A method and system for predicting the speed of bulk carriers.

Abstract

This invention relates to the field of ship performance prediction technology, and discloses a method and system for predicting the speed of bulk carriers. The method includes: acquiring route planning data of the target bulk carrier and dividing it into different segments; acquiring sea area characteristic data and navigation time-series data for each segment to establish a sea state spatiotemporal sequence dataset; and extracting sea state feature vectors for each segment through multi-dimensional time-series feature extraction; determining the current operating condition based on the target bulk carrier's design parameters and cargo information, and calculating the hull wetted surface area and displacement volume under the current operating condition; analyzing the total resistance and effective propulsion power of each segment by combining the sea state feature vectors; calculating the predicted speed value of each segment based on the total resistance and effective propulsion power; and generating segmented speed prediction curves along the route and the total voyage time by combining the voyage distance of each segment. This invention improves the reliability of long-voyage maritime speed prediction.

Description

Technical Field

[0001] This invention relates to the field of ship performance prediction technology, and more specifically, to a method and system for predicting the speed of bulk carriers. Background Technology

[0002] With the continued growth of global trade and the rapid development of the shipping industry, the operational efficiency and cost control of bulk carriers, as an important carrier of international cargo transportation, are receiving increasing attention. Bulk carrier speed is affected by a variety of complex factors, including marine environmental conditions, the ship's own condition, cargo load, and weather conditions, among which complex nonlinear relationships exist. Traditional speed forecasting methods often rely on empirical formulas or simplified models, which struggle to fully capture the interactions between these complex factors, resulting in limited forecast accuracy. In recent years, with the rapid development of information technology and artificial intelligence, intelligent speed forecasting of bulk carriers using data processing techniques and machine learning methods has become a research hotspot in the shipping field, providing a new technical approach to improve the accuracy and reliability of speed forecasts.

[0003] Currently, speed forecasting for bulk carriers mainly employs methods based on physical models and methods based on statistical learning. Physical model-based methods typically establish mathematical relationships between ship speed and factors such as propulsion power and sea conditions based on ship hydrodynamics principles and resistance theory, obtaining speed forecast results through theoretical calculations. Statistical learning-based methods utilize historical navigation data and employ machine learning algorithms such as regression analysis and neural networks to uncover statistical patterns between speed and various influencing factors, thereby achieving speed prediction. Some studies have also attempted to combine physical models with data-driven methods, integrating expert knowledge and historical data to improve forecast performance.

[0004] However, existing speed forecasting methods still have significant shortcomings when dealing with the actual operating scenarios of bulk carriers. On the one hand, most existing methods are based on static forecasts using specific sea state conditions or average environmental parameters of the route, without considering that ocean routes usually cross multiple sea areas and experience significantly different marine environments (for example, there are fundamental differences in wave characteristics, ocean current direction, and monsoon influence from the Indian Ocean to the Pacific Ocean). Bulk carrier voyages typically last 15 to 45 days, and the dynamic changes in sea state along the way have a continuous impact on speed. Existing static forecasting models cannot reflect the comprehensive effect of this spatiotemporal evolution of sea state on speed in real time, resulting in a significant decrease in forecast accuracy as the voyage lengthens.

[0005] On the other hand, existing technologies mainly focus on building forecasting models for single operating conditions. Due to their high cargo density and large cargo capacity, ocean-going bulk carriers can experience draft differences of over 10 meters between full load and empty load (e.g., a 200,000-ton bulk carrier has a full load draft of approximately 18 meters but an empty draft of only about 8 meters). This extreme difference in operating conditions leads to fundamental changes in parameters such as the ship's wetted surface area, drag components, propeller immersion depth, and propulsion efficiency. Existing models fail to effectively capture the impact of these operating condition transitions on speed, making it difficult to accurately predict the speed performance of bulk carriers under different cargo conditions during round trips. These shortcomings limit the applicability and forecasting accuracy of existing speed forecasting methods in the complex and ever-changing actual operating environment of ocean-going bulk carriers, making it difficult to provide sufficient and reliable decision support.

[0006] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0007] In view of the problems in the related technologies, the present invention proposes a method and system for predicting the speed of bulk carriers in maritime transport, so as to overcome the above-mentioned technical problems existing in the existing related technologies.

[0008] Therefore, the specific technical solution adopted by the present invention is as follows: According to one aspect of the present invention, a method for predicting the speed of a bulk carrier is provided. The method includes: acquiring route planning data of a target bulk carrier and dividing it into different segments; acquiring sea area characteristic data and navigation time series data for each segment to establish a sea state spatiotemporal sequence dataset; and obtaining sea state feature vectors for each segment through multi-dimensional time series feature extraction; determining the current operating condition based on the ship design parameters and cargo information of the target bulk carrier, and calculating the wetted surface area and displacement volume of the hull under the current operating condition; analyzing the total resistance value and effective propulsion power of each segment by combining the sea state feature vectors of each segment; solving for the predicted speed value of each segment based on the total resistance value and effective propulsion power of each segment; and generating segmented speed prediction curves along the route and the total voyage time by combining the voyage distance of each segment.

