This PDF file contains the front matter associated with SPIE Proceedings Volume 13286, including the Title Page, Copyright information, Table of Contents, and Committee Page.

The new power IGBT module is widely used in new application scenarios such as high voltage and high frequency, which is an important research direction of the development of information science. In this paper, the influence of uneven temperature between multiple chips on the short-circuit operation of IGBT module is analyzed, a three-dimensional model is established and analyzed by finite element method. After creating the temperature condition of uneven temperature distribution caused by the degradation of some heat dissipation conditions, the temperature distribution and heat flux distribution of IGBT module under this temperature condition are simulated, This paper analyzes the influence of uneven temperature between multiple chips in IGBT module on the short-circuit operation of IGBT module, and puts forward some reliability suggestions for the wide application of IGBT in high-frequency and high-voltage occasions.

Based on the 0.18 micron CMOS process, a two-stage Miller compensation operation transimpedance amplifier circuit was designed, the influence of temperature change on its performance parameters was studied, the temperature simulation analysis within the limit operating temperature range was completed, and the statistical performance analysis of the device parameter changes was carried out by using the process angle simulation method.The simulation results show that the main performance parameters of the op amp can meet the design indicators in the extreme temperature range, which proves the temperature adaptability of the circuit. Process angle simulation improves the reliability and performance range of the circuit.

To proactively address the disruptions in chip supply, the State Grid Corporation of China (SGCC) has undertaken comprehensive measures in line with the requirements set forth in the Regulations on the Protection of Critical Information Infrastructure by the State Council. Starting from November 2019, the National Power Dispatch and Communication Center initiated efforts to localize and replace substation automation systems. This involved conducting research on domestic chip options, achieving breakthroughs in key alternative technologies, developing equipment with entirely domestic components, conducting specialized testing, integrating with the power grid, and carrying out large-scale deployment. These endeavours have significantly enhanced the autonomy and controllability of the substation automation system, thereby comprehensively improving the level of prevention and control of grid security risks.

This paper discusses a tower inclination measurement system based on inclination sensors, the system can be collected to the tower real-time tilt angle, and then through the fiber optic transmission to the ground base station, so as to provide a certain guarantee for the maintenance personnel to better analyze the tower force situation. Through the tower inclination measurement system based on the inclination sensor, using the magnetic field data collected by the sensor, using Newton iteration method to analyze the sensor attitude information in real time. According to the simulation experiment results, the average error of the three attitude angles is less than 1°, the error is small, which can be widely used in the tower inclination measurement.

Combined with the local climatic conditions in Hainan, the theoretical data simulated using PVsyst software is compared with the data of the empirical project. Under the irradiation value and temperature conditions, the power generation of different models of modules is analyzed, and the experimental results obtained from the simulation of PVsyst software show that: for different models of double-sided double-glazed photovoltaic modules, under the same conditions, the Ntype modules, compared with the P-type modules, have an annual power generation gain of 5.1%. The simulation data (5.1%) is similar to the power plant operation data (5.2%), which is supportive of the theoretical experimental results.

With cooperative vehicle infrastructure system, information exchange can occur between vehicles and road side units (RSU), and there is potential for the improvement of communication benefits brought about by vehicle road collaboration to be explored. This article reduces the cost of roadside unit information dissemination through vehicle to vehicle (V2V) communication. The roadside units participating in the cooperation each distribute a portion of the message, while the vehicles exchange information during aggregation, thereby achieving information sharing. However, cooperation between roadside units can also lead to increased costs. To address this issue, this article establishes a collaborative vehicle road communication scenario and proposes a core network side alliance partitioning method based on message passing neural network (MPNN). The effectiveness of the algorithm is verified through simulation.

Conventional dynamic scheduling methods for power system query engine tasks mainly use MEC (Mobile Edge Computing) multi mobile edge computing method for scheduling decisions, which is vulnerable to changes in the scale of the scheduling network, resulting in poor dynamic scheduling performance. Therefore, it is necessary to design a new dynamic scheduling method for power system query engine tasks based on open source Hongmeng. That is to say, we use open source Hongmeng to build the dynamic collaborative scheduling architecture of the power system query engine task, and design the dynamic scheduling algorithm of the power system query engine task, thus realizing the dynamic scheduling of the power system query engine task. The experimental results show that the designed dynamic task scheduling method for the open source Hongmeng query engine in the power system has low scheduling load, MAPE prediction error, scheduling delay, scheduling task start time, and a high number of scheduling tasks, which proves that the designed dynamic task scheduling method for the query engine in the power system has good scheduling effect, reliability, and certain application value, It has made certain contribution to improving the comprehensive performance of power system operation.

With the increasing integration of renewable energy into the power grid, the demand for efficient communication systems for managing and controlling distributed power sources has become urgent. Here, we introduce the design and implementation of a 5G communication terminal tailored specifically for the unique requirements of distributed power sources. The proposed terminal aims to improve communication reliability, latency, and throughput, thereby optimizing the overall performance of the distribution network, and designs a set of self-feedback fault elimination. The system architecture, key features, and experimental results were discussed in detail, demonstrating the feasibility and effectiveness of the proposed solution.

A branch prediction scheme for the LoongArch five-stage pipeline architecture is proposed. This scheme is compatible with cache-equipped designs, predicting the branch direction and target address of branch instructions after fetching them from the high-speed cache during the instruction fetch stage. It can recover from branch prediction failures while ensuring the correct execution of the instruction pipeline. After software simulation and FPGA board testing, this scheme meets the design requirements, has high prediction efficiency, and can significantly enhance the performance of the LoongArch five-stage pipeline CPU.

