[Objective] Alien species invasion has emerged as a global ecological security problem. The resulting biodiversity loss and ecosystem degradation pose a serious threat to the sustainable development of human society. Traditional biological control methods often suffer from limited control accuracy and poor robustness when faced with environmental uncertainties and random disturbances. To address the dynamic characteristics of the stochastic biological system, this study proposed an adaptive fuzzy control (AFC) strategy based on Lyapunov stability theory. By building a stochastic dynamic model incorporating white noise disturbances and unknown nonlinearities, the study focused on solving the dual-objective problem of coordinated native species protection and invasive species suppression under coupled environmental uncertainty and stochastic perturbations. [Methods] Based on stochastic differential equation theory, a stochastic biological system dynamics model for alien species invasion was constructed. A fuzzy logic system (FLS) was employed to approximate the uncertain nonlinear term in the model. The backstepping method was integrated with adaptive fuzzy control and applied to stochastic biological systems. A fuzzy backstepping controller and an adaptive law with parameter self-tuning capabilities were designed using an appropriately chosen Lyapunov function. [Results] The proposed adaptive controller exhibits intelligent adjustment capabilities, allowing the population density of native species to effectively track the desired reference trajectory within a finite time. The tracking error converges to a neighborhood of zero, demonstrating the controller′s ability to monitor and analyze errors in real time. By dynamically adjusting control inputs, the system keeps tracking errors within acceptable bounds. Moreover, all system states under alien species invasion are proven to be semi-globally uniform and ultimately bounded. Notably, the system also maintains stable convergence characteristics and demonstrates strong adaptability under varying intensities of stochastic disturbance. [Conclusions] By combining FLS with nonlinear control theory, the proposed AFC strategy effectively addresses uncertainty control in stochastic biological systems. Numerical simulations further verify the strategy′s effectiveness in both protecting native species and managing invasive species, offering novel insights for intelligent regulation of complex ecosystems.

[Objective] With the longer service life of oil and gas pipelines, the failure to identify the pipeline material due to data deficiency has become increasingly prominent. Traditional detection methods are unable to meet the engineering requirements for material identification and aging status assessment. A novel non-destructive testing method is expected to be proposed based on the magnetic compression effect. By analyzing the base value variation of magnetic flux leakage (MFL) signals in different steels under an applied external magnetic field, a correlation model between material and magnetic signal was established to provide a theoretical basis and technical approach for achieving rapid, accurate, and non-contact identification of pipeline materials. [Methods] The magnetic charge theory was integrated with the magnetic compression effect to develop a mathematical model describing the MFL field on the surface of steel under the influence of an external magnetic field. A theoretical derivation was performed to establish the functional relationship among the MFL signal base values, the external magnetic field intensity, and the material′s magnetization intensity. The degree of the magnetic compression effect was proposed to be quantitatively characterized by the magnetic compression coefficient. Through the systematic experimental design, six representative structural and pipeline steels were selected as test materials, including Q235, Q345, X52, X65, X70, and X80. The specimens were machined into a standardized dimension of 270 mm×140 mm×10 mm. On a custom-built high-precision MFL testing platform, the external magnetic field was incrementally increased from 0 to 48 kA/m in steps of 1 kA/m, and the corresponding MFL signal base values were recorded in real time. To further validate the stability of the method, comparative tests were conducted between the signals obtained from original state steel plates and those obtained from polished plates with a surface roughness of 0.8 μm. All experiments were replicated to ensure both repeatability and accuracy of the results. [Results] According to the experimental results, with the enhancement of external magnetic field intensity, the MFL signal base values in all tested steel materials exhibit an initial increase followed by a subsequent decrease. The peak position varies significantly with the magnetic properties of the material. Specifically, the signal base values for Q235 and X65 peak at 24 kA/m, for X70 and X80 at 25 kA/m, while for Q345 and X52 with higher magnetization intensity, the peaks occur at 26 kA/m. This variation pattern highly coincides with the saturation magnetic characteristics of each steel′s M-H curve, which indicates the close relation between the critical field strength for the onset of the magnetic compression effect and the material′s magnetic properties. Furthermore, tests conducted under varying surface conditions of the steel plates reveal that, although the signal amplitudes differ, the critical point at which the base value begins to decline remains consistent for the same material. This finding confirms that the method is insensitive to surface conditions, with good anti-interference capability and adaptability to diverse operational environments. [Conclusions] A non-destructive testing method was proposed for pipeline material identification based on the magnetic compression effect. By establishing the correspondence between the MFL signal base values and the external magnetic field intensity, an effective differentiation between various structural and pipeline steels was achieved. This method exhibits not only good repeatability but also strong engineering applicability. The detection results are unaffected by complex factors such as internal pipeline surface roughness or corrosion state, making it suitable for complex working conditions such as internal pipeline detection. The research results provide a new technical means for material identification and safety assessment of aging pipelines, holding significant theoretical importance and engineering application value.

[Objective] DDoS attacks, as a highly destructive network threat, seriously threaten the stable operation of the power system. Due to the complexity and variability of data traffic in the power monitoring local area network (LAN), DDoS attack traffic and normal traffic have many similarities in their manifestations, making it difficult to effectively distinguish between the two. Although traditional static threshold methods can achieve traffic monitoring to a certain extent, misjudgments often occur due to their inability to adapt to the dynamic changes of traffic. The detection effect of DDoS attacks is thus weakened and reliable security guarantee cannot be provided for power monitoring LANs. Therefore, a DDoS attack matching detection method for power monitoring local area networks was proposed based on dynamic thresholds. [Methods] Real time network traffic data in the power monitoring local area network were collected through network traffic collection devices. The entropy value of these flows was calculated using information entropy theory. The chaos degree in data could be reflected by information entropy. Normal traffic usually had a certain regularity with relatively stable entropy values. Due to the influx of a large number of abnormal packets, however, DDoS attack traffic had significant fluctuations in entropy values. Based on this characteristic, a dynamic threshold was set, and the entropy value of the traffic was determined as abnormal when exceeding this dynamic threshold. The six-tuple feature set of the abnormal traffic was then extracted, including average flow packet count, average byte count, source IP address growth, flow table survival time variation, port growth, and convection ratio, and input into a pre-trained least squares support vector machine (LSSVM) classifier. The LSSVM classifier learned from the existing samples to establish the mapping relationship between features and classes. The abnormal traffic was then classified and judged to determine whether it was DDoS attack traffic. [Results] According to the test result, the proposed method shows significant improvements on both the ROC and PR curves, with higher receiver operating characteristic curve (ROC-AUC) and accuracy recall curve (PR-AUC) values than the traditional method. This fully demonstrates that the method, with higher accuracy and recall rate in detecting DDoS attacks, can effectively identify DDoS attack traffic hidden in normal traffic and reduce the misjudgment rate. [Conclusions] The detection method based on dynamic thresholds and LSSVM classifier can effectively overcome the difficulty in distinguishing DDoS attack traffic from normal traffic in power monitoring local area networks. By improving the accuracy and reliability of DDoS attack detection, it provides a more effective DDoS attack detection method for power monitoring local area networks, helps improve the security and stability of power systems, ensures the reliable operation of the power supply, and has important practical application value for network security protection in the power industry.