[0009] Furthermore, the route planning data of the target bulk carrier is acquired and divided into different segments. For each segment, segment sea area feature data and segment navigation time series data are acquired to establish a sea state spatiotemporal sequence dataset. The sea state feature vectors for each segment are obtained through multidimensional time series feature extraction. This includes: acquiring the route planning data of the target bulk carrier, which includes the origin port, destination port, and planned route trajectory; dividing the planned route trajectory into different segments based on its geographical span, with each segment corresponding to a different sea area interval; acquiring segment sea area feature data and segment navigation time series data for each segment; establishing a sea state spatiotemporal sequence dataset for the entire route based on the segment sea area feature data and segment navigation time series data; and performing multidimensional time series feature extraction on the sea state spatiotemporal sequence dataset to obtain the sea state feature vectors for each segment.

[0010] Furthermore, based on the geographical span of the predetermined route trajectory, the predetermined route trajectory is divided into different segments, each segment corresponding to a different sea area interval, including: determining the geographical range traversed by the predetermined route trajectory according to the geographical locations of the origin and destination ports, and obtaining the sea area distribution information within this geographical range; this sea area distribution information includes the boundary coordinates and sea area names of each sea area; spatially matching the predetermined route trajectory with the boundary coordinates of each sea area to identify the various sea areas that the predetermined route trajectory traverses sequentially from the origin port to the destination port, and recording the entry and exit points of the predetermined route trajectory in each sea area; dividing the predetermined route trajectory into multiple segments based on the entry and exit points of the predetermined route trajectory in each sea area, with the starting point of each segment being the entry point into a certain sea area and the ending point being the exit point from that sea area; for each segment, determining the corresponding sea area interval based on the geographical coordinates of the starting and ending points of the segment, and calculating the voyage distance of the segment.

[0011] Furthermore, the sea area characteristic data of the navigation segment includes historical sea state records and real-time sea state forecasts of the sea area where the navigation segment is located. The navigation time series data of the navigation segment includes the expected entry time and expected navigation duration of the segment. Both the historical sea state records and the real-time sea state forecasts include wave height, wave period, ocean current speed, wind speed and wind direction parameters.

[0012] Furthermore, multi-dimensional temporal feature extraction is performed on the sea state spatiotemporal sequence dataset to obtain the sea state feature vector for each segment, including: for each segment, extracting the wave height time series, wave period time series, ocean current velocity time series, wind speed time series, and wind direction time series from the sea state spatiotemporal sequence dataset; performing temporal dynamic feature extraction on the wave height time series, ocean current velocity time series, and wind speed time series respectively to obtain the mean feature and wave intensity feature of each time series; extracting the periodic feature of the wave period time series to obtain the dominant wave period feature; extracting the directional feature of the wind direction time series to obtain the mean wind direction angle feature; and concatenating the mean feature and wave intensity feature of each time series, as well as the dominant wave period feature and mean wind direction angle feature, to generate the sea state feature vector for that segment.

[0013] Furthermore, the temporal dynamic feature extraction includes: based on each time series, dividing it into several time windows according to the expected flight duration of the current segment, with each time window corresponding to a fixed time interval; for each time window, calculating the local mean of the data points within that time window, and then taking a weighted average of the local means of all time windows to obtain the mean feature of the time series; the weight of the weighted average is determined based on its proximity to the expected entry time of the current segment, with time windows closer to the expected entry time having a larger weight; for each time window, calculating the standard deviation of the data points within that time window relative to the local mean, and then taking the average of the standard deviations of all time windows to serve as the fluctuation intensity feature of the time series.

[0014] Furthermore, the periodicity feature extraction includes: performing frequency domain analysis on the wave period time series to extract the wave period value with the highest frequency of occurrence as the dominant wave period feature of the flight segment; the directional feature extraction includes: converting each wind direction angle in the wind direction time series into a unit circle vector, calculating the average vector of all unit circle vectors, and taking the azimuth angle of the average vector as the average wind direction angle feature.

[0015] Furthermore, the ship design parameters include design draft, lightship displacement volume, hull line parameters, main engine rated power, and shafting efficiency; cargo information includes cargo type, cargo volume, and cargo density; based on the target bulk carrier's ship design parameters and cargo information, the current operating condition is determined, and the hull wetted surface area and displacement volume under the current operating condition are calculated, including: calculating the cargo volume based on the cargo volume and cargo density; adding the cargo volume to the lightship displacement volume to obtain the total displacement volume; and obtaining the actual draft through hydrostatic calculation based on the total displacement volume and hull line parameters. The hull form parameters include the deck area curve and the cross section area curve; the ratio of the actual draft to the design draft is calculated; if the ratio is greater than the full load threshold, the current operating condition is determined to be a full load condition; if the ratio is less than the empty load threshold, the current operating condition is determined to be an empty load condition; if the ratio is between the full load threshold and the empty load threshold, the current operating condition is determined to be a partially loaded condition; based on the actual draft, the wetted surface area of ​​the hull below the current waterline is calculated by integration according to the deck area curve, and the displacement volume corresponding to the current draft is calculated by integration according to the cross section area curve.