As the usage of wireless devices has increased exponentially, deployments of IEEE 802.11 WLANs has been significantly increasing and hence is bound to cause higher interference and reduced throughputs, particularly in dense-deployments. Though a network simulator like Komondor captures comprehensive deployment details of WLANs to provide accurate performance metrics, the high compute time and the concomitant complexity of such simulators cannot be disregarded. Deep learning techniques prove to be the ideal alternative to be able to quickly compute the throughput of a WLAN deployment with reduced compute time and complexity. In this paper we implement a Graph Isomorphism Network (GIN) for predicting the throughputs of BSSs in dense deployments. The regression model is trained on the data obtained from the Komondor network simulator for a several comprehensive set of deployments. As the topology of the deployments mimic a graph network, the capability of GIN is explored for mapping the deployments and to accurately predict the throughput of a densely deployed WLAN.

In this paper, a low power startup circuit is designed to ensure the circuit can be started with a supply voltage of 2.8-6V. The performance of the circuit is guaranteed by the use of a series of trimming circuits. The circuit adopts 180nm BCD technology and operates at a 5V power supply voltage. Simulation results show that the temperature coefficient is 1.3ppm/°C from -40 °C to 125 °C. In low frequencies, the power supply rejection ratio is 110dB. In the range of 2.8V-6V power supply voltage, the output reference voltage value is 2.501V.

Cognitive radio technology is a new communication technology to cope with the increasing demand and inefficient use of radio spectrum. This paper summarizes the historical development of CR technology, from its initial concept to the present research status and shortcomings. This paper mainly introduces spectrum detection and allocation technology in CR network and some applications of CR technology in real life. This paper makes up for the lack of review papers on the development and application of the latest spectrum technology of CR in recent years. In terms of future challenges, this paper analyzes the environmental and ethical issues faced by this technology in recent years, some research bottlenecks in 6G technology and some breakthroughs in the field of satellite communication.

Multi classification tasks are an important direction for the application of classical convolutional neural networks. However, in the application research of quantum convolutional neural networks, due to limitations such as quantum bits, a large amount of research has focused on binary and ternary classification tasks, while few have studied the classification of four or more categories. This article presents an improved quantum neural network and uses it to accomplish multi classification tasks. We use the Pennylane library to implement our hybrid quantum classical neural network model, and experimental results show that our model performs slightly better than classical convolutional neural networks with comparable parameter quantities. We hope that our findings can provide some inspiration for the application research of quantum convolutional neural networks in the future.

Wireless Power transfer (WPT) technology has received a lot of attention in recent years for its potential to revolutionize the way electronic devices are charged and powered. The compensation topology composed of inductance and capacitance is the key part of wireless charging system. In this paper, a wireless charging circuit based on LCC-S compensation topology is proposed, and its output characteristics under resonant conditions are analyzed, and a simulation model is built for simulation verification. The results show that the circuit can realize constant current charging under resonant conditions.

In this paper, we designed a compact printed dipole antenna with multiple bands. This novel triple-band dipole antenna appears preeminent radiation performance and working bandwidth. Compare with the related early work, the dipole antenna is not merely a multi-band design, but also, a miniaturize result. It consists of two long branches and one short branch, which induce resonances at 2GHz, 2.37GHz and 3.3GHz band. The low band is generated by two long branches. Moreover, the short branches generates other two resonances with different long branch respectively. These complex resonances finally provide extended frequency bands. The design in present paper is suitable for most 5th generation mobile terminals and WLAN terminals.

The patch shape of microstrip patch antenna is an important factor affecting the performance of the antenna, in order to make the microwave antenna miniaturization, high gain performance requirements, respectively, k-band operating frequency at 24GHz rectangular and circular coaxial fed microstrip antenna simulation, compare the performance and size of the circular and rectangular microstrip antenna, the dielectric substrate using Rogers 4350, the dielectric constant of 3.66, the thickness of substrate is 0.5mm. The simulation results show that the circular coaxial fed microstrip antenna operating at 24GHz is superior to the rectangular microstrip antenna in terms of size and performance, and this structure can be used in the development of small microwave sensors for the transmission and reception of microwave signals.

This paper discusses the application of digitalization technology for electrical substation meters based on the Internet of Things (IoT). With the advancement of new power systems and the development of smarter, greener energy internet, smart substations and intelligent operation and maintenance are key future directions. Digital remote metering technology is crucial for achieving smart substations, offering real-time monitoring and management of power equipment, enhancing safety, stability, reliability, and economic efficiency. The paper outlines the design of digital meters, including mechanical and digital components. The paper presents A system architecture design and selection of communication protocols. The article concludes with the advantages of IoT-based digital meters, such as flexible wireless networking, bi-directional communication, and preciser monitoring, and suggests areas for future research. The proposed solution in this article serves as a valuable reference for the digital upgrade of current substations and offers a clear technological path for the construction of new substations.

This paper constructs a construction management system of overhead transmission line based on B/S architecture, and uses Internet technology to improve the efficiency of construction management. The system updates data in real time and optimizes resource allocation through a variety of key modules. The results show that the system significantly improves construction efficiency and safety through a user-friendly interface, strong data security and system reliability.

This paper explores high-voltage switching equipment secondary circuit technology based on domain-based control, aimed at overcoming the technical challenges present in current traditional and intelligent substations. Traditional high-voltage switch secondary control circuits employ a multitude of discrete electromagnetic components, which not only increases system complexity but also reduces reliability and complicates maintenance due to the numerous connecting wires involved. Moreover, the variety of product specifications for high-voltage switches, lacking standardization, adds to the design and operational difficulties. By integrating a domain-based control system, a segmented control network system based on circuit breaker control domains and isolator control domains is proposed. This research demonstrates how reducing the use of discrete components and wiring can enhance the overall reliability and rapid-protective response of the system. Experimental results indicate that domain-based control not only optimizes control strategies and response speeds but also significantly reduces maintenance costs and complexity through standardized design. The findings of this study hold significant theoretical and practical relevance for the design and operation of modernized substations.

