Method and system for minimizing European Union carbon emission penalty of LNG (Liquefied Natural Gas) dual-fuel ship

Abstract

The invention relates to the technical field of ship greenhouse gas emission reduction optimization, in particular to a method and system for minimizing the European Union carbon emission penalty of an LNG dual-fuel ship. The method comprises the following steps: collecting data, processing the data to construct a data set, dividing the data set into a training set and a test set according to collection time, learning a relationship between fuel consumption and ship dynamic data and meteorological data, constructing an artificial neural network model, predicting ship fuel consumption, setting an upper limit and a lower limit of energy consumption, and calculating the ship fuel consumption. And the optimal energy consumption distribution of the European Union carbon emission penalty minimization problem is sought by using a genetic algorithm. According to the method, the problem that in the prior art, no effective method for optimizing the carbon emission of the ship in the European leg at the departure port or the destination port exists so as to meet the requirements of new regulations of the European Union is solved through technical innovation, an economic and reasonable operation strategy is provided for ship owners, and decarburization and environmental protection development of the marine transportation industry can be promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of ship greenhouse gas emission reduction optimization, and in particular to a method and system for minimizing EU carbon emission fines for LNG dual-fuel ships. Background Art

[0003] The continuous increase of greenhouse gas emissions poses a serious challenge to climate change. Current technologies face the following major problems:

[0004] Rising greenhouse gas emissions: Greenhouse gas emissions from the maritime industry continue to rise, especially in the EU economy, posing a threat to the achievement of climate change targets. Existing technologies are not effective in countering the rising greenhouse gas emissions.

[0005] Impact of new regulations on the shipping industry: The promulgation of new regulations such as Fuel EU Maritime has imposed stricter greenhouse gas emission limits on shipping companies, requiring the shipping industry to adopt more environmentally friendly fuels and technologies. Existing technologies have not yet provided comprehensive and innovative solutions to meet the requirements of these new regulations.

[0006] The challenge of energy optimization: The shipping industry needs to achieve more efficient energy use to reduce carbon emissions and lower carbon penalties. However, there is currently no comprehensive and intelligent system to optimize the carbon emissions of ships in the European leg of the voyage at the port of departure or destination and comply with the requirements of the new regulations. Summary of the invention

[0007] The present invention solves the problem that there is no effective method in the prior art to optimize the carbon emissions of ships in the European section at the departure port or the destination port to comply with the requirements of new regulations, and provides a method and system for minimizing the EU carbon emission fines for LNG dual-fuel ships.

[0008] The technical solution claimed in the present invention is as follows:

[0009] A method for minimizing EU carbon emission fines for LNG dual-fuel ships, comprising the following steps:

[0010] S1: Data collection: Obtain ship dynamic data, ship fuel consumption message data, and meteorological data, and process the data to construct a data set; the ship fuel consumption message data includes: fuel consumption and fuel quality; the fuel quality includes: traditional fuel quality and liquefied natural gas (LNG) quality;

[0011] S2: Data processing: Divide the data set described in S1 according to the acquisition time to obtain a training set and a test set;

[0012] S3: Model construction and training: construct an artificial neural network model, and use the training set in S2 to train the model, and use the test set in S2 to test the model; the artificial neural network model learns the relationship between fuel consumption and ship dynamic data and meteorological data;

[0013] S4: Energy consumption prediction and range setting: Use the artificial neural network model trained in S3 to simulate the operation of all ships whose departure ports or destination ports are in the European section in the future, predict the fuel consumption of ships in each section, and set reasonable upper and lower limits based on the fuel consumption, where the upper and lower limits are the maximum and minimum values ​​of fuel consumption respectively;

[0014] S5: Genetic algorithm optimization: Based on the maximum and minimum values ​​of fuel consumption set in S4, a genetic algorithm is used to minimize the EU carbon emission fines as the goal, and fuel consumption, traditional fuel quality and LNG quality are used as variables to construct an objective function, seek the optimal energy consumption allocation for the EU carbon emission fine minimization problem, and obtain the optimal solution of the objective function.

[0015] Preferably, the fuel consumption includes traditional fuel consumption and LNG consumption; the ship dynamic data includes: collection time, position, navigation speed, heading, daily main engine speed, navigation distance, navigation time, ship draft; the meteorological data includes: wind speed, sea conditions, upload time, longitude and latitude; the wind speed includes wind force; the sea conditions include wave size and surge size; the ship fuel consumption message data also includes reporting time.

[0016] Preferably, the data set in S1 includes: historical data of the LNG dual-fuel ship and historical data of its sister LNG dual-fuel ships.

[0017] Preferably, the processing in S1 includes: data cleaning, data fusion, data clustering and data encoding; the data cleaning refers to deleting some abnormal values ​​in the ship dynamic data; the data fusion refers to aligning the ship fuel consumption message data with the ship dynamic data according to the reporting time in the ship fuel consumption message data, and matching the meteorological data with the ship dynamic data according to the upload time and longitude and latitude in the meteorological data; the data clustering is to divide different loading states according to the ship draft in the ship dynamic data using the K-Means clustering algorithm, including: empty, full and half load; the data encoding is the encoding processing of the original three segments of data; the original three segments of data include: ship dynamic data, ship fuel consumption message data, and meteorological data;

[0018] Preferably, S3 comprises the following steps:

[0019] S31: Input feature selection: variables strongly related to fuel consumption in ship dynamic data and meteorological data are used as input features of the model; the variables strongly related to fuel consumption in the ship dynamic data include: sailing speed, daily main engine speed, sailing distance, full load, no load and states between full load and no load; the variables strongly related to fuel consumption in the meteorological data include: wind force and wave size;

[0020] S32: Output feature setting: The fuel consumption of the ship is used as the output feature of the model;

[0021] S33: Model training: using the training set to train the artificial neural network model, and learning the nonlinear relationship between fuel consumption and basic ship attributes, ship dynamic data and meteorological data;

[0022] S34: Model testing: Use the test set to test the artificial neural network model to test whether the model training is complete. If the test fails, return to S33 to continue training.

