Marine fuel supply data monitoring method and system based on sea area environment

Abstract

The invention relates to the field of ship fuel, and discloses a ship fuel supply data monitoring method and system based on a sea area environment, and the method comprises the steps: obtaining the real-time navigation data, weather forecast data and ship log data of a ship; based on the real-time navigation data, the weather forecast data, the ship log data and the trained multi-task learning model, obtaining a corresponding energy consumption mode and a predicted energy consumption trend of a future first duration; when the high energy consumption mode is identified, calculating a fuel depletion time and a minimum replenishment demand based on a current fuel stock, a remaining voyage, and a predicted energy consumption trend; according to the supply demand, the fuel depletion time and a preset route range, screening out candidate supply places; performing priority ranking on the candidate supply places; and generating a replenishment scheme based on the replenishment place with the highest priority in the priority sequence. According to the technical scheme, the navigation risk caused by insufficient supply is effectively avoided.

Description

Technical Field

[0001] This application relates to the field of marine fuel, and in particular to a method and system for monitoring marine fuel supply data based on marine environment. Background Technology

[0002] During navigation, ships are affected by various factors such as sea state, wind and waves, ocean currents, load, and main engine operating conditions, resulting in highly nonlinear and time-varying energy consumption characteristics. In harsh sea areas or under high-load conditions, ships are prone to entering a high-energy-consuming state. If timely identification and replenishment are not planned, fuel may be depleted prematurely, seriously affecting the ship's schedule.

[0003] Currently, mainstream ship energy efficiency management systems or electronic chart display and information systems typically estimate remaining voyage and fuel demand based on fixed fuel consumption models or historical average data. Some systems have introduced meteorological routing technology, combining forecast wind and wave information to optimize routes and reduce energy consumption; however, such methods cannot provide timely warnings in the early stages of sudden increases in energy consumption.

[0004] Solving this technical problem is a technical challenge that needs to be overcome by those skilled in the art. Summary of the Invention

[0005] This application provides a method for monitoring ship fuel supply data based on marine environment, so as to at least partially solve the above-mentioned technical problems.

[0006] To achieve the above objectives, according to a first aspect of this application, a method for monitoring ship fuel supply data based on marine environment is provided, comprising: Acquire real-time navigation data, weather forecast data, and ship log data; Based on the real-time navigation data, weather forecast data, and ship log data, as well as the trained multi-task learning model, the corresponding energy consumption pattern and the expected energy consumption trend for the first time period are obtained; the energy consumption pattern includes a high energy consumption pattern and a normal energy consumption pattern. When a high-energy-consumption mode is identified, the fuel depletion time and minimum replenishment requirement are calculated based on the current fuel inventory, remaining range, and expected energy consumption trends. Candidate resupply locations are selected based on the resupply requirements, fuel depletion time, and preset route range. Acquire service availability data, fuel price data, port condition data, and estimated berthing time for each candidate resupply location; based on the service availability data, fuel price data, port condition data, and estimated berthing time, prioritize the candidate resupply locations while meeting fuel and time constraints; the fuel constraint is that the amount of fuel after resupply must meet the needs of reaching the next destination; the time constraint is that the resupply operation must not cause the arrival time at the destination to exceed the maximum allowable delay time; wherein, the maximum allowable delay time is obtained according to the ship's operation plan; A resupply plan is generated based on the highest priority resupply location in the priority ranking, which includes a recommended route, estimated arrival time, and required resupply amount.

