitoring center for a railway network. To tackle these problems, a method by using a smooth weighted round-robin scheduling based on backpressure flow control (BF-SWRR) is proposed. The method is developed based on a model for message queues and real-time streaming computing. By using telemetry data flow as input data sources, the fields of data sources are segmented into different sets by using a distributed model of stream computing parallel processing. Moreover, the round-robin (RR) scheduling method for the distributed server is improved. The parallelism, memory occupancy, and system delay are tested by taking a high-speed train section of a certain line as an example. The result showed that the BF-SWRR method for clusters can control the delay to within 1 s. When the parallelism of distributed clusters is set to 8, occupancy rates of the CPU and memory can be decreased by about 15%. In this way, the overall load of the cluster during stream computing is more balanced.

Trying to provide a medical data visualization analysis tool, the machine learning methods are introduced to classify the malignant neoplasm of lung within the medical database MIMIC-III (Medical Information Mart for Intensive Care III, USA). The K-Nearest Neighbor (KNN), Support Vector Machine (SVM) and Random Forest (RF) are selected as the predictive tool. Based on the experimental result, the machine learning predictive tools are integrated into the medical data visualization analysis platform. The platform software can provide a flexible medical data visualization analysis tool for the doctors. The related practice indicates that visualization analysis result can be generated based on simple steps for the doctors to do some research work on the data accumulated in hospital, even they have not taken special data analysis training.

Buildings account for a significant portion of the world’s electricity consumption, including industrial, commercial, residential, hospital buildings, etc. During building operation, significant energy consumption and economic loss may be caused due to various forms of electricity theft. To this end, an ensemble model for electricity theft detection method based on genetic optimization is developed. At first, synthetic samples that have an approximate distribution to that of true samples of electricity theft are prepared through use of a synthetic minority oversampling technique (SMOTE). Afterwards, features of anomalous electricity consumption are extracted with the aid of dimension reduction through principal component analysis (PCA). Finally, an ensemble deep learning network based on AdaBoost is established to mine implicit information in continuous time series data and detect anomalous electricity consumption of labelled users. Moreover, the hyperparameters of the deep neural network are optimized based on a genetic algorithm (GA). By taking data pertaining to the electricity consumption of users collected from smart meters of the State Grid Corporation of China as examples, the results show that our model is superior to other detection methods (such as support vector machine (SVM), random forest (RF), and traditional artificial neural network (ANN)) in the sensitivity and the area under the curve (AUC).

Electric power makes a significant contribution to society. Predicting power generation is becoming increasingly important for electric power planning and energy utilization. A reliable forecasting model is necessary for accurate planning of electricity generation. The main goal of this study is to develop effective and realistic solutions for the full-load power generation prediction of combined cycle power plants. According to 9568 items of data pertaining to a combined cycle power plant in six years of its full-load operation, a prediction method based on stacking ensemble hyperparameter optimization is established. The results demonstrate that this method provides high prediction accuracy for the power plant under multiple complex environmental variables. Besides, the predictions generated using this method are compared with those of traditional machine learning methods, random forest, and other ensemble methods, as well as those cited in the literature using the same dataset. The predictions show that the proposed method offers more accurate predictions of the power generation from a combined cycle plant, which opens up a new idea for power planning and energy utilization.

In order to obtain the information of the vehicle tags in adverse traffic conditions, we proposed a novel reservation framework named reservation to cancel idle-dynamic frame slotted ALOHA (RTCI-DFSA) algorithm. Firstly, the framework employed reservation mechanism to remove idle slot, and thus improve the system identification efficiency. Secondly, the vehicle information was identified by the tag serialization polling identification method. The experimental results showed that the proposed RTCI-DFSA algorithm performed better than the traditional frame slotted ALOHA (FSA) and dynamic frame slotted ALOHA (DFSA) algorithms. More specifically, the tag loss rate of the proposed framework is significantly lower than the frame length fixed and conventional dynamic vehicle identification algorithms. In addition, the experimental results demonstrated that the throughput rate of the proposed algorithm increased from 0.368 to 0.6. Besides, the identification efficiency and applicability of the proposed framework were both higher than other tag identification algorithms.

Accurate traffic flow data is crucial for traffic control and management in an intelligent transportation system (ITS), and thus traffic flow prediction research attracts significant attention in the transportation community. Previous studies have suggested that raw traffic flow data may be contaminated by noises caused by unexpected reasons (e.g., loop detector damage, roadway maintenance, etc.), which may degrade traffic flow prediction accuracy. To address this issue, we proposed an ensemble framework via ensemble empirical mode decomposition (EEMD) and artificial neural network (ANN) to predict traffic flow under different time intervals ahead. More specifically, the proposed framework firstly employed the EEMD model to suppress the noises in the raw traffic data, which were then processed to predict traffic flow at time steps under different time scales (i.e., 1, 2, and 10 min). We verified our model performance on three loop detectors&rsquo, data, which were supported by the Department of Transportation, Minnesota. The research findings can help traffic participants collect more accurate traffic flow data and thus benefits transportation practitioners by helping them to make more reasonable traffic decisions.

