With the development of DC (direct current ) distribution networks and the large-scale integration of distributed energy storage and photovoltaics into the distribution network, the structure of the distribution network has undergone revolutionary changes. After a short circuit fault occurs in the DC distribution network, the short circuit voltage drops sharply, the short circuit current rises rapidly, and the stability of the power system operation is disrupted. To address this issue, a model for inter pole and single pole short circuit faults in DC systems was proposed. Firstly, by sampling voltage data at both ends of the DC line, the voltage equation was written, and the transition resistance was eliminated. Then, a fitness function was constructed, and the adaptive optimization red fox algorithm with faster convergence speed and higher positioning accuracy was used to calculate the distance from the fault point to the protection installation site for fault location in the DC distribution network. Based on the red fox algorithm, combined with the isolation forest algorithm to remove abnormal data, the algorithm performance and accuracy were improved by improving adjustable feedback factors and introducing genetic crossover operators. When the sampling frequency is low, the accuracy of fault localization is improved through adaptive interpolation. Simulation verification was conducted in Simulink, and the results show that the method has strong resistance to transition resistance, small positioning error, is not affected by system parameters, and can effectively reduce the impact of low sampling frequency on fault localization, and has good robustness.

To comprehensively and deeply analyze the overall benefits of provincial-level construction of new type of power load management system, a study on the comprehensive evaluation system for such systems was conducted. Firstly, an evaluation index system that considers technical, social, environmental, and economic benefits was established. Secondly, the subjective weights of the comprehensive benefits were determined through an improved analytic hierarchy process. Given that the index system encompasses both qualitative and quantitative indicators, the objective weights were determined based on the concept of intuitionistic fuzzy numbers, and the comprehensive weights were determined through cooperative game theory. Finally, the concept of perpendicular distance was introduced to enhance the traditional technique for order preference by similarity to ideal solution (TOPSIS) method, thereby improving the accuracy of the evaluation results. By applying the proposed evaluation system to evaluate the construction of new power load management systems in different provinces, the results show that the studied system exhibits strong systematicity, scientificity, accuracy, and feasibility in the comprehensive and multidimensional benefit evaluation of the construction of new power load management systems.

Lightning is the main cause of active distribution line fault. It is of great significance to study the lightning risk assessment of active distribution network. The distribution line with distributed photovoltaic system in Nanjing area was taken as the research object. The calculation model of lightning overvoltage on distribution lines with distributed photovoltaic system was established, and the interaction between photovoltaic side and distribution line side during lightning strike was analyzed. The electrical geometric model on both sides was constructed. The trip rates of the photovoltaic side and the line side were calculated, and the risk was evaluated according to the calculation results. The results show that when lightning strikes the nearest tower on the photovoltaic side, the lightning trip-out rate on the photovoltaic side increases from 27.52 times/(100 km·a) to 29.63 times/(100 km·a), and the lightning risk is higher. When the photovoltaic side is struck by lightning, the tripping rate of the adjacent three towers affected by the lightning intrusion wave is doubled, and the risk level is also higher.

To address the challenges of high-dimensional features, large computational demand, and difficulty in improving the accuracy of secondary return water temperature prediction models for heat stations, a secondary return water temperature prediction model based on the xtreme gradient boosting-artifical neural network(XGBoost-ANN) was proposed. The feature screening layer uses XGBoost algorithm to calculate the importance scores of the original data features and determine the main features that affect the secondary backwater temperature, thus reducing the complexity of the model and improving the computational efficiency. Three layers of feedforward ANN were trained by Bayesian regularization algorithm as the secondary backwater temperature prediction layer, and the initial weights and thresholds of the ANN model were optimized by grey wolf optimizer (GWO) algorithm. The weights and thresholds of the ANN model were represented by grey wolf position vector. The fitness function was introduced to evaluate the performance of each set of weights and thresholds to help the model avoid falling into local optimality at the initial stage of training, so as to improve the performance and generalization ability of the model. Experimental results demonstrate that the constructed XGBoost-GWO-ANN secondary return water temperature prediction model achieved significant improvements. Compared to the model before feature filtering, the root mean squared error(RMSE) is reduced by 26.8%, the R² is increased by 11.3%, and the model inference time is shortened by 46.1%. Furthermore, the optimization of the initial ANN weights and thresholds using the GWO algorithm improve the RMSE by 20.0% and the R² by 3.4% compared to the unoptimized ANN model. These results indicate that the accuracy and generalization ability of the proposed prediction model are effectively enhanced.

Aiming at the unmanned autonomous positioning task of mobile robot during the maintenance of nuclear reactor internals, an underwater geometric positioning method combining inertial measurement unit(IMU) and monocular visual distance measurement was proposed. Combining with the actual environment demand, the principle of distance measurement with monocular camera was described briefly. Aiming at the limitation of the underwater nuclear radiation environment on the use of sensors, a method based on monocular vision correction of inertial measurement unit position estimation integral error and geometric positioning was proposed. Simulation experiments show that the positioning method can effectively correct the accumulated error generated in the positioning process of the inertial measurement unit, and improve the positioning accuracy of the robot in underwater motion.