[0007] According to a second aspect of this application, a ship fuel supply data monitoring system based on marine environment is provided, comprising: The first processing module is used to: acquire real-time navigation data, weather forecast data and ship log data of the ship; The second processing module is used to: obtain the corresponding energy consumption pattern and the expected energy consumption trend for the first time period based on the real-time navigation data, weather forecast data, ship log data and the trained multi-task learning model; the energy consumption pattern includes a high energy consumption mode and a normal energy consumption mode. The third processing module is used to: when a high energy consumption mode is identified, calculate the fuel depletion time and minimum replenishment requirement based on the current fuel inventory, remaining range and expected energy consumption trend; The fourth processing module is used to: filter candidate resupply locations based on the resupply demand, fuel depletion time, and preset route range; The fifth processing module is used to: acquire service availability data, fuel price data, port condition data, and estimated berthing time for each candidate resupply location; based on the service availability data, fuel price data, port condition data, and estimated berthing time, prioritize the candidate resupply locations while meeting fuel and time constraints; the fuel constraint is that the amount of fuel after resupply must meet the needs of reaching the next destination; the time constraint is that the resupply operation must not cause the arrival time at the destination to exceed the maximum allowable delay time; wherein, the maximum allowable delay time is obtained according to the ship's operation plan. The sixth processing module is used to generate a resupply plan that includes a recommended route, estimated arrival time, and required resupply amount based on the highest priority resupply location in the priority ranking.

[0008] In summary, the embodiments of this application, through the above technical solutions, effectively avoid navigation risks caused by insufficient supplies.

[0009] Other features and advantages of this application will be described in detail in the following detailed description section. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating the steps of a method for monitoring ship fuel supply data based on marine environment, provided in an exemplary embodiment of this application. Figure 2 This is a schematic diagram of a ship fuel supply data monitoring system based on a marine environment, provided in an exemplary embodiment of this application. Explanation of reference numerals in the attached drawings: 01, First processing module; 202, Second processing module; 203, Third processing module; 204, Fourth processing module; 205, Fifth processing module; 206, Sixth processing module. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the protection scope of this application.

[0013] This application provides a method for monitoring ship fuel supply data based on marine environment. Please refer to [link / reference]. Figure 1 The present application provides a method for monitoring ship fuel supply data based on marine environment, which includes the following steps: Step 101: Obtain the ship's real-time navigation data, weather forecast data, and ship log data.

[0014] Step 102: Based on the real-time navigation data, weather forecast data, ship log data, and the trained multi-task learning model, obtain the corresponding energy consumption pattern and the expected energy consumption trend for the first time period; the energy consumption pattern includes a high energy consumption pattern and a normal energy consumption pattern.

[0015] Step 103: When a high-energy-consumption mode is identified, calculate the fuel depletion time and minimum resupply requirement based on the current fuel inventory, remaining range, and expected energy consumption trend.

[0016] Step 104: Based on the supply demand, fuel depletion time, and preset route range, select candidate supply locations.

[0017] Step 105: Obtain service availability data, fuel price data, port condition data, and estimated berthing time for each candidate resupply location; based on the service availability data, fuel price data, port condition data, and estimated berthing time, prioritize the candidate resupply locations while meeting fuel and time constraints; the fuel constraint is that the amount of fuel after resupply must meet the needs of reaching the next destination; the time constraint is that the resupply operation must not cause the arrival time at the destination to exceed the maximum allowable delay time; wherein, the maximum allowable delay time is obtained according to the ship operation plan.

[0018] Step 106: Generate a resupply plan that includes recommended routes, estimated arrival times, and required resupply amounts based on the highest priority resupply location in the priority ranking.

[0019] Specifically, real-time navigation data refers to operational data collected in real time during a ship's voyage, including ship position, speed, course, main engine speed, load status, and fuel inventory; meteorological forecast data includes weather factors affecting ship energy consumption and navigation, such as wind speed, wind direction, wave height, ocean current speed, and visibility; and ship log data includes energy consumption records for past voyages and historical replenishment data. Minimum replenishment requirement refers to the minimum amount of fuel required to ensure a ship's safe voyage from a replenishment point to its next destination.

