Current Issue

  • Select all
    |
    Electrical Engineering
  • Electrical Engineering
    Design of partial discharge monitoring algorithms for high-voltage power terminals based on power Beidou system
    ZHANG Xiaodong, ZHANG Yiming, YANG Qifan, WANG Guanlin, GU Bingling
    2026, 48(4): 1-9. https://doi.org/10.7688/j.issn.1000-1646.2026.04.01
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] There are problems of high misjudgment and false alarm rates of partial discharge patterns in existing partial discharge monitoring devices for high-voltage power terminal lines, and it is difficult to realize real-time partial discharge monitoring in areas without communication signal coverage. To this end, a study was conducted on partial discharge monitoring algorithms for high-voltage power terminals based on the power Beidou system and an improved deep belief network (DBN) model. [Methods] The proposed partial discharge monitoring algorithm mainly consists of two parts:data acquisition and synchronization based on the power Beidou system, and partial discharge pattern recognition and positioning based on the gray histogram moment-krill herd algorithm (KHA)-DBN. In the data acquisition and synchronization stage, the Beidou short message communication and positioning functions were employed to realize real-time transmission of monitoring data from power terminals in areas without communication coverage, with the geographic location information of monitoring equipment carried synchronously. Meanwhile, the Beidou synchronous timing function was adopted to unify and calibrate the acquisition time of multi-terminal sensors, and improve the synchronization of multi-region monitoring data sharing, thereby enhancing the partial discharge positioning accuracy by relying on the spatio-temporal characteristics of data. The partial discharge pattern recognition and positioning stage mainly includes three steps of data acquisition, feature extraction, and pattern recognition. By analyzing the partial discharge patterns and their generation mechanisms, four typical partial discharge patterns of internal discharge, surface discharge, corona discharge, and floating discharge were summarized. Additionally, the high-frequency current method was employed to collect signals of the above four discharge patterns and the normal state signals of the equipment. A spatial relation feature enhancement method was employed to divide the time-frequency distribution map of monitoring data into several regions and extract three low-order gray feature parameters respectively, aiming to alleviate the influence of mode aliasing in multiple discharge patterns. Based on the multi-scale permutation entropy theory, appropriate scale parameters were selected to refine the strong time-series feature data and calculate the entropy value, further enhancing the independence of multi-scale data features. The improved KHA was adopted to optimize the DBN parameters, and the optimized results were taken as the initial parameters of DBN. Specifically, the initial population of the KHA was generated by Logistic chaotic mapping, and the Gaussian mutation operation was adopted to update the IK fitness value and thus complete the KHA parameter optimization. After determining the initial parameters of DBN, the DBN model was trained with the training sample set, and the influence of different hidden layer number and node number on the recognition accuracy was analyzed to finally determine the optimal structure of the gray histogram moment-KHA-DBN partial discharge pattern recognition algorithm. [Results] A test platform for model training data acquisition was built, and tests were carried out with samples equivalent to the data scale of actual scenarios. The influence of hidden layer number and node number on the recognition accuracy of partial discharge patterns was analyzed. The optimal hidden layer number is determined as 5 and the node number as 26. Compared with the partial discharge pattern recognition algorithm based on the extreme learning machine (ELM) and support vector machine (SVM), the recognition accuracy of the proposed algorithm is improved by 4.87％ and 3.22％ respectively, exhibiting better pattern recognition performance. [Conclusions] The algorithm can recognize a wide range of partial discharge patterns and meet the requirements of practical engineering applications. Integrated with the power Beidou system, the functions of the Beidou short message communication, positioning and synchronous timing can realize automatic transmission of power terminal monitoring data, terminal positioning and acquisition time calibration of multi-region sensors, providing theoretical support for improving the positioning accuracy of partial discharge faults.
  • Electrical Engineering
    High-voltage insulator fault detection and disposal technology based on AR digital model
    LI Jingxiang, LAI Hao, JIANG Zhibo, PAN Libang
    2026, 48(4): 10-18. https://doi.org/10.7688/j.issn.1000-1646.2026.04.02
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] The distribution network is a core link in power distribution, and its operational stability directly affects the continuity and reliability of power supply. However, as key components in the distribution network, high-voltage insulators take the responsibility of electrical isolation and supporting wires. During long-term operation, they are susceptible to faults under the influence of environmental factors, climate change, and material aging. Once a fault occurs, it not only results in degraded operational stability of the distribution network and affects the power supply quality, but also may cause accidents such as short circuits and trips, bringing serious economic losses and social impacts to power enterprises and users. Therefore, the timely and accurate detection and disposal of high-voltage insulator faults are of great significance for ensuring the safe and stable operation of distribution networks. [Methods] A high-voltage insulator fault detection and disposal technology based on the auto regressive (AR) digital model was proposed. A module specifically for real-time acquisition of high-voltage insulator status data in distribution networks was designed and implemented. This module is capable of efficiently processing differential signals in ultra-high voltage direct current transmission systems, providing a high-quality data foundation for subsequent data analysis. On this basis, an AR digital model was built to extract key information in the signals by parameter estimation and filtering, thereby effectively suppressing noise interference. Furthermore, by comparing the filtered residual sequence and its variance, accurate identification of abnormal data was realized. In the fault detection stage, a high-voltage insulator fault detection model was built by adopting an extreme learning machine. By inputting the residual sequence of abnormal data into the model, the fault type, fault location, and fault time can be quickly and accurately determined, achieving accurate detection. Based on this, a fast processing model for high-voltage insulator faults was built. By adopting preset disposal processes and algorithms, rapid maintenance and replacement of faults were achieved. [Results] The experimental results show that the proposed method can realize real-time monitoring and fault early warning of high-voltage insulator status, effectively improve the accuracy and efficiency of fault detection, and reduce the fault disposal time and cost. At the same time, this technology has strong adaptability and scalability, which can provide reliable technical support for the safe and stable operation of distribution networks. [Conclusions] According to the research results, the proposed fault detection and disposal method for high-voltage insulators based on the AR digital model not only improves the accuracy and efficiency of fault detection but also provides a feasible technical solution for the safe and stable operation of the distribution networks. This method holds great significance in practice, helping to reduce power outages caused by insulator faults and improve the overall stability and power supply quality of the power systems.
