In this paper, a novel virtual negative inductor stabilizing strategy is proposed for the dc microgrid with constant power loads. It is known that in the dc-based power system, the constant power load will generate a virtual negative incremental resistance, which may deteriorate the whole system stability. The situation will be more serious for the dc power system with a large line inductance. In the proposed stabilizing strategy, a virtual negative inductor is built on the source-side converter through the droop control method. The built virtual negative inductor counteracts the large line inductance, thus enhancing the system damping effect. Small-signal models of the studied dc microgrid system under the proposed stabilizing strategy are carefully derived. A root-locus based parameter designing approach is proposed for obtaining the optimal parameter value for the stabilizer. An explicit Nyquist stability criterion for the studied dc microgrid system is proposed, with the system minor loop gain carefully derived. Several comparative stability analyses are taken for showing the system robustness to the parameter perturbations. Detailed numerical simulations are also conducted for validating the effectiveness of the proposed stabilizing strategy.

Path planning is an important problem in autonomous control technology. This paper aims to overcome the shortcomings of the wolf pack algorithm (WPA), such as slow rate of convergence and low convergence precision, by improving the three intelligent behaviors of the WPA, namely, scouting, summoning, and beleaguering. To improve the scouting behavior, interactive scouting is proposed to increase the interactivity among wolf pack. Furthermore, to improve the summoning behavior, a prey-based adaptive step model is established to improve the searching ability. Finally, calculation rules of new beleaguering behavior are designed, which enhance the local fine search ability considerably. A fast path planning method based on dubins path was proposed, which applied the dubins path planning to meet angle control constraint and tunes the turning radius to meet control constraint. The dubins path planning method based on the modified WPA is proposed by establishing the underwater environment threat model under the condition of autonomous underwater vehicle constraint. The path between the path points is the shortest, the threat is minimal, and the energy consumption is the least without the consideration of ocean current. Simulation results show that the modified WPA has a high rate of convergence and good local search capability in the high-precision, high-dimensional, and multi-peak function; moreover, it does not converge prematurely.

To solve the problem in which the auxiliary white-noise parameters need to be artificially selected in a noise-assisted multivariate empirical mode decomposition (NA-MEMD) and considering the fact that obtaining a large number of typical fault samples in practical engineering is difficult, a rolling bearing fault-diagnosis method based on velocity modified-mutation particle swarm optimization (PSO)-optimized NA-MEMD and improved functional neural fuzzy network (FNFN) is proposed. First, the original vibration signal is processed using the velocity modified-mutation PSO–NA-MEMD method to decompose it into a series of intrinsic mode functions (IMFs) with different characteristic time scales. Owing to the distribution characteristics of the IMF with the time scales and because the energy in different states of a bearing in different frequency bands changes, we calculate the energy moment of each IMF as a fault feature vector. Second, the fault feature vectors are then used as input to construct the improved FNFN structure. Finally, the method is validated using the data from the bearing data center of Case Western Reserve University and the vibration signals of the propulsion-motor bearings in the rudder paddle compartment during normal ship navigation and is compared with other neural network. The results show that the method proposed in this paper can quickly and more accurately diagnose rolling-bearing faults using limited training samples.

The visual signal object detection technology of deep learning, as a high-precision perception technology, can be adopted in various image analysis applications, and it has important application prospects in the utilization and protection of marine biological resources. While the marine environment is generally far from cities where the rich computing power in cities cannot be utilized, deploying models on mobile edge devices is an efficient solution. However, because of computing resource limitations on edge devices, the workload of performing deep learning-based computationally intensive object detection on mobile edge devices is often insufficient in meeting high-precision and low-latency requirements. To address the problem of insufficient computing resources, this paper proposes a lightweight process based on a neural structure search and knowledge distillation using deep learning YOLOv8 as the baseline model. Firstly, the neural structure search algorithm was used to compress the YOLOv8 model and reduce its computational complexity. Secondly, a new knowledge distillation architecture was designed, which distills the detection head output layer and NECK feature layer to compensate for the accuracy loss caused by model reduction. When compared to YOLOv8n, the computational complexity of the lightweight model optimized in this study (in terms of floating point operations (FLOPs)) was 7.4 Gflops, which indicated a reduction of 1.3 Gflops. The multiply–accumulate operations (MACs) stood at 2.72 G, thereby illustrating a decrease of 32%; this saw an increase in the AP50, AP75, and mAP by 2.0%, 3.0%, and 1.9%, respectively. Finally, this paper designed an edge computing service architecture, and it deployed the model on the Jetson Xavier NX platform through TensorRT.

The propulsion systems of hybrid electric ship output and load demand have substantial volatility and uncertainty, so a hierarchical collaborative control energy management scheme of the ship propulsion system is proposed in this paper. In a layer of control scheme, the traditional perturbation algorithm is improved. Increasing the oscillation detection mechanism and establishing the dynamic disturbance step length realizes the real-time stability of maximum power point tracking control. In the second-layer control scheme, the power sensitivity factor and voltage and current double closed-loop controller is introduced. By designing a two-layer coordinated control strategy based on the dynamic droop coefficient, the problem of voltage and frequency deviation caused by load switching is solved. In the third-layer control scheme, due to the need of the optimal scheduling function, the multi-objective particle swarm optimization algorithm was improved through three aspects: introducing the mutation factor, improving the speed formula, and re-initializing the strategy. Compared with other algorithms, this algorithm proves its validity in day-ahead optimal scheduling strategy. The superiority of the hierarchical collaborative optimization control schemes proposed was verified, in which power loss was reduced by 39.3%, the overall tracking time was prolonged by 15.4%, and the environmental cost of the diesel generator was reduced by 8.4%. The control strategy solves the problems of the steady-state oscillation stage and deviation from the tracking direction, which can effectively suppress voltage and frequency fluctuations.

Unsupervised 2D image-based 3D model retrieval aims at retrieving images from the gallery of 3D models by the given 2D images. Despite the encouraging progress made in this task, there are still two significant limitations: (1) feature alignment of 2D images and 3D model gallery is still difficult due to the huge gap between the two modalities. (2) The important view information in the 3D model gallery was ignored by the prior arts, which led to inaccurate results. To alleviate these limitations, inspired by the success of vision transformers (ViT) in a great variety of vision tasks, in this paper, we propose an end-to-end 3D model retrieval architecture on top of ViT, termly transformer-based 3D model retrieval network (T3DRN). In addition, to take advantage of the valuable view information of 3D models, we present an attentive module in T3DRN named shared view-guided attentive module (SVAM) to guide the learning of the alignment features. The proposed method is tested on the challenging dataset, MI3DOR-1. The extensive experimental results have proved the superiority of our proposed method to state-of-the-art methods.