[0020] The above scheme acquires real-time navigation data, weather forecast data, and ship log data during the data collection phase. These three types of data are then input into a trained multi-task learning model, which outputs the ship's current energy consumption pattern and the projected energy consumption trend for the first time in the future, enabling the identification and prediction of energy consumption status and trends. When a high-energy-consumption pattern is identified, the model calculates the fuel depletion time and minimum replenishment requirement based on the current fuel inventory, remaining range, and projected energy consumption trend, combined with energy consumption characteristic parameters from historical high-energy-consumption scenarios. This yields the replenishment timeline and fuel quantity requirement. A geographically feasible region is constructed using replenishment requirement, fuel depletion time, and a preset route range as constraints. Candidate replenishment locations with replenishment capabilities and meeting navigation restrictions are selected. Multi-dimensional evaluation data of the candidate locations is acquired and comprehensively scored. These candidate locations are then ranked under the premise of meeting fuel and time constraints. Based on the highest-priority replenishment location, a complete replenishment plan is generated, including a recommended route, estimated arrival time, and required replenishment quantity, effectively avoiding navigation risks caused by insufficient replenishment.

[0021] In some embodiments, the multi-task learning model includes a feature extraction layer, a dual-task sub-model layer, and a fusion layer; The feature extraction layer is configured to extract general features of ship navigation; the feature extraction layer includes a data preprocessing module, a feature engineering module, and a feature extraction network; the data preprocessing module normalizes the input data; the feature engineering module constructs ship navigation feature vectors; the feature extraction network adopts a convolutional neural network architecture to extract high-level abstract features from the ship navigation feature vectors; The dual-task sub-model layer includes a parallel energy consumption prediction sub-model and an energy consumption pattern recognition sub-model. The energy consumption prediction sub-model uses a hybrid structure of temporal convolutional networks and bidirectional LSTM to predict the future energy consumption trend of the ship. The energy consumption pattern recognition sub-model uses a cascaded structure of an improved clustering algorithm and a multi-head self-attention classifier to identify the ship's current energy consumption pattern. The fusion layer is used to integrate the output results of the dual-task sub-models to generate a comprehensive energy consumption analysis result.

[0022] Specifically, the feature extraction layer is the foundation of the model, comprising a data preprocessing module, a feature engineering module, and a feature extraction network. The data preprocessing module normalizes real-time ship navigation data, weather forecast data, and ship log data to ensure data quality. The feature engineering module constructs ship navigation feature vectors, converting the raw data into feature representations suitable for model processing. The feature extraction network adopts a convolutional neural network architecture, extracting high-level abstract features from the feature vectors through multiple convolutional operations. The dual-task sub-model layer comprises two parallel sub-models: an energy consumption prediction sub-model and an energy consumption pattern recognition sub-model. The energy consumption prediction sub-model employs a hybrid structure of a temporal convolutional network and a bidirectional LSTM. The temporal convolutional network captures short-term fluctuations and medium-term trends in energy consumption, while the bidirectional LSTM captures long-term dependencies. The combination of these two technologies enables the prediction of future energy consumption trends for ships. The energy consumption pattern recognition sub-model uses a cascaded structure of an improved clustering algorithm and a multi-head self-attention classifier. First, the improved clustering algorithm discovers latent patterns in the data, and then the multi-head self-attention classifier identifies these patterns, effectively distinguishing between high-energy-consuming and normal-energy-consuming patterns. The fusion layer is the result integration layer of the model, responsible for integrating the output results of the dual-task sub-models.

[0023] The above scheme employs a convolutional neural network architecture for feature extraction, which effectively extracts abstract features from the data. The hybrid structure of the energy consumption prediction sub-model combines the advantages of temporal convolutional networks and bidirectional LSTM, enabling it to simultaneously capture short-term fluctuations, medium-term trends, and long-term dependencies, thus making the prediction results more accurate. The cascaded structure of the energy consumption pattern recognition sub-model, through the synergistic effect of improved clustering algorithms and multi-head self-attention classifiers, can accurately identify different energy consumption patterns.