[Objective] As the power system continues to expand, transmission lines, being a crucial channel for power transmission, require safe and stable operation. However, transmission lines, long exposed to the complex and variable natural environment, face multiple safety risks such as external force damage and equipment aging. To enable high-precision and high-efficiency automatic detection of potential hazards in transmission lines, this study proposed an intelligent identification technology for external damage risks in transmission lines, based on deep learning. [Methods] This study developed an integrated technical framework of “geometric correction-image enhancement-intelligent recognition”, systematically addressing key technical challenges in transmission line image recognition. In the geometric correction stage, a polynomial geometric correction model based on the least squares method was employed. By establishing an accurate coordinate mapping, this model effectively eliminated geometric distortions caused by factors such as shooting angles and lens distortion. In the image enhancement stage, a new image processing algorithm, combining bilateral filtering and the maximum between-class variance method, was proposed. This algorithm effectively removed image noise while retaining the edge features of transmission lines, providing high-quality data for subsequent recognition. In the intelligent recognition stage, a dual-optimized convolutional neural network (CNN) model was designed. The feature extraction process was optimized by dynamically adjusting the convolution kernel weights, and sparse constraints were introduced to enhance feature discriminability. Finally, precise recognition was achieved by integrating the support vector machine classifier. This method overcame the limitations of traditional technologies, such as insufficient geometric distortion correction and feature extraction, offering a comprehensive solution for intelligent identification of hidden dangers of transmission lines. [Results] Tested on real datasets containing multiple types of damage, this method demonstrates significantly higher recognition accuracy compared to mainstream algorithms such as YOLOv4 and Mask R-CNN. It shows greater robustness, especially in complex backgrounds. The method achieves an average positional offset of only 0.013 meters, fully meeting engineering application requirements. The floating point operations for processing 1 000 images reduce to 3.24×10⁹, significantly enhancing the real-time processing capability. [Conclusions] The proposed intelligent recognition technology for external damage risks in transmission lines has made significant improvements in recognition accuracy, positioning precision, and computational efficiency through innovative technical approaches and systematic optimizations. The theoretical contributions of this research include establishing a complete image processing system for transmission lines, providing a new approach for related studies; introducing a dual optimization mechanism that offers a viable solution for feature extraction in complex environments; adopting the lightweight network design, which serves as an important reference for applying deep learning models in engineering.

[Objective] With the expanding user group of social software, multi-label annotation has been increasingly adopted for text information. How to analyze the behavior and psychology of the user group through data mining of multi-label text information has become a research hotspot. A data mining algorithm for multi-label implicit knowledge based on a deep topic feature extraction model was utilized to enhance text classification accuracy and data mining efficiency. [Methods] To deeply understand the implicit knowledge in text information, the socialization, externalization, combination, and internalization (SECI) theory was employed to convert the implicit knowledge into explicit knowledge. The short-term memory capability of recurrent neural networks was utilized to improve the conversion efficiency. Considering the complexity of text information, local and global features were analyzed separately, and feature fusion was used to improve data mining efficiency. Due to the strong correlation between the context of text information, the gate mechanism of the long short-term memory (LSTM) model was applied to extract contextual dependencies, while the unsupervised latent Dirichlet allocation (LDA) topic model was selected to model the topic structure of the text to mitigate standard differences from manual labeling. Combining LDA-derived global features and LSTM-derived local features, feature stitching was performed to reduce information loss during the feature extraction. A theme controller was introduced to narrow down the inference scope, which obtained more effective text features. Simultaneously, a Gaussian decoder-based contextual topic layer was constructed to calculate the conditional probability matrix of each vocabulary under a given topic, and a Gaussian mixture decoder was used to obtain the conditional probability of the vocabulary. Topic modeling optimization and content expansion were achieved through a Gaussian mixture decoder. Finally, multi-label classification was implemented using the Softmax function to calculate label probabilities. [Results] During model training, perplexity was used as a criterion for evaluation. The proposed model exhibited better perplexity than the control groups (LDA topic model and LSTM model), demonstrating the effectiveness of feature concatenation combining the LDA topic model and LSTM model. By comparing with NVDM, LSTM, LDA, and VAETM models, with precision and recall as evaluation metrics, the proposed model improves precision and recall by 5.05% and 2.75%, respectively. [Conclusion] The comparative experimental results show that the proposed model can significantly improve the performance of text classification. Compared with the LDA topic model and the LSTM model, it outperforms in processing multi-label texts. It can efficiently mine the implicit knowledge in multi-label text data, providing an efficient and accurate solution for tasks such as text classification, semantic analysis, and information retrieval.

[Objective] Due to the influence of the limited regulated direct current (DC) power supply, amplitude control of each variable of the chaotic system, that is, variable compression, has become an essential prerequisite for chaotic circuit design and implementation. Currently, geometric control of the attractors of chaotic systems, such as amplitude control and bias control, is a hot research direction in the field of chaotic systems. Based on existing methods, a new amplitude control method was proposed in this paper in the expectation of exploring more potential applications of chaotic systems. [Methods] A five-dimensional chaotic system was developed, and its chaos was verified by using a three-dimensional phase diagram and Lyapunov exponents. After the absolute values of state variable-u in the two equations of the system were taken, two new switched chaotic systems were obtained. Compared with the phase diagram of the chaotic system, the amplitudes of these two new systems changed, and their shapes were highly similar, namely that global amplitude control was achieved. After the absolute value of-u in the second equation was taken, it became a memristive chaotic system. The existence of the memristor was verified by the pinched hysteresis loops of three frequencies. Further analysis of the memristive chaotic system was carried out. By adding the parameter k to the three nonlinear terms of the memristive chaotic system, it was found that the average amplitudes of the attractor on five dimensions changed accordingly, which indicated that the memristive chaotic system had a global amplitude control parameter. The existence of multi-stability in the memristive chaotic system was verified by the Lyapunov exponent spectrum changed with the memristive parameter a. Moreover, the absolute mean value of the signal and the phase diagram changed with a proved that when an appropriate value of the memristive parameter a was selected, global amplitude control could also be achieved. [Results] The simulation circuit equations, equivalent circuit diagram of the memristive chaotic system, and the simulated phase diagram of the chaotic system on the oscilloscope are highly similar to the computer simulation results, which indicates that the chaotic circuit design is of reliability. [Conclusion] The proposed five-dimensional chaotic system has strong chaotic property. The switching system with switching amplitude variation was proposed, providing a new direction for the research of memristive chaotic systems. In future work, it is possible to attempt to use a curved surface as the switching surface. Additionally, through computer simulation experiments, whether the phenomenon of switching amplitude variation widely exists in memristive chaotic systems will be further studied, and further work will be carried out to explore the principle of its existence. The phase diagram on the oscilloscope is highly consistent with the computer simulation experiment in five dimensions. The system has the characteristics of high dimensionality, strong chaos, and switching amplitude control, which make it have good application prospects in engineering.

[Objective] Industrial control protocol parsing is a critical component of industrial internet security. However, traditional methods suffer from poor universality and low accuracy. These issues lead to a low efficiency in protocol parsing, making it difficult to meet the demands for high precision and adaptability in real-world industrial scenarios. [Methods] A deep learning-based reverse engineering method was proposed for industrial control protocols by integrating a bidirectional encoder representations from transformers (BERT) pre-trained model, a bidirectional long short-term memory (BiLSTM) network, and conditional random fields (CRF). The goal is to enhance the universality and accuracy of protocol parsing, thereby providing technical support for security analysis and vulnerability mining in industrial control systems. First, the BERT pre-trained model was employed to dynamically encode industrial control protocol data into high-dimensional word vector representations, so as to capture the semantic information of the protocol data. Leveraging the powerful contextual understanding capabilities of BERT, the model effectively handled the complexity and diversity of protocol data. Subsequently, a BiLSTM network was utilized to model the relationships between protocol data as well as between protocol data and label data. The BiLSTM network captured long-range dependencies within the protocol data, enabling a better understanding of the structure and semantics of the protocol. Finally, CRF were introduced as constraints to optimize the prediction of protocol formats and semantics. By incorporating transition probabilities between labels, CRF further enhanced prediction accuracy and consistency. The combination of the BERT pre-trained model, BiLSTM network, and CRF enabled the format extraction and semantic analysis of industrial control protocols. Additionally, the proposed method was optimized for large-scale protocol data, which ensured efficiency and stability in complex industrial scenarios. [Results] Experiments were conducted on three typical industrial control protocols. The results demonstrate that the proposed method achieves an accuracy of over 96％ in both format extraction and semantic analysis, outperforming traditional methods. The method exhibits high adaptability and accuracy across different protocols, effectively identifying field boundaries and semantic information. [Conclusion] The proposed method significantly improves the universality and accuracy of industrial control protocol parsing, providing reliable technical support for security analysis in industrial control systems. Future work will focus on further optimizing the model, expanding its application scenarios, and enhancing its practicality.