[0023] Preferably, the future period of time in S4 is the next year.

[0024] Preferably, S4 comprises the following steps:

[0025] S41: Route segmentation: segment the routes according to the starting and ending ports of the ships, and count and filter out the routes whose departure ports or destination ports are in the European section;

[0026] S42: Calculation of energy consumption for each voyage in the next year: For the voyages where the departure port or the destination port selected by S41 is in the European voyage, the artificial neural network model trained by S3 is used to simulate the ship operation status of each voyage in the next year, and calculate the fuel consumption of the corresponding voyage;

[0027] S43: Energy consumption range setting: The upper and lower limits of energy consumption of the corresponding flight segment are set through the fuel consumption of each flight segment described in S42, so that the fuel energy consumption of each flight segment is between the upper and lower limits of energy consumption of the corresponding flight segment; the upper limit of energy consumption of each flight segment is equal to the value of fuel consumption of the corresponding flight segment multiplied by 1.2, and the lower limit of energy consumption of each flight segment is equal to the value of fuel consumption of the corresponding flight segment multiplied by 0.8.

[0028] Preferably, the objective function formula in S5 is:

[0029] M i2 =491*(M i -M i1 ) / 405

[0030]

[0031]

[0032] Where: M i is the fuel consumption of the i-th flight segment, i=1,2,3,…,n; M i1 is the conventional fuel mass of the i-th flight segment, i = 1, 2, 3, ..., n; M i2 is the LNG mass of the i-th flight segment, i = 1, 2, 3, ..., n; α is the carbon emission factor of the whole life cycle; required is the fuel intensity limit; GHGIE actual_i is the annual average greenhouse gas intensity of energy use on board calculated during the relevant reporting period; Feul_EU_Penalty is the carbon emission penalty of the EU marine fuel regulations; the constraints of the objective function are:

[0033] min i ≤M i ≤max i

[0034] 0≤M i1 ≤M i

[0035] 0≤M i2 ≤M i

[0036] M i1 +M i2 =M i

[0037] Where: min i is the minimum value of fuel consumption; max i is the maximum value of fuel consumption; M i is the fuel consumption of the i-th flight segment, i=1,2,3,…,n; M i1 is the conventional fuel mass of the i-th flight segment, i = 1, 2, 3, ..., n; M i2 is the LNG mass of the i-th segment, i = 1, 2, 3, ..., n; the M i 、M i1 and M i2 is the solution set of the objective function; the min i 、max i The value of fuel consumption is set according to the prediction of the trained artificial neural network model.

[0038] Preferably, the population optimization of the genetic algorithm in S5 comprises the following steps:

[0039] Step 1: Initialize the population: randomly generate a certain number of individuals as the initial population; each individual contains two elements, namely fuel consumption and traditional fuel mass;

[0040] Step 2: Fitness evaluation: Calculate the fitness of each individual in the initial population;

[0041] Step 3: Selection operation: through a certain selection algorithm, select individuals with higher fitness in the population as parents to produce individuals of the next generation;

[0042] Step 4: Crossover operation: select a pair of parent individuals and generate offspring individuals through crossover operation;

[0043] Step 5: Mutation operation: perform mutation operation on the offspring individuals and introduce some randomness of mutation;

[0044] Step 6: Replacement operation: replace the individuals with poor fitness in the original population with the newly generated offspring to form a new population;

[0045] Step 7: Repeat iteration: Repeat steps S3-S6 until the stop condition is reached;

[0046] Step 8: Output results: Output the individual with the highest fitness in the population, and then obtain the optimal solution of the objective function.

[0047] The present invention also provides a system for minimizing EU carbon emission fines for LNG dual-fuel ships, comprising a data acquisition module connected in sequence for acquiring data and processing the data to construct a data set, a data processing module for dividing the data set obtained by the data acquisition module according to the acquisition time and obtaining a training set and a test set, a model building and training module for building an artificial neural network model and using the training set and test of the data processing module to train and test the model respectively, an energy consumption prediction and energy consumption range setting module for simulating the operation of all ships in the European section at the departure port or the destination port in the future period of time using the artificial neural network model trained by the model building and training module to predict the fuel consumption of the ships in each section and set a reasonable upper and lower limits of energy consumption according to the fuel consumption, and a genetic algorithm optimization module for using the upper and lower limits of fuel consumption set by the energy consumption prediction and energy consumption range setting module and using a genetic algorithm to minimize the EU carbon emission fine as a goal and to construct an objective function as variables to seek the optimal energy consumption allocation for the EU carbon emission fine minimization problem and obtain the optimal solution of the objective function; the ship fuel consumption message data includes: fuel consumption and fuel quality; the fuel quality includes: traditional fuel quality and LNG quality.