[0024] In some embodiments, the temporal convolutional network portion of the energy consumption prediction sub-model adopts a multi-layer cascaded structure, with each layer employing causal convolution and the expansion factors of each layer being progressively increased. The first temporal convolutional network layer uses a first preset expansion factor to capture short-term energy consumption fluctuations; the intermediate temporal convolutional network layers use a second preset expansion factor to capture medium-term energy consumption trends; and the top temporal convolutional network layer uses a third preset expansion factor to capture long-term energy consumption patterns. The first preset expansion factor is smaller than the second preset expansion factor, and the second preset expansion factor is smaller than the third preset expansion factor. Each temporal convolutional network layer is followed by a batch normalization layer and a regularization layer. The bidirectional LSTM part of the energy consumption prediction sub-model includes multiple bidirectional LSTM layers; the front bidirectional LSTM layer returns a complete sequence output to preserve the time-by-time variation details of energy consumption data; the back bidirectional LSTM layer returns the final output vector to analyze the energy consumption trend of the entire navigation segment; when a significant change in sea state is detected, the forget gate weights of the LSTM are adjusted to adapt to the impact of environmental changes on energy consumption; the output layer adopts a fully connected layer structure to output the energy consumption trend prediction sequence for the first time period in the future.

[0025] Specifically, the temporal convolutional network employs a multi-layer cascaded structure. The first layer uses a first preset dilation factor to capture short-term energy consumption fluctuations; the middle layers use a second preset dilation factor to capture medium-term energy consumption trends; and the top layer uses a third preset dilation factor to capture long-term energy consumption patterns, thus enabling the simultaneous capture of features at different time scales. Each temporal convolutional network layer is followed by a batch normalization layer and a regularization layer. The batch normalization layer accelerates training convergence, while the regularization layer prevents overfitting. The bidirectional LSTM component comprises multiple bidirectional LSTM layers. The first bidirectional LSTM layer returns the complete sequence output, preserving the time-period variation details of energy consumption data. The second bidirectional LSTM layer returns the final output vector, used to analyze the energy consumption trend throughout the entire voyage. When a significant change in sea state is detected, the forget gate weights of the LSTM are adjusted to adapt to the impact of environmental changes on energy consumption. The output layer employs a fully connected layer structure, integrating the features extracted from previous layers to output a predicted energy consumption trend sequence for the first time segment in the future. This sequence contains the projected energy consumption for each time segment within the future period. The above scheme can provide predictions of energy consumption trends under different sea conditions and navigation conditions.

[0026] In some embodiments, the energy consumption pattern recognition sub-model includes a clustering layer, a feature transformation layer, and a classification layer; The clustering layer uses an improved K-means algorithm to determine the optimal number of clusters; The feature transformation layer employs a multi-head self-attention mechanism. Different numbers of attention heads are set to process numerical features, categorical features, and time-series features respectively. A different feature mapping strategy is used for each attention head. Linear transformations are performed on the input features to generate a query matrix, a key matrix, and a value matrix. The contribution of different features to the energy consumption pattern is calculated based on the attention mechanism, and the attention weights are determined based on the correlation between features. The value matrix is ​​weighted and summed according to the attention weights to generate the attention output. The outputs of multiple attention heads are concatenated and linearly transformed to generate the final feature representation. The classification layer uses a support vector machine algorithm to output the probability distribution of various energy consumption patterns.

[0027] Specifically, the feature transformation layer employs a multi-head self-attention mechanism, which effectively captures the dependencies between features. Specifically, different numbers of attention heads are set to process numerical features, categorical features, and time-series features respectively, with each attention head using a different feature mapping strategy. Linear transformations are performed on the input features to generate query matrices, key matrices, and value matrices. The contribution of different features to energy consumption patterns is calculated based on the attention mechanism, with attention weights determined based on the correlation between features. The value matrices are weighted and summed according to the attention weights to generate the attention output. The outputs of multiple attention heads are concatenated and linearly transformed to generate the final feature representation. This design allows the model to focus on important features, improving the accuracy of pattern recognition. The classification layer uses a support vector machine algorithm, and its output is the probability distribution of various energy consumption patterns, including high-energy-consuming patterns and normal-energy-consuming patterns. Through the above scheme, the clustering layer can discover potential pattern structures in the data by improving the K-means algorithm; the feature transformation layer adopts a multi-head self-attention mechanism, which can focus on the importance of different types of features and effectively capture the dependencies between features by calculating attention weights, thereby improving the feature expression ability; thus, the sub-model can accurately identify the high energy consumption mode and normal energy consumption mode of the ship and detect abnormal energy consumption in a timely manner.