The natural orifice intervention using continuum robots faces challenges such as tortuous and narrow intervention paths, as well as compressive forces exerted by soft tissues in the orifice. To address the issue in the delivery process where existing planning methods struggle to balance multiple control objectives, resulting in difficulty in reaching deeper positions, an autonomous planning scheme based on residual reinforcement learning was proposed. The method enables the autonomous delivery of continuum robots through natural orifices. A feedback deviation model between the delivery posture of the continuum robot and the spatial state of the natural orifice was established to control the posture target during the delivery process. Simultaneously, a Markov model of the overall motion process of the continuum robot was constructed to train the reinforcement learning algorithm. A residual strategy, generated by combining posture feedback control with reinforcement learning control, was used to output the optimal actions for the continuum robot's delivery process. Experiments conducted in a simulated bronchial orifice show that the proposed method converges over 60% faster than existing methods and can plan smooth, collision-free trajectories for the continuum robot's intervention through the orifice, outperforming existing methods in several key metrics.

In the field of unmanned systems, heterogeneous cross-domain collaboration is recognized as an effective solution to current bottlenecks, integrating functional redundancy and complementary capabilities across different platforms. However, traditional rope-hook recovery systems, due to their high stiffness, result in significant instantaneous impact forces during aircraft hooking, which leads to undesired landing attitudes and an inability to land precisely. Furthermore, these systems are less flexible and have slower response times compared to robotic arms in achieving altitude tracking. To address these challenges, admittance control for robotic arm compliance was introduced, with the goal of enhancing landing stability, preventing structural damage caused by excessive lateral forces, and ensuring stability. In addition, the roll compensation process was optimized through a combination of short-time Fourier transform-fast Fourier transform(STFT-FFT) prediction algorithms and active disturbance rejection control, leading to smoother and more stable compensation responses. Simulation tests confirmed the effectiveness of the admittance control strategy and the optimization of the compensation response algorithm, resulting in improvements in both landing stability and system flexibility. This research presents novel methods for unmanned aircraft recovery and offers valuable insights for the development of future unmanned systems.

To address the cumbersome calibration process of fisheye cameras and its inapplicability to everyday scene images, a novel convolutional neural network(CNN)-based method was proposed that simultaneously calibrates the intrinsic parameters of fisheye lenses and corrects image distortion. The accuracy of fisheye camera calibration and image distortion correction was improved by predicting the displacement of pixel points under different distortion parameters. A coordinate attention module was introduced in the encoding part to enhance the model's accuracy and generalization ability to increase attention to image position information. Additionally, a cross-scale fusion module was designed in the skip connections to enhance image detail features. To address the issues of dataset scarcity and incomplete distortion parameter distribution, a new large-scale dataset labeled with corresponding distortion parameters and images after distortion correction was created. Experimental results show that compared to other fisheye camera calibration methods, this method achieves a reprojection error of 0.312 pixel, indicating the highest calibration accuracy. Additionally, compared to other image distortion correction methods, a peak signal to noise ratio(PSNR) of 38.055 dB and an structural similarity(SSIM) of 0.874 are achieved, indicating the best quality of image distortion correction.

As a large transportation hub, airport terminals have complex structures, and the evacuation efficiency becomes extremely important when an emergency occurs. To improve the evacuation efficiency of the terminal building, an improved A* algorithm has been proposed for selecting the optimal evacuation path based on the actual distribution of personnel. Firstly, the flow of personnel in the terminal building was simulated, and data on the distribution of personnel was obtained. Then the time-varying distribution of personnel in each area was considered in the cost calculation of path selection. Finally, the A* algorithm was improved in terms of traversal methods, network weights, other factors, and congested paths were replanned to avoid congestion. The results indicate that considering the distribution of passengers in the terminal improves the evacuation paths at each node, which leads to shorter evacuation time compared to traditional A* algorithm. It also allows for the avoidance of congested paths. This study can provide theoretical and methodological support for the rapid evacuation of passengers in the terminal under emergencies.

Currently, mainstream gait recognition methods often rely on stacked convolutional layers to gradually expand the receptive field and integrate local features. These methods mostly use shallow networks, which have limitations in extracting global features from gait images and lack attention to temporal cycle feature information. Therefore, a deep neural network algorithm combining Transformer and 3D convolution, named 3D convolutional gait recognition network based on AdaptFormer and Spect-Conv (3D-ASgaitNet)was proposed. Firstly, the initial residual convolution layer converts the binary contour data into a floating-point encoded feature map to provide dense low-level structural features. On this basis, the spectral layer enhances the feature extraction ability through the joint processing of frequency domain and time domain, and uses the pseudo-3D residual convolution module to further extract advanced spatio-temporal features. Finally, AdaptFormer module was integrated to provide flexible feature transformation capability through lightweight down-sampling and up-sampling network structure to adapt to different data distribution and task requirements. 3D-ASgaitNet was carried out on four publicly available indoor datasets (CASIA-B, OU-MVLP) and outdoor datasets (GREW, Gait3D), and achieved recognition accuracy rates of 99.84%, 87.83%, 45.32% and 72.12%, respectively. Experimental results show that the recognition accuracy of the proposed method in CASIA-B and Gait3D data sets is close to the performance of SOTA.