In order to ensure the safe and stable operation of the power system after large-scale grid-connection of new energy, it is necessary to quantitatively assess the vulnerability of the power grid before and after grid-connection. Therefore, a composite new energy power grid vulnerability analysis method based on complex networks and electrical characteristics was proposed, and the IEEE-39 node was simulated. Firstly, a vulnerability assessment system was constructed by selecting degree centrality, proximity centrality, eigenvector centrality, electric power flow interval, voltage stability and power balance, and then solving each secondary index according to the formula. Secondly, entropy weight method was used to calculate the index weight, TOPSIS method was used to rank the vulnerability of nodes before and after grid connection, and comparative analysis was made. Finally, the obstacle factor model was used to analyze the influence degree of each index on the vulnerability of power grid. The research results indicate that the number and location of new energy sources connected to the grid affect the vulnerability of the power grid. Among the influencing factors of grid vulnerability, the order of influence is voltage stability indicator > closeness centrality indicator > active power balance indicator.

To explore the feasibility of using AI (artificial intelligence) three-dimensional motion analysis for analyzing the influencing factors on the distance of ski jumping and optimizing athletes’ technical movements, a study was conducted during the 2022 FIS Continental Cup Beijing event. Sixteen athletes’ take-off phase motions were captured within a fixed range using AI-based three-dimensional motion analysis system. This system automatically parsed the videos to obtain biomechanical parameters of the athletes’ take-off phase. By comparing the correlation coefficients and differences between manually processed data and AI-generated data of the three-dimensional coordinates of body joints over time, the validation of the equipment for ski jumping was conducted. The multiple correlation coefficient was found to be greater than 0.91, with an average difference value of less than 1.48 cm, indicating the reliability of the system. Furthermore, a comparison was made between the technical parameters of high-level foreign athletes and domestic athletes. Using t-tests and Pearson correlation coefficient analysis, the relationship between body posture parameters during the take-off phase and sports performance was examined. The results reveal correlations between take-off speed and angles of ankle and knee during the take-off phase, and between the final score and ankle angle during take-off, suggesting that Chinese athletes should focus on achieving full extension during take-off and timing their jumps appropriately to significantly enhance sports performance. Overall, the system demonstrated precise feedback for ski jumping technique analysis. Additionally, it enabled the acquisition of biomechanical parameters from world champion athletes to construct a champion model, providing valuable training references for Chinese athletes.

Aiming at the problem of poor model accuracy caused by poor stability and strong randomness of wind power output. A short-term prediction model of wind power based on quadratic decomposition error compensation was proposed. Firstly, BiLSTM (bidirectional long short-term memory) prediction model is established to predict wind power and output prediction errors. Secondly, an IDBO (improved dung beetle optimizer) algorithm was used to initialize the population by using chaotic mapping, update the position of rolling dung beetles by introducing golden sine strategy, and update the position of thieving dung beetles by adding dynamic adaptive weight coefficient to optimize the parameters of the prediction model. Prevent the network from falling into the local optimal solution, and adaptively search the optimal parameter combination. Then, using the decomposition-reconstruction-decomposition strategy, CEEMDAN (complete ensemble empirical mode decomposition with adaptive noise) was used for the first decomposition. In addition, SE(sample entropy) and K-means are introduced to reconstruct the sequence according to frequency, and the high-frequency error sequence was decomposed into error sequences of different frequency bands by VMD(variational mode decomposition). Improve the prediction efficiency and accuracy of subsequent models. Finally, the input error compensation model of each component was used to predict and the Attention mechanism was introduced to learn the feature relationship of different time steps and give different weight values to enhance the attention to key information. Through the measured data of a wind farm in Xinjiang, the prediction accuracy of the proposed model is proved to be high and has significant advantages.

With China’s Manufacturing 2025 Plan, the military industry to implement the unmanned production line, and AGV (automated guided vehicle) as a fully automated production line of the main logistics carrier, the scheduling of its strengths and weaknesses directly determines the capacity and efficiency of the entire production line. Due to the security requirements of military industrial places, wireless communication and other means cannot be used, and only point-to-point optical communication can be used, which also worsens the real-time communication of AGVs. Based on Plant Simulation software, a simulation system model was established, the real-time data interaction channel between the logistics simulation software and the field controller was opened, the synchronous operation of the simulation system and the reality was realized, and the seamless connection between the logistics simulation software and the field controller was completed, which effectively solves the difficult problem of poor real-time AGV scheduling caused by the lack of wireless power in the military industry. Experiments have proven that this method effectively simplifies the difficulty of writing the scheduling system and improves the overall real-time performance of the system by 0.058 seconds. Compared with the traditional method, the writing time is shortened by 9.7 times, and the debugging time is even shorter by 22 times. This study lays the foundation for the realization of full automation of military production lines and provides technical support for the use of pulsation production lines in hazardous places.

In order to improve the path planning ability and efficiency of AUV (autonomous underwater vehicle), an AUV path planning algorithm based on the community information transmission mechanism was proposed. Firstly, based on the community information transmission mechanism, the global short and long connection operators were designed to achieve the optimal search of the neighborhood of the planned path points and the probabilistic search outside the neighborhood. Then, the local short and long connection operators were designed, which implements the search for four boundary derived points of the path center point and the connections of feasible paths outside the derived points. Finally, the AUV path planning algorithm flow was completed. Six simulation and two seabed map simulation tests show that, compared with other algorithms, the algorithm has the advantages of strong planning ability, high planning efficiency, and smooth planning path.