[0028] In some embodiments, candidate resupply locations are selected based on the resupply demand, fuel depletion time, and preset flight path range, including: Starting from the current ship position and with the furthest navigable distance corresponding to the fuel depletion time as the radius, a geographically feasible domain is constructed within a strip area with a width threshold set on both sides of the preset route. Within the geographically feasible region, ports or offshore bunkering points with ship refueling capabilities are selected as the initial candidate set. For each location in the initial candidate set, calculate the fuel consumption required to travel from the current position to that location, and eliminate locations that meet any of the following conditions: the required fuel consumption is greater than the current fuel supply, the time to reach the location is later than the fuel depletion time minus the safety buffer time, and the lateral deviation from the preset route exceeds the maximum allowable yaw distance. The remaining locations will be retained as final candidate resupply locations.

[0029] Specifically, a geographically feasible region is constructed within a strip-shaped area with a width threshold set on both sides of the preset route, starting from the current ship position and using the furthest navigable distance corresponding to the fuel depletion time as the radius. Fuel constraints and route constraints effectively limit the search range of candidate replenishment locations, ensuring that the selected locations are reachable by the ship before fuel depletion and are located near the preset route. Within the constructed geographically feasible region, all ports or offshore bunkering points with ship refueling capabilities are selected as the initial candidate set, excluding locations without refueling capabilities. For each location in the initial candidate set, the required fuel consumption to travel from the current ship position to that location is calculated, and locations meeting any of the following conditions are eliminated: the required fuel consumption is greater than the current fuel stock, ensuring the ship has sufficient fuel to reach the location; the arrival time is later than the fuel depletion time minus the safety buffer time, ensuring the ship has sufficient time to reach the location and reserve a safety buffer; the lateral deviation from the preset route exceeds the maximum allowable yaw distance, ensuring the replenishment location is within a reasonable route deviation range. Locations meeting all conditions after multiple rounds of screening are retained as the final candidate replenishment locations.

[0030] The above scheme effectively narrows the search range and improves screening efficiency by using the dual constraints of fuel depletion time and preset route. By using three indicators—fuel consumption, arrival time, and route deviation—infeasible locations are eliminated. This allows for the selection of suitable candidate resupply locations, thereby improving the efficiency of resupply decision-making.

[0031] In some embodiments, when a high-energy-consumption mode is identified, the fuel depletion time and minimum replenishment requirement are calculated based on the current fuel inventory, remaining range, and projected energy consumption trends, including: Based on the confirmed high energy consumption patterns, energy consumption characteristic parameters of historical high energy consumption scenarios are extracted from the historical database. The projected energy consumption trend is broken down into fuel consumption forecasts for multiple time periods; Based on the current fuel inventory, the time-period fuel consumption forecasts are accumulated and corrected in chronological order to generate a cumulative fuel consumption time series and determine the fuel depletion time. Calculate the remaining distance from the ship's current position to its next destination; The average fuel consumption level under high energy consumption mode is determined based on the energy consumption characteristic parameters; the minimum amount of fuel required to complete the segment is calculated based on the average fuel consumption level and the remaining range. Based on the minimum fuel quantity, the safe reserve fuel quantity, and the current actual fuel inventory, calculate the minimum replenishment requirement.