Time headway distributions play an important role in analyzing driver behavior prediction, road service evaluation, traffic state estimation, et al. Thus, less attention was paid on the start-up time headway distributions at urban signalized intersections. To bridge the gap, we proposed a novel framework with three-parameter burr model and log-logistic distribution model to analyze vehicle start-up time headway distribution characteristics at signalized intersections. First, we collected a large amount of start-up time headway samples at the signalized intersection located at Guangzhou city in China. Second, all the vehicle start-up sample data have been processed and analyzed with three-parameter Burr model and log-logistic distribution curves. Third, we explored vehicle start-up time headway distribution characteristics by considering and regard less of queuing positions. Last but not least, we have applied our model to analyze saturated time headway distributions, and estimate minimum green time for vehicle queues passing through intersections. The findings of our research can help transport authorities set more reasonable green time at intersections, and improve urban transportation commuting efficiency.

The capacity and the scale of smart substation are expanding constantly, with the characteristics of information digitization and automation, leading to a quantitative trend of data. Aiming at the existing processing shortages in the big data processing, the query and analysis of smart substation, a data compression processing method is proposed for analyzing smart substation and Hive. Experimental results show that the compression ratio and query time of RCFile storage format are better than those of TextFile and SequenceFile. The query efficiency is improved for data compressed by Deflate, Gzip and Lzo compression formats. The results verify the correctness of adjacent speedup defined as the index of cluster efficiency. Results also prove that the method has a significant theoretical and practical value for big data processing of smart substation.

Aiming at the existing problem of monitoring data volume blowout and lack of valid real-time distributed fault tolerant strategy in power distribution network, a new fault tolerant strategy based on Spark micro batch computing model is proposed. Depending on the design technique of resilient distributed data set renewal and lineage mechanism, the lineage chain retrospection in fault tolerance of distributed data is realized. This strategy can meet the real-time fault tolerant performance requirements of monitoring big data, and it is of evident practical value for improving the fault tolerant performance of dispatching system in intelligent power distribution network.

Big data in smart grid dispatch monitoring systems is susceptible to interference from processing delays and slow response times. Hence, a new fault-tolerant flexible lossless cluster compression method is proposed. This paper presents the Five-tuples (S, D, O, T, M) model, and builds a monitoring data processing platform based on hive. By deploying the dispatch host and monitoring servers under the cloud computing environment, where data nodes are respectively transformed by Deflate, Gzip, BZip2 and LZO lossless compression method. Taking the power dispatch automation system of Long-hai line as example, experimental results show that the cluster lossless compression ratio of BZip2 is greater than 81%; when data records reach twelve million, the compression ratio can be further improved to certain extent by using RCFile storage hive format, which has significant flexible features. Therefore, the new method proposed in this paper can improve the flexibility and fault-tolerant ability of big monitoring data processing in smart grid.

This paper investigates least-cost QoS multicast routing problem in IP/DWDM optical networks, and proposes an improved evolutionary algorithm (AEQEA). Based on quantum-inspired evolutionary algorithm (QEA) with quantum rotation gate strategy, AEQEA introduces adaptive evolution mechanism (AEM), which allows each chromosome in a population to update itself to a fitter position according to its own situation. In term of this mechanism, AEQEA can significantly improve its capability of exploration and exploitation, since every chromosome is able to be allocated with suitable evolutionary parameters before each update. A repair method is applied to eliminate illegal graphs as many as possible hence more excellent solutions will appear in each evolutionary generation. Simulations are carried out over a number of network topologies. And the results show that, for the QoS multicasting problem, AEQEA outperforms other existing heuristic algorithms and is characterized by robustness, high success ratio, fast convergence and excellent capability on global searching.

This paper concentrates on the problem of minimizing network coding resources with end-to-end delay and delay variation constraints in multimedia network. Network coding resources optimization problem has proved to be NP-hard, and the introduction of delay constraints makes it more complicated to be solved, since essentially the single problem of constructing such delay constrained tree without combining coding resources optimization, is NP-complete. In the process of tackling this problem, the basic model of the multiple-choice knapsack problem (MCKP) is used as an optimized entity for the proposed genetic algorithm (GA) called routing-based encoding GA (REGA). Then numeric results over three network scenarios demonstrate the effectiveness of the presented algorithm, which shows some potential and helpful capacity on topology designing with higher QoS level in practical multimedia network.