[0048] Beneficial effects:

[0049] The present invention provides a method and system for minimizing EU carbon emission fines for LNG dual-fuel ships, the method comprising: acquiring ship dynamic data, ship fuel consumption message data, and meteorological data, and processing the data to construct a data set, wherein the ship fuel consumption message data comprises fuel consumption and fuel quality; the fuel quality comprises traditional fuel quality and LNG quality, LNG is a clean energy, the main component of which is methane, and less pollutants are generated during combustion. Compared with traditional ships that only use traditional fuel, LNG dual-fuel ships can effectively reduce greenhouse gas emissions; dividing the data in the data set into a training set and a test set, respectively training and testing the artificial neural network model, learning the relationship between fuel consumption and the ship dynamic data and meteorological data, and improving the accuracy of subsequent model predictions; predicting the fuel consumption of each departure port or destination port in the European section, and setting the upper and lower limits of energy consumption for the corresponding section according to the predicted value of the fuel consumption, wherein the method for setting the upper and lower limits of energy consumption takes into account the uncertainty of model prediction, provides a more reliable reference for ship operation, and helps to conduct subsequent effective and more realistic energy consumption predictions. allocation; by seeking the optimal energy consumption allocation for the EU carbon emission penalty minimization problem through genetic algorithms, the carbon emissions of LNG dual-fuel ships whose departure or destination ports are in the European section can be minimized, so that the ship can achieve the optimal energy consumption allocation in each section, effectively reduce greenhouse gas emissions, meet the greenhouse gas emission cap of the newly promulgated regulations, adapt to the progressive emission reduction targets of the new regulations, and solve the problem that there is no effective method in the existing technology to optimize the carbon emissions of ships in the European section at the departure or end port to meet the requirements of the new EU regulations; the energy utilization efficiency is optimized through the above method, so that ships can use green energy more effectively, accelerate the decarbonization process, provide strong support for the sustainable development of the shipping industry, and meet the emission reduction strategy of the International Maritime Organization and the climate goals of the EU; in addition, the optimal energy consumption of each section is obtained by comprehensively using artificial neural networks and genetic algorithms, and the best traditional fuel quality and LNG quality are determined to minimize the EU carbon emission fines, which not only improves the technical level of ships, but also reduces the operating costs of ships, improves the overall economic benefits, and provides a feasible way for enterprises to maintain their competitive advantages under the new regulations.

[0050] The data set includes data of LNG dual-fuel ships and data of its sister LNG dual-fuel ships, which can effectively improve the accuracy of the artificial neural network model and reduce the impact of the scarcity of single ship data. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the artificial neural network model structure of an embodiment of the present invention.

[0052] Figure 2 The data fusion flow chart of the embodiment of the present invention.

[0053] Figure 3 It is a schematic diagram of the fitting effect between the predicted value and the actual value of the artificial neural network model of the embodiment of the present invention; wherein, (a), (b), and (c) are the fitting effect diagrams of three sister LNG dual-fuel ships; (d) is the fitting effect diagram of the LNG dual-fuel ship - "Yuan Rui Yang"; light purple represents the ship energy consumption prediction line based on the artificial neural network model; red represents the actual ship energy consumption line.

[0054] Figure 4 It is a schematic diagram of the fitting effect between the predicted value and the actual value of the artificial neural network model after correction according to the embodiment of the present invention. Specific implementation methods

[0056] In order to make the purpose, technical solution and advantages of the present invention more clear, the technical solution is further clearly and completely described below in conjunction with the accompanying drawings of the present invention.

[0057] First Group of Embodiments: Methods for Minimizing EU Carbon Emission Penalties for LNG Dual-Fuel Ships

[0058] This group of embodiments provides a method for minimizing the EU carbon emission fines for LNG dual-fuel ships, including the following steps:

[0059] S1: Data collection: Obtain ship dynamic data (AIS data), ship fuel consumption message data (ship MRV daily report), and meteorological data, and process the data to construct a data set; in a specific embodiment of the present invention, the above data sources include AIS data and MRV data provided by COSCO SHIPPING and meteorological data provided by Shanghai Meteorological Bureau. These three data sets are integrated into a data set for dual-energy ship energy consumption modeling. The data set was collected from January 17, 2021 to September 10, 2023.

[0060] The ship fuel consumption message data includes: fuel consumption (ship energy consumption), fuel quality and reporting time, which are used to establish a historical fuel consumption database; the fuel consumption includes traditional fuel consumption and LNG consumption; the fuel quality includes: traditional fuel quality and LNG quality; the ship dynamic data includes: collection time, location, navigation speed, heading, daily main engine speed, navigation distance, navigation time, ship draft, which are used to monitor the operation status of the ship in real time; the meteorological data includes: wind speed, sea conditions, upload time, longitude and latitude, which are used to consider the impact of the external environment on the ship's energy consumption; the wind speed includes the wind force; the sea conditions include the size of the waves and the size of the surge;

[0061] The processing includes: data cleaning, data fusion, data clustering and data coding;

[0062] The data cleaning refers to deleting some abnormal values ​​in the ship dynamic data;

[0063] The data fusion, such as Figure 2 As shown, specifically, the ship fuel consumption message data is aligned with the ship dynamic data according to the reporting time in the ship fuel consumption message data, that is, the two data with the same reporting time in the ship fuel consumption message data and the same collection time in the ship dynamic data are merged into one data, that is, all field data in the two data are merged into one data, and the meteorological data is matched with the ship dynamic data according to the upload time and longitude and latitude fields in the meteorological data, and the fields in the matched meteorological data are taken out and put into the above-mentioned merged data to form the merged data; specifically, in this embodiment, the matching process of meteorological data and ship dynamic data is: according to the ship dynamic data According to the longitude and latitude fields in the data, it is determined which grid corresponding to the meteorological data the ship dynamic data falls into, and then according to the collection time in the ship dynamic data, it is determined which meteorological data update time period the ship dynamic data is in, and the ship dynamic data is matched with the latest meteorological data in the time period; the grid is the grid of the same size divided by the Shanghai Meteorological Bureau when collecting meteorological data. The data used in this embodiment is collected in a grid of 0.25 longitude and latitude * 0.25 longitude and latitude, and is collected every three hours. The Shanghai Meteorological Bureau will perform weighted processing on the data in the grid to obtain a meteorological data representing the grid;

[0064] The data clustering is to divide different loading states according to the ship draft in the ship dynamic data using the K-Means clustering algorithm, including: empty, fully loaded, and a state between full and empty (half loaded); specifically, the ship draft is recorded in the ship dynamic data and the ship fuel consumption message data. The ship draft data used in this embodiment is taken from the ship dynamic data. The reason is that the ship dynamic data is updated more frequently, and the daily average can better reflect the draft condition of the ship when sailing on the same day.