[0032] Specifically, based on confirmed high-energy-consuming patterns, energy consumption characteristic parameters under similar historical high-energy-consuming scenarios are extracted from historical databases. These parameters include average energy consumption rate, energy consumption fluctuation range, and energy consumption growth trend. The expected energy consumption trend is decomposed into fuel consumption forecasts for multiple time segments. Using the current fuel inventory as a benchmark, the time-segmented fuel consumption forecasts are accumulated and corrected in chronological order to generate a cumulative fuel consumption time series. By analyzing this time series, the point in time when the cumulative fuel consumption equals the current fuel inventory is determined, which is the fuel depletion time. The ship's current position is then calculated. The remaining flight distance to the next destination is the basic parameter for calculating the minimum fuel requirement. Based on energy consumption characteristic parameters extracted from historical data, the average fuel consumption level under high energy consumption mode is determined. Combining this average consumption level and the remaining flight distance, the minimum amount of fuel required to complete this segment is calculated. Taking into account the minimum fuel requirement, the safety reserve fuel quantity, and the current actual fuel inventory, the minimum replenishment requirement is calculated. The calculation formula is: Minimum replenishment requirement = Minimum fuel quantity + Safety reserve fuel quantity - Current actual fuel inventory. When the calculation result is negative, it means that the current fuel inventory is sufficient and no replenishment is required. The above scheme can predict the fuel depletion time and calculate a reasonable minimum replenishment requirement when the ship is in a high-energy-consuming mode, thus avoiding navigation risks caused by insufficient fuel.

[0033] In some embodiments, based on the service availability data, fuel price data, port condition data, and estimated berthing time, and under the premise of satisfying fuel constraints and time constraints, the candidate resupply locations are prioritized, including: For each candidate resupply location, obtain service availability data, fuel price data, port condition data, and estimated docking time. The service availability data, fuel price data, port condition data, and estimated berthing time are converted into corresponding scoring items. Based on the vessel's current operational objectives, each scoring item is assigned a corresponding weight; these operational objectives include priority of cost, time, or service reliability. Based on each scoring item and its corresponding weight, a comprehensive score is calculated for each candidate supply location; Candidate locations that simultaneously meet both fuel and time constraints are retained; the retained candidate locations are prioritized according to their comprehensive scores from highest to lowest.

[0034] Reference Figure 2 The second embodiment of the present invention provides a ship fuel supply data monitoring system based on marine environment, comprising: The first processing module 201 is used to: acquire real-time navigation data, weather forecast data and ship log data of the ship; The second processing module 202 is used to: obtain the corresponding energy consumption pattern and the expected energy consumption trend for the first time period based on the real-time navigation data, weather forecast data, ship log data and the trained multi-task learning model; the energy consumption pattern includes a high energy consumption mode and a normal energy consumption mode. The third processing module 203 is used to: when a high energy consumption mode is identified, calculate the fuel depletion time and minimum replenishment requirement based on the current fuel inventory, remaining range and expected energy consumption trend; The fourth processing module 204 is used to: filter candidate resupply locations based on the resupply demand, fuel depletion time, and preset route range; The fifth processing module 205 is used to: acquire service availability data, fuel price data, port condition data, and estimated berthing time for each candidate resupply location; based on the service availability data, fuel price data, port condition data, and estimated berthing time, prioritize the candidate resupply locations while meeting fuel and time constraints; the fuel constraint is that the amount of fuel after resupply must meet the needs of reaching the next destination; the time constraint is that the resupply operation must not cause the arrival time at the destination to exceed the maximum allowable delay time; wherein, the maximum allowable delay time is obtained according to the ship operation plan; The sixth processing module 206 is used to generate a resupply plan that includes a recommended route, estimated arrival time and required resupply amount based on the resupply location with the highest priority in the priority ranking.

[0035] It should be noted that the marine environment-based ship fuel supply data monitoring system provided in this embodiment of the invention is used to execute all the process steps of the marine environment-based ship fuel supply data monitoring method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0036] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0037] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0038] The embodiments, implementation methods, and related technical features of this application can be combined and substituted for each other without conflict.

[0039] The above are merely preferred embodiments of this application and are not intended to limit this application in any way. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application without departing from the scope of the technical solution of this application shall still fall within the scope of the technical solution of this application.