[0016] Furthermore, by combining the sea state characteristic vectors of each segment, the total resistance value and effective propulsion power of each segment are analyzed, including: for each segment, extracting the average wave height, average current velocity, and average wind speed characteristics from the sea state characteristic vector of that segment; calculating the frictional resistance of that segment based on the wetted surface area of ​​the hull under the current operating conditions; calculating the wave-making resistance of that segment based on the displacement volume and average wave height characteristics under the current operating conditions; and calculating the additional resistance from the current and the additional resistance from wind pressure based on the average current velocity and average wind speed characteristics, respectively. Frictional drag, wave-making drag, ocean current drag, and wind pressure drag are superimposed to obtain the total drag value of this segment under the current operating conditions. Based on the current operating conditions, the propulsion efficiency of this segment is determined: if it is a fully loaded condition, the full-load propulsion efficiency is used; if it is an empty condition, the empty propulsion efficiency is used; if it is a partially loaded condition, the full-load propulsion efficiency and the empty propulsion efficiency are linearly interpolated according to the ratio of the actual draft to the design draft. Combining the main engine rated power, shaft system efficiency, and the propulsion efficiency of this segment, the effective propulsion power of this segment is calculated.

[0017] According to another aspect of the present invention, a speed prediction system for bulk carriers is also provided. The system includes: a route and sea state analysis module, used to acquire route planning data of a target bulk carrier and divide it into different segments; for each segment, it acquires segment sea area characteristic data and segment navigation time series data to establish a sea state spatiotemporal sequence dataset; and extracts sea state feature vectors for each segment through multi-dimensional time series feature extraction; a working condition resistance calculation module, used to determine the current working condition based on the target bulk carrier's ship design parameters and cargo information, and calculate the hull wetted surface area and displacement volume under the current working condition; and analyze the total resistance value and effective propulsion power of each segment by combining the sea state feature vectors of each segment; and a speed prediction generation module, used to solve for the predicted speed value of each segment based on the total resistance value and effective propulsion power of each segment, and generate segmented speed prediction curves along the route and the total voyage time by combining the voyage distance of each segment.

[0018] The beneficial effects of this invention are as follows: (1) This invention divides the shipping route into segments of different sea areas according to geographical span, establishes a sea state spatiotemporal sequence dataset containing historical sea state records and real-time sea state forecasts for each segment, and extracts multi-dimensional temporal features of parameters such as wave height, ocean current speed, and wind speed, so that the speed forecasting system can fully reflect the spatiotemporal dynamic evolution of sea state when the ocean route crosses multiple sea areas, effectively solves the problem that the existing static forecasting method cannot capture the comprehensive effect of continuous changes in sea state along the route on speed, and significantly improves the reliability of long-voyage maritime speed forecasting.

[0019] (2) Based on the cargo capacity, cargo density and hull line parameters of the bulk carrier, the present invention calculates the actual draft and determines the current operating condition type through hydrostatic calculation. Then, based on the actual draft, it dynamically calculates the wet surface area and displacement volume of the hull under the current operating condition, and determines the corresponding propulsion efficiency in combination with the operating condition type. This makes the calculation of frictional resistance, wave-making resistance and effective propulsion power able to accurately reflect the fundamental changes in hull parameters and propulsion performance of the bulk carrier under different cargo conditions, effectively solving the problem that the existing single operating condition model is difficult to adapt to the extreme operating condition differences in the round trip of the bulk carrier.

[0020] (3) This invention integrates traditional ship hydrodynamics theory with machine learning methods. First, it uses a physical model to calculate the total resistance and effective propulsion power of each segment under the current operating conditions. Then, it inputs these physical calculation results along with sea state feature vectors into a machine learning model trained based on historical navigation samples to solve for the predicted speed value. This combination ensures that the prediction process conforms to the basic principles of ship engineering and makes full use of the ability of data-driven methods to mine complex nonlinear relationships. Finally, it generates segmented speed prediction curves along the route and the total voyage time, providing decision support for bulk carrier route planning and operation scheduling. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a method for predicting the speed of a bulk carrier according to an embodiment of the present invention. Figure 2 This is a specific implementation diagram of dividing a predetermined route trajectory into different segments in a method for predicting the speed of a bulk carrier according to an embodiment of the present invention; Figure 3 This is a detailed implementation diagram of obtaining the sea state characteristic vectors for each segment in a method for predicting the speed of a bulk carrier according to an embodiment of the present invention. Figure 4 This is a schematic diagram of a speed prediction system for bulk carriers according to an embodiment of the present invention.

[0023] In the picture: 1. Route and sea state analysis module; 2. Operating condition resistance calculation module; 3. Speed ​​forecast generation module. Detailed Implementation

[0024] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0025] According to an embodiment of the present invention, a method and system for predicting the speed of bulk carriers are provided.

[0026] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a method for predicting the speed of a bulk carrier is provided, the method comprising: S1. Obtain the route planning data of the target bulk carrier and divide it into different segments. For each segment, obtain the segment sea area feature data and segment navigation time series data to establish a sea state spatiotemporal sequence dataset. Then, obtain the sea state feature vector of each segment through multi-dimensional time series feature extraction. S2. Based on the target bulk carrier's design parameters and cargo information, determine the current operating condition and calculate the wetted surface area and displacement volume of the hull under the current operating condition; combine the sea state characteristic vectors of each segment to analyze the total resistance value and effective propulsion power of each segment. S3. Based on the total resistance and effective propulsion power of each segment, calculate the predicted speed value of that segment, and combine it with the distance of each segment to generate segmented speed prediction curves along the route and the total voyage time.

