In order to improve the efficiency of tugboat operation scheduling and energy utilization,

an improved intelligent load forecasting algorithm integrating dynamic data preprocessing and online learning is proposed. Based on a hybrid LSTM-Adaboost architecture, the algorithm addresses the issues of temporal feature degradation and multimodal data fusion through the integration of differentiated LSTM weak predictors and dynamic weight allocation (error sensitivity penalty mechanism), and designs an online learning trigger mechanism (automatic retraining based on prediction error threshold) to achieve dynamic model updating. Additionally, an environmental data collaborative optimization module, including tidal information, is introduced to enhance the adaptability of load forecasting to port conditions. The algorithm is compared with conventional LSTM-Adaboost to validate its effectiveness.

The results indicate that after iterative optimization, the mean squared error of the improved algorithm is reduced by 40.8% compared to the conventional LSTM-Adaboost algorithm, demonstrating higher prediction accuracy and environmental adaptability.

The research results can provide a reference for tugboat energy optimization, safety management, and intelligent scheduling in ports.

Offshore wind speed observations often suffer from data gaps, limiting the accuracy of wind resource assessment and wind farm operation.

A ratio-based interpolation method for reconstructing missing wind speed data using ERA5 reanalysis and floating LiDAR observations is proposed. Taking 100 m wind speed data from a coastal buoy as a case study, the method is evaluated across annual scale, seasonal variability, wind speed levels, and typical extreme weather events.

Results show that the method effectively captures temporal wind speed trends, with an annual average correlation coefficient of 0.839. However, it tends to underestimate wind speed magnitudes, with errors increasing notably under high wind conditions, especially during convective summer periods and typhoon events. Compared to traditional linear regression methods, the ratio method performs better in maintaining trends and controlling errors, and it demonstrates greater stability and robustness under conditions of severe wind speed fluctuations or extreme weather.

Overall, the ratio method demonstrates good applicability in stable wind environments and is suitable for long-term wind resource evaluation and data reconstruction. Nevertheless, its accuracy under extreme weather remains limited, suggesting the need for integration with high-resolution simulations or multi-source data fusion approaches.

To systematically review the technological evolution of unmanned surface vehicles (USVs) and explore the path of their convergence with intelligent ships, aiming to overcome the performance bottlenecks of individual USVs regarding endurance, computing power, and communication.

It reviews the centennial evolution of USVs, tracing the transition from radio remote control to fully autonomous navigation, and from single-agent operation to swarm collaboration. It provides an in-depth analysis of four core technologies: environmental perception, decision planning, motion control, and communication links. On this basis, the study focuses on the convergence trend between USVs and large intelligent ships, analyzing the "mothership-drone" cross-domain collaborative operational mode and the cloud-based management system driven by digital twins.

It indicates that current USV technology is undergoing an intelligent transition from "perception-avoidance" to "cognition-gaming". Furthermore, the "mothership-drone" collaborative mode, by combining the platform advantages of large ships with the high maneuverability of USVs, effectively resolves the challenges of individual USV operations in complex deep-sea environments and the "last mile" maneuvering difficulties for large intelligent ships entering and leaving ports, thereby achieving complementary advantages.

Collaborative mode represents a mainstream paradigm for future maritime operations. However, continuous breakthroughs are still required in areas such as regulatory adaptability, communication network security, and green energy propulsion. The findings provide theoretical references for constructing a new integrated air-surface-underwater intelligent maritime equipment system.

To optimize the propulsion efficiency of trailing suction hopper dredgers (TSHD) in two typical operating conditions: low-speed operation and self high-speed navigation,

The ducted pitch propeller and the ducted pitch propeller are designed based on the graph method, and the performance difference of the two cases is compared. A multi-objective optimization platform is established, utilizing the Reynolds-Averaged Navier-Stokes (RANS) method and the non-dominated sorting genetic algorithmⅡ (NSGA-Ⅱ) to conduct an optimization study of the fixed-pitch ducted propeller that balances both operating conditions.

The results show that the pitch ratios of the ducted propellers obtained based on the graph method are very close for both operating conditions, allowing a compromise propeller design to achieve good efficiency in both conditions. Furthermore, compared to the ducted fixed-pitch propeller, the ducted controllable-pitch propeller has higher requirements for the disk area ratio, and under dredging conditions, the fixed-pitch propeller exhibits higher efficiency. Through optimization, the two optimal ducted propeller designs obtained show efficiency improvements of 5.65% and 5.59% under dredging condition, and increases of 7.70% and 8.09% under high-speed navigation condition, respectively.

It provides assistance and reference for subsequent research.

To overcome the limitations of the traditional random sample consensus (RANSAC) algorithm in cylindrical segmentation, a novel method is developed for segmenting point clouds of ring-ribbed shells by integrating structural features and statistical methods.

Initially, the model surface area feature is utilized to estimate the proportion of inliers, thereby enhancing the accuracy of initial parameters. Subsequently, principal component and radius constraints are introduced to enhance the accuracy of cylinder identification and reduce the number of iterations. Then, a weight function-based correction method is applied to mitigate outlier interference, thereby improving the accuracy of cylinder fitting. Finally, the DBSCAN algorithm clustered the point clouds of ring-ribs, and an improved RANSAC algorithm identified localized features, thus achieving precise measurement of component dimensions.