Based on the analysis of the current situation and management requirements of geological data, the modeling process of spatial data is elaborated. Based on the characteristics of spatial data, the logical model of geological vector data is abstracted and designed. A physical model is proposed, and problems related to the physical storage of data such as schema partitioning, security, index and hash partitioning are explored. It realizes the function of geological feature model management and service based on cloud native, provides model support for the integrated management and service of geological data, and improves the ability of feature-based geological data management and productization service.

The Power Grid Business Data Center integrates diverse data sources, offering efficient processing and service capabilities for grid operations. This article discusses the center’s architecture, covering the data collection, storage, processing, and output layers, and explores key technologies like real-time processing and data security. Implementing this center has enhanced grid efficiency and reliability, reduced fault response times, and cut costs, providing a valuable reference for big data applications in the power grid sector.

When the temperature of a submarine cable spikes unexpectedly, prompt repairs are crucial to prevent a thermal breakdown accident that could compromise production safety and result in substantial economic losses. Monitoring and diagnosing real-time temperature data of submarine cables holds significant practical engineering importance and economic value. To this end, we propose an innovative method for detecting abnormal temperature rises that leverages eigenvalue amplification. This technique effectively enhances the visibility of abnormal values, making their characteristics more pronounced. Our approach enables early detection of abnormal temperature spikes in submarine cables, ensuring a high level of accuracy in identifying temperature differences. Furthermore, our exponential characteristics-based monitoring and simulation system for submarine cables offers robust protection, meeting the stringent requirements for both accuracy and real-time monitoring.

This research aims to solve the problems of traditional temperature measurement systems such as cumbersome operation, slow response time, and inability to detect remotely, and designs a wireless temperature measurement system based on STM32. This system integrates a non-contact temperature sensor, LCD display and voice broadcast module. The system uses the STM32 microcontroller as the core control unit, which can monitor ambient temperature, humidity and object temperature data in real time to achieve comprehensive monitoring of the environment. When the detected temperature value exceeds the preset threshold, the system proactively issues an alarm. In addition, the system can upload environmental parameters to the cloud through network connection to achieve remote monitoring. The system realizes the detection and alarm functions of abnormal conditions in the industrial product environment, effectively improves the monitoring level and safety of the traditional environmental temperature measurement system, and provides technical support for the intelligent management of the industrial product environment.

A solar-powered wireless fire monitoring system based on LoRa (Long Rage) technology is designed to solve the problems of high laying cost and difficult construction of communication cable of traditional fire monitoring system, which can not meet the special occasions such as outdoor. The system includes power module, controller, communication module, probe node, sink node, server and display terminal. The power circuit is self-powered by solar energy, and the detection node is used to collect data, and the collected data is sent to the controller, then sent to the aggregation node through the communication module, and finally sent to the server through Wireless Fidelity, and the summary display. The system can realize outdoor reliable monitoring and improve the intelligence of fire prevention.

This paper introduces a multivariate time series forecasting model that combines a sliding window technique with machine learning, incorporating a convolutional neural network to extract spatio-temporal features. This integration boosts predictive accuracy and robustness. Applied to the Jilin credit card installment dataset, our SHRF method improves AUC PRECISION and F1 scores. Validations on the BeijingAirQuality dataset using the SHL model show our model reduces RMSE per time step to 15.18, demonstrating excellent performance in multivariable time series forecasting.

Brake system is a crucial component for aircraft landing deceleration, and its malfunction poses a direct threat to landing safety. Currently, brake system is predominantly diagnosed post-event using the fault tree analysis method, which has limited effectiveness in preventing failures. Based on vast amount of experimental data accumulated from aircraft, this paper utilizes the LOF (Local Outlier Factor) algorithm to calculate the local reachability density of the experimental data points, thereby obtaining the LOF distribution for the normal operation of the antiskid brake system. On this basis, the local reachability density of new experimental points is calculated to identify whether they are abnormal outliers, thus enabling the detection of abnormal states in the antiskid brake system. The results of the case study demonstrate that this method can effectively identify abnormal states in the antiskid brake system. By isolating and rectifying these detected abnormalities using engineering expertise, the frequency of failures during brake system operation can be significantly reduced, thereby enhancing fault prevention measures and offering promising application prospects.

This study introduces a novel approach for wind power forecasting, addressing the unpredictability of wind energy production and the issue of power curtailment. We developed a hybrid model combining the strengths of Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Transformer models, tailored to enhance the forecasting accuracy of wind power time series data. Through extensive experiments with data from two wind farms in Gansu Province, our model demonstrated superior performance over traditional models (ANN, SVR, RR, RF, LR) across four seasons, evidenced by lower RMSE and MAE values. Additionally, we proposed a systematic approach to manage power curtailment, effectively recovering curtailed power output values and ensuring data integrity for predictive modeling. Our findings not only contribute to the advancement of wind power forecasting methodologies but also highlight the potential for integrating advanced machine learning techniques to improve renewable energy management and grid stability.

To understand the research on big data monitoring technology in power systems, a study focusing on power system big data monitoring technology based on digital twins is proposed. This paper begins by introducing the concept, architecture, characteristics, and core technologies of digital twins, followed by a comprehensive analysis of the application system of a digital twin power grid. Additionally, it explores the potential applications of an electric power digital twin. The primary objective of an electric power digital twin is to enhance the effective use of data streams and integrate real-time situational awareness. This approach aims to achieve a thorough understanding of the power system, thereby aiding in regulatory decision-making processes.

In this paper, an artificial vision-based measurement method of concentrated light spot is proposed, and the experimental study is carried out on the solar furnace in Yanqing base. The experimental results show that the energy flow distribution measured by the indirect measurement method based on artificial vision is more accurate, and the internal energy of the spot conforms to Gaussian distribution. At the same time, the indirect measurement method based on artificial vision considers taking real-time gray photos directly, and the results obtained by visual processing are not affected by the specific climate and environment. The system is simple and easy to operate, and it is more suitable for popularization and application.