  • Electrical Engineering
    Power-energy-flexible ramping market equilibrium model considering demand uncertainty
    SUN Tian, MIN Rui, GUO Zhuochen, LIU Xue, CAI Yuanji
    2026, 48(4): 19-28. https://doi.org/10.7688/j.issn.1000-1646.2026.04.03
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] With the increasing penetration of renewable energy, the “duck curve” problem is becoming more severe. The rapid rise in the system′s net load curve over a short time period will become more pronounced, placing higher demands on the flexibility of the power system. As a novel ancillary service product, the flexible ramping product (FRP) can incentivize units to reserve certain flexible ramping capacity during the current period, in order to address changes in net load in the next period. In return, units providing this product can also receive corresponding compensation. However, existing research has paid limited attention to the uncertainty of the flexible ramping demand price curve, and there is insufficient discussion on how this product impacts market equilibrium. Moreover, the distinct characteristics of different resource providers have not been fully incorporated, and these issues still require further research to resolve. [Methods] A flexible ramping demand price curve calculation model was proposed that considered the uncertainties of both supply and demand. The model combined the characteristics of system net load, where load and renewable energy power forecast errors were influenced by different factors. A quantile regression method was employed to fit the upper and lower bound curve equations for the system′s load forecast errors, from which the system′s flexible ramping demand was calculated. Furthermore, the flexible ramping demand price curve was determined by incorporating the probability density function of the system′s net load forecast errors. Built on this calculation model, the paper further developed a power-energy-flexible ramping market equilibrium model, which extended the original mathematical model that minimized only power generation costs to one that minimized the total costs of power-energy and FRP supply. [Results] Using the IEEE 30-bus system as an example for simulation, and combining actual photovoltaic and load data from a specific region for the simulation analysis. First, the net load forecast errors are obtained by calculating the error regression curve, and the system′s flexible ramping demand price curve is plotted, providing a basis for pricing the FRP. Second, a multi-scenario market equilibrium analysis shows that the energy clearing price under the constructed model effectively reflects the system′s flexible ramping supply-demand relationship, achieving the economic optimization of both the units and the system. Furthermore, the analysis of the photovoltaic output curves in various scenarios demonstrates that the proposed model can effectively reduce the phenomenon of curtailing photovoltaic power. Additionally, the validation results show that the integration of independent energy storage can replace some high-cost gas-fired units, thereby reducing the system′s total electricity procurement cost. [Conclusions] Through the constructed model, the system can effectively adjust the procurement volume of FRPs. On one hand, this reduces the loss costs caused by the failure to procure sufficient flexible ramping capacity, and on the other hand, it significantly enhances the economic operation of the system. Under the guidance of market prices, FRPs can incentivize more flexible resources to connect to the grid and provide flexible ramping capacity, thereby creating conditions for further improving the capacity for renewable energy consumption. Additionally, independent energy storage can better match the system′s flexibility requirements, alleviating the supply-demand pressure of system flexibility while securing more profit opportunities for itself.
  • Electrical Engineering
    Optimization technology for EV charging and energy scheduling based on multi-objective particle swarm
    ZHU Yanjie, GAO Yudou, LI Yuan
    2026, 48(4): 29-35. https://doi.org/10.7688/j.issn.1000-1646.2026.04.04
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] In response to the widespread utilization of electric vehicles (EVs), which has resulted in problems such as an expanded peak-to-valley difference in power systems and localized overloads in power grids, a multi-objective particle swarm optimization-based charging and energy scheduling strategy for EVs was proposed. This method not only ensures the lowest electricity costs for users and safe operation of the power grid but also reduces the peak-to-valley difference of grid loads by optimal scheduling of EV charging strategies, thereby effectively ensuring the safe and reliable operation of power systems. [Methods] The multi-objective particle swarm optimization algorithm was adopted to solve the optimal scheduling function of EVs. The evaluation indexes of the particle swarm state were constructed, and the inertia weight and learning factor of particle swarm optimization were dynamically adjusted based on this, with the Sigmoid function mapping method employed to ensure that the inertia weight and learning factor were dynamically adjusted with the change of evaluation indexes. On this basis, Gaussian mutation was utilized to prevent the algorithm from falling into premature convergence or stagnation, further improving the reliability and stability of the algorithm. [Results] The objective function of EV optimal scheduling was established, and the multi-objective particle swarm optimization algorithm was adopted to solve the function. The performance of the algorithm was compared and analyzed from the aspects of convergence speed and standard deviation of the objective function. Finally, the adaptive weight optimization combined with the Gaussian mutation particle swarm optimization algorithm was selected as the solution. Simulation analyses were carried out on optimization strategies under fixed electricity price and time-of-use tariffs. The simulation results show that the proposed scheduling strategy can realize “peak shaving and valley filling” under both the time-of-use tariff and fixed electricity price. [Conclusions] The results show that the Gaussian mutation particle swarm optimization algorithm with adaptive weights has advantages in stability and reliability, although the convergence speed is slightly lower. After optimal scheduling, the peak-to-valley difference of the fixed electricity price load curve is 37.26% lower than that of the original load, and the peak-to-valley difference of the time-of-use tariff load curve is 53.62% smaller than that of the original load. The proposed algorithm is also applicable to different time-of-use tariff strategies. The highlights of this paper lie in the following aspects. First, an evaluation index of the convergence state of particle swarms was proposed to evaluate the convergence state of the population, and based on this, the inertia weight and learning factor of the particle swarm optimization algorithm were dynamically adjusted to give full play to the optimization performance of the particle swarm optimization algorithm. Second, for the optimization of multi-objective functions, a particle mutation method based on Gaussian mutation was introduced, thus avoiding the population from falling into premature convergence or stagnation, and improving the stability and reliability of the algorithm.