Claims

1. A method for monitoring ship fuel supply data based on marine environment, characterized in that, include: Acquire real-time navigation data, weather forecast data, and ship log data; Based on the real-time navigation data, weather forecast data, and ship log data, as well as the trained multi-task learning model, the corresponding energy consumption pattern and the expected energy consumption trend for the first time period are obtained; the energy consumption pattern includes a high energy consumption pattern and a normal energy consumption pattern. When a high-energy-consumption mode is identified, the fuel depletion time and minimum replenishment requirement are calculated based on the current fuel inventory, remaining range, and expected energy consumption trends. Candidate resupply locations are selected based on the resupply requirements, fuel depletion time, and preset route range. Obtain service availability data, fuel price data, port condition data, and estimated docking time for each candidate resupply location; Based on the service availability data, fuel price data, port condition data, and estimated berthing time, the candidate resupply locations are prioritized while meeting fuel and time constraints. The fuel constraint is that the amount of fuel after resupply must be sufficient to reach the next destination. The time constraint is that the resupply operation must not cause the arrival time at the destination to exceed the maximum allowable delay time. The maximum allowable delay time is obtained according to the ship's operation plan. A resupply plan is generated based on the highest priority resupply location in the priority ranking, which includes a recommended route, estimated arrival time, and required resupply amount.

2. The method according to claim 1, characterized in that, The multi-task learning model includes a feature extraction layer, a dual-task sub-model layer, and a fusion layer; The feature extraction layer is configured to extract general features of ship navigation; the feature extraction layer includes a data preprocessing module, a feature engineering module, and a feature extraction network; the data preprocessing module normalizes the input data; the feature engineering module constructs ship navigation feature vectors; the feature extraction network adopts a convolutional neural network architecture to extract high-level abstract features from the ship navigation feature vectors; The dual-task sub-model layer includes a parallel energy consumption prediction sub-model and an energy consumption pattern recognition sub-model; the energy consumption prediction sub-model adopts a hybrid structure of temporal convolutional network and bidirectional LSTM to predict the future energy consumption trend of ships. The energy consumption pattern recognition sub-model adopts a cascaded structure of an improved clustering algorithm and a multi-head self-attention classifier to identify the ship's current energy consumption pattern. The fusion layer is used to integrate the output results of the dual-task sub-models to generate a comprehensive energy consumption analysis result.

3. The method according to claim 2, characterized in that, The temporal convolutional network part of the energy consumption prediction sub-model adopts a multi-layer cascaded structure, with each layer using causal convolution and the expansion factor of each layer being set in an increasing manner; the first temporal convolutional network layer uses a first preset expansion factor to capture short-term energy consumption fluctuation characteristics; the middle temporal convolutional network layer uses a second preset expansion factor to capture medium-term energy consumption trend characteristics. The top-level temporal convolutional network layer uses a third preset dilation factor to capture long-term energy consumption pattern characteristics; wherein, the first preset dilation factor is smaller than the second preset dilation factor, and the second preset dilation factor is smaller than the third preset dilation factor; each temporal convolutional network layer is followed by a batch normalization layer and a regularization layer; The bidirectional LSTM part of the energy consumption prediction sub-model includes multiple bidirectional LSTM layers; the front bidirectional LSTM layer returns a complete sequence output to preserve the time-by-time variation details of energy consumption data; the back bidirectional LSTM layer returns the final output vector to analyze the energy consumption trend of the entire navigation segment; when a significant change in sea state is detected, the forget gate weights of the LSTM are adjusted to adapt to the impact of environmental changes on energy consumption; the output layer adopts a fully connected layer structure to output the energy consumption trend prediction sequence for the first time period in the future.

4. The method according to claim 3, characterized in that, The energy consumption pattern recognition sub-model includes a clustering layer, a feature transformation layer, and a classification layer; The clustering layer uses an improved K-means algorithm to determine the optimal number of clusters; The feature transformation layer employs a multi-head self-attention mechanism. Different numbers of attention heads are set to process numerical features, categorical features, and time-series features respectively. A different feature mapping strategy is used for each attention head. Linear transformations are performed on the input features to generate a query matrix, a key matrix, and a value matrix. The contribution of different features to the energy consumption pattern is calculated based on the attention mechanism, and the attention weights are determined based on the correlation between features. The value matrix is ​​weighted and summed according to the attention weights to generate the attention output. The outputs of multiple attention heads are concatenated and linearly transformed to generate the final feature representation. The classification layer uses a support vector machine algorithm to output the probability distribution of various energy consumption patterns.