The detection of defects and status changes in substation equipment has always been one of the important tasks of substation inspection. With the gradual expansion of substation scale, detecting defects and status changes in substation equipment, accurately and timely discovering safety hazards in substations, has become an urgent problem to be solved. The article introduces a siamese YOLO network based algorithm for detecting equipment defects and changes to tackle the challenges posed by the variety of equipment types, the intricacies of defect patterns, and the challenge of distinguishing between background and foreground in substation equipment inspections. Firstly, a siamese YOLO feature extraction module is proposed to extract target features from the normal image and test image of the device to be recognized, ignoring background information. Subsequently, a feature fusion module is devised to merge the target features with the differential features. Finally, an upsampling module is proposed to upsample multi-scale fusion features, utilize adaptive attention methods for fusion, generate differential masks between normal images and test images, and detect and identify substation equipment defects or state changes. Using on-site image data collected from substations for verification, the experimental results show that the proposed method can fully utilize the difference information before and after equipment defects or status changes in substations, achieve high-precision detection of equipment defects and status changes, and apply it to the field of intelligent inspection in substations to effectively improve inspection efficiency and accuracy.

The routing, depth, and location of submarine cables are the key to ensure the maintenance and inspection of submarine cables, so this paper designs a submarine cable localization system based on a dual triaxial fluxgate sensor array. The system uses the AC magnetic field method for submarine cable positioning, and the relationship between the size of the induced electromotive force generated and the distance of the sensor is analyzed by establishing the electromagnetic field model of the energized cable. In this paper, a dual triaxial fluxgate sensor array scheme is adopted, on the basis of which the spatial model of the sensor array and the sea cable is constructed, and the horizontal and vertical offsets of the sea cable relative to the center position of the array are deduced. The overall architecture of the system consists of the sensor array, the multi-channel data acquisition module and the software of the upper computer, which work together to realize the conversion, acquisition, transmission and processing of magnetic signals, and finally accurately solve the position of the sea cable. And the positioning accuracy and positioning range test, its measurement error are satisfied with the system requirements.

In the last decade, network attacks frequently happened in power system, due to significant economic losses to global industrial production. Malware plays an important role in cyber attacks against power system. This paper puts forward EGemini- V2, which is a developed version of our previous research E-Gemini. E-Gemini-V2 model not only expands the contents of E-ACFG proposed in E-Gemini to describe malware, but also improves the detection ability of GAT network. Experiments show that our development’s effectiveness compared to earlier versions. The best accuracy can achieve 91.5%, which is sharply promoted compared to E-Gemini and Gemini models.

This paper investigates a signal source system for countering civilian unmanned aerial vehicles (UAVs). The system is designed based on a software-defined radio platform to modulate signals. It generates comb spectrum signals through a frequency multiplier system, providing broad-spectrum noise signals covering the entire frequency band (ISM-2.4GHz) used for communication between UAVs and ground stations, as well as pseudo-GPS signals with preset location information. This paper mainly focuses on the system design of the interference source, software-defined radio design, and frequency multiplier system design. The proposed system solution is feasible and can be applied to UAV countermeasure systems based on radio interference, and can also be extended to the research of other radio interference sources.

This paper proposes to build a high performance and reconfigurable face detection acceleration system, and uses the method of hardware and software collaboration to give full play to the advantages of ARM and FPGA. It accelerates the MTCNN cascaded deep convolutional neural network framework while ensuring recognition accuracy, and ultimately completes a face detection system based on ZYNQ. This design is mainly divided into two parts: software and hardware. In terms of software, facial detection adopts the MTCNN cascaded deep convolutional neural network framework. It uses a carefully designed three-layer cascaded deep convolutional network to predict facial positions and facial keypoint coordinates from rough to fine. The MTCNN model is trained on the PC using the Caffe framework, OpenCV visual library, and C++ language, using the WIDER FACE dataset for training. Afterwards, the system will be written and deployed on the ZYNQ 7020 SOC platform using standard C language (C99 standard) on the Xilinx SDK software. The hardware part combines the ADV7511 controller and video buffer to achieve real-time image display, displaying the images stored in the SD card. Using ARM Cortex-A9 hardcore processor and VDMA as AXI slave device, direct memory access of video data is achieved.

After an earthquake, rapid detection of collapsed building distribution from remote sensing images is crucial for government agencies, as it facilitates effective rescue operations and minimizes casualties. Traditional methods of damage assessment methods, such as on-site surveys and manual interpretation of remote sensing data, are often hindered by their labor-intensive and time-consuming nature, thus necessitating a more efficient approach. In this study, we address this challenge by employing a deep learning-based methodology for detecting collapsed buildings from satellite imagery. Specifically, we developed a Deep Learning model based on the U-net architecture, chosen for its high flexibility and performance. Utilizing a custom weighted loss function, our model achieved an Average Precision of 0.871 and an Average Recall of 0.893 on the test set. Application of the model to the Turkey earthquake case demonstrated rapid and accurate segmentation of most buildings. This study suggests that integrating U-net-based deep learning with satellite images can provide the precise distribution of collapsed buildings needed for earthquake emergency management.

Parabolic trough solar thermal power system (PTSTPS) is a kind of renewable energy technology, which can not only bear a large proportion of the basic power load, but also bear the flexible peak regulation of the grid. Its generation prediction is an important basis for the design, investment, and operation of power station. Based on the hourly meteorological resources data of a 50MW PTSTPS in China from 2017 to 2021, this paper firstly conducted abnormal data elimination and normalization operations, and then conducted correlation analysis to determine the final training samples and test samples. In this paper, an LSTM neural network is constructed to predict hourly power generation of the PTSTPS, and the model is tested and optimized by using test samples. Finally, the predicted results are compared with hourly, daily, and monthly power generation calculated by SAM software. The relative errors are -8.96%-8.88%, -3.23%-4.07% and -0.35%- 0.33%, respectively. Therefore, the constructed LSTM neural network is suitable for the prediction of the electricity generation of the PTSTPS, and provides a new technical way for the real-time and accurate electricity generation prediction of the solar thermal power system.