[0027] In one embodiment, the route planning data of the target bulk carrier is acquired and divided into different segments. For each segment, segment sea area feature data and segment navigation time series data are acquired to establish a sea state spatiotemporal sequence dataset. The sea state feature vector of each segment is obtained through multi-dimensional time series feature extraction, including: S11. Obtain the route planning data of the target bulk carrier, including the port of origin, the port of destination, and the planned route trajectory; based on the geographical span of the planned route trajectory, divide the planned route trajectory into different segments, each segment corresponding to a different sea area. S12. For each segment, acquire the segment's sea area characteristic data and segment navigation time sequence data; S13. Based on the sea area feature data and navigation time series data of the route segment, establish a sea state spatiotemporal sequence dataset for the entire route; extract multi-dimensional time series features from the sea state spatiotemporal sequence dataset to obtain the sea state feature vector for each route segment.

[0028] In one embodiment, such as Figure 2 As shown, based on the geographical span of the predetermined route trajectory, the route trajectory is divided into different segments, each segment corresponding to a different sea area, including: S111. Based on the geographical locations of the originating and destination ports, determine the geographical range traversed by the planned route and obtain the sea area distribution information within that geographical range; the sea area distribution information includes the boundary coordinates and sea area names of each sea area. S112. Spatial match the predetermined route trajectory with the boundary coordinates of each sea area, identify the sea areas that the predetermined route trajectory passes through sequentially from the starting port to the destination port, and record the position points where the predetermined route trajectory enters and leaves each sea area. S113. Based on the positions of entering and leaving each sea area according to the predetermined route trajectory, the predetermined route trajectory is divided into multiple segments. The starting point of each segment is the position of entering a certain sea area, and the ending point is the position of leaving the sea area. S114. For each segment, based on the geographical coordinates of the starting and ending points of the segment, determine the corresponding sea area range for the segment and calculate the voyage distance for the segment.

[0029] It should be noted that the route planning data is obtained through the target bulk carrier's ship management system or electronic chart system, and the predetermined route trajectory is stored in the form of a series of latitude and longitude coordinate points; the sea area distribution information is obtained from the sea area delineation database published by the International Hydrographic Organization or the marine geographic information service platform. This database contains the polygon boundary coordinates and standard sea area names of major sea areas around the world (such as the Mediterranean Sea, Arabian Sea, Bay of Bengal, South China Sea, Pacific Ocean, etc.); in this invention, spatial matching is achieved through the ray method or the point-on-polygon determination algorithm. Specifically, the method is to traverse each coordinate point of the predetermined route trajectory and determine whether the point is located within the polygon boundary of a certain sea area. When consecutive coordinate points enter the boundary from outside the boundary of a sea area, the entry position point is recorded, and when they leave the boundary from inside the boundary to outside the boundary, the departure position point is recorded; the voyage distance is obtained by summing the great circle distances between adjacent coordinate points within the voyage segment. The great circle distance is calculated based on the Havelsein formula according to the latitude and longitude of the two points.

[0030] In one embodiment, the sea area feature data of the navigation segment includes historical sea state records and real-time sea state forecasts of the sea area where the navigation segment is located, and the navigation time series data of the navigation segment includes the expected entry time and expected navigation duration of the navigation segment. Both the historical sea state records and the real-time sea state forecasts include wave height, wave period, ocean current speed, wind speed, and wind direction parameters.

[0031] Specifically, historical sea state records are obtained from global ocean data assimilation systems (such as the ECMWF's ERA5 reanalysis dataset) or commercial ocean data service platforms (such as Copernicus Marine Service). These databases provide hourly or 6-hourly sea state observation data for the sea area where the voyage segment is located in the recent period (e.g., the past 30 to 90 days). Real-time sea state forecasts are obtained from national meteorological and oceanographic forecasting centers (such as the US NOAA's WAVEWATCH III model and the European ECMWF's wave forecasting system) or commercial meteorological service providers. The forecast data covers sea state change trends for the next 3 to 10 days. The estimated entry time is determined based on the target bulk carrier's voyage plan and current sailing progress. In this invention, if the vessel has already departed, the entry time is estimated based on the current position, the speed already traveled, and the remaining distance from the start of the voyage segment. If the vessel has not yet departed, the estimated entry time is calculated by summing the planned departure time and the estimated sailing time of the preceding voyage segment. The estimated sailing time is initially estimated by dividing the voyage distance of the segment by the target bulk carrier's design speed or historical average speed. This initial estimate is used for the time window division of subsequent sea state data.

[0032] In one embodiment, such as Figure 3 As shown, multi-dimensional temporal feature extraction is performed on the sea state spatiotemporal sequence dataset to obtain the sea state feature vectors for each leg of the journey, including: S131. For each segment, extract the wave height time series, wave period time series, ocean current speed time series, wind speed time series, and wind direction time series from the sea state spatiotemporal sequence dataset. S132. Perform time-series dynamic feature extraction on the wave height time series, ocean current velocity time series, and wind speed time series respectively to obtain the mean features and wave intensity features of each time series; The mean characteristics of each time series include: mean wave height, mean ocean current velocity, and mean wind speed; the fluctuation intensity characteristics of each time series include: wave height fluctuation intensity, ocean current velocity fluctuation intensity, and wind speed fluctuation intensity.

[0033] S133. Extract periodic features from the wave period time series to obtain the dominant wave period features; extract directional features from the wind direction time series to obtain the average wind direction angle features. S134. The mean and fluctuation intensity features of each time series, as well as the dominant wave period and mean wind direction features, are spliced ​​together to generate the sea state feature vector for this section.