In this study, cholesteric liquid crystal (Ch-LCs) composite with a double-handed circularly polarized light reflection band that contains hydrogen-bonded (H-bonded) polymer network was fabricated by a polymer template method, in which the environmental sensitivity and reversibility of hydrogen bond and memory effect of the polymer template were incorporated into the polymer stabilized cholesteric liquid crystal (PSCLCs) film. The thermal and electrical tunability of the Ch-LCs composite before and after irradiation were investigated. It has been demonstrated that reflection characteristics of the Ch-LCs composite derived from H-bonded chiral monomer (HBCM) after irradiation exhibited strong temperature sensitivity with temperature change, which we believe is due to that helical twisting power value of both HBCM and cholesteryl additives increased with an increasing temperature. Additionally, the reflective wavelength of the Ch-LCs composite before and after irradiation can be electrically switched to reflect green from the initial state reflecting a red color. The modulating mechanism is due to Helfrich deformation in which the tilt of helices axis in the Ch-LCs composite was induced when a voltage was applied. The technique developed in this study has great applications such as tunable lasers, optical sensors and LCs displays.

Summary form only given. Optical-layer multicast will be a promising solution for high-speed and high-bandwidth multicast applications. While network coding can achieve maximum multicast rate by allowing intermediate network nodes to combine the data received from different incoming links. Therefore, how to apply network coding to optical-layer multicast has been attracting much attention. Due to the limitation of optical buffering and optical processing capabilities, several critical issues should be addressed, e.g., constructing an optical multicast tree with minimum coding resource and finding a network coding scheme realized by practical optical hardware. In this talk, we introduce our research and development activities on the optical-layer multicast based on networking coding, which is partly supported by NSFC (No. 60772024). Firstly, we discuss an improved quantum-inspired evolutionary algorithm that allows different chromosomes to select different updating parameter at each generation. Its efficiency is verified with high success ratio and small number of coding nodes in constructing a multicast tree. Then, we present a network coding scheme that recoveries the original signals at each coding node. The kernel function can be realized by logical bit shifting and bitwise XOR in optical domain. Finally, we build an experimental network, on which optical multicast via network coding is successfully demonstrated.

In order to solve main technology and application problems of human machine interface in power automation system, this paper studies scalable vector graphics and AJAX engine and designs new power automation monitor human machine interface system based on Web2.0 network architecture. Interactive cases, tele-signal process sequence, tele-metering collaborative process, tele-control interactive process of supervisory control and data acquisition system has been designed using unified modeling language and incremental iteration model, activity diagram of alarm function has been completed. At last, this paper completes architecture of this system, and achieves web remote terminal system that has been actually put into use, and resolves confusion of graphic format and real-time data inefficient of monitoring system. SVG is an expansion of two-dimensional vector graphics format based on eXtensible Marker Language (XML), SVG is designed to meet the growing web application for the dynamic, scalable complexity of content, and realizes interactive and flexible demands. In this paper, SVG will be used in human machine interface power automation system to upgrade graphics performance and operation ability of web-based power application system, SVG provides a network standard for electric power automated monitoring system

In this study, novel photopolymerizable chiral hydrogen-bonded self-assembled complexes (PCHSCs) were fabricated, which were derived from photopolymerizable 4-(6-acryloyloxyhexyloxy) benzoic acid (AHBA, proton donor) and chiral pyridine derivatives (proton acceptor). Their structures were characterized by fourier transform infrared (FT-IR) and the proton nuclear magnetic resonance (1H-NMR) spectoscopy. The thermal stability, phase behaviors and helical twisted power (HTP) characteristics of the PCHSC were investigated by measuring the variable-temperature FT-IR spectrum, differential scanning calorimetry (DSC), polarizing optical microscopy (POM) and the Cano wedge. The results demonstrate that all the PCHSCs have good thermal stability within a temperature range, and the pitch length of all the cells containing the PCHSCs increases with increasing temperature, which is due to the fact that the HTP values of all the PCHSCs decrease with an increase of temperature. What's more, the introduction of AHBA leads to chiral enhancement of the PCHSCs. Based on the above results, a polymer stabilized cholesteric liquid crystals (PSCLCs) composite with the above PCHSCs was prepared and the thermal-optical characteristics of the PSCLCs film were investigated. The results confirm that the reflective wavelength of the PSCLCs film before and after irradiation can be thermally switched to reflect green and red color from the initial state reflecting a blue/green color with the temperature increasing from 30 °C to 75 °C. On the basis of this mechanism, the novel material in this study can be used as optical/photonic paper, optical sensors and LCs displays, etc.