[0065] Due to the particularity of ship transportation, when transporting liquid and dry bulk cargoes, the ship is usually fully loaded and fully unloaded, that is, it is fully loaded and unloaded at the same time, so there are only two states: empty and fully loaded. Containers are relatively special. There are operations of unloading and loading during the transportation of containers. Therefore, when transporting containers, the ship has three states: empty, fully loaded and half loaded.

[0066] Specifically, in this embodiment, in order to better distinguish these states, it is necessary to calculate a clear state dividing line; since the loading state of each type of ship is determined, K-Means clustering is used to divide the states of different types of ships. Among them, dry bulk and liquid bulk transport ships correspond to two loading states, and the K value is 2; container ships correspond to three loading states, and the K value is 3. After using K-Means to cluster the categories of each state, the maximum and minimum values ​​of the draft in each category are counted and sorted according to the value. Taking the container as an example: the mean of the maximum value of the empty draft and the minimum value of the half-loaded draft constitutes the dividing line between empty and half-loaded; the mean of the maximum value of the half-loaded draft and the minimum value of the full-loaded draft constitutes the dividing line between full and half-loaded. By analogy, the state dividing line of different types of ships can be calculated. Different loading states are determined by dividing lines.

[0067] The data encoding is an encoding process of the original three-segment data; the original three-segment data includes: ship dynamic data, ship fuel consumption message data, and meteorological data; specifically, in this embodiment, the data of some fields in the ship dynamic data and meteorological data after data encoding include: the collection time corresponds to Input_Data; the voyage duration corresponds to Duration_voyage; the voyage distance corresponds to Distance; the wind force corresponds to Wind_val (m / s); the wave size corresponds to Wave_val (m); the swell size corresponds to Stream_val (m); the ship draft corresponds to Draught; the daily main engine speed corresponds to Average_Engine_Speed_Daily (kn); no load corresponds to No_load; full load corresponds to Full_load; half load corresponds to Half_load; daily fuel consumption corresponds to Daily_Fuel_Mrv (tons / day);

[0068] The data set includes historical data of LNG dual-fuel ships and historical data of its sister LNG dual-fuel ships, which can effectively improve the accuracy of the artificial neural network model and reduce the negative impact of the scarcity of single ship data;

[0069] Specifically, in this embodiment, the LNG dual-fuel ship is the LNG dual-fuel ship "Yuan Rui Yang"; it is a liquid bulk carrier, so it has only two states: empty or fully loaded, and no half-load state; its sister LNG dual-fuel ships are three ships built using the same drawings as the LNG dual-fuel ship "Yuan Rui Yang";

[0070] Furthermore, the point-by-point fusion data of similar ship dynamic data obtained after data processing is shown in Table 1:

[0071] Table 1. Point-by-point fusion data after data processing

[0072]

[0073] S2: Data processing: Divide the data set described in S1 according to the acquisition time to obtain a training set and a test set;

[0074] In this embodiment, the LNG dual-fuel ship - "Yuan Ruiyang" was put into use at the end of February 2022, and the fuel consumption data in the ship's fuel consumption report was recorded from April 2022; in addition, the daily fuel consumption data only recorded the daily usage of traditional fuel, and the daily usage of LNG was not recorded until November 7, 2022; therefore, this embodiment divides the data before April 1, 2022 in the data set into a training set, and divides the data from April 7, 2022 to November 7, 2022 into a test set.

[0075] S3: Model construction and training: construct an artificial neural network model, and use the training set in S2 to train the model, and use the test set in S2 to test the model; the artificial neural network model ( Figure 1 ) Learn the relationship between fuel consumption and ship dynamic data and meteorological data;

[0076] The specific steps of S3 are as follows:

[0077] S31: Input feature selection: Input feature selection: Variables in the ship dynamic data and meteorological data that are strongly related to fuel consumption are used as input features of the model; the variables in the ship dynamic data that are strongly related to fuel consumption include: navigation speed, daily main engine speed, navigation distance, full load, no load, and the state between full load and no load (half load); the variables in the meteorological data that are strongly related to fuel consumption include: wind force and wave size; specifically, in this specific embodiment, wind force, wave size, navigation speed, daily main engine speed, navigation distance, no load, and full load are used as input features of the model (if it is a container ship, the input features should also include The input features can be subdivided into four categories, namely: ship load status, engine operation status, ship operation status and meteorological conditions; specifically, the fields of the meteorological conditions class include: Wind_val (m / s), Wave_val (m); the fields of the ship operation status class include: Average_Engine_Speed_Daily (kn), Distance (km); the fields of the engine operation status class include: Engine_speed (rpd / s); the fields of the ship load status class include: No_load, Full_load;

[0078] S32: Output feature setting: The fuel consumption of the ship is used as the output feature of the model;

[0079] Specifically, in this embodiment, the field of daily ship fuel consumption is: Daily_Fuel_Mrv (tons / day);