Electric vehicle charging pile is one of the energy Internet entrances, carrying the important missions of charging power supply, charging metering and charging, vehicle charging safety and charging data interconnection, etc. According to the characteristics of its embedded operating system, an improved fuzzy testing method based on blot analysis is proposed. The method firstly analyzes the attack surface of the firmware from the Angle of vulnerability exploitation, then derives the corresponding security rules according to the attack surface, carries on the stain analysis, and finally generates the fuzzy test case set according to the results of the stain analysis.

Heating furnace is an important heating equipment in the modern industrial field, is a nonlinear, time-varying, hysteresis and strong coupling of the controlled object, the traditional PID control and fuzzy control theory based on fuzzy control of the rectification of the PID control method in the temperature control of the adaptability of the poor, the effect is not ideal. Based on the improvement of genetic algorithms of fuzzy PID control and the simulation comparison with the conventional PID, fuzzy PID, and achieve a better control effect.

Aiming at the limitations of a single sensor in complex environments, this paper utilizes the joint calibration and data fusion techniques of LIDAR and camera to make up for the shortcomings of a single sensor. The physical points of LiDAR in the 3D coordinate system are associated with the image points under the pixel coordinate system of the camera through the checkerboard grid calibration method, and then based on the parameters of the joint calibration, the 3D point cloud is projected to the 2D plane and the coordinate mapping is established with the image data of the camera to obtain the point cloud with color information. Finally, the experimental validation of projecting the point cloud data onto the imaging plane confirms the remarkable feasibility and accuracy of the data fusion method for LiDAR and optical imaging sensors.

The silicon drift detector, known for its high resolution and low noise as a semiconductor detector, is extensively utilized in fields like X-ray spectroscopy analysis and particle energy detection. The performance of the detector is directly impacted by the magnitude of its leakage current. Illumination is one of the significant factors affecting the leakage current in silicon drift detectors. This work designs a 1000um×300um concentric silicon drift detector, employs Sentaurus TCAD to analyze the potential distribution within the SDD, and methodically investigates the impact of factors including light intensity, wavelength, and angle of incidence on the SDD's leakage current. Additionally, the study analyzes the detector's light pulse response and examines the physical mechanisms underlying the detector's light response, offering a theoretical foundation for further optimizing the detector's design and enhancing its performance.

Nowadays, large language models (LLM) have made significant progress in understanding and generating natural language and have become an important focus of research in the field of artificial intelligence. The capabilities of these models have been validated in a variety of applications, including machine translation, question-answer systems, and text generation. However, there are still challenges in understanding and processing spatial data, e.g., perception of abstract graph data. This study proposes a method to convert abstract graph representations into concrete object representations to enhance the impact of LLM comprehension. In the study, correspondences between abstract graph elements and concrete objects were first identified, such as mapping nodes in the graph to concrete objects and transforming the properties of edges into distances or connectivity between these spaces. Then, the graph data are transformed into easy-to-understand textual descriptions by constructing concrete scene descriptions. The experimental results show that, compared with the direct input of abstract graph, the concrete object representation can significantly improve the understanding and information processing ability of the model. This suggests that by converting abstract data into concretized descriptions, the language comprehension capability of LLM can be utilized more effectively, thus improving its performance in processing complex data. This study not only provides new perspectives for understanding the behavior of LLM when dealing with different types of data, but also provides practical strategies for improving model performance.

The increasing incidence of traffic accidents triggered by "road rage" poses a significant threat to public safety. Therefore, an in-depth analysis of its mechanisms and influencing factors is imperative. Previous research indicates that factors inducing driving anger include personal characteristics (such as age, gender, and temperament) and situational stimuli (such as traffic lights, hostile behaviors from surrounding vehicles, and overall traffic conditions). Building on this, the current study utilized naturalistic driving experiments with taxi drivers as subjects and employed a structural equation model to quantitatively analyze the interrelationships among these factors and to calculate the impact weights of personal characteristics and situational stimuli on driving anger. The results indicate that temperament, traffic flow conditions, behaviors of other traffic participants, and traffic signals are positively correlated with driving anger, whereas gender shows a negative correlation. Age, while not significantly correlated with anger, is negatively associated with temperament. Additionally, traffic conditions, traffic signals, and participant behaviors exhibit significant negative correlations with anger. These findings provide theoretical support for traffic management authorities in addressing road rage.

With the increasingly serious aging problem in our country, elderly care has become an urgent social problem to be solved. This article mainly studies the key technologies and 6G technology of the "migratory bird style elderly care" smart service platform to design a comprehensive, efficient, and convenient elderly care service platform, including data collection and processing, service provision, data analysis, etc., to provide personalized and intelligent elderly care services for the elderly. Through the analysis of platform data, the effectiveness of key technologies in improving the quality and satisfaction of elderly care services has been verified.

In modern drone applications, precise positioning in formation flying is one of the primary issues that needs to be addressed. Among these, passive positioning technology is a commonly used and effective method for precise formation flying. However, most such methods require the transmission and reception of radio signals, which is inconvenient in scenarios requiring electromagnetic silence. This study proposes a drone formation positioning correction strategy based solely on azimuth passive positioning. This strategy, through geometric analysis and mathematical visualization techniques combined with polar coordinate positioning and a 0-1 planning model, corrects the positions of drones with deviations, optimizes flight efficiency, and enhances safety. Experimental results show that this method can effectively improve the positioning accuracy of drone formations under conditions of electromagnetic silence, demonstrating substantial practical value and prospects for wider application.