  • Electrical Engineering
    Ultra-short-term prediction algorithm for power generation at photovoltaic power stations based on spatiotemporal irradiation distribution characteristics
    LI Hui, CHEN Shaoliang
    2026, 48(4): 36-42. https://doi.org/10.7688/j.issn.1000-1646.2026.04.05
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] To address the issue of insufficient power generation prediction accuracy caused by the uneven spatial irradiation distribution within photovoltaic power stations, and to break through the limitations of traditional methods in capturing spatiotemporal dynamic characteristics, an ultra-short-term prediction algorithm integrating spatiotemporal irradiation distribution characteristics was proposed to enhance the prediction reliability under complex conditions such as sunny and cloudy days and meet the real-time dispatching requirements of the power grid. [Methods] A calculation system for irradiance was constructed through three key dimensions (extraterrestrial solar irradiance, atmospheric mass, and atmospheric transparency). The incident angle was determined by combining hour angle correction and solar altitude angle. The solar constant and the atmospheric attenuation coefficient were introduced to quantify the outer radiation intensity. The atmospheric attenuation effect was characterized by parameters of atmospheric mass thickness and transparency. Direct radiation and scattered radiation were integrated to form a complete irradiance calculation framework. To improve prediction accuracy, the concept of similar days was introduced. By calculating the day similarity and the similarity of the previous trend, an objective function was established by combining the two to select historical dates with a similar irradiance distribution to the prediction day. In terms of power generation prediction, the multi-layer perceptron (MLP) and the deep belief network (DBN) were combined to construct the MLP-DBN model, and the prediction samples were constructed by using the similar day selection method and solar irradiance data. The solar irradiance samples were input into the MLP module to obtain the initial predicted value of power generation and calculate the residuals. The DBN module achieved precise prediction based on the residuals. The outputs of the hidden layer and the output layer were calculated through relevant formulas. An objective function was established to adjust the network weights. The MLP was trained until the mean square error was minimized, and finally the ultra-short-term prediction results of power generation at the photovoltaic power station were output. [Results] The experimental results show that under sunny and cloudy conditions, the irradiance calculated by the proposed algorithm is highly consistent with the actual value, while the traditional method exhibits obvious deviations. Furthermore, the power generation prediction curve shows that the prediction results of this algorithm are closer to the actual values, thanks to its efficient screening mechanism for historical similar radiation days. Further quantitative analysis shows that the determination coefficients of this algorithm are superior to those of the comparison methods on both sunny and cloudy days. Especially under cloudy conditions, its stability is outstanding, verifying its strong robustness. It can learn complex nonlinear relationships and high-level features from samples, thereby accurately capturing the subtle differences under different meteorological conditions. [Conclusions] By integrating the spatiotemporal distribution characteristics of irradiance and deep learning techniques, this study not only accurately depicts the atmospheric radiation transfer process, but also effectively extracts weather-related nonlinear features through deep networks. This method provides reliable technical support for photovoltaic power stations to participate in power market bidding and grid dispatching, and has important practical value for promoting the consumption of new energy.
  • Electrical Engineering
    Cost prediction model for power transmission and transformation projects based on voting ensemble learning
    LIU Yun, ZHANG Wenjing, WANG Dongchao, ZHOU Bo, HAN Jinpeng
    2026, 48(4): 43-50. https://doi.org/10.7688/j.issn.1000-1646.2026.04.06
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] As data technology advances, vast amounts of multi-source heterogeneous data are accumulated in the field of power transmission and transformation projects. Meanwhile, the diversity and nonlinear characteristics of factors influencing power grid project costs contribute to the generally low accuracy of traditional cost prediction models. To this end, this paper aims to enhance the cost prediction accuracy for power transmission and transformation projects, thereby providing reliable decision-making basis for project cost management. [Methods] A cost prediction method integrating a strong-correlation sample balancing algorithm and voting ensemble learning was proposed. Firstly, the Pearson correlation coefficient was employed to select key factors affecting the costs of power transmission and transformation projects, followed by data normalization. Secondly, a sample balancing algorithm was adopted to randomly generate multiple training subsets of varying sizes, and the convolutional neural network (CNN) sub-models were built respectively based on each subset. Finally, based on the performance of each sub-model in the validation phase, the weights of prediction results across different categories were calculated, and a weighted voting ensemble strategy was adopted to obtain the final prediction results. The design principle of this method is as follows: the Pearson correlation coefficient can precisely identify key influencing factors, reducing redundant information. Additionally, the sample balancing algorithm can solve the problem of imbalanced data distribution, ensuring balanced identification ability of the model across various categories of samples. Voting ensemble learning can integrate the advantages of multiple CNN sub-models, significantly enhancing the generalization ability of the model. [Results] 30 sets of project cost sample data of 220 kV substations owned by a regional power enterprise were taken as the experimental subjects, with the first 20 sets as the training dataset and the remaining 10 sets as the test dataset for verification. Under the premise of identical parameter conditions, the CNN voting ensemble model based on Pearson correlation analysis outperforms both the Pearson-CNN algorithm and the single CNN model in training and testing phases. The experimental results show that the proposed model achieves the training accuracy and testing accuracy of 97％ and 96.7％, respectively, with the loss function values of 0.35 (training dataset) and 0.45 (test dataset). Additionally, the average prediction accuracy reaches 97.2％, generally demonstrating sound prediction performance. [Conclusions] The designed model can effectively integrate imbalanced multi-source heterogeneous data. By combining Pearson correlation analysis, the sample balancing algorithm, and voting ensemble learning, the prediction performance was notably improved, and the problem of low prediction accuracy caused by data imbalance in the field of power transmission and transformation projects was solved. This advancement helps optimize cost management processes of the projects and thus provides more accurate and reliable data support for the decision-making of related projects.