[Objective] With the continuous expansion of oil and gas pipeline transportation, the importance of pipeline safety inspection has become increasingly prominent. Stress concentration at pipeline defects is the main cause of crack propagation and fracture accidents. However, existing detection methods struggle to achieve quantitative stress evaluation. [Methods] This study proposed a pipeline stress detection method based on dual-field stress-magnetic coupling. By incorporating changes in the Jiles-Atherton (J-A) model parameters under different pipeline stress states, a magnetic stress detection model was built. The effects of elastic stress, plastic strain, and external magnetic fields on magnetization intensity and magnetic signal characteristics were systematically analyzed. The study was grounded in the principles of magnetic stress detection, the J-A model, and magnetic charge theory. By examining the influence of stress at different stages and external magnetic fields on magnetization intensity and magnetic signals, the relationship between hysteresis loops and magnetization intensity under varying conditions was established. In addition, the variation patterns of axial and radial signals under different stress and magnetic field conditions were identified. A proportional coefficient was introduced to develop a dual-magnetic field stress detection model, and separate models for elastic and plastic stress detection were built. Finally, experiments were conducted to verify the theory. Equivalent magnetic field strength formulas for the elastic stress and plastic strain stages were derived, clarifying the variation laws of the pinning coefficient k, shape coefficient a, and domain wall coupling coefficient α with stress. Experimental validation was conducted using X80 pipeline steel specimens subjected to tensile loads ranging from 10 to 80 kN and external magnetic fields from 0 to 10A/m, with magnetic signal characteristics measured. [Results] The axial component of magnetic signals under different magnetic fields and stress levels exhibits distinct peaks, with peak positions remaining stable despite variations in external fields or stress. Tangential peaks increase with the external magnetic field, aligning with theoretical calculations. Experimental data indicate that the model closely matches measured results under high stress, with minimal error, while low-stress scenarios show slight deviations due to parameter fitting limitations. [Conclusion] In the elastic stage, tensile stress causes the hysteresis loop to rotate counterclockwise initially and then clockwise. Magnetization changes significantly under weak magnetic fields, whereas stress effects become negligible under strong fields. During the plastic stage, plastic strain reduces the slope of the magnetization curve, and both the initial magnetization curve and hysteresis loop rotate clockwise. Magnetization intensity is proportional to magnetic signals, with the ratio of strong magnetic signals to magnetization intensity serving as a proportionality coefficient dependent solely on defect size. The dual-magnetic field stress detection model demonstrates high accuracy under high stress, confirming its capability for stress detection. This study innovatively integrates the dual-magnetic field method with J-A theory, proposing a proportional coefficient-based model for separating elastic and plastic stresses. The approach resolves the issue of overlapping defect and stress signals in traditional methods, providing a high-precision, quantifiable technical solution for stress detection at pipeline defects. This advancement holds significant value for preventing pipeline failures and ensuring safe energy transportation.

[Objective] In the monitoring system of large substations, the target recognition of glass insulators is an important step to ensure the safe operation of power equipment. However, due to the complexity of the environment and the limitation of image acquisition conditions, glass insulator images often have problems such as insufficient clarity and similar color interference, which leads to the difficulty of target recognition and directly affects the safety monitoring effect of substations. [Methods] To solve this problem, a target recognition algorithm was proposed for glass insulators in large substations under similar color interference. The original image was converted from RGB space to HSV space to address insufficient image sharpness and similar color interference. By fine decomposition of hue H, saturation S, and brightness V components in HSV space, the feature difference was calculated to enhance the color performance and visual effect of the image, so as to effectively eliminate similar color interference. An adaptive threshold segmentation technique, combined with the color features of HSV space, was used to accurately segment the image, and the glass insulator target region and complex background were separated. A dual-scale classification convolutional neural network (CNN) was designed to realize high-precision target recognition of glass insulators under complex background through multi-scale feature extraction and classification. The network combined local details and global context information to further improve the robustness and accuracy of recognition. [Results] The experimental results show that the proposed algorithm has significant advantages in application. In terms of color enhancement, the feature difference calculation in HSV space significantly improves the color contrast and visual effect of the image and effectively eliminates similar color interference. In terms of image segmentation performance, the adaptive threshold segmentation technique can accurately separate the glass insulator target region and the complex background, and the segmentation accuracy reaches a high level. In the aspect of target recognition, the dual-scale classification CNN shows strong anti-interference ability under complex background, and the recognition accuracy of glass insulators is significantly higher than that of traditional methods. [Conclusion] The target recognition algorithm proposed in this study for glass insulators in large substations under similar color interference successfully solves the target recognition problems including insufficient image sharpness and similar color interference through the organic combination of color enhancement, adaptive threshold segmentation, and dual-scale classification CNN. The algorithm has excellent performance in color enhancement, segmentation performance, and anti-interference ability and can recognize glass insulator targets efficiently and accurately, which provides a reliable technical guarantee for the safety monitoring of large substations.

[Objective] With the continuous expansion of network scale and the evolving complexity of attack techniques, network traffic anomaly detection has become a critical link in ensuring network security and maintaining the stable operation of key information infrastructure. However, traditional machine learning methods generally face bottlenecks such as slow convergence and insufficient feature representation accuracy when handling complex network traffic feature extraction, which limits their effectiveness in practical anomaly detection scenarios. To address these challenges, an innovative spatiotemporal fusion deep learning model, C²-GRU, was proposed in this paper, which was based on a convolutional neural network (CNN)-enhanced learner with a gated recurrent unit (GRU). The proposed model aims to enhance the multi-dimensional detection performance for abnormal traffic. [Methods] A dual-fusion deep learning framework was designed, leveraging the strength of CNN in spatial feature extraction and the capability of GRU in temporal feature modeling. A C-GRU model was constructed to achieve preliminary spatiotemporal feature fusion. It was then cascaded with CNN to form the C²-GRU model, which extracted spatiotemporal features through dual parallel convolution operations. This approach effectively captured the multidimensional features of abnormal traffic in complex network environments. [Results] The experimental results demonstrate that the proposed model achieves optimal overall performance on the KDD99 dataset. Specifically, the fused model attains an accuracy of 99.89％ and an area under curve (AUC) of 0.990 2, significantly outperforming individual CNN and GRU models. Furthermore, compared to traditional anomaly detection models, the proposed model not only achieves high recognition performance but also exhibits a relatively short model runtime, which highlights its superior engineering applicability. [Conclusion] The proposed C²-GRU hybrid model, employing a dual-convolution fusion strategy, effectively enhances spatiotemporal feature learning, suitable for abnormal traffic detection in complex network environments. It has dual advantages in anomaly recognition accuracy and computational efficiency, capable of offering technical support for securing key information infrastructure and mitigating the economic losses caused by network attacks. The model is of significant practical reference value for ensuring network information security.