The objective of news recommendation pertains to the facilitation of Personalized news delivery amid the vast expanse of available news. However, existing methods inadequate exploitation of textual information inherent in news articles, and secondly, when building user interest models from historical clicked news, the interaction behavior between clicked news words is ignored. In this paper, we propose a method using Transformer combined with candidate awareness. This method is adept at capturing the contextual nuances between words within user-generated text and exhibits superior efficacy in handling lengthy text sequences, to improve the ability of user modeling to capture user interests. We use a Transformer-based news encoder to learn semantic features at different levels of abstraction of the input text sequence. In addition, we use Fastformer, a variant of the Transformer network, in conjunction with candidate awareness modeling to capture the interactive behavior of users clicking on word-level news. Empirical analyses conducted on authentic datasets substantiate the efficacy of our approach in enhancing the efficacy of news recommendation systems.

In order to study the parameter selection of the information collaboration, this paper analyzes the model control effect and model complexity under these two kinds of information synergy representations by constructing the traffic state-aware interaction and reward information synergy experiments, and finally selects the optimal model information synergy parameters according to the experimental results, which provides a certain research basis for the subsequent parameter settings.

Digital up-conversion (DUC) and digital down-conversion (DDC) are widely used in digital and analog circuits. In order to satisfy the design of DUC/DDC in different application scenarios and simplify the design process of DUC/DDC RTL code, this paper proposes an automatic platform for automatically generating DUC/DDC RTL code. The platform is designed by Python language, and can call MATLAB function to realize the design of half-band filter. An interactive GUI is designed based on Tkinter tool. By inputting the required parameters in the GUI, the platform can automatically generate the corresponding RTL code, and the generated code meets the design requirements through simulation and verification, which greatly improves the design efficiency of DUC and DDC.

Crayfish optimization algorithm (COA) is a swarm-based metaheuristic algorithm proposed in July 2023. The fact that the optimal individuals are not processed makes the COA algorithm less capable of exploring in the early optimization phase and converging in the later phase. In this paper, an improved COA algorithm based on non-monopoly search called NOCOA was proposed. The optimal individuals are perturbed using a non-monopoly search strategy, and the strategy is further optimized using Gaussian and Cauchy operators, which enhances the algorithm optimization ability. Finally, the superior performance of NOCOA is verified by the CEC2017 test suite.

With the rapid development of cross-border e-commerce, the logistics system behind it faces various challenges such as complex routes, diverse transportation modes, and unpredictable customs clearance procedures. To address these challenges, this study proposes a design and simulation of an intelligent cross-border e-commerce logistics system based on genetic algorithm technology of artificial intelligence.The proposed system utilizes genetic algorithm techniques to optimize the logistics route planning, transportation mode selection, and customs clearance process. The genetic algorithm is employed to search for the optimal combination of routes, transportation modes, and customs procedures, considering various factors such as cost, time, and reliability. The system takes into account real-time data, including transportation costs, transit times, customs regulations, and historical shipment information, to make informed decisions and continuously improve the logistics efficiency.The simulation of the proposed system is conducted using historical cross-border e-commerce shipment data. The system is evaluated based on performance metrics such as transportation cost, delivery time, and customer satisfaction. The results demonstrate that the intelligent cross-border e-commerce logistics system based on genetic algorithm technology can effectively optimize the logistics process, reduce costs, and improve customer satisfaction.

Multi-agent pathfinding aims to find a set of optimal paths for multiple agents to reach their respective destinations without colliding with each other. Conflict-Based Search (CBS) is an algorithm that prioritizes conflict resolution, utilizing a two-level search framework to find optimal solutions. However, its performance significantly deteriorates as the number of agents increases. Various enhancements have been proposed, such as Enhanced CBS (ECBS), which introduces focused search in its two-level framework, resulting in a notable speed improvement. In this paper, we propose improvements to the bottom-level solving framework of ECBS, by designing a bottom-level two-level A* solving method and enhancing the heuristic function of the bottom level algorithm. These enhancements greatly reduce the computation time of the bottom-level algorithm and are applicable to most map instances, allowing for faster solution generation. Experimental results show that the improved AAECBS algorithm maintains a significantly faster speed compared to ECBS while incurring only minimal quality loss, with experimental results indicating a maximum loss of no more than 5%, and sometimes no loss in quality. In numerous experiments varying the suboptimality factor, changing maps, and adjusting the number of agents, AAECBS consistently outperforms ECBS in terms of speed.

Symptom-based automated disease diagnosis involves an iterative process of gathering potential symptoms through natural language interactions with patients. Previous research has made use of Reinforcement Learning (RL) techniques to optimize policy networks for symptom inquiry and disease diagnosis, yielding promising results. However, reinforcement learning methods encounter various challenges such as the risk of getting stuck in suboptimal solutions, limited efficacy in training, and difficulties in designing appropriate reward functions, especially when confronting large decision spaces. To address these challenges, this paper converts policy learning in automated disease diagnosis into generative task-QA by utilizing language models in a fully supervised way. We further introduced beam search decoding and an early stop mechanism to facilitate efficient interactive symptom queries and improve the efficiency of the diagnostic process, respectively. To assess the efficacy of our approach, we conducted extensive experiments on three Chinese real-world medical dialogue datasets: Dxy, MUZHI, and IMCS. Preliminary experimental results demonstrate that our method is comparable to or better than previous RL methods in terms of symptom recall and diagnostic accuracy.

Hairstyle is an essential part of appearance and one of the most important criteria for people to judge the image of others. Choosing the right hairstyle can enhance the charm of an individual's appearance and increase his or her self-confidence. The lack of visualization effect and personalized consideration when choosing hairstyles using traditional methods often leads to a significant deviation of the final result from the expectation. We designed a hairstyle generation system for hairstyle simulation and selection, which provides users with high-quality images and 3D visualizations of target hairstyles, thus realizing all-around personalized hairstyle design. The system's front end uses the Uni-app framework to build the user interface, an SDK to connect the front-end and back-end, and a MySQL database to store data. The system realizes the steps of face detection, image segmentation, hairstyle matching, and 3D reconstruction based on the adversarial generative network model and 3D reconstruction model. Finally, it generates precise 3D virtual hairstyles for users. Experimental results show that the system can help users preview and simulate the target hairstyles, providing convenience and intuition for users in hairstyle design and selection.