    • Materials Science & Engineering
  • Materials Science & Engineering
    Texture evolution and VPSC simulation of extruded AZ31 magnesium alloy under different loading paths
    MAO Pingli, LIU Weidi
    2026, 48(4): 51-57. https://doi.org/10.7688/j.issn.1000-1646.2026.04.07
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] Leveraging abundant magnesium resources, industries such as defense, medicine, and digital electronics have seen a significant increase in demand for the processing and forming of lightweight alloys, including magnesium alloys. The hexagonal close-packed (HCP) crystal structure of magnesium alloys results in poor room-temperature plastic deformation ability, and strong textures are easily formed during processing, leading to anisotropy, which further limits their application range. To mitigate the anisotropy of magnesium alloys and enhance their mechanical properties, it is necessary to clarify the influence rules of different loading paths and strains on their texture evolution. [Methods] In this study, commercial AZ31 (Mg-3％Al-1％Zn) magnesium alloy hot-extruded bars were used as the research object, and experimental and simulation analyses were carried out focusing on the correlation mechanism of “loading path-strain-texture evolution”. After annealing at 400 ℃ for 8 h, deformation samples with strains of 3.5％, 6.5％, and 9.5％ were prepared by compression in the extrusion direction (ED-C) and tension in the extrusion radial direction (ERD-T) under quasi-static loading conditions at 0.001 s-1. Electron backscatter diffraction (EBSD) technology was used to characterize the texture evolution characteristics, and the viscoplastic self-consistent (VPSC) model was combined to set up two deformation mechanism combinations for equal-strain simulation of texture evolution. [Results] The dominant deformation mechanism under both loading paths is basal slip, and the secondary deformation mechanism is ｛1012｝ tensile twinning. At the same strain, the nucleation and growth of ｛1012｝ tensile twins in the ED-C loading path are more significant, and its contribution is much greater than that in the ERD-T loading path. Under the ED-C loading path, the texture undergoes a significant transformation with increasing strain. At 9.5％ strain, the maximum texture intensity reaches 16.25, and the c-axis of the grains is oriented toward the ED direction. By contrast, the texture intensity under the ERD-T loading path shows a slight increase and slow evolution with increasing strain. In the VPSC simulation, the simulation results of combination 1 under the ED-C loading path are more consistent with the experimental pole figures and texture intensity changes, while the simulation results of combination 2 under the ERD-T loading path are closer to the experimental reality. [Conclusions] The extensive activation of ｛1012｝ tensile twinning changes the grain orientation, and its contribution determines the degree of texture transformation, resulting in a characteristic that the maximum density region of the texture of AZ31 magnesium alloy tends to be parallel to the loading force axis. The established VPSC model can effectively characterize the texture evolution process of AZ31 magnesium alloy under different loading paths and strains, and the model has a strong dependence on the combination of deformation mechanisms.
  • Materials Science & Engineering
    Effects of thermal exposure time at 650 ℃ on creep properties of Ti65 alloy
    WANG Yuanchen, LIU Jianrong
    2026, 48(4): 58-64. https://doi.org/10.7688/j.issn.1000-1646.2026.04.08
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] The microstructural evolution and surface oxidation of high-temperature titanium alloys during long-term high-temperature thermal exposure can adversely affect their mechanical properties, thereby threatening the service safety at high temperatures. The creep properties of alloys are critical to the service reliability of key engineering components under high-temperature service conditions. [Methods] To investigate the effect of high-temperature thermal exposure on the creep properties of high-temperature titanium alloys, Ti65 alloy was used as the research object. Thermal exposure experiments were conducted at 650 ℃ for 10, 20, 55, 100, and 200 h, and the microstructures and surface oxidation layer morphologies during thermal exposure were observed and analyzed. Creep tests were carried out on the thermally exposed Ti65 alloy under the condition of 650 ℃ /100 MPa/100 h to examine the influences of thermal exposure time on its microstructures, surface oxidation behavior, and creep properties. [Results] At 650 ℃ , with the extension of thermal exposure time, elemental diffusion is intensified, the size and content of α2 phases within the lamellar αs phase and silicides near the α/β phase boundaries both increase gradually, and the thickness of the surface oxide layer also increases. Under the condition of 650 ℃ /100 MPa/100 h, the total creep strain of Ti65 alloy decreases gradually with prolonged thermal exposure time, in which the elastic strain remains basically unchanged while the plastic strain decreases progressively. This is because α2 phases and silicides have a significant hindrance effect on dislocation motion during creep, and the increase of their contents improves the creep resistance of Ti65 alloy after thermal exposure, further leading to a gradual reduction of its total creep strain with prolonged thermal exposure time. [Conclusions] The research results provide a theoretical reference for clarifying the property evolution law of Ti65 alloy components under long-term high-temperature service and improving their service safety.