[Objective] With the rapid development of power generation from renewable energy, photovoltaic power generation is widely adopted due to its merits of safety, reliability, flexible adjustment, and clean production. Due to the real demand for large-scale photovoltaic power generation, multiple inverters connected in parallel and grid-connected inverters are often adopted in photovoltaic power stations to enhance the power generation efficiency. However, with the expansion of the grid-connected scale, the inductive impedance under the weak grid poses a threat to the stability and reliability of the grid, leading to poor global resonance suppression as well as a high risk of uncontrollable system stability. The aim of this study is to propose a global resonance suppression strategy for photovoltaic (PV) multi-inverter parallel system to guarantee the stable operation of the system and improve its power quality. [Methods] Firstly, a Norton equivalent model of the PV multi-inverter parallel system was constructed. Based on this model, this paper analyzed in depth the resonance characteristics of the multi-inverter parallel system under a weak grid, and it was found that the coupling resonance frequency was negatively correlated with the number of inverters. Secondly, based on the control theory, the optimal control strategy combining capacitor current feedback and grid voltage feed-forward was applied to solve the global coupling resonance problem in the multi-inverter system. At the same time, the global resonance suppression strategy of paralleling virtual admittance at the point of common coupling (PCC) was designed to realize the effective suppression of global resonance from the system level. Finally, comparative simulation experiments before and after adopting the strategy proposed in this paper were conducted under two-inverter parallel system and four-inverter parallel system. In addition, simulation experiments were also carried out to compare the suppression effect under the same system by using other methods reported previously and the strategy proposed in this paper. The correctness and effectiveness of the proposed strategy were verified through simulation. [Results] Theoretical analysis and simulation results show that the proposed global resonance suppression strategy can significantly improve the stability of the system. The rationality of the control strategy and its parameters are validated and optimized by the Nyquist criterion. Simulation test results show that after the application of the proposed strategy, the harmonic content in the system is reduced from 17.32％ to 1.71％. This indicates that the proposed strategy can effectively suppress the global resonance of the system and enhance the stability of the system operation. [Conclusion] In this paper, a Norton equivalent model of PV multi-inverter parallel system was constructed. Innovatively, the resonance characteristics of the multi-inverter parallel system under a weak grid were analyzed, and a global resonance suppression strategy of paralleling virtual admittance at the PCC was proposed on the basis of the optimal control of capacitor current feedback and grid voltage feed-forward. The strategy effectively improves the stability of the system operation in the presence of a large number of parallel inverters and high inductive reactance of the grid. The comparative simulation verification further demonstrates that the proposed strategy can suppress the global resonance of the system effectively, providing important reference for the efficient operation of PV power generation grid-connected system.

[Objective] With the rapid development of power engineering, construction site safety has become increasingly critical. Traditional manual inspection methods are time-consuming and prone to errors. In recent years, advancements in computer vision, deep learning, and knowledge graph technologies have made it possible to automatically recognize unsafe operation behaviors. However, existing computer vision methods have limitations in detecting small objects and lack high-quality databases for unsafe operation inference. To address these issues, knowledge graphs, ontology models, graph databases, and computer vision techniques were integrated to detect unsafe operations through entity detection, scene analysis, and spatial relation reasoning. An improved self-attention mechanism was also introduced to enhance small object detection capabilities. [Methods] The proposed method mainly involved ontology model construction, knowledge extraction, and knowledge reasoning. First, an ontology model of construction safety was built based on engineering documents, historical accident reports, and safety hazard reports, with information categorized into six types:entities, attributes, time, space, events, and attribute values, which were represented by normative knowledge. Second, computer vision techniques were employed to detect entities and their attributes and extract spatial relationships between entities. A Mask region-based convolutional neural network (Mask R-CNN) was used for object detection, with an improved self-attention mechanism incorporated to improve small object detection accuracy. As a result, model performance was optimized, and computational complexity was reduced. Finally, a Neo 4j graph database was utilized to store entities and their relationships, enabling automatic recognition of unsafe operations through database queries. In this way, structured reasoning for construction safety knowledge was achieved, and the intelligent level of recognizing unsafe operations was enhanced. [Results] In the experiments, a power engineering construction site was used as the test environment, and six kinds of unsafe operations that could lead to high-altitude falling were selected for simulation experiments. The simulation results indicate that the proposed method outperforms existing approaches in both detection accuracy and training efficiency. Particularly, the improved model demonstrates superior accuracy in small object detection. Additionally, scene segmentation was conducted using a feature pyramid network (FPN) and a unified perceptual parsing (UPP) method, which significantly improved the scene understanding capability of the model. Furthermore, the knowledge reasoning approach based on the Neo 4j graph database effectively integrates entity attributes and spatial relationships, enhancing the automation of unsafe operation recognition. [Conclusion] The proposed method can accurately detect unsafe operations in complex construction environments, thereby improving the intelligence level of construction site safety management. The key innovations of this research are as follows:integrating computer vision with an ontology model to enhance automation in construction safety management; improving the self-attention mechanism by modifying convolutional kernels and introducing a global max-pooling layer, which enhances the small object detection capability of the Mask R-CNN; incorporating the Neo 4j graph database for structured storage and reasoning of construction safety knowledge. This study provides an efficient and scalable solution for the automatic recognition of unsafe operations on construction sites.

[Objective] With the continuous expansion of power grid scale and the increasing complexity of the operating environment, surface corrosion of transmission equipment has become a critical factor threatening the safe operation of power grids. Traditional manual inspection methods are not only inefficient but also struggle to accurately identify subtle corrosion features on equipment surfaces, especially in complex natural environments where the boundaries of corrosion areas are often blurred, posing significant challenges for precise recognition. To address this, a fine-grained recognition method for corrosion areas on the surface of transmission equipment based on image semantic segmentation was proposed, aiming to achieve precise detection and recognition of corrosion areas through deep learning technology. [Methods] The core of this method was the construction of a semantic segmentation network integrated with an attention mechanism. This network, by introducing both channel attention and spatial attention mechanisms, could effectively capture the subtle features and precise boundaries of corrosion areas. Specifically, the channel attention mechanism enhanced the response to channels with prominent corrosion features by analyzing the relationships among various channels in the feature map. Meanwhile, the spatial attention mechanism strengthened the spatial feature representation of corrosion areas by focusing on the spatial location information in the feature map. After the initial segmentation, the K-means++clustering algorithm was employed to perform clustering analysis on the RGB values of the pixels in the segmented images. By optimizing the selection of initial clustering centers, this algorithm effectively avoided the issue of local optimum that could arise with the traditional K-means algorithm, thereby more accurately dividing corroded and uncorroded areas. To further improve recognition accuracy, the structural similarity index was introduced to evaluate each clustered area, and fine-grained recognition of corrosion areas was achieved at the pixel level by calculating the structural similarity between areas. [Results] Experimental results demonstrate that the proposed method exhibits remarkable performance on a dataset of transmission equipment images in complex natural environments, achieving a significantly improved corrosion area recognition accuracy and an obvious improvement in boundary localization accuracy compared to traditional methods. [Conclusion] In summary, the semantic segmentation network integrated with an attention mechanism, combined with the K-means++clustering algorithm and SSIM evaluation, pioneers an efficient and precise new approach for fine-grained recognition of corrosion areas on the surface of transmission equipment. By incorporating the attention mechanism, the proposed method effectively addresses the challenges posed by complex corrosion features and blurred boundaries, significantly enhancing recognition accuracy. Meanwhile, the combination of the clustering algorithm and SSIM evaluation enables pixel-level detailed differentiation, further improving the fineness and practicality of recognition and providing solid technical support for the safe monitoring and maintenance of power grids. Not only does the proposed method ensure the safe and stable operation of power grids, but it also offers valuable insights and inspiration for the application of image recognition and segmentation technologies in other fields.