The integration of distributed massive renewable energy, energy storage, and electric vehicles has altered the original operation mode of the distribution network, transforming it from deterministic to stochastic, and from unidirectional to bidirectional power flow. Phenomena such as voltage exceeding limits, forward and reverse overloads, and reverse power flow in transformer zones have become increasingly frequent. Therefore, it is crucial to accurately and promptly grasp the operational status of the distribution network, eliminate potential safety hazards and operational risks, and ensure the safe and stable operation of the power grid. However, the distribution network faces challenges such as a wide distribution of points, complex structure, and incomplete coverage of measurement and acquisition points. With the construction of a new power system, the number of electrical equipment, generator nodes, and load nodes has multiplied, resulting in a surge in the amount of data and increasing demands for computational efficiency. Additionally, the complexity of the distribution network and the uncertainty of distributed power sources pose challenges for data identification and analysis. To address these issues, this paper proposes a high-dimensional data analysis technology based on digital twins, which is used to analyze the operational status of the distribution network and the interactive and collaborative relationship between power sources, loads, and storage. This approach enhances the management and control capabilities of optimal scheduling, maximum consumption, and safe and stable operation of the distribution network.

Chat models, such as ChatGPT, have shown outstanding performance in understanding and responding to human instructions, with far-reaching impacts on natural language questioning and answering. However, in vertical domains, harmful false facts can be generated due to the lack of targeted optimization during question answering. To address the generation of harmful false facts and improve the accuracy rate of large language models in answering questions in vertical domains, this paper proposes a retrieval-enhanced generative inference method based on large language models. Firstly, utilizing large language models to extract entity information from question sentences, enhancing database retrieval; secondly, through semantic encoding and embedding vectorization of question sentences, enhancing knowledge base retrieval; finally, we propose a new technique called self-consistency prompt word engineering, and strengthening the retrieval results, generating multiple independent thought chains, and summarizing to obtain the highest consistency answer. Compared with existing benchmark models on a self-built dataset, the experimental results show that the retrieval-enhanced generative inference method based on large language models proposed in this paper can effectively improve the accuracy of large language models in answering questions in the vertical domains.

Driven by natural language processing technology, topic models have shown a wide range of application value in fields such as text analysis, classification and clustering. The GSDMM topic model has been widely recognized in the field of topic modeling due to its ability to automatically infer the number of clusters, fast convergence, and efficient processing of sparse and high-dimensional short texts. This paper aims to explore the text evaluation research based on natural language processing and GSDMM (Gibbs Sampling and Distributed Memory Model) topic model, and carries out relevant practice in the text evaluation of project-based teaching course.

Temporal knowledge graph (TKG) can abstract the temporal information of entities and relations in the real world. If we can infer unknown temporal quadruples, we can predict the developmental trends of events in the long run. However, current TKG reasoning methods are difficult to model the relative temporal relations between quadruples well, and there is also an issue of insufficient reasoning information. Therefore, we propose a TKG reasoning model named TKBK, which combines formalized temporal knowledge and generative background knowledge. TKBK retrieves temporal knowledge from TKG and generates background knowledge from large language models (LLM). It uses a masking strategy to train a pre-trained language model and transforms the complex reasoning task into a masked token prediction task. We evaluated our proposed model on two datasets. The results show that TKBK outperforms the baseline model on most metrics, proving the effectiveness of this model in TKG reasoning tasks.

At present, robot target detection is widely used in tremendous fields including industrial, home, and commercial robots. However, the existing detection methods have some problems, such as low recognition accuracy, slow processing speed, and poor adaptability to the environment. In this work, an object detection model based on deep learning was designed, which mainly included a feature extraction network, a regional suggestion network, and an object classification and localization network. The feature extraction network uses the pre-trained ResNet-50 as the backbone network to extract the high-dimensional features of the image. The region proposal network (RPN) is responsible for generating high-quality candidate regions, which are then fed into subsequent classification and location networks. The target classification and localization network utilize multilayer perceptron (MLP) to perform accurate target category classification and bounding box regression for candidate regions. Experimental results show that the proposed model is significantly better than the traditional target detection methods in terms of accuracy and real-time performance.

Lacquerware as a treasure of Chinese craft heritage, integrates rich cultural connotations and practical value. However, the complexity of the production process and the conservatism of the design make it not competitive in the diversified market, thus facing the challenge of being marginalized. In recent years, the rapid development of generative artificial intelligence technology has profoundly transformed many industry sectors. Especially within the design field, AIGC as an assistive tool can greatly improve design efficiency. However, there are fewer studies in the field related to the design innovation of traditional artifacts styling by researchers using AIGC, Therefore, it is necessary to explore an effective ai-assisted design path for the innovation of traditional artifacts styling. Taking Song Dynasty lacquerware as an entry point, this paper aims to explore the application potential of AIGC technology in promoting the styling innovative design of traditional cultural products, introduce an effective application path based on AIGC. According to compare the generation effects of the three most popular image generation software on the market, and the AIGC tool with the best performance is selected to generate modeling samples from three perspectives. Selection of the most representative image samples by means of a questionnaire and construction of 3D models. The experiment shows that the effectiveness and superiority of AIGC in the design of traditional cultural products can greatly improve the efficiency of cultural and creative design, optimize the design process of traditional cultural products, and provide a new path for traditional artifacts to achieve modern transformation.