    • Mechanical Engineering
  • Mechanical Engineering
    Vehicle path tracking control strategy based on gain-scheduling output feedback
    WANG Xiaowei
    2026, 48(4): 65-73. https://doi.org/10.7688/j.issn.1000-1646.2026.04.09
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] As autonomous driving technology rapidly develops, high-accuracy path tracking control has become a core technology for autonomous ground vehicles (AGVs). However, in practical applications, it is usually difficult to directly measure the lateral velocity of vehicles, which limits the performance of traditional control methods relying on full-state feedback. Meanwhile, factors such as the strong nonlinearity of vehicle dynamics, time-varying driving conditions, and external disturbances further pose severe challenges to the tracking accuracy and robustness of the controllers. To solve the above problems, this paper designed a path tracking controller that could guarantee high accuracy, superior stability, and strong robustness for AGVs when the lateral velocity was unavailable for measurement. [Methods] Firstly, a vehicle dynamic model was built. On this basis, a vehicle path tracking control strategy integrating gain scheduling and robust H∞ theory was proposed. The strategy achieved a synergistic effect through three core designs. Specifically, an output feedback structure only utilizing measurable system outputs was adopted to eliminate the dependence on unmeasurable state variables such as lateral velocity. Gain scheduling technology was introduced, and the controller gain was dynamically adjusted online by constructing a parameter-dependent function with longitudinal velocity as the scheduling variable to adapt to the dynamic characteristic variations of vehicles under different driving states. Additionally, H∞ control theory was embedded to suppress the adverse effects of model uncertainty, parameter perturbation, and external disturbances by optimizing system performance indices, thus ensuring sound robustness of the closed-loop system. [Results] Hardware-in-the-loop (HIL) simulation tests in typical driving conditions were designed to verify the effectiveness of the designed controller. The results show that compared with traditional controllers, the lateral error and heading angle error of the designed controller are reduced by more than 30％, which greatly improves the control accuracy. The responses of key vehicle state variables (the lateral velocity and yaw rate) are smooth, and their peak values are strictly constrained within the stable physical limit, which effectively guarantees the lateral stability during driving. [Conclusions] The combination of gain scheduling strategy and robust output feedback control is an effective way to address the problems of unmeasurable states, time-varying parameters, and external disturbances in the path tracking control of AGVs. The designed controller not only theoretically guarantees the closed-loop stability of the system but also significantly improves the comprehensive path tracking performance, providing reliable technical support for the practical engineering application of AGVs.
  • Mechanical Engineering
    Bifurcation control of new quasi-zero-stiffness system under joint controller
    SUN Ziqiang, LI Zhenghong, ZHAO Guiren, YAN Ming, ZHAO Pengduo
    2026, 48(4): 74-81. https://doi.org/10.7688/j.issn.1000-1646.2026.04.10
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] In the field of modern engineering, quasi-zero-stiffness systems have gradually become a cutting-edge direction in controller research and development due to their excellent dynamic characteristics and high sensitivity. By reducing stiffness, the systems can maintain a high degree of stability and reliability under load changes and external disturbances, and thus have been widely applied in the vibration isolation of ship equipment. However, the existing quasi-zero-stiffness vibration isolators generally suffer from the problem of a narrow zero-stiffness interval. Meanwhile, due to their geometric structure and stiffness-damping nonlinear characteristics, bifurcation phenomena are likely to occur during the vibration isolation process, leading to system instability and thus limiting their effectiveness in engineering applications. [Methods] To address the above problems, a novel quasi-zero-stiffness vibration isolator was designed, which was composed of a negative stiffness mechanism formed by inclined connecting rods and tension springs and a positive stiffness mechanism formed by vertical double compression springs in parallel. Firstly, the structural parameters and zero-stiffness conditions of this vibration isolator were determined through static analysis, and a comparative analysis was conducted with the traditional three-spring quasi-zero-stiffness vibration isolator. The mechanical expression of the system was simplified through the application of the third-order Taylor series expansion. Subsequently, a dynamic analysis of the novel quasi-zero-stiffness vibration isolation system was conducted. The system′s equations were solved using the harmonic balance method. Moreover, the vibration isolation performance was evaluated based on the force transmissibility, with a comparison made to the corresponding linear system. In addition, a joint linear and nonlinear controller was introduced to manage the saddle-node bifurcation region of the system. Finally, the changes in the amplitude-frequency characteristics and force transmissibility of the system were studied under various control parameters. [Results] For the new quasi-zero-stiffness vibration isolator, upon satisfying the zero-stiffness condition, its zero-stiffness interval measures 0.4, while that of the traditional three-spring system is only 0.2. In comparison, the zero-stiffness interval of the novel structure is doubled, and it can effectively isolate vibrations within a wider frequency range. At the same time, under the regulation of the joint linear and nonlinear controller, the system can effectively eliminate the saddle-node bifurcation phenomenon, which significantly improves the system stability. [Conclusions] The core innovation of this work lies in proposing a novel quasi-zero-stiffness structure and a linear-nonlinear joint control strategy. This approach significantly expands the system′s quasi-zero-stiffness region while substantially enhancing robustness. The designed vibration isolation system enhances operational reliability and service life for shipboard equipment. The introduction of the joint controller effectively addresses multi-variable coupling effects, ensuring superior dynamic response performance and stability margin. These findings provide a solid theoretical foundation and practical guidance for performance optimization and engineering applications of quasi-zero-stiffness systems.