[Objective] Convolutional neural networks, as an important technology in the field of deep learning, have demonstrated outstanding performance in multiple areas such as image recognition, object detection, and natural language processing. However, with the increase in model depth and complexity, the size and computational requirements of convolutional neural network models increase sharply, which poses severe challenges to model deployment and real-time applications. [Methods] Therefore, to reduce the size and computational complexity of neural network and improve the efficiency and deployability of the models, a convolutional neural network compression method based on knowledge distillation was proposed. Model compression and acceleration could be achieved by transferring knowledge from large complex models (teacher network models) to small simplified models (student network models). A high-performance teacher network and a student network with a simpler structure and fewer parameters were established. Abundant feature representations and accurate prediction results were provided by the teacher network, while the student network learned the behavior of the teacher network to approach its performance. A standard loss function was used, and its parameters were iteratively updated through the back propagation algorithm to ensure that it achieved good performance on the training dataset. An improved knowledge distillation method was adopted to obtain a comprehensive threshold function, which was used to evaluate the knowledge differences between the teacher network and the student network and guide the learning process of the student network. During the training process, the student network was supervised by using this comprehensive threshold function, gradually approaching the output of the teacher network while maintaining a small model size and computational complexity. In this way, the compression processing of the convolutional neural network was realized. [Results] The results show that the proposed method exhibits good model compression performance on both ImageNet and Labelme datasets. The proposed method has a high degree of fitting for the output results of the convolutional neural network before and after compression, which indicates that the student network has successfully learned the key features of the teacher network. The cross entropy loss value is relatively low, around 1.0, further verifying its good predictive performance. The compression time for completing the convolutional neural network model is relatively short, between 79.8 and 89.4 s, indicating that the proposed method has high computational efficiency. [Conclusion] From the above results, it can be seen that the convolutional neural network compression method based on knowledge distillation can effectively reduce model size, decrease computational complexity, and maintain or even improve model performance. The proposed method provides not only a new approach for model compression but also strong support for the deployment and application of deep learning models. In addition, the proposed method is improved on the basis of the knowledge distillation method. A comprehensive threshold function is introduced to more comprehensively evaluate and guide the learning process of the model, which enhances the effectiveness and efficiency of knowledge distillation to some extent. Therefore, the proposed method has not only theoretical value but also important practical significance.

[Objective] Remote sensing images, as an important means of Earth observation, are widely used in various fields such as environmental monitoring, resource exploration, and disaster warning. However, remote sensing images are easily affected by sensor noise, atmospheric interference, and other factors during the acquisition process, which leads to a decrease in image quality and blurry details and thus poses significant challenges to subsequent image analysis and target classification. In the task of multi-label remote sensing image classification, traditional supervised learning methods are inadequate with significant classification errors as multiple categories of targets exist in the image, and there may be complex correlations and dependencies between these targets. [Methods] Therefore, to effectively address the impact of remote sensing image noise, accurately capture image features, and improve classification accuracy, a multi-label remote sensing image classification method based on semi-supervised learning was proposed. The remote sensing images were preprocessed using the perceptual loss function. By searching for pixel positions with missing details and blurs in the images, the signal-to-noise ratio residuals of the original and defective images were calculated, and the degree of degradation in remote sensing image quality was determined. A residual mapping based image denoising algorithm was designed, which adjusted the spectral values of noise positions according to the residual mapping values. By adjusting the relationship between high and low frequencies of pixels, the signal-to-noise ratio was improved, and the detailed information in the image was restored. The semi-supervised learning method was used to update and improve the image classifier, which improved the processing efficiency and classification accuracy of remote sensing images, thus achieving the classification of multi-label remote sensing images. [Results] To verify the effectiveness of the proposed method, image classification experiments were conducted at different resolutions and principal component numbers, and classification experiments were designed for different types of remote sensing images. The test results show that the proposed method performs well in denoising and image detail restoration and can clearly distinguish the color blocks in each region, restoring key detail information in the image. In terms of landform feature extraction, its result has a high degree of consistency with the actual landform distribution with only small errors, which proves its advantages in remote sensing image feature extraction. In terms of image classification accuracy, the proposed method achieves a classification accuracy of 0.88 at an image resolution of 70×80 and the principal component number of 12, demonstrating high classification accuracy. Meanwhile, when classifying different types of remote sensing images, the proposed method has a classification accuracy above 0.9 with a maximum of 0.98, which fully verifies its wide applicability and high classification accuracy. [Conclusion] The above results indicate that the proposed method achieves multi-label remote sensing image classification by utilizing an image denoising algorithm that combines the perceptual loss function and residual mapping and a semi-supervised learning method. It not only improves the efficiency and accuracy of remote sensing image classification but also provides new ideas and technical supports for the field of remote sensing image processing, which has higher theoretical significance and practical application value.

[Objective] In the context of the rapid development of the aviation industry, the scale and level of air transportation have significantly improved, air transportation becomes an indispensable mode of transportation in economic activities. However, the issue of cargo loading path planning in air transportation limits the optimization of transportation efficiency and cost. To address the challenges of enhancing operational efficiency and optimizing costs in air transportation, this paper proposed an air transportation loading path optimization algorithm based on an adaptive genetic algorithm. [Methods] To elucidate the loading path optimization algorithm for air transportation, this study analyzed the actual needs of air transportation loading and the computational conditions of the path planning platform and explored the transportation cost factors influencing the optimization of air transportation loading paths. On this basis, an improved genetic algorithm with adaptive capabilities was employed, utilizing adaptive fitness functions, crossover probabilities, and mutation probabilities to circumvent the issues of poor stability and slow convergence speeds inherent in traditional algorithms. The essence of this algorithm was the dynamic adjustment of crossover probabilities and mutation probabilities to align with the evolutionary state of the population, thereby augmenting the algorithm′s global search capability and convergence speed. During the research, the encoding method of the adaptive genetic algorithm, the establishment of the fitness function, and the calculation method and control principle of crossover probabilities and mutation probabilities were detailed, along with the specific execution steps of the loading path optimization algorithm. The algorithm was implemented on the MATLAB platform and tested using actual distribution data from an air transportation airport. [Results] The simulation results demonstrate that compared to traditional genetic algorithm, intelligent water drop algorithm, and improved ant colony algorithm, the air loading path optimization algorithm based on the adaptive genetic algorithm exhibits significant advantages in both transportation efficiency and overall transportation cost. In other words, the air loading path optimization algorithm can effectively reduce the average transportation cost and enhance transportation efficiency. However, actual air transportation loading processes are influenced by complex environment factors, such as the size limitation of aircraft cargo hold and the complex road conditions during delivery. These problems have not been deeply considered in the algorithm, which indicates that there is still room for improvement in the algorithm. [Conclusion] In summary, the air transportation loading path planning algorithm based on adaptive genetic algorithm introduces an improved genetic algorithm with adaptive mechanism, which shows better global search ability and convergence speed in solving the air transportation loading path planning problem. This paper provides a new idea for air transportation loading path planning and is also of important theoretical and practical value for the field of air logistics. Future studies will aim to take into account more actual operating environment factors to further optimize algorithm performance.