[0080] S33: Model training: using the training set to train the artificial neural network model, and learning the nonlinear relationship between fuel consumption and basic ship attributes, ship dynamic data and meteorological data;

[0081] Specifically, in this embodiment, an artificial neural network (ANN) model is selected, Distance, Wind_val, Wave_val, Stream_val, Average_Engine_Speed_Daily, No_load, and Full_load are selected as input features, and Daily_Fuel_Mrv is selected as an output feature to perform model training. Figure 3 As shown in (a)-(c), the ship energy consumption prediction line based on the artificial neural network model is highly matched with the actual ship energy consumption line, and the cumulative error is very small;

[0082] S34: Model testing: Use the test set to test the artificial neural network model to test whether the model training is complete. If the test fails, return to S33 to continue training;

[0083] It is worth noting that in this embodiment, the actual fuel consumption message data of the LNG dual-fuel ship "Yuan Rui Yang" does not contain the LNG mass on that day. Starting from November 7, 2022, the actual fuel consumption of the LNG dual-fuel ship "Yuan Rui Yang" includes LNG consumption. Therefore, Figure 3 (d) The fitting effect is poor; starting from November 7, 2022, the actual fuel consumption data of the LNG dual-fuel ship "Yuan Rui Yang" includes traditional fuel consumption and LNG consumption. Therefore, when the data after November 7, 2022 are introduced, it is necessary to convert the LNG mass into the traditional fuel (petroleum) mass according to the calorific value for unified comparison. The corrected results are as follows Figure 4 As shown in the figure, after supplementing the LNG consumption, the predicted value is highly matched with the actual value, and the cumulative error is about 7%, which is acceptable in practical applications; overall, the trained artificial neural network model performs well after integrating meteorological data and ship dynamic data.

[0084] S4: Energy consumption prediction and range setting: Use the artificial neural network model trained in S3 to simulate the operation of all ships whose departure ports or end ports are in the European section in the future, predict the fuel consumption of ships in each section, and set a reasonable upper limit and lower limit based on the fuel consumption, the upper limit and lower limit are the maximum value and minimum value of the fuel consumption respectively; in a specific embodiment of the present invention, the future period is the next year;

[0085] The specific operation steps of S4 are as follows:

[0086] S41: Route segmentation: segment the route according to the ship's starting and ending ports, and count and filter out the segments where the departure port or destination port is in Europe;

[0087] Specifically, in this embodiment, the route of the LNG dual-fuel ship "Yuan Rui Yang" from November 7, 2022 to September 10, 2023 is divided according to the starting and ending ports of the ship, and there are 23 segments in total after segmentation, of which 15 segments have departure ports or destination ports in Europe; some results of the route segmentation of the LNG dual-fuel ship "Yuan Rui Yang" are shown in Table 2:

[0088] Table 2. Route segmentation results of LNG dual-fuel ship "Yuan Rui Yang"

[0089]

[0090] S42: Calculation of energy consumption for each voyage in the next year: For the voyages whose departure ports or destination ports are located in Europe selected by S41, the artificial neural network model trained by S3 is used to simulate the ship operation status of each voyage in the next year, and calculate the fuel consumption of the corresponding voyage;

[0091] Furthermore, in this embodiment, the trained artificial neural network model is used to predict the fuel consumption of the LNG dual-fuel ship "Yuan Rui Yang" from November 7, 2022 to September 10, 2023. Some of the prediction results are shown in Table 3:

[0092] Table 3. Prediction results of artificial neural network model

[0093]

[0094]

[0095] Among them: the daily actual traditional fuel mass and daily actual LNG mass are the records after the actual use of LNG, and the fuel consumption prediction value is the daily fuel consumption predicted by the artificial neural network model, in which the daily fuel consumption is based on traditional fuel as a unified standard; the fuel consumption of each departure port or destination port in the European segment is calculated based on the fuel consumption prediction value in Table 3

[0096] S43: Energy consumption range setting: The upper and lower limits of energy consumption of the corresponding flight segment are set by the fuel consumption of each flight segment described in S42, so that the value of the fuel consumption of each flight segment is between the upper and lower limits of the energy consumption of the corresponding flight segment; the upper limit of energy consumption of each flight segment is equal to the value of the fuel consumption of the corresponding flight segment multiplied by 1.2, and the lower limit of energy consumption of each flight segment is equal to the value of the fuel consumption of the corresponding flight segment multiplied by 0.8; specifically, the formulas for the upper and lower limits of energy consumption are:

[0097]

[0098]

[0099] Where: M_min is the lower limit of energy consumption; M_max is the upper limit of energy consumption; the value of fuel consumption for each departure or end port in the European segment Both are between M_min and M_max;

[0100] Setting reasonable upper and lower limits of energy consumption based on the predicted value of fuel consumption can help to effectively allocate energy consumption and make it closer to the actual situation. Such energy consumption range setting not only takes into account the uncertainty of model prediction, but also provides a more reliable reference for ship operations, ensuring better control and adjustment of energy consumption levels in actual operations.