Experimental results show that the proposed method effectively addresses the intelligent recognition and dimensions measurement of components in various parts of the ring-ribs, significantly improving the recognition speed and accuracy of cylindrical shell and ring-ribs. The precision, recall, and overall accuracy of cylindrical shell reach 96.9%, 99.5% and 96.4% respectively, with a computational speed increase of approximately 4.6 times. The measurement error for ring-rib component dimensions is within 0.2%.

Compared with traditional methods, the proposed method offers significant advantages in the accuracy and computational efficiency of point cloud segmentation.

To review the current state of research on autonomous navigation decision-making and control technologies for intelligent unmanned surface vehicles, and to clarify the technical bottlenecks and development trends under scenarios of varying complexity,

a systematic investigation is conducted into the development history of key technologies for unmanned surface vehicles both domestically and internationally. It review addresses the differing technical requirements between low-to-medium complexity and high-complexity application scenarios, covering path planning, line-of-sight guidance, autonomous collision avoidance, automatic docking and undocking, multi-agent cooperative control, and autonomous recovery. It evaluates existing technological shortcomings and provides recommendations for future development.

Analysis indicates that autonomous navigation technology for open waters has matured and is gradually being implemented in engineering applications. However, core technologies for complex waters and complex missions still face developmental bottlenecks.

Looking further ahead, we propose establishing a standardized simulation and real-vessel testing evaluation system tailored to real-world scenarios. It will accelerate the rapid iteration and implementation of key technologies, thereby supporting the advancement of autonomous navigation decision-making and control technologies for unmanned surface vehicles in China.

In order to study the potential application of the diesel fuel direct coal liquefaction diesel (DDCL) and polyoxymethylene dimethyl ethers (PODE) mixed fuel in marine diesel engines,

the volume of fluid (VOF) method is adopted to simulate and study the influence of different fuels, nozzle hole conical and the angle between the nozzle hole and the needle valve axis on the cavitation flow in the nozzle.

The results show that the cavitation intensity in the nozzle hole with a larger angle between the needle valve axis is greater, while there is no obvious cavitation in the nozzle hole with an angle less than 60°. The gradually converging nozzle hole can effectively suppress cavitation and has good flowability and low turbulence intensity. With the increase of PODE content, the density of the mixed fuel increases, the cavitation, turbulence intensity and flow loss in the nozzle hole decrease, and the effective flow area increases. The mass flow rate of DDCL is lower than that of petrochemical diesel. After blending the same volume of PODE, the mass flow rate of the former increases by 6.2% and is higher than that of petrochemical diesel.

The blended fuel of PODE-coal direct liquefaction diesel can reduce the internal flow loss of nozzle orifices, improve the flow performance of the orifices, and increase the mass flow rate.

In order to apply the dynamic inclinometers based on low-cost micro electro mechanical systems (MEMS) inertial measurement units to ships conviniently, it is necessary to overcome the significant impact of the ship sway and surge motion in moored condition, which are periodic acceleration changes of several seconds to tens of seconds, on inclinometer measurements.

The algorithm for inclinometers in the situation is studied. Gyroscope measurements are used for attitude quaternion update. Then the horizontal accelerometer measurement values without the effects of roll and pitch are obtained. On this basis, the error analysis and impact analysis of sway and surge is carried out. The horizontal accelerometer measurement values are put through a low-pass filter with zero-phase-delay. Then Kalman filtering is performed using the low frequency component of horizontal accelerometer measurement values as the observation of the Kalman filter, and horizontal attitude errors and angular velocity measurement errors as state variables. Attitude closed-loop correction is conducted to make the MEMS dynamic inclinometer keeping the expected accuracy in a long time when the ship is moored.

An experiment is conducted using a certain type of MEMS dynamic inclinometer to validate the algorithm of reducing the effect of sway and surge. The measurement accuracy reached 0.2°(1σ) with an acceleration amplitude of 0.8g in the experiment,

verifying the effectiveness of the algorithm. Key words: low pass filter; Kalman filter; dynamic inclinometer; sway and surge

To enhance the real-time computational capability of diesel generator set simulation models under dynamic conditions such as sudden load changes, and to address the issues of computational complexity and insufficient dynamic response timeliness in traditional mechanistic models during ship deployment,

a physics-mechanism-inspired multilayer perceptron (MLP) data-driven modeling method is proposed. By constructing a dual-hidden-layer network topology mapped to the electromagnetic-electromechanical transient process of generators, the approach achieves coordinated rapid calculation of the DC bus voltage and current of diesel generator sets.

The model effectively captures the nonlinear dynamic characteristics of diesel generator sets, improving computational efficiency while maintaining the accuracy of mechanistic models.

The research providing rapid-deployable technical support for real-time situational awareness and intelligent management of ship power systems.

Aiming at the problem of unstable bus voltage output caused by Marine condition disturbance switching during the operation of ship direct current (DC) microgrid photovoltaic power generation units,

a photovoltaic power generation control strategy based on double integral sliding mode controller is proposed, including establishing an engineering model of photovoltaic cells and adopting a new exponential approach law and hyperbolic tangent switching function.

Simulation verification shows that the controller shortens the startup time to 0.002 s, reduces the overshoot to 2%, and effectively eliminates the steady-state error. Compared with traditional proportional-integral (PI) control, the startup time is reduced by 67% and the ability to resist load disturbances is enhanced by 41.2%.

The strategy significantly enhances the dynamic response speed and robustness, and is applicable to the dynamic working conditions of ships, addressing the shortcomings of traditional control methods under sudden light changes and load disturbances.