[0034] In one embodiment, temporal dynamic feature extraction includes: Based on each time series, the current flight segment is divided into several time windows according to the estimated flight time, and each time window corresponds to a fixed time interval; For each time window, the local mean of the data points within that time window is calculated, and the local means of all time windows are weighted and averaged to obtain the mean characteristics of the time series. The weight of the weighted average is determined based on the proximity of the time window to the expected entry time of the current flight segment, with a larger weight for time windows closer to the expected entry time. For each time window, the standard deviation of the data points within that time window relative to the local mean is calculated, and the average of the standard deviations of all time windows is taken as the characteristic of the volatility intensity of the time series.

[0035] Specifically, in the above embodiments, the fixed time interval is set to 6 hours or 12 hours, so that voyages with a longer expected sailing time (such as transoceanic voyages that may last more than 10 days) are divided into a sufficient number of time windows (such as 20 to 40) to capture fine-grained changes in sea state, while voyages with a shorter expected sailing time (such as near-shore voyages that may only last 2 to 3 days) are also divided into at least 8 to 12 time windows to ensure the effectiveness of feature extraction; the weights of the weighted average are calculated using an exponential decay function, the first... i The weight expression for each time window is: ; In the formula, t i For the first iThe center moment of a time window t 0 represents the estimated entry time for the current flight segment. λ The attenuation coefficient (valued from 0.01 to 0.05 in this embodiment) is used. This weighting method ensures that the weight of time windows closer to the expected entry time increases exponentially, making the mean characteristic more reflective of the sea state at the actual navigation time of the ship; the standard deviation is calculated using the Bessel correction formula. ; In the formula, x j For the first i The first time window j Data points, μ i This is the local mean for that time window. n The number of data points within the time window; the average standard deviation of each time window is used as the volatility characteristic of the time series.

[0036] In one embodiment, periodic feature extraction includes: Frequency domain analysis was performed on the wave period time series to extract the wave period value with the highest frequency to serve as the dominant wave period feature for this flight segment. Directional feature extraction includes: Each wind direction angle in the wind direction time series is converted into a unit circle vector. The average vector of all unit circle vectors is calculated, and the azimuth of the average vector is taken as the feature of the average wind direction angle.

[0037] Specifically, frequency domain analysis is achieved by performing a Fast Fourier Transform (FFT) on the wave period time series. The resulting spectrum is used to identify the frequencies corresponding to energy peaks. The reciprocal of these frequencies is the wave period value with the highest frequency of occurrence. This value represents the dominant wave period characteristics of the sea area and reflects the concentrated distribution of wave energy. The unit circle vector conversion method involves converting the wind direction angle θ (with true north as 0 degrees and clockwise as positive) into a unit vector (cos(θ), sin(θ)). The average vector is obtained by summing the unit circle vectors at all times. The azimuth angle of this average vector is calculated using the two-parameter arctangent function, i.e., the average wind direction angle φ = atan2. This invention avoids the error caused by directly averaging wind angles near 0 degrees (such as 350 degrees and 10 degrees) through this calculation method, so as to correctly reflect the dominant wind direction characteristics of the flight segment.

[0038] In one embodiment, the ship design parameters include design draft, empty displacement volume, hull line parameters, main engine rated power, and shafting efficiency; cargo information includes cargo type, cargo volume, and cargo density. Based on the target bulk carrier's design parameters and cargo information, determine the current operating condition and calculate the hull's wetted surface area and displacement volume under this condition, including: Calculate the cargo volume based on the cargo capacity and cargo density; add the cargo volume to the empty ship's displacement volume to obtain the total displacement volume. The actual draft is obtained through hydrostatic calculation based on the total displacement volume and hull line parameters; the hull line parameters include the platform area curve and the cross section area curve. Calculate the ratio of the actual draft to the design draft; if the ratio is greater than the full load threshold, the current operating condition is determined to be a full load condition; if the ratio is less than the empty load threshold, the current operating condition is determined to be an empty load condition; if the ratio is between the full load threshold and the empty load threshold, the current operating condition is determined to be a partially loaded condition. Based on the actual draft, the wetted surface area of ​​the hull below the current waterline is calculated by integral based on the surface area curve, and the displacement volume corresponding to the current draft is calculated by integral based on the cross-sectional area curve.

[0039] Specifically, ship design parameters are obtained from the target bulk carrier's technical manual or classification society certification documents. The design draft is the draft value under full-load design conditions, and the lightship displacement volume is obtained by dividing the ship's displacement under light conditions by the seawater density (taken as 1.025 tons / cubic meter). Cargo information is obtained through the ship cargo management system or bill of lading records. Cargo types include bulk cargoes such as iron ore, coal, and grain. Cargo density is obtained from the standard bulk cargo density table according to the cargo type. Cargo volume is calculated by dividing the cargo load by the cargo density. The hull form parameters include a surface area curve describing the distribution of the cross-sectional area of ​​the hull at different longitudinal positions from bow to stern, and a cross-sectional area curve describing the relationship between the cross-sectional area and draft at different draft depths. These two curves are extracted from the ship's form drawings and stored in the form of numerical tables or fitting functions. Hydrostatic calculations are achieved through iterative solutions. In this invention, an initial assumption is made. The draft depth is calculated by integrating the cross-sectional area curve to determine the drainage volume at that draft depth. The calculated drainage volume is compared with the total drainage volume. If the difference is greater than the convergence threshold (0.1 cubic meters in the above embodiment), the draft depth is adjusted and recalculated. This process is repeated until the difference is less than the convergence threshold, at which point the actual draft depth is obtained. The full-load threshold is set to 0.90 to 0.95, and the empty-load threshold is set to 0.40 to 0.50. These thresholds are preset based on the actual operating experience of the target bulk carrier. The integration of the wetted surface area of ​​the hull adopts the trapezoidal rule or Simpson's rule. The wetted surface area curve is numerically integrated along the length of the ship for the portion below the waterline. The integration interval is from the bow to the stern, and the integration step size is 1% to 5% of the ship's length to ensure calculation accuracy. The integration of the drainage volume also adopts the trapezoidal rule or Simpson's rule. The cross-sectional area curve is numerically integrated along the draft direction from the baseline to the actual draft depth.