[Objective] The iron and steel industry, as one of the pillar industries of economic development in China, has an irreplaceable position in the entire manufacturing industry. Hot rolled strip steel has the advantages of strong covering capacity, easy processing, and material saving and is the raw material for producing other steel products. Improving the surface quality of strip steel products is an important part of improving the quality of steel products. Due to the influence of many factors of production, processing, shooting, etc., the brightness of the surface defect image of the original strip steel is uneven, and the contrast between the defect area and the non-defect area is low. As a result, the defect information is not clear enough for easy detection. To solve the above problems, a method for surface defect image enhancement of strip steel based on wavelet denoising and improved homomorphic filtering was proposed. [Methods] The original image was decomposed into low-frequency component and high-frequency component by two-level wavelet transform. The low-frequency component contained the main information of the original image, which was enhanced to improve the overall effect of the image. The improved homomorphic filtering algorithm and the contrast limited adaptive histogram equalization (CLAHE) algorithm were used to enhance the low-frequency component, equalizing the image brightness and improving the overall contrast. Moreover, the low-frequency images after being processed by the above two algorithms were fused with appropriate weights to obtain the enhanced low-frequency component. The high-frequency component contained the detail information of the image and noise. The improved threshold function was used to improve the denoising effect of the high-frequency component, and the edge details were well preserved. Finally, the processed low-frequency and high-frequency components were subjected to wavelet reconstruction to obtain the final enhanced image. [Results] Multiple sets of comparative analysis were conducted on the processing results of the algorithm through subjective visual evaluation and objective evaluation indicators. Brightness is significantly improved for all kinds of surface defect images of strip steel which are enhanced by the proposed algorithm compared with other algorithms, and the overall brightness remains balanced. At the same time, the contrast is improved, and the texture details and defect information of the images are more obvious. The mean square error (MSE), peak signal-to-noise ratio (PSNR), and image entropy (IE) were used for evaluating the algorithms, and each parameter was comprehensively analyzed. The proposed algorithm has a remarkable effect on improving contrast and reducing noise. It retains more details and leads to less distortion. [Conclusion] Experimental results show that the proposed algorithm can effectively enhance the brightness evenness and overall contrast of surface defect images of strip steel, strengthen the denoising effect, and significantly enhance defect information and edge details. It is suitable for the detection of various types of strip steel surface defects.

[Objective]High-value agricultural products such as precious medicinal materials and organic fruits generally have requirements for high-quality control and preservation and usually require initial processing before entering the market circulation. Therefore, the location of the initial processing center plays an important coordinating role in balancing the dispersed rural procurement logistics on the production end and the urban distribution logistics with dense distribution points. Given the common characteristics of poor coordination between rural procurement logistics and urban distribution logistics and a high proportion of transportation costs for high-value agricultural products, how to reduce costs and increase efficiency while ensuring customer satisfaction is a key issue that urgently needs to be addressed in the location-path planning of high-value agricultural products. [Methods]A two-stage logistics location-path planning model was proposed with the goals of minimizing total cost and maximizing customer satisfaction. The first stage focused on the location of the drying center, considering construction costs, transportation convenience, service radiation range, etc., to construct a location model and optimize the initial processing center location that matches the Chinese herbal medicine production area and users' locations. In the second stage, logistics transportation paths were planned depending on the selected initial processing center location, with vehicle capacity, speed, and time windows taken as constraints. A multi-objective path planning model was constructed by integrating transportation, penalties, cargo damage costs, and customer satisfaction. To solve the above model, particle swarm algorithm, differential evolution concept, and population evolution factors were integrated into the bacterial foraging algorithm, and a hybrid multi-objective bacterial foraging optimization-niche multi-objective particle swarm optimization (MOBFO-NMOPSO) algorithm was proposed for multi-objective optimization. The designed algorithm improved solution accuracy by introducing niche multi-objective particle swarm optimization (NMOPSO). Differential evolution was introduced in replication operations to preserve population diversity. Population evolution factors were introduced into migration operations to improve the convergence speed of the algorithm. For the verification of the effectiveness of the model and algorithm, the proposed MOBFO-NMOPSO algorithm was compared with nondominated sorting genetic algorithm II (NSGA-II), multi-objective bacterial foraging optimization (MOBFO), NMOPSO, grey wolf optimizer with estimation of distribution algorithms (GWOEDA), genetic algorithm (GA), and other algorithms, which verified the advantages of the proposed algorithm in solving performance and speed. Then, with the actual data of S enterprise's Chinese herbal medicine supply chain as a research example, the two-stage location-path planning problem was comprehensively solved by considering the construction cost of the drying center, vehicle transportation cost, time penalty cost, and cargo damage cost. [Results]The simulation results show that the optimized transportation cost of the enterprise is reduced by 10.26％, and customer satisfaction is improved by 44.84％, which verifies the effectiveness of the model in solving high-value agricultural product logistics planning problems. Finally, considering the quantity of Chinese herbal medicine production areas, logistics costs, and customer satisfaction, actual logistics path solutions were designed for S enterprise's Chinese herbal medicine supply chain, taking into account different extreme and compromise solutions, for the enterprise to choose from. [Conclusion]The two-stage location-path planning model and the improved MOBFO-NMOPSO algorithm constructed in this study can effectively enhance the competitiveness of the supply chain by reducing the total cost of the supply chain, stabilize the supply-demand cooperation relationship by improving customer satisfaction, and effectively promote the coordinated and steady development of the supply chain of high-value agricultural products by constructing a two-stage logistics planning system to improve the operation efficiency of high-value agricultural products.

[Objective]In the application process of deep neural networks, their huge computing requirements and storage overhead have become bottlenecks that restrict their widespread application on edge devices. Edge devices are limited by deficient computing resources and storage space, which makes it particularly difficult for them to efficiently run complex deep neural network models. Therefore, how to reduce the complexity and computational load of deep neural networks while maintaining model accuracy to meet the requirements of edge devices for lightweight edge resources has become an important research topic at present. To improve the performance of deep neural networks in edge devices, an optimization method for deep neural networks was proposed which combines the ant colony algorithm and dual-angle parallel pruning. [Methods]The structural characteristics of deep neural networks were analyzed, and a deep neural network model with multiple hidden layers was constructed. The ant colony algorithm was utilized to search for approximate optimal solutions in complex spaces by simulating the pheromone transmission mechanism in the process of ants foraging. Similar nodes in the hidden layers of the constructed model were clustered to identify highly similar neuron nodes and group them into the same category, which reduced the scale and complexity of the network. On this basis, dual-angle parallel pruning processing was further carried out on redundant nodes and free nodes after clustering. On the one hand, from the perspective of the sparsity of the weight matrix, nodes with small weights were pruned to reduce computational overhead. On the other hand, from the perspective of node contribution, the contribution of each node to the overall output result was evaluated, and nodes with small contribution were pruned. [Results]The experimental results show that compared to the original model without pruning, the deep neural network pruned using the proposed method has a higher data volume of 120 MB, an average network complexity of 88.32％, and scalability of 99％ while maintaining a high accuracy within the same computation time. This means that under limited resource conditions, deep neural networks can run more efficiently and better adapt to the needs of edge devices with the help of the proposed method. The experimental results not only validate the effectiveness of the proposed method but also provide new ideas for the deployment and application of deep neural networks on edge devices. [Conclusion]The proposed method applies ant colony algorithm to the pruning process of deep neural networks, achieving effective clustering of similar nodes in the hidden layers and providing accurate targets for subsequent pruning. At the same time, the dual-angle parallel pruning strategy further improves the efficiency and effectiveness of pruning, ensuring the balance between accuracy and scalability of the pruned model. The proposed method can not only promote the widespread application of deep neural networks on edge devices but also provide useful reference and guidance for complex network optimization problems in other fields.