[0101] S5: Genetic algorithm optimization: Based on the maximum and minimum values ​​of fuel consumption set in S4, a genetic algorithm is used to minimize the EU carbon emission fines as the goal, and fuel consumption, traditional fuel quality and LNG quality are used as variables to construct an objective function, seek the optimal energy consumption allocation for minimizing the EU carbon emission fines, and obtain the optimal solution of the objective function;

[0102] The objective function is expressed as follows:

[0103] M i2 =491*(M i -M i1 ) / 405

[0104]

[0105]

[0106] Where: M i is the fuel consumption of the i-th flight segment (i=1,2,3,…,n); M i1 is the conventional fuel mass of the i-th flight segment (i=1,2,3,…,n); M i2is the LNG mass of the i-th flight segment (i=1,2,3,…,n); α is the carbon emission factor of the whole life cycle, in gCO2e / gFuel, which represents the greenhouse gas emission of the whole life cycle of unit weight fuel. This factor is related to the fuel type and consumption device; required is the fuel intensity limit; GHGIE actual_i is the annual average GHG intensity of energy use on board ships calculated during the relevant reporting period; Feul_EU_Penalty is the penalty imposed by the EU Maritime Fuel Regulation;

[0107] For LNG, the escape rates of four different fuel consumption devices are different, resulting in multiple different full life cycle carbon emission factors for LNG. When calculating, the corresponding full life cycle carbon emission factor can be used according to the fuel consumption device actually selected for the LNG dual-fuel ship; the reference values ​​of the full life cycle carbon emission factor are shown in Table 4:

[0108] Table 4. Reference values ​​of carbon emission factors for the entire life cycle

[0109] LNG Otto (dual-fuel medium-speed engine) 4.37986382 LNG Otto (dual-fuel low-speed engine) 4.06882274 LNG Diesel (dual-fuel low-speed engine) 3.73556444 LBSI 4.26877772 No Escape 3.6611

[0110] Among them: LNG Otto means a medium-speed Otto cycle engine using liquefied natural gas (LNG) as one of the fuels; the low-speed Otto cycle engine using liquefied natural gas (LNG) as one of the fuels; the LNG Diesel means a low-speed engine using liquefied natural gas (LNG) and diesel as fuels; the LBSI means a lean burn gas engine; LNG dual-fuel ships have the phenomenon of unburned methane escaping when using LNG as fuel, and the non-escaping means that the fuel consumption device adopts an internal combustion engine with zero methane escaping;

[0111] The reference values ​​of the fuel intensity limits are shown in Table 5:

[0112] Table 5. Reference values ​​for fuel intensity limits

[0113] Emission reduction targets 2025 2030 2035 2040 2045 2050 Required greenhouse gas intensity (gCO2e / MJ) 89.3 85.7 77.9 62.9 34.6 18.2

[0114] Wherein: the required greenhouse gas (GHC) intensity is expressed in carbon dioxide equivalent (gCO2e) produced per megajoule (MJ); the emission reduction target is the reduction target of the required greenhouse gas intensity from 2025 to 2050;

[0115] The constraint condition is expressed as follows:

[0116] min i ≤M i ≤max i

[0117] 0≤Mi1 ≤M i

[0118] 0≤M i2 ≤M i

[0119] M i1 +M i2 =M i

[0120] Where: min i is the minimum value of fuel consumption; max i is the maximum value of fuel consumption; M i is the fuel consumption of the i-th flight segment (i=1,2,3,…,n); M i is the fuel consumption of the i-th flight segment (i=1,2,3,…,n); M i1 is the conventional fuel mass of the ith flight segment; M i2 is the LNG mass of the i-th segment (i=1,2,3,…,n); the M i 、M i1 and M i2 is the solution set of the objective function; the min i 、max i The value of fuel consumption is set according to the prediction of the trained artificial neural network model.

[0121] Specifically, in this embodiment, the fuel consumption is required to set an upper limit and a lower limit on the reference value given by the artificial neural network model, and at the same time, it is required that the traditional fuel mass and the LNG mass of the i-th flight segment are less than or equal to the fuel consumption of the i-th flight segment, and the fuel consumption of the i-th flight segment is equal to the sum of the traditional fuel mass and the LNG mass;

[0122] The S52 specifically includes the following steps:

[0123] Step 1: Initialize the population: randomly generate a certain number of individuals as the initial population; each individual contains two elements, namely fuel consumption and traditional fuel quality; specifically, the individuals in this embodiment are represented by vectors as X i =(M i ,M i1 );

[0124] Step 2: Fitness evaluation: Calculate the fitness of each individual in the initial population; the value of the fitness is the value of the objective function f(X i ), indicating the degree of individual superiority or inferiority in the problem;

[0125] Step 3: Selection operation: through a certain selection algorithm, individuals with higher fitness are selected as parents to generate individuals of the next generation; the goal of the selection algorithm is to increase the probability of individuals with higher fitness being selected, thereby retaining excellent solutions;

[0126] Step 4: Crossover operation: select a pair of parent individuals and generate offspring individuals through crossover operation; the crossover operation simulates gene crossover in biology and combines part of the gene information of the two parents to generate a new individual;

[0127] Specifically, in this embodiment, for the intersection problem of two elements, single-point intersection and multi-point intersection are selectively considered;

[0128] Step 5: Mutation operation: perform mutation operation on the offspring individuals and introduce some randomness of mutation to enable more extensive exploration in the search space;

[0129] Specifically, in this embodiment, the mutation operation includes modifying certain elements of an individual and exchanging the positions of elements in the individual;

[0130] Step 6: Replacement operation: replace the individuals with poor fitness in the original population with the newly generated offspring to form a new population;

[0131] Step 7: Repeat iteration: Repeat steps S3-S6 until the stop condition is reached;

[0132] Step 8: Output results: Output the individual with the highest fitness in the population, and then obtain the optimal solution of the objective function;

[0133] Specifically, in this embodiment, when the fuel intensity limit requied=62.9 and the carbon emission factor α of the whole life cycle=3.753556444, the optimization result using the genetic algorithm is shown in Table 6:

[0134] Table 6. Genetic algorithm optimization results

[0135]

[0136] Among them: the optimized carbon emission penalty calculation process is: after obtaining the distribution results of LNG consumption and traditional fuel consumption in the EU section of the departure port or destination port according to the fuel consumption prediction value through the genetic algorithm, the optimized traditional fuel quality and LNG quality are calculated, and the cumulative traditional fuel quality prediction value and the cumulative LNG quality prediction value are respectively substituted into the objective function calculation formula described in S5, where required is queried in Table 5 according to the year of calculation, and α is queried in Table 4 according to the actual usage, and the EU maritime fuel regulation penalty value is calculated, which is the optimized carbon emission penalty;

[0137] The calculation process of the actual value of the carbon emission penalty is as follows: just substitute the actual value of the accumulated traditional fuel mass and the actual value of the accumulated LNG mass into the objective function calculation formula described in S5, where required is queried in Table 5 according to the year of calculation, and α is queried in Table 4 according to the actual usage, and the EU maritime fuel regulation penalty value is calculated, which is the actual value of the carbon emission penalty.

[0138] As shown in Table 6, after optimization, the EU carbon emission fines were reduced from the original 7.47 million euros to 5.38 million euros, a reduction of 27.98%. Therefore, LNG is a more environmentally friendly choice compared with traditional fuels.

[0139] This embodiment achieves accurate prediction and optimal control of fuel consumption of LNG dual-fuel ships whose departure or destination ports are in the European section, provides shipowners with an economical and reasonable operating strategy under the EU's carbon emission restrictions, and helps promote the decarbonization and environmental protection development of the shipping industry.

[0140] Second Group of Embodiments: System for Minimizing EU Carbon Emission Penalties for LNG Dual-Fuel Ships

[0141] This group of embodiments provides a system based on minimizing the EU carbon emission fines for LNG dual-fuel ships, the system comprising a data acquisition module connected in sequence for acquiring data and processing the data to construct a data set, a data processing module for dividing the data set obtained by the data acquisition module according to the acquisition time and obtaining a training set and a test set, a model building and training module for building an artificial neural network model and using the training set and test set of the data processing module to train and test the model respectively, an energy consumption prediction and energy consumption range setting module for simulating the operation of all ships in the European section at the departure port or the destination port in the future period of time using the artificial neural network model trained by the model building and training module to predict the fuel consumption of the ships in each section and set a reasonable upper and lower limit of energy consumption according to the fuel consumption, and a genetic algorithm optimization module for minimizing the EU carbon emission fines based on the upper and lower limits of the fuel consumption set by the energy consumption prediction and energy consumption range setting module and using a genetic algorithm to take minimizing the EU carbon emission fines as a goal and taking fuel consumption, traditional fuel quality and LNG quality as variables to construct an objective function to seek the optimal energy consumption allocation for minimizing the EU carbon emission fines and obtain the optimal solution of the objective function; the ship fuel consumption message data includes: fuel consumption and fuel quality; the fuel quality includes: traditional fuel quality and LNG quality;

[0142] The system described in this embodiment can be used to execute the method of the first group of embodiments.

Claims

1. A method for minimizing EU carbon emission penalties for LNG dual-fuel ships, characterized in that: The steps include: S1: Data collection: Obtain ship dynamic data, ship fuel consumption message data, and meteorological data, and process the data to build a data set; The ship fuel consumption message data includes: fuel consumption and fuel quality; the fuel quality includes: traditional fuel quality and liquefied natural gas quality; S2: Data processing: Divide the data set described in S1 according to the acquisition time to obtain a training set and a test set; S3: Model construction and training: construct an artificial neural network model, and use the training set in S2 to train the model, and use the test set in S2 to test the model; the artificial neural network model learns the relationship between fuel consumption and ship dynamic data and meteorological data; S4: Energy consumption prediction and range setting: Use the artificial neural network model trained in S3 to simulate the operation of all ships whose departure ports or destination ports are in the European section in the future, predict the fuel consumption of ships in each section, and set reasonable upper and lower limits based on the fuel consumption, where the upper and lower limits are the maximum and minimum values ​​of fuel consumption respectively; S5: Genetic algorithm optimization: Based on the maximum and minimum values ​​of fuel consumption set in S4, a genetic algorithm is used to minimize the EU carbon emission fines as the goal, and fuel consumption, traditional fuel quality and LNG quality are used as variables to construct an objective function, seek the optimal energy consumption allocation for the EU carbon emission fine minimization problem, and obtain the optimal solution of the objective function.

2. The method for minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 1, characterized in that: The fuel consumption includes traditional fuel consumption and LNG consumption; the ship dynamic data includes: collection time, location, sailing speed, heading, daily main engine speed, sailing distance, sailing time, and ship draft; the meteorological data includes: wind speed, sea conditions, upload time, longitude and latitude; the wind speed includes wind force; the sea conditions include wave size and surge size; the ship fuel consumption message data also includes reporting time.

3. The method for minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 1, characterized in that: S1 The data set includes: historical data of LNG dual-fuel ships and historical data of its sister LNG dual-fuel ships.

4. The method for minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 2, characterized in that: The processing described in S1 includes: data cleaning, data fusion, data clustering and data encoding; the data cleaning refers to deleting some abnormal values ​​in the ship dynamic data; the data fusion refers to aligning the ship fuel consumption message data with the ship dynamic data according to the reporting time in the ship fuel consumption message data, and matching the meteorological data with the ship dynamic data according to the upload time and longitude and latitude in the meteorological data; the data clustering is to use the K-Means clustering algorithm to divide different loading states according to the ship draft in the ship dynamic data, including: empty, full and half load; the data encoding is the encoding processing of the original three segments of data; the original three segments of data include: ship dynamic data, ship fuel consumption message data, and meteorological data.