[0040] In one embodiment, the analysis of the total resistance and effective propulsion power of each segment, based on the sea state characteristic vectors of each segment, includes: For each segment, mean wave height, mean ocean current velocity, and mean wind speed are extracted from the sea state feature vector of that segment. Calculate the frictional resistance of this section of the voyage based on the wetted surface area of ​​the hull under the current operating conditions; calculate the wave-making resistance of this section based on the displacement volume and average wave height characteristics under the current operating conditions; calculate the additional resistance from ocean currents and the additional resistance from wind pressure based on the average characteristics of ocean current velocity and wind speed, respectively. The total resistance value of this section under the current operating conditions is obtained by superimposing the frictional resistance, wave-making resistance, ocean current-induced resistance, and wind pressure-induced resistance. Based on the current operating conditions, determine the propulsion efficiency for this segment: if it is a fully loaded condition, take the full-load propulsion efficiency; if it is an empty condition, take the empty propulsion efficiency; if it is a partially loaded condition, perform linear interpolation on the full-load propulsion efficiency and the empty propulsion efficiency based on the ratio of the actual draft to the design draft. Calculate the effective propulsion power for this leg of the journey by combining the rated power of the main engine, shaft efficiency, and propulsion efficiency of the leg.

[0041] It should be noted that the frictional resistance in this invention R f According to the ITTC-1957 frictional resistance formula, the expression is: ; In the formula, ρ The density of seawater is 1025 kg / m³. V The initial value of the ship's speed is (using the design speed or historical average speed of the target bulk carrier as the initial estimate); S is the wetted surface area of ​​the hull under the current operating conditions. C f The coefficient of frictional resistance. C f =0.075 / (log 10 (Re)-2) 2 Re is the Reynolds number, Re = V × L / ν, where L is the ship's length and ν is the kinematic viscosity of seawater (taken as 1.19 × 10⁻⁶). -6 (square meters per second).

[0042] Specifically, the wave resistance in this invention R w Based on Havelock's wave-making resistance theory or regression formula, the expression is: ; In the formula, g The acceleration due to gravity is taken as 9.81 m / s². This represents the drainage volume under current operating conditions. H s This represents the average wave height characteristic of this flight segment. C 1 and C 2 is a coefficient related to ship type ( C Take 0.01 to 0.05, C 2 (take values ​​from 0.1 to 0.3), these coefficients were all obtained from ship model test data.

[0043] Specifically, the ocean current-induced resistance in this invention R c Based on the relative velocity of ocean currents, the expression is: ; In the formula,V c This represents the average characteristic of ocean current velocity. α The angle between the direction of the ocean current and the ship's course (calculated from the average wind direction angle characteristic in the sea state characteristic vector or assumed to be the most unfavorable angle of 0 degrees, i.e., headwind). A m This is the underwater cross-sectional area of ​​the hull (approximately taken as the displacement volume under current operating conditions divided by the ship's length). C dc The current drag coefficient is taken as 0.5 to 0.8.

[0044] Specifically, the wind pressure additional drag in this invention R a Based on wind speed and the projected area on the water, the expression is: ; In the formula, ρ a The density of air is taken as 1.225 kg / m³. V w The characteristic of wind speed is the mean. A a The projected area of ​​the ship on the water (estimated based on the ship's length, width, and superstructure dimensions). C da The drag coefficient is denoted as ηp (0.6 to 1.0, with the smaller value for full load and the larger value for empty load). Full load and empty load propulsion efficiencies are obtained from the ship's sea trial report or propeller design data. The full load propulsion efficiency is 0.6 to 0.7, and the empty load propulsion efficiency is 0.5 to 0.6. The propulsion efficiency under partial cargo loading conditions is ηp = ηp 空载 +(ηp 满载 -ηp 空载 )×(T 实际 / T 设计 -T 空载 / T 设计 ) / (1-T 空载 / T 设计 In the formula, T 实际 For the actual draft, T 设计 For design draft, T 空载 The draft is the unloaded draft (obtained from the ship's design parameters); the effective propulsion power P e =P m ×η s ×η p In the formula, P m η is the rated power of the host. s For shaft system efficiency (0.95 to 0.98 in this embodiment), η p This refers to the propulsion efficiency of this flight segment.