    • Architectural Engineering
  • Architectural Engineering
    Multi-scale analysis of material degradation mechanism and load-bearing capacity of reinforced concrete in historic buildings
    WANG Xiaolong, SHANG Huaishuai , XIE Zonglong, MENG Weiguang, GU Lilong, XU Shaohui, DONG Peisong
    2026, 48(4): 82-90. https://doi.org/10.7688/j.issn.1000-1646.2026.04.11
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] This study aims to systematically evaluate the aging condition of reinforced concrete materials in historic buildings in the Qingdao area and its impact on structural load-bearing capacity. Due to long-term exposure to coastal environments characterized by high humidity and salinity, these buildings exhibit significant material degradation, posing potential threats to structural safety. The research focuses on the changes in material microstructure, the attenuation of mechanical properties, and their correlation with macroscopic load-bearing capacity, thereby providing a theoretical basis and data support for the scientific conservation and reinforcement of historic structures. [Methods] Three typical historic buildings in Qingdao on Beijing Road, Huangdao Road, and Pingdu Road were selected as case studies. A combined approach of on-site sampling and laboratory testing was employed, conducting analyses at both macro and micro levels. Macroscopically, concrete compressive strength tests and steel reinforcement tensile tests were performed. Microscopically, scanning electron microscope (SEM) and X-ray diffractometer (XRD) were used to analyze the microstructure and phase composition of concrete, while metallographic microscope was utilized to examine the microstructure of the steel reinforcement. Furthermore, finite element models of beam and slab components were established using ABAQUS software to simulate the load-displacement responses of components with different reinforcement strengths under various displacement conditions, quantifying the impact of material property differences on component capacity. [Results] Experimental results indicate a significant reduction in the compressive strength of concrete in some buildings; for instance, the column strength in the Pingdu Road building is only 8.8 MPa, far below the design expectations. Severe reinforcement corrosion is observed, with the yield strength of some samples failing to meet the current HPB235 standard. The chloride ion content in the concrete is measured between 0.56％ and 0.84％, substantially exceeding the standard limit. Microstructural analysis reveals significant carbonation of the concrete, increased porosity, and a loose C-S-H gel structure. Non-metallic inclusions and inhomogeneous microstructures are found in the steel reinforcement. Finite element simulations further reveals that components utilizing the measured historical reinforcement strengths have significantly lower ultimate load-bearing capacities compared to those using current HRB400 reinforcement. For example, the capacity difference in beam components reaches 18 to 19 kN, confirming the substantial impact of material aging on structural safety. [Conclusions] This research confirms that material aging and environmental erosion are key factors leading to the reduced load-bearing capacity of historic buildings in the Qingdao area. Based on the experimental and simulation results, it is proposed that greater emphasis should be placed on the detection and assessment of the actual strength of reinforcement in historic buildings. The use of modern testing techniques such as SEM, XRD, and infrared thermography for regular structural health monitoring is recommended. The study emphasizes that repair and reinforcement strategies should be developed based on the actual material properties and microscopic degradation mechanisms, implementing targeted interventions. This comprehensive evaluation framework not only provides a scientific basis for the structural safety assessment and maintenance of historic buildings but also offers important references for improving relevant conservation policies and technical standards.
  • Architectural Engineering
    Influence of wedge angle on settlement of tapered pile-net structure subgrade
    DENG Yousheng, CHEN Zhuo, WU Along, DONG Chenhui, ZHUANG Ziying
    2026, 48(4): 91-99. https://doi.org/10.7688/j.issn.1000-1646.2026.04.12
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] Tapered piles demonstrate superior load-bearing capacity as single piles, while their performance within pile-net structure subgrades, particularly the influence of the wedge angle on settlement, pile-soil interaction, and the soil arching effect, remains inadequately characterized. This study aims to systematically investigate the influence mechanism of the wedge angle on the settlement behavior of tapered pile-net structure subgrades, providing a theoretical basis for design optimization. [Methods] A combined method of laboratory model testing and finite element numerical simulation was employed. The model test was conducted in a large-scale test tank, with loess as the foundation soil and custom-made beech piles to simulate tapered piles. Instrumentation monitored settlements, stresses, and strains under incremental loading. Concurrently, eight full-scale finite element models featuring different wedge angles (ranging from 0° to 1.9°) were built. Following model validation, a systematic analysis of the wedge angle′s influence was performed. [Results] An increase in the wedge angle induces greater settlement in the underlying composite foundation, leading to increased overall settlement compared to constant-diameter pile structures. Conversely, embankment settlement decreases with a larger wedge angle. The application of overburden load enhances pile-soil interaction, diminishing the wedge angle′s relative influence. The pile-soil stress ratio of tapered pile foundations is significantly lower than that of their constant-diameter counterparts, demonstrating the wedge angle′s efficacy in mitigating stress concentration at the pile top and improving the load-sharing capacity of the surrounding soil. The stress ratio generally increases with decreasing wedge angle, whereas a reversal of this trend may occur near 0°. Under a 200 kPa load, both the range and variance of settlements across the pile top plane decrease with increasing wedge angle, indicating that larger wedge angles promote surface evenness and better control differential settlement. The wedge angle inhibits the development of the soil arching effect within the embankment by moderating pile-soil differential settlement, which consequently alters the shear stress redistribution. This inhibitory effect is more pronounced in the directly loaded area. [Conclusions] The wedge angle is a critical parameter governing the performance of tapered pile-net structure subgrades. Although a larger wedge angle may increase overall settlement, it effectively reduces embankment settlement, alleviates stress concentration, controls differential settlement, and suppresses the soil arching effect. A comprehensive evaluation suggests that a wedge angle of approximately 1.43° offers superior overall performance. This research provides enhanced insight into the operational mechanisms of tapered pile-net structures and delivers valuable references for optimized engineering design.