Establishing an accurate rolling bearing performance degradation prediction model plays a crucial role in subsequent processing such as bearing fault classification and life prediction. In order to solve the problem of inaccurate prediction of bearing performance degradation model, an IBA(improved bat algorithm) was proposed to improve the accuracy of degradation model prediction. Firstly, Cat chaotic mapping was applied to the initial position of the population to enhance the traversability of the population and improve the quality of the initial solution. Secondly, an inverse tangent-like control factor was added in the iterative process to improve the algorithm’s accuracy in finding the optimum. Finally, the position updating strategy was improved to prevent from falling into the local optimum. By comparing the results with those obtained from SVR(support vector regression machine) optimized by BA(bat algorithm), SVR optimized by particle swarm optimization algorithm, and SVR optimized by gray wolf optimization algorithm, the results show that the absolute mean error of the prediction model optimized by the IBA decreases by 70.60%, 67.19%, 55.56%, and the root-mean-square error decreases by 76.64%, 76.12%, and 76.12%, respectively. 76.64%, 76.12%, and 30.29%, respectively, further proving the accuracy of the improved prediction model.

To address the issues of long stitching time due to numerous mismatched feature points and insufficient stitching accuracy when using all feature points directly in image stitching tasks, an optimized image stitching method combining a matching point increasing strategy with RANSAC(random sample consensus) was proposed. The method initially screened feature points to prevent numerous ineffective samples, thus improving computational efficiency. Then, a progressive sampling strategy was employed to incrementally increase matching points and repeatedly sample for precise results. Finally, the optimal model was obtained by utilizing a new loss function based on root mean square error to filter the results. The experimental results indicate that, without a noticeable increase in time consumption, the interior point rate of the algorithm in this paper is further enhanced, the mean and root mean square errors of feature points have decreased significantly, the accuracy of image stitching is improved, the misalignment phenomenon at the stitching seam is effectively improved, and the stitching errors in image stitching tasks are significantly reduced.

Content Aiming at the problems of single convolution model, insufficient Receptive field and inaccurate feedback information of single discriminant network in current face image super-resolution reconstruction algorithm, an algorithm based on adaptive convolution and joint Loss function was designed. A generation adversarial network architecture was used by the model. On the generator side, adaptive convolution was used to construct dual path residual blocks and further form efficient residual groups. It can independently learn feature weights extracted under different receptive fields and supplement missing information from a single branch. The subpixel convolution layers were used to complete quadruple reconstruction of face images. In terms of discriminators, Vgg and U-net architecture networks were used as dual discriminant networks, and dual discriminant results were used to calculate adversarial losses. The losses, content losses, and perceptual losses form a joint loss function. Experiments on the Celeba dataset show that compared with RWSA, this algorithm improves PSNR by 1.166 dB, SSIM by 0.037, LPIPS by 0.033, and PI by 0.119, compared with other mainstream algorithms, it has advantages in image detail clarity.

For the motion control of four-rotor UAV in attitude, the main method is the application of ADRC (active disturbance rejection control) system. For this system to deal with the complex interference with sensor noise, the previous fal function design still has many defects in application. Under the function of traditional fal function, the ESO(extended state observer) has the problems of insufficient observation accuracy and high chattering rate. Therefore, a new nonlinear smooth tfal function was improved on the basis of the previous fal function, and ESO was studied with this function. Finally, other ADRC methods were compared with this method in MATLAB/Simulink software, and the newly designed tfal function shows better convergence. The new ESO based on tfal design has obvious improvement in error estimation and error following performance. At the same time, compared with the improved function galn and the traditional function fal, the tracking capability of the new ESO attitude active disturbance rejection control system is improved by 2.3% and 4% respectively, and the anti-interference performance is improved by 50% and 67% respectively.

The significant advantages of point cloud data are presented in domains such as architectural reverse modeling, 3D reconstruction, and construction progress monitoring. Vast amounts of data are typically involved in the collection of point clouds for architectural structures, with the point clouds of components like beams and columns being particularly crucial. The challenges faced by current semantic segmentation methods for 3D point clouds when processing large-scale data include insufficient extraction of local features and suboptimal recognition accuracy. An enhanced approach for the semantic segmentation of large-scale point clouds of key architectural components using the RandLA-Net deep learning network was proposed. In this regard, the robustness of segmentation results was improved by incorporating a coordinate attention module in the local spatial encoding section. Furthermore, an extended channel attention module has been developed to strengthen the model’s capability in feature discernment, and a focal loss function has been introduced to effectively train the network, while addressing class imbalance issues within architectural point cloud scenes. Consequently, the efficient processing of architectural structure point cloud data and the extraction of key components are enabled. The performance comparisons and analyses conducted through experiments demonstrate that the original RandLA-Net model is outperformed by our model in terms of overall accuracy and component extraction precision in semantic segmentation of large-scale point clouds, thereby confirming the enhanced performance and practical value of the proposed method.