5. The method for minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 1, characterized in that: The S3 comprises the following steps: S31: Input feature selection: variables strongly related to fuel consumption in ship dynamic data and meteorological data are used as input features of the model; the variables strongly related to fuel consumption in the ship dynamic data include: sailing speed, daily main engine speed, sailing distance, full load, no load and states between full load and no load; the variables strongly related to fuel consumption in the meteorological data include: wind force and wave size; S32: Output feature setting: The fuel consumption of the ship is used as the output feature of the model; S33: Model training: using the training set to train the artificial neural network model, and learning the nonlinear relationship between fuel consumption and basic ship attributes, ship dynamic data and meteorological data; S34: Model testing: Use the test set to test the artificial neural network model to test whether the model training is complete. If the test fails, return to S33 to continue training.

6. The method for minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 1, characterized in that: The future period of time referred to in S4 is the next year.

7. The method for minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 6, characterized in that: S4 includes the following steps: S41: Route segmentation: segment the routes according to the starting and ending ports of the ships, and count and filter out the routes whose departure ports or destination ports are in the European section; S42: Calculation of energy consumption for each voyage in the next year: For the voyages where the departure port or the destination port selected by S41 is in the European voyage, the artificial neural network model trained by S3 is used to simulate the ship operation status of each voyage in the next year, and calculate the fuel consumption of the corresponding voyage; S43: Energy consumption range setting: The upper and lower limits of energy consumption of the corresponding flight segment are set through the fuel consumption of each flight segment described in S42, so that the fuel energy consumption of each flight segment is between the upper and lower limits of energy consumption of the corresponding flight segment; the upper limit of energy consumption of each flight segment is equal to the value of fuel consumption of the corresponding flight segment multiplied by 1.2, and the lower limit of energy consumption of each flight segment is equal to the value of fuel consumption of the corresponding flight segment multiplied by 0.

8.

8. The method of minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 1, characterized in that: The objective function formula in S5 is: M i2 =491*(M i -M i1 ) / 405 Where: M i is the fuel consumption of the i-th flight segment, i=1,2,3,…,n; M i1 is the conventional fuel mass of the i-th flight segment, i = 1, 2, 3, ..., n; M i2 is the LNG mass of the i-th flight segment, i = 1, 2, 3, ..., n; α is the carbon emission factor of the whole life cycle; required is the fuel intensity limit; GHGIE actual_i is the annual average greenhouse gas intensity of energy use on board calculated during the relevant reporting period; Feul_EU_Penalty is the carbon emission penalty of the EU marine fuel regulations; the constraints of the objective function are: min i ≤M i ≤max i 0≤M i1 ≤M i 0≤M i2 ≤M i M i1 +M i2 =M i Where: min i is the minimum value of fuel consumption; max i is the maximum value of fuel consumption; M i is the fuel consumption of the i-th flight segment, i=1,2,3,…,n; M i1 is the conventional fuel mass of the i-th flight segment, i = 1, 2, 3, ..., n; M i2 is the LNG mass of the i-th segment, i = 1, 2, 3, ..., n; the M i 、M i1 and M i2 is the solution set of the objective function; the min i 、max i The value of fuel consumption is set according to the prediction of the trained artificial neural network model.

9. The method of minimizing EU carbon emission fines for LNG dual-fuel ships according to claim 1, characterized in that: The population optimization of the genetic algorithm described in S5 comprises the following steps: Step 1: Initialize the population: randomly generate a certain number of individuals as the initial population; each individual contains two elements, namely fuel consumption and traditional fuel mass; Step 2: Fitness evaluation: Calculate the fitness of each individual in the initial population; Step 3: Selection operation: through a certain selection algorithm, select individuals with higher fitness in the population as parents to produce individuals of the next generation; Step 4: Crossover operation: select a pair of parent individuals and generate offspring individuals through crossover operation; Step 5: Mutation operation: perform mutation operation on the offspring individuals and introduce some randomness of mutation; Step 6: Replacement operation: replace the individuals with poor fitness in the original population with the newly generated offspring to form a new population; Step 7: Repeat iteration: Repeat steps S3-S6 until the stop condition is reached; Step 8: Output results: Output the individual with the highest fitness in the population, and then obtain the optimal solution of the objective function.

10. A system to minimize EU carbon emission penalties for LNG dual fuel ships, characterized in that It includes a data acquisition module for acquiring data and processing the data to construct a data set, a data processing module for dividing the data set obtained by the data acquisition module according to the acquisition time and obtaining a training set and a test set, a model building and training module for building an artificial neural network model and using the training set and test set of the data processing module to train and test the model respectively, an energy consumption prediction and energy consumption range setting module for using the artificial neural network model trained by the model building and training module to simulate the operation of all ships in the European section at the departure port or the destination port in the future period of time, predict the fuel consumption of the ships in each section and set a reasonable upper and lower limit of energy consumption according to the fuel consumption, and a genetic algorithm optimization module for using the upper and lower limits of fuel consumption set by the energy consumption prediction and energy consumption range setting module and using a genetic algorithm to take minimization of EU carbon emission fines as a goal and use fuel consumption, traditional fuel quality and LNG quality as variables to construct an objective function to seek the optimal energy consumption allocation for minimizing the EU carbon emission fine problem and obtain the optimal solution of the objective function; The ship fuel consumption message data includes: fuel consumption and fuel quality; the fuel quality includes: traditional fuel quality and LNG quality.

Images