    • Artificial Intelligence
  • Artificial Intelligence
    Multi-scale trajectory prediction based on spatial-temporal decoupling
    LAI Yue, CAI Ziman, WANG Xiaoxu, ZHANG Zirui, WU Maoqiang, YU Rong
    2026, 48(4): 100-108. https://doi.org/10.7688/j.issn.1000-1646.2026.04.13
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] In complex and heterogeneous traffic scenarios, autonomous vehicles must simultaneously consider the trajectories of multiple types of traffic entities, including motor vehicles, non-motor vehicles, and pedestrians. The precise prediction of the trajectories is crucial for safe decision-making and path planning. However, existing methods still face limitations in modeling long-term motion trends and capturing local details, restricting prediction accuracy and generalization capability. [Methods] To address these issues, a spatial-temporal transformer model (STTM) was proposed. A spatial-temporal decoupling mechanism was introduced and combined with multi-scale modeling mechanism to cover both local and global features. First, a multi-head self-attention (MHSA) mechanism was employed to model interactions among traffic entities and extract spatial correlations at each time step. Second, a multi-scale temporal convolution network (MS-TCN) was incorporated to achieve multi-granularity modeling from short-term perturbations to long-term trends. To further enhance temporal feature representation, a past-aware decomposed mixing (PDM) module was developed to achieve cross-scale feature fusion through local and global modeling pathways. Finally, a future-multipredictor-mixing (FMM) module was constructed to generate stable and diverse prediction trajectories through multi-predictor collaboration and dynamic fusion mechanisms. [Results] Experiments conducted on the ApolloScape dataset show that the STTM achieves a weighted average displacement error of 1.212 7 m and a weighted final displacement error of 2.244 6 m for different traffic entities, significantly superior to the existing methods. [Conclusions] STTM effectively models complex spatial-temporal interaction through the collaboration of MHSA and MS-TCN, and integrates the PDM and FMM modules to further enhance the stability and diversity of prediction results. The proposed method excels in prediction accuracy, generalization capability, and adaptability in complex environments, providing a feasible and effective solution for trajectory prediction in multi-entity traffic scenarios.
  • Artificial Intelligence
    Active information security protection technology for ubiquitous power Internet of Things
    LOU Fei, ZHANG Yifan, CHEN Qirui, ZHONG Zhenyuan, CHEN Bo
    2026, 48(4): 109-117. https://doi.org/10.7688/j.issn.1000-1646.2026.04.14
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] The massive electricity consumption data collected by user-side smart terminals in the ubiquitous power Internet of Things (UPIoT) face privacy leakage risks. Traditional centralized data processing models suffer from hidden security dangers in data transmission and storage, while existing federated learning schemes struggle to balance security and performance due to fixed privacy budgets, inadequate adaptation to heterogeneous data, and a lack of business integration optimization. This study proposed a hierarchical aggregation federated learning architecture to establish an active protection system spanning the entire data lifecycle, addressing data privacy, training efficiency, and system trustworthiness to support the digital transformation of the power sector. [Methods] A three-tier architecture comprising the “user terminal, edge aggregation node and power grid center” was designed. Localized data processing was conducted on data by user-side devices and edge nodes, with statistical features and frequency-domain features extracted by adopting a lightweight long short-term memory (LSTM) model. Meanwhile, only desensitized gradients and weight parameters were transmitted to the cloud, avoiding centralized exposure of sensitive data. A differential privacy enhancement mechanism was introduced, dynamically injecting Laplace-distributed noise and adaptively adjusting the privacy budget based on data sensitivity. Furthermore, sparsification and quantization technologies were combined to reduce parameter transmission volume by 92%. The paper constructed a trusted aggregation mechanism and utilized the SM2 elliptic curve cryptographic algorithm for bidirectional device authentication and blockchain evidence-storage technology to record parameter update logs, ensuring verifiable data sources and transparent processes. [Results] The experimental results demonstrate that the hierarchical aggregation scheme outperforms traditional centralized and baseline federated learning schemes in multiple dimensions. In privacy protection, compared with the traditional centralized and baseline federated learning schemes, the hierarchical aggregation federated learning scheme reduces the success rate of inversion attacks by 33% and 16% respectively, while increasing data reconstruction error by 53% and 37%. In terms of model accuracy, the proposed scheme achieves the lowest mean squared error (MSE) and mean absolute error (MAE) across multiple regions, with faster MSE and MAE decrease during training. This means the proposed scheme has a faster convergence speed. In communication efficiency, the proposed scheme reduces single-iteration data transmission to 128 kB and shortens communication time to 4.5 s, significantly outperforming traditional centralized and baseline federated learning schemes. In trusted verification, the scheme achieves a 100% identification rate for identity forgery attacks and the average anomaly tracing time is just 2.3 min, meeting the real-time requirements of power systems. [Conclusions] The hierarchical aggregation federated learning architecture effectively addresses the collaborative optimization of data security and model performance for user-side power IoT by conducting the collaborative design of edge preprocessing, dynamic privacy protection, and trusted aggregation. It demonstrates advantages in privacy protection strength, training convergence efficiency, communication overhead control, and system trustworthiness, providing a technical scheme for data security protection of smart grids. Future work should focus on exploring its integration with services such as demand response and renewable energy consumption to advance the secure, large-scale application of UPIoT.