[Objective]In the power system, overhead transmission lines are a critical link in the transmission of electrical energy, and their safe and stable operation is crucial. However, with the continuous changes in the natural environment and rapid growth of vegetation, trees in transmission line corridors have become one of the main hidden dangers affecting line safety. The high proximity between trees and transmission lines may not only cause faults such as short circuits and tripping but also lead to fires in severe cases, posing a serious threat to the safety of the power grid and people's lives and property. Therefore, to improve the accuracy of identifying tree obstacles, this paper designed a method to identify tree obstacles for overhead transmission lines based on unmanned aerial vehicle (UAV) inspection images. [Methods]To improve the quality of UAV inspection images, histogram equalization was used to enhance the contrast of the images, making the detailed information in the images clearer. The use of transformation functions further enhances the edge features of the images, laying the foundation for subsequent feature extraction. The FROST filter was used to remove image noise, ensuring accuracy of subsequent processing while preserving edge details. The images were smoothed using binarization methods, and the color features of tree obstacles and the texture features of conductor sag of the transmission line were extracted from the inspection images. In response to the missing edge information in images due to factors such as shooting angle and lighting, an interpolation algorithm was used to supplement the missing image edge values, ensuring the integrity of feature extraction. On this basis, the Euclidean distance between adjacent data was calculated to obtain the annotation results of feature fusion. Consequently, hidden dangers in overhead transmission line corridors were identified. [Results]The experimental results show that the proposed method performs well in the task of identifying hidden dangers brought by tree obstacles for the overhead transmission lines. It not only accurately identifies 5 areas of hidden dangers from tree obstacles but also has a small error between the identification results and the actual number of hidden dangers due to tree obstacles, demonstrating excellent identification ability. In the accuracy analysis of the location coordinates of hidden dangers, the proposed method identifies coordinates of (1.43 m, 8.3 m) and (1.49 m, 9.8 m) in areas b and d, respectively, which are closest to the actual data. This proves the high accuracy of the proposed method in identifying the actual distance of tree obstacle areas. In addition, compared to the other methods, the proposed method has more accurate identification results for various levels of hidden dangers brought by tree obstacles, and the values are closer to the actual situation in the experimental area, which verifies its superiority and reliability in practical applications. [Conclusion]The proposed method can effectively identify the hidden danger areas of transmission lines and accurately judge the number and characteristics of hidden dangers, having high practicability. From the above results, it can be seen that the proposed method combines UAV inspection images with advanced image processing technology to achieve automated and intelligent identification of tree obstacles in overhead transmission line corridors. In addition, by integrating color features and texture features, the accuracy and robustness of identification have been improved. This research achievement is of great significance for improving the safety and stability of the power system and has made positive contributions to promoting the construction and development of smart grids.

[Objective]In the context of rapidly advancing network technology, due to the insufficient efficiency and accuracy, traditional network detection techniques struggle to meet the complex demands of network management. Particularly in power communication networks, the statistics and management of network traffic, structure, and load become intricate, which makes it difficult for network management technicians to quickly propose effective remedial measures when cyber security events occur. This affects the quality of Internet services and the stability of social order. Therefore, a network cooperative detection system based on a multi-Agent model was proposed to enhance the efficiency and accuracy of network detection. [Methods]The efficiency and accuracy of network detection were significantly improved by integrating active and passive detection functions into a network topology algorithm and incorporating various agents and dynamic decision-making mechanisms. The active detection technology used the Traceroute algorithm to discover active devices and open ports in the network, while the passive detection technology collected detailed information from network traffic in line with protocols such as simple network management protocol (SNMP). A more comprehensive view of network assets was obtained by the combination of the two. In the specific research, a module deployment and technical architecture that integrated active and passive network detection technologies was designed, and an organizational structure of the distributed detection system was established. [Results]Simulation experiments and analysis show that under the same testing environment and process, compared to single passive and active network detection systems, the network cooperative detection system has stronger communication performance and shorter detection time while consuming less time. [Conclusion]In summary, the network cooperative detection system demonstrates superior communication performance and detection efficiency in simulation experiments, capable of detecting more hosts in a short time with greater data traffic and a broader coverage range, which further validates the feasibility and effectiveness of the proposed system. During actual tests, in complex network environments that include multiple operating systems, the network cooperative detection system based on the multi-Agent model detects the most hosts and is able to clearly identify the composition of host operating systems. This system not only improves the efficiency and accuracy of network detection but also has significant importance for applications requiring real-time responses, which enhances the response speed and processing capabilities of network management. It holds important theoretical and practical value for network security and optimization. There is still room for optimization and improvement in the theoretical mechanisms and detection time of network cooperative detection systems that can meet a wide range of engineering requirements, which remains a core issue in the field of network detection research.

[Objective] The power monitoring system carries a large amount of sensitive information, including the operation data of power equipment, power load conditions, and user power consumption information. Since these data are crucial to ensuring the safety and stability of the power system, it is a key issue in the power monitoring system to protect the security of these data. As a core component of the power monitoring system, the distributed database has advantages such as fast processing speed and large data storage capacity, but it also faces security risks such as data leakage and illegal access. [Methods] To enhance the security of the power monitoring system, this paper proposed a chaotic encryption control method for secure access to distributed databases of power monitoring based on a binary Trie tree. The SRP-6 protocol was used for identity authentication to ensure that only authorized users could access the power monitoring system. The SRP-6 protocol effectively prevented access of malicious attackers through forging identity by encrypting identity information. In the environment of distributed databases, key management was a highly challenging task. Since a distributed system involves multiple nodes, and different users have different permissions and keys, an efficient and secure key management mechanism was required. This paper took a binary Trie tree as the key management structure. Binary Trie trees are a kind of efficient data structure that can quickly retrieve and store keys. Through the hierarchical structure of a binary Trie tree, keys can be quickly located and allocated. Each node represents a step in key management, and leaf nodes store the specific encryption keys. Based on this tree structure, the system can easily manage a large number of keys, reduce the redundancy of key storage, and improve the efficiency of key distribution. The use of the binary Trie tree can also reduce the risk of key leakage and enhance the overall security of the system. On the basis of key management, chaotic encryption technology was introduced, combining two chaotic models, Tent mapping and Logistic mapping, to generate random numbers and round keys required for encrypted data. Tent mapping mapped input values to the interval [0, 1) through folding and expansion operations, while Logistic mapping generated pseudo-random sequences by iteration. The two together formed the basis of chaotic encryption. The data in the distributed database of power monitoring were encrypted into unpredictable ciphertext, which greatly increased the difficulty for attackers to deduce plaintext by analyzing ciphertext. [Results] The experimental results show that the proposed method has good performance while enhancing encryption strength and improving system security. By controlling the frequency of ciphertext distribution and ensuring the uniformity of encrypted data, the distribution frequency of ciphertext can be maintained above 2 800 times. The avalanche effect value increases to above 0.524, which proves the effectiveness of the proposed method in data encryption. [Conclusion] Compared with traditional methods, this method reduces the consumption of computing and resource storage while ensuring data security, which is applicable to large-scale distributed databases.

[Objective] With the rapid development of social network technology, obtaining the topology of large-scale complex networks has become an urgent problem in various disciplines such as electronics, networks, biology, and medicine. Typically, large-scale complex networks consist of participating nodes and virtual connections, where nodes represent roles such as individuals, families, and society, and connections depict the complex relationships between these roles. Generally, there is a phenomenon of extremely high homology in complex networks, namely that there are a large number of repetitive or similar architectures, which greatly increases the difficulty of discovering the dynamic structure of networks. [Methods] Based on the principle of temporal locality, a heuristic network community discovery algorithm was proposed to further optimize the accuracy and running time of topology discovery. By modifying the calculation rules of nodes within adjacent time ranges and using the cosine similarity criterion, the topology discovery algorithm deeply described the predictability of complex relationships among multiple participating nodes in the network. Specifically, the algorithm was based on the classical Louvain algorithm, optimizing the accuracy and running time of community detection by calculating incremental modularity and cosine similarity. In addition, the algorithm used the concept of modularity to accurately measure network topology, and the calculation formula of incremental modularity indicators was introduced to grasp the changes in topology discovery algorithm indicators in real time. [Results] To verify the effectiveness of the proposed algorithm, simulation was conducted using an actual communication dataset of a smart grid, which included 616 pieces of communication connection data of 115 power-using units. The simulation results show that compared with the classical Louvain algorithm, the proposed algorithm has significant advantages in detection efficiency and running time. The comparative analysis of normalized mutual information indicators shows that the proposed algorithm has higher normalized mutual information and lower average running time when the number of participating nodes is large. This indicates that the new algorithm has superiority in large-scale networks, although its performance is slightly inferior in small-scale networks. Simulation with the actual dataset reveals that the topology discovery algorithm based on temporal locality has obvious advantages in the precise discovery of large-scale smart distribution networks, able to provide strategies for optimizing network topology discovery in fields such as smart grids. [Conclusion] In summary, the innovation of the topology discovery algorithm based on temporal locality lies in applying the principle of temporal locality to network community discovery. This paper provides a new perspective and method for the study of complex network community discovery algorithms, having reference significance for researchers in related fields. Future research will address the application issues of the algorithm in small-scale networks and further analyze the robustness of the algorithm.