[0045] In one embodiment, the predicted speed value for each segment is determined based on the total drag and effective propulsion power of that segment, including: Machine learning models are trained based on historical navigation samples; Among them, the historical navigation samples include the hull wet surface area, displacement volume, total resistance value of the segment, effective propulsion power of the segment, sea state characteristic vector of the segment and corresponding actual speed value recorded in different voyages under the current operating conditions. The machine learning model employs at least one of gradient boosting decision trees, random forests, or neural networks. The total resistance value, effective propulsion power, and sea state feature vector of the voyage segment are input into the machine learning model, which outputs the predicted speed value of the voyage segment.

[0046] Specifically, historical voyage samples are obtained through the historical trajectory data of the Automatic Identification System (AIS) of the target bulk carrier or sister ships of the same type, the ship's navigation log and engine room log. Each historical voyage sample corresponds to a completed voyage segment, and the hull wetted surface area, displacement volume, total resistance value and effective propulsion power of the voyage segment are calculated back according to the calculation method in S2 of this invention based on the historical cargo information and sea state data of the voyage segment. The sea state feature vector of the voyage segment is generated according to the feature extraction method in S13 of this invention based on the historical sea state records of the corresponding time period of the voyage segment. The actual speed value of the voyage segment is calculated through the timestamps of the start and end points of the voyage segment and the voyage distance recorded by the AIS.

[0047] Specifically, in this embodiment, the Gradient Boosting Decision Tree (XGBoost) model is selected as the machine learning model. This model is trained using supervised learning. The input feature vector includes 12 feature dimensions: hull wetted surface area, displacement volume, total resistance, effective propulsion power, and sea state feature vector (mean wave height, wave height fluctuation intensity, mean current velocity, current velocity fluctuation intensity, mean wind speed, wind speed fluctuation intensity, dominant wave period, and mean wind direction angle). The output target is the actual speed value. The XGBoost model training uses 80% of the samples as the training set and 20% as the validation set. During training, the root mean square error (RMSE) is used as the loss function. After training, the model is saved as a serialized file. In the forecast stage, the total resistance, effective propulsion power, and sea state feature vector of the current segment are assembled into an input vector according to the feature order during training. The trained XGBoost model is then called for inference. The value output by the model is the forecast speed value of the segment, and the unit of the forecast speed value is knots or meters per second.

[0048] like Figure 4 As shown, according to another embodiment of the present invention, a speed prediction system for bulk carriers is also provided, the system comprising: The route and sea condition analysis module 1 is used to acquire the route planning data of the target bulk carrier and divide it into different segments. For each segment, it acquires the segment sea area feature data and the segment navigation time series data to establish a sea condition spatiotemporal sequence dataset. The sea condition feature vector of each segment is obtained through multi-dimensional time series feature extraction. The operating condition resistance calculation module 2 is used to determine the current operating condition based on the ship design parameters and cargo information of the target bulk carrier, and to calculate the wetted surface area and displacement volume of the hull under the current operating condition; combined with the sea state characteristic vector of each segment, it analyzes the total resistance value and effective propulsion power of each segment. The speed forecast generation module 3 is used to calculate the forecast speed value of each segment based on the total resistance value and effective propulsion power of each segment, and generate segmented speed forecast curves along the route and the total voyage time by combining the distance of each segment.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the speed of a bulk carrier, characterized in that, include: Obtain the route planning data of the target bulk carrier and divide it into different segments. For each segment, obtain the sea area feature data and the navigation time series data of the segment to establish a sea state spatiotemporal sequence dataset. For each leg of the journey, wave height time series, wave period time series, ocean current speed time series, wind speed time series, and wind direction time series are extracted from the sea state spatiotemporal sequence dataset. Temporal dynamic features were extracted from the wave height time series, ocean current velocity time series, and wind speed time series to obtain the mean features and wave intensity features of each time series. The temporal dynamic feature extraction includes: dividing each time series into several time windows according to the expected flight duration of the current segment, with each time window corresponding to a fixed time interval; calculating the local mean of the data points within each time window, and then taking a weighted average of the local means of all time windows to obtain the mean feature of the time series; the weight of the weighted average is determined based on the proximity to the expected entry time of the current segment, with time windows closer to the expected entry time having a larger weight; and calculating the standard deviation of the data points within each time window relative to the local mean, and then taking the average of the standard deviations of all time windows to serve as the fluctuation intensity feature of the time series. Periodic features are extracted from the wave period time series to obtain the dominant wave period features; directional features are extracted from the wind direction time series to obtain the average wind direction angle features. The mean and fluctuation intensity features of each time series, as well as the dominant wave period and mean wind angle features, are spliced ​​together to generate the sea state feature vector for each segment. Based on the target bulk carrier's design parameters and cargo information, determine the current operating conditions and calculate the wetted surface area and displacement volume of the hull under the current operating conditions. The process of determining the current operating condition based on the target bulk carrier's design parameters and cargo information, and calculating the hull wet surface area and displacement volume under the current operating condition, includes: calculating the cargo volume based on the cargo load and cargo density; adding the cargo volume to the empty ship displacement volume to obtain the total displacement volume; calculating the actual draft using hydrostatic calculations based on the total displacement volume and hull line parameters; the hull line parameters include the deck area curve and the cross section area curve; calculating the ratio of the actual draft to the design draft; if the ratio is greater than the full load threshold, the current operating condition is determined to be a full load condition; if the ratio is less than the empty load threshold, the current operating condition is determined to be an empty load condition; if the ratio is between the full load threshold and the empty load threshold, the current operating condition is determined to be a partially loaded condition; and calculating the hull wet surface area below the current waterline based on the actual draft and the deck area curve, and calculating the displacement volume corresponding to the current draft based on the cross section area curve. By combining the sea state characteristic vectors of each segment, the total resistance and effective propulsion power of each segment are analyzed; The analysis of total resistance and effective propulsion power for each segment, based on the sea state characteristic vectors, includes: extracting mean wave height, mean current velocity, and mean wind speed features from the sea state characteristic vector for each segment; calculating the frictional resistance of the segment using the wetted surface area of ​​the hull under current operating conditions; calculating the wave-making resistance of the segment based on the displacement volume and mean wave height features under current operating conditions; and calculating the additional resistance from current and wind pressure based on the mean current velocity and mean wind speed features, respectively. Frictional drag, wave-making drag, ocean current drag, and wind pressure drag are superimposed to obtain the total drag value of this segment under the current operating conditions. Based on the current operating conditions, the propulsion efficiency of this segment is determined: if it is a fully loaded condition, the full-load propulsion efficiency is used; if it is an empty condition, the empty propulsion efficiency is used; if it is a partially loaded condition, the full-load propulsion efficiency and the empty propulsion efficiency are linearly interpolated according to the ratio of the actual draft to the design draft. Combining the main engine rated power, shaft efficiency, and the propulsion efficiency of this segment, the effective propulsion power of this segment is calculated. Based on the total resistance and effective propulsion power of each segment, the predicted speed value of that segment is calculated. Combined with the distance of each segment, segmented speed prediction curves along the route and the total voyage time are generated.