  • Artificial Intelligence
    Stochastic SIQRS infectious disease models with time-delay effect and optimal tracking control
    ZHANG Yi, LIU Lu, JIN Zhenghong
    2026, 48(4): 118-126. https://doi.org/10.7688/j.issn.1000-1646.2026.04.15
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] To simulate the complex dynamical behavior of infectious disease transmission and achieve effective containment of infectious disease spread, a stochastic SIQRS infectious disease model incorporating multiple realistic factors is constructed, and an optimal tracking control method based on stochastic nonlinear time-delay systems is proposed. [Methods] Firstly, based on the traditional SIQRS infectious disease models, a stochastic SIQRS infectious disease dynamics model more consistent with actual spreading scenarios was constructed by incorporating key realistic factors such as media coverage effect, time-delay effect, vaccination strategy, saturation incidence and white Gaussian noise (WGN) to accurately describe the spreading characteristics of infectious diseases in complex social environments. Secondly, an optimal tracking control strategy based on the stochastic time-delay system neural network was proposed by combining the system model characteristics, optimal control theory, and tracking control method. In this strategy, adaptive dynamic programming (ADP) techniques and neural network algorithms were introduced, and a critical neural network model was built to further design an ADP neural network optimal tracking controller applicable to continuous time stochastic nonlinear state time-delay systems. This controller has excellent adaptability and can flexibly cope with the nonlinear time-delay characteristics of the system and external stochastic disturbances, thus ensuring the effectiveness and stability of the control strategy. [Results] Numerical simulation experiments verify the effectiveness of the built stochastic SIQRS infectious disease dynamics model and optimal tracking control strategy. The results show that the model can accurately simulate the spreading process of infectious diseases in complex environments. Additionally, the optimal tracking control strategy based on the stochastic time-delay system neural network not only drastically reduces the proportion of infected and quarantined people in the total population, but also notably accelerates the effective control process of the epidemic, and realizes the rapid and efficient control of the infectious disease spread. [Conclusions] A stochastic SIQRS infectious disease dynamics model with media coverage effect, time-delay effect, vaccination, saturation incidence, and WGN was built, and an optimal tracking controller based on a stochastic nonlinear time-delay system with the ADP neural network was designed. This controller achieves remarkable results in controlling the spread of infectious diseases, thus providing a new perspective and effective means for the prevention and control of infectious diseases. By innovatively integrating the stochastic SIQRS infectious disease model, optimal control theory, and tracking control theory, the optimal tracking control strategy based on the stochastic time-delay system neural network was proposed. Meanwhile, by simulating and optimizing the control parameters, the dynamic monitoring and effective intervention of epidemic development were realized to reduce the spread rate and narrow the infection scope of infectious diseases.
  • Artificial Intelligence
    Optimization method for data transmission and collaborative control of inspection UAV swarm operations
    LIU Lan, QIN Ping, ZHANG Chenrui
    2026, 48(4): 127-136. https://doi.org/10.7688/j.issn.1000-1646.2026.04.16
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    [Objective] To address the issues of rigid spectrum resource allocation, low energy efficiency, and susceptibility to interference-induced data transmission interruptions faced by unmanned aerial vehicle (UAV) swarm control in complex electromagnetic interference and dynamic topology industrial inspection scenarios, a data transmission optimization model was proposed based on a dynamic bandwidth scaling factor, along with a collaborative control method for UAV swarm operations. This study aims to enhance data transmission integrity and provide reliable communication support for swarm collaboration, thus achieving efficient collaborative operations. [Methods] First, based on Shannon′s theorem, a dynamic bandwidth scaling factor was introduced to sense available spectrum resources in real-time and adaptively adjust the bandwidth allocation ratio, jointly optimizing spectrum utilization and transmission energy consumption. Second, a cluster collaborative control function was constructed with multi-objective effectiveness, involving regional coverage efficiency maximization, environmental uncertainty attenuation, an information pheromone revisit mechanism, and deviation and collision cost quantification. Then, by combining a coupled dynamic model with a sliding mode control strategy, a collaborative controller was designed to ensure stable tracking of the desired trajectories in dynamic environments. Finally, ablation experiments were conducted to evaluate the contributions of the three methods, namely, the dynamic bandwidth optimization, the multi-objective collaborative control, and the coupled dynamics-sliding mode control. The comparative tests were also carried out with the existing control methods. [Results] The experimental results indicate that dynamic bandwidth optimization increases the control success rate by 9.2％, reduces trajectory error by 17.8％, and improves the consistency index by 9.7％. The multi-objective collaborative control contributes to a 13.4％ improvement in success rate, a 30.8％ reduction in trajectory error, and a 21.0％ enhancement in consistency. Coupled dynamics-sliding mode control achieves a 10.8％ rise in success rate, a 25.9％ decrease in trajectory error, and a 27.4％ boost in consistency. The complete method demonstrates overall superiority to all ablation groups in dynamic scenarios. Further comparative testing shows that the proposed method significantly outperforms control methods with anti-collision safety constraints and those considering search-reward balance in terms of success rate, trajectory tracking accuracy, and consistency. [Conclusions] The proposed data transmission optimization model based on the dynamic bandwidth scaling factor and the collaborative control method for UAV swarm operations effectively overcomes the data transmission bottleneck in dynamic electromagnetic environments through the synergistic effect of dynamic bandwidth optimization, multi-objective collaborative control, and coupled dynamics-sliding mode control. The transmission integrity is improved, and efficient collaborative operation is achieved. The experimental results fully validate the outstanding performance of this method in improving the overall reliability and stability of UAV swarms, providing a highly robust and strongly adaptive solution for high-dynamic industrial inspection scenarios.