Community Question Answering (CQA) has become increasingly prevalent in recent years. However, the large volume of answers poses a challenge for users in identifying the most pertinent ones, thereby making answer selection a vital subtask within CQA. In this paper, we introduce the Question-Answer cross attention networks (QAN) with pre-trained models for improved answer selection and leverage large language models (LLMs) for enhanced answer selection with knowledge augmentation. Specifically, we utilize the BERT model as the encoder layer to pre-train on question subjects, question bodies, and answers separately. The cross attention mechanism is then used to identify the most relevant answers for different questions. Our experimental results demonstrate that the QAN model attains state-of-the-art performance on the SemEval2015 and SemEval2017 datasets. Additionally, we use LLMs to generate external knowledge from questions and correct answers, enhancing the answer selection task. By optimizing the LLM prompts in various aspects, we found that incorporating external knowledge improves the correct answer selection rate on both the SemEval2015 and SemEval2017 datasets. Moreover, optimized prompts enable LLMs to select the correct answers for a greater number of questions.

A flexible fruit classification system based on machine vision was designed to cope with the issues of low manual classification efficiency, skin damage-prone from machine grading, and fixed classification standards and types in the conventional classification methods for cherry tomato. Firstly, the principle of cherry tomatoes flexible classification system was determine and the technical roadmap was determined including offline training and online detection. Secondly, the standards was determined for the classification experiment of the cherry tomato based on the classification theoretical model and market demands, while machine vision was applied to extract the image information by preprocessing; Thirdly, the range of feature threshold including shape, size, and color were obtained by feature recognition research using roundness value method, two-dimensional feature method, and color space threshold method respectively; Finally, experiments was conducted after camera calibration. The single feature recognition experiment shows that three types of features can be effectively recognized separately, and the threshold is adjustable, which verifies the flexibility of the recognition system; The comprehensive feature recognition experiment has verified that the similarity between machine recognition and manual recognition is 93.3%, which is close to manual recognition and has practical and promotional significance.

With the rapid development of artificial intelligence technology, its application in aircraft design has become a critical means of enhancing design efficiency and innovation. In order to explore the specific application pathways of AI technology in the domain of aircraft design and evaluate its actual impact on design process optimization, this paper provides an overview of the current application status of key AI technologies in aircraft design. It analyzes specific cases where AI technology has improved the efficiency of aircraft design and forecasts the future challenges and development trends in AI aircraft design. The results of the study indicate that AI is currently widely applied in aircraft control systems, configuration optimization, and fault prediction and health management. Data-driven design scheme generation can significantly enhance the accuracy of aircraft design, shorten the development cycle, and to some extent, predict and mitigate potential risks.

In order to provide a higher level of home security protection for women living alone and reduce potential safety hazards, the Smart Doorway System is designed. The Smart Doorway System consists of a main control node, a camera shooting node, a PTZ face tracking shooting node, a cloud platform, and a WeChat applet. The main control node monitors the abnormal information at the doorway in real time, such as someone passing by, staying or unlocking the door. When an abnormal situation occurs, it reports the abnormal information to the cloud platform via Wi-Fi, and at the same time pushes it to the user via WeChat applet. In addition, the node supports fingerprint, RFID and password unlocking methods, and the system automatically dials a distress call when an emergency occurs. The camera shooting node detects whether there is someone at the door, automatically takes photos when there is someone, and uploads the photos to the cloud platform through the Wi-Fi module. The face tracking and shooting node of the cloud platform realizes the function of automatically tracking and shooting and saving the faces in the detected area. The cloud platform is responsible for storing the data uploaded by each node, and the storage duration is one month. The WeChat applet obtains the latest door lock status, alarm information and other data in real time and displays it to the user. This allows users to control the security dynamics of the home portal in real time no matter where they are, greatly enhancing the convenience and effectiveness of security. In short, the Smart Doorway system allows women living alone to get allaround security protection and reduce potential security risks.

The compliance and integrity of transaction data is paramount in power markets, yet the presence of missing or corrupted data can compromise market analysis and decision-making. To address this, we introduce GEDR-Net, a framework that combines Graph Neural Networks (GNNs) with Generative Adversarial Networks (GANs) for the recovery of missing transaction data. By harnessing the spatial relationships within the power market's graph structure, GEDR-Net predicts missing values while ensuring the authenticity of the recovered data through adversarial training. Our experimental findings suggest that GEDR-Net offers a promising approach to data recovery, showing improved performance compared to existing methods, especially in scenarios with substantial data incompleteness.

With the rapid development of China's power industry, the safe and stable operation of substations has become an important guarantee for the power system. The emergence of intelligent inspection robots provides a new solution for the inspection work of substations. This article first introduces the structure, system composition, and main functions of intelligent inspection robots, including recognition and detection, intelligent inspection classification, positioning and navigation, and analyzes their key technologies, such as navigation positioning, detection and recognition, image recognition, and automatic charging technology. Next, this article explores the relevant design of the inspection robot system, including the backend management system, communication system, and power supply system, as well as functional construction, such as video recognition, environmental detection, and warning linkage. Then, this article analyzes the application practice of intelligent inspection robots in substation operation, including single station mode and centralized control mode, and evaluates their application effects.

In recent times, artificial intelligence has experienced rapid advancement and is finding its applications across diverse sectors. Within the medical domain, lung X-rays play a pivotal role in diagnosing lung ailments. Considering this, we introduce a classification algorithm tailored for lung X-ray images. This algorithm merges the strengths of DenseNet and VGG through feature fusion. To further enhance deep feature extraction, we've incorporated two attention mechanisms: the global attention block and the category attention block. Through rigorous five-fold cross-validation, it was observed that a single model could achieve a maximum accuracy of approximately 0.843. By integrating the pertinent features from both DenseNet and VGG, we noted a marginal enhancement in the overall performance. Remarkably, the introduction of the attention mechanisms onto this integrated model elevated the accuracy to 0.873. This refined model demonstrates superior diagnostic precision for 14 distinct disease types compared to other models.