[Objective] Industrial Internet is an important part of the national key infrastructure, and its security is directly related to national security, economic stability, and social order. In recent years, with the widespread application of industrial Internet, network attacks targeting industrial control systems have occurred frequently, causing serious economic losses and social impact. Therefore, it is particularly important to develop efficient real-time intrusion detection systems. Traditional intrusion detection systems often fail to effectively distinguish between normal traffic and abnormal traffic when processing high-dimensional network traffic data, especially in the absence of abnormal traffic samples. [Methods] To solve this problem, this study proposed a real-time anomaly detection method for industrial networks which combined Suricata and density clustering based on a sliding window through analyzing the real network traffic characteristics of an oil and gas pipeline industrial control system. This method utilized the open-source and extensible nature of Suricata and the dynamic detection capability of the density clustering algorithm based on a sliding window to establish a full-process intrusion detection model from traffic collection and analysis to real-time intrusion detection. Through analyzing the network traffic characteristics in the real industrial control system environment, this study finds that industrial network traffic has a certain periodicity. By using the GINI coefficient to select features that can reflect the heterogeneity of industrial network traffic characteristics, the present study realizes dimensionality reduction of industrial network traffic. The reduced-dimensional data were grouped using a sliding window to construct the threshold of normal traffic characteristics in industrial networks. By rewriting Suricata to realize real-time traffic collection and analysis and inputting the real-time analysis results into the constructed density clustering intrusion detection algorithm based on a sliding window, this study quickly screened absolute normal traffic groups and absolute abnormal traffic groups through comparison using the threshold of normal traffic characteristics in industrial networks. Abnormal traffic was separated by the density clustering algorithm for groups encompassing normal and abnormal traffic, through which abnormal traffic detection was completed. [Results] The above intrusion detection method was applied in the attack and defense range of the entire process of oil and gas gathering and transportation in an industrial scenario, and a large number of experiments were carried out. This method can effectively identify abnormal traffic with a detection rate exceeding 96％ and a false positive rate below 3％. This proves that the proposed method can meet the needs of high-efficiency, reliable, and real-time detection of abnormal traffic in industrial networks. [Conclusion] The innovation of this study lies in providing a new method for detecting abnormal traffic in industrial networks, which combines Suricata and the density clustering algorithm based on a sliding window to establish a full-process intrusion detection model from traffic collection and analysis to real-time intrusion detection. The method has important practical value for the security protection of industrial Internet and provides a new research idea for real-time intrusion detection of industrial networks.

To solve the problem of poor accuracy and low efficiency of vibration damper defect detection due to too many external influencing factors of transmission lines, an intelligent recognition technology based on image representation and pixel coordinates for damper defects was proposed. The details and contour of damper were described by nonlinear mapping, and the change of damping force, stiffness, mass and displacement caused by stress deformation of vibration damper was estimated. The transformation matrix was used to establish the identification coordinate system, and to calculate the rotation angle and dynamic position of key points on each axis of vibration damper. The calculated node inconsistent with the parameter expression of the identified node was confirmed as the defect part. The experimental results show that the as-proposed technology can also achieve accurate defect recognition in complex background sceneries, and the processing effect of noise and distortion is also better, indicating its good robustness and applicability.

In order to improve the image reconstruction quality of electrical capacitance tomography (ECT), an image reconstruction algorithm based on an improved forward problem model and a memetic algorithm, abbreviated as Im-MA algorithm, was proposed. The forward problem model was improved by utilizing smooth filtering and dimension reduction of the sensitivity matrix. An objective function was constructed to transform the image reconstruction problem into an optimization problem by using the L₂ norm as data fitting measurement and the total variation as regular term. The objective function was solved by using the memetic algorithm formulated by the combination of sparrow search algorithm and beetle antenna search algorithm. Simulation and experimental results show that the reconstruction error of the Im-MA algorithm is smaller, compared with the commonly used Landweber algorithm, and its reconstructed image is much closer to the real distribution. The Im-MA algorithm provides an effective new way to solve the ECT inverse problem.

To address the inconvenience of developing specific smart contracts and relational data storage operations for multiple business scenarios in blockchain data forensic systems, a smart contract layered architecture, design pattern and methods for operation on relational database based on smart contracts for multiple business scenarios were proposed. Through reducing the coupling between different smart contracts and the coupling between smart contracts and data tables, the extensibility of smart contract was enhanced. Based on blockchain systems, which use key-value pair databases as the underlying storage, the data forensic operations on relational database tables were effectively supported to realize the data forensic operation of existing governmental businesses. The experimental results show that the feasibility and effectiveness of the as-proposed solution is demonstrated, and the amount of work on developing smart contracts can be significantly reduced.

Pipeline detection technology is one of the most effective methods for internal inspection of long-distance oil and gas pipelines. Pipeline detection technology is an important means for defect detection, evaluation and integrity evaluation of long-distance oil and gas pipelines. With the rapid development of China's energy industry, pipelines become the most effective carrier of energy transportation. Thus, pipeline detection technology faces to new challenges and a variety of new detection methods and solutions have emerged. This paper briefly introduces the development of detection technology in long-distance oil and gas pipelines, expounds the research status of detection technology in long-distance oil and gas pipelines at home and abroad, summarizes the principles and applications of new technologies and methods in the face of the existing problems in long-distance oil and gas pipelines, and finally puts forward the future prospects of the pipeline detection industry.

In order to solve the consensus problem of parabolic multi-agent systems (MASs) under denial of service (DoS) attacks, the leader-follower consensus control protocol based on output feedback method was proposed for MASs to deal with DoS attacks initiated by malicious attackers, in order to achieve leader-follower consistency on undirected communication topology of agents. In addition, the frequency and duration of DoS attacks were analyzed, and the necessary and sufficient conditions for the stability of parabolic partial differential systems were derived by Lyapunov stability theorem. It is confirmed that the time interval between events is not zero, eliminating the occurrence of Zeno behavior, and the effectiveness of theoretical results is verified through simulation examples.

In order to address the problems in target matching and localization by unmanned aerial vehicles (UAVs), such as the difficulties of feature extraction caused by image rotation variations between UAV images and target reference images and small image sizes from the UAV perspective, a UAV target image matching algorithm combining the candidate region detection and the SE-Hardnet feature extraction network was proposed. The Edge Boxes algorithm was used to detect candidate regions, and the SE-Hardnet network was employed for regional feature extraction, for precise target image matching by comparing feature similarities. Experimental results demonstrate that the proposed algorithm exhibits higher matching accuracy and robustness under the condition of image angle and size variations. In close-range scenarios, the matching accuracy within the image dataset surpasses that of current image matching algorithms by 8％ to 11％, providing a feasible and effective solution for UAV target positioning.