2. The method for predicting the speed of a bulk carrier according to claim 1, characterized in that, The process of acquiring route planning data for the target bulk carrier and dividing it into different segments, and then acquiring segment sea area feature data and segment navigation time series data for each segment to establish a sea state spatiotemporal sequence dataset includes: Obtain the route planning data of the target bulk carrier, which includes the port of origin, the port of destination, and the planned route trajectory; Based on the geographical span of the predetermined route trajectory, the predetermined route trajectory is divided into different segments, and each segment corresponds to a different sea area. For each segment, acquire the segment's sea area characteristics data and the segment's navigation time sequence data; Based on the sea area characteristic data and the navigation time sequence data of the route segment, a sea state spatiotemporal sequence dataset for the entire route is established.

3. The method for predicting the speed of a bulk carrier according to claim 2, characterized in that, The geographical span based on the predetermined route trajectory divides the predetermined route trajectory into different segments, each segment corresponding to a different sea area interval, including: Based on the geographical locations of the origin and destination ports, the geographical range traversed by the predetermined route is determined, and the sea area distribution information within that geographical range is obtained; the sea area distribution information includes the boundary coordinates and sea area names of each sea area. Spatial matching is performed between the predetermined route trajectory and the boundary coordinates of each sea area to identify the sea areas that the predetermined route trajectory passes through sequentially from the starting port to the destination port, and the position points where the predetermined route trajectory enters and leaves each sea area are recorded. Based on the positions of entering and leaving each sea area according to the predetermined route trajectory, the predetermined route trajectory is divided into multiple segments. The starting point of each segment is the position of entering a certain sea area, and the ending point is the position of leaving that sea area. For each segment, based on the geographical coordinates of the starting and ending points of the segment, the corresponding sea area is determined, and the voyage distance of the segment is calculated.

4. The method for predicting the speed of a bulk carrier according to claim 2, characterized in that, The sea area characteristic data of the navigation segment includes historical sea state records and real-time sea state forecasts of the sea area where the navigation segment is located. The navigation time series data of the navigation segment includes the expected entry time and expected navigation duration of the navigation segment. Both the historical sea state records and the real-time sea state forecasts include wave height, wave period, ocean current speed, wind speed and wind direction parameters.

5. The method for predicting the speed of a bulk carrier according to claim 2, characterized in that, The periodic feature extraction includes: performing frequency domain analysis on the wave period time series and extracting the wave period value with the highest frequency to serve as the dominant wave period feature of the flight segment; the directional feature extraction includes: converting each wind direction angle in the wind direction time series into a unit circle vector, calculating the average vector of all unit circle vectors, and taking the azimuth angle of the average vector as the average wind direction angle feature.

6. The method for predicting the speed of a bulk carrier according to claim 1, characterized in that, The ship design parameters also include design draft, empty displacement volume, hull line parameters, main engine rated power and shafting efficiency; the cargo information also includes cargo type, cargo volume and cargo density.

7. A speed prediction system for bulk carriers, used to implement the speed prediction method for bulk carriers according to any one of claims 1-6, characterized in that, The system includes: The route and sea condition analysis module is used to acquire route planning data of the target bulk carrier and divide it into different segments. For each segment, it acquires segment sea area feature data and segment navigation time series data to establish a sea condition spatiotemporal sequence dataset, and obtains sea condition feature vectors for each segment through multi-dimensional time series feature extraction. The operating condition resistance calculation module is used to determine the current operating condition based on the target bulk carrier's ship design parameters and cargo information, and to calculate the hull wetted surface area and displacement volume under the current operating condition; combined with the sea state characteristic vectors of each segment, it analyzes the total resistance value and effective propulsion power of each segment. The speed forecast generation module is used to calculate the forecast speed value for each segment based on the total resistance value and effective propulsion power of each segment, and generate segmented speed forecast curves along the route and the total voyage time by combining the distance of each segment.

Images