5. The method according to claim 4, characterized in that, Based on the resupply requirements, fuel depletion time, and preset route range, candidate resupply locations are selected, including: Starting from the current ship position and with the furthest navigable distance corresponding to the fuel depletion time as the radius, a geographically feasible domain is constructed within a strip area with a width threshold set on both sides of the preset route. Within the geographically feasible region, ports or offshore bunkering points with ship refueling capabilities are selected as the initial candidate set. For each location in the initial candidate set, calculate the fuel consumption required to travel from the current position to that location, and eliminate locations that meet any of the following conditions: the required fuel consumption is greater than the current fuel supply, the time to reach the location is later than the fuel depletion time minus the safety buffer time, and the lateral deviation from the preset route exceeds the maximum allowable yaw distance. The remaining locations will be retained as final candidate resupply locations.

6. The method according to claim 5, characterized in that, When a high-energy-consumption mode is identified, the fuel depletion time and minimum resupply requirement are calculated based on the current fuel inventory, remaining range, and projected energy consumption trends, including: Based on the confirmed high energy consumption patterns, energy consumption characteristic parameters of historical high energy consumption scenarios are extracted from the historical database. The projected energy consumption trend is broken down into fuel consumption forecasts for multiple time periods; Based on the current fuel inventory, the time-period fuel consumption forecasts are accumulated and corrected in chronological order to generate a cumulative fuel consumption time series and determine the fuel depletion time. Calculate the remaining distance from the ship's current position to its next destination; The average fuel consumption level under high energy consumption mode is determined based on the energy consumption characteristic parameters; the minimum amount of fuel required to complete the segment is calculated based on the average fuel consumption level and the remaining range. Based on the minimum fuel quantity, the safe reserve fuel quantity, and the current actual fuel inventory, calculate the minimum replenishment requirement.

7. The method according to claim 6, characterized in that, Based on the service availability data, fuel price data, port condition data, and estimated berthing time, and under the premise of meeting fuel and time constraints, the candidate resupply locations are prioritized, including: For each candidate resupply location, obtain service availability data, fuel price data, port condition data, and estimated docking time. The service availability data, fuel price data, port condition data, and estimated berthing time are converted into corresponding scoring items. Based on the vessel's current operational objectives, each scoring item is assigned a corresponding weight; these operational objectives include priority of cost, time, or service reliability. Based on each scoring item and its corresponding weight, a comprehensive score is calculated for each candidate supply location; Candidate locations that simultaneously meet both fuel and time constraints are retained; the retained candidate locations are prioritized according to their comprehensive scores from highest to lowest.

8. A ship fuel supply data monitoring system based on marine environment, characterized in that, include: The first processing module is used to: acquire real-time navigation data, weather forecast data and ship log data of the ship; The second processing module is used to: obtain the corresponding energy consumption pattern and the expected energy consumption trend for the first time period based on the real-time navigation data, weather forecast data, ship log data and the trained multi-task learning model; the energy consumption pattern includes a high energy consumption mode and a normal energy consumption mode. The third processing module is used to: when a high energy consumption mode is identified, calculate the fuel depletion time and minimum replenishment requirement based on the current fuel inventory, remaining range and expected energy consumption trend; The fourth processing module is used to: filter candidate resupply locations based on the resupply demand, fuel depletion time, and preset route range; The fifth processing module is used to: acquire service availability data, fuel price data, port condition data, and estimated berthing time for each candidate resupply location; based on the service availability data, fuel price data, port condition data, and estimated berthing time, prioritize the candidate resupply locations while meeting fuel and time constraints; the fuel constraint is that the amount of fuel after resupply must meet the needs of reaching the next destination; the time constraint is that the resupply operation must not cause the arrival time at the destination to exceed the maximum allowable delay time; wherein, the maximum allowable delay time is obtained according to the ship's operation plan. The sixth processing module is used to generate a resupply plan that includes a recommended route, estimated arrival time, and required resupply amount based on the highest priority resupply location in the priority ranking.

Images