Optimization solution-based carbon emission penalty minimization system and method for European leg LNG (Liquefied Natural Gas) dual-fuel ship

Abstract

According to the European leg LNG dual-fuel ship carbon emission penalty minimization method and system based on optimization solution, an energy consumption prediction model based on an artificial neural network (ANN) is established, and ship basic attributes, dynamic data and meteorological data are comprehensively considered, so that energy consumption prediction is more accurate. Through optimization algorithms such as the quasi-Newton method, the optimal combination of the sum of energy consumption, the traditional ship fuel oil quality and the LNG quality is achieved, and the utilization of ship energy is more efficient. By optimizing the energy consumption configuration, the purpose of minimizing the carbon penalty is achieved. According to the method, on the premise that environmental protection laws and regulations are met, the expenditure of carbon penalty can be reduced to the maximum extent, economic benefits are improved, and more flexible, economical and reasonable operation strategies are provided for ship owners. Ship owners can make more intelligent operation decisions conforming to environmental protection standards based on accurate prediction and optimization decisions by using the intelligent system.

Description

Technical Field

[0001] The present invention relates to the field of ship energy consumption, and in particular to a system and method for minimizing carbon emission fines for LNG dual-fuel ships in a European section based on optimization solution. Background Art

[0002] Carbon emissions from shipping are a huge and growing environmental problem. According to a report by the International Maritime Organization (IMO), between 2012 and 2018, the carbon intensity of international shipping decreased by about 11%, but annual greenhouse gas emissions increased to 1.076 billion tons. It is expected that by 2050, as the demand for shipping grows, carbon dioxide emissions will increase further, seriously threatening the global temperature control goals set out in the Paris Agreement.

[0003] To address this problem, IMO adopted the 2023 IMO Ship Greenhouse Gas (GHG) Emission Reduction Strategy in July 2023, requiring international shipping to achieve peak greenhouse gas emissions as soon as possible and achieve net zero emissions before or around 2050. In order to promote the reduction of carbon emissions from ships in Europe, the EU adopted the FuelEU Maritime Regulation on September 22, 2023, which will come into effect on January 1, 2025. The regulation aims to reduce carbon emissions from ships by limiting the greenhouse gas intensity of fuels used by ships and encouraging shipowners to adopt sustainable marine fuels and zero-emission technologies.

[0004] At present, the shipping industry mainly relies on the use of highly polluting fossil fuels, making maritime transport a significant source of greenhouse gas emissions in the EU as a whole. Existing emission reduction methods cannot effectively predict ship energy consumption, thereby achieving intelligent decision-making on carbon emissions of shipping companies. Existing technologies mainly face three problems:

[0005] 1) Technical issues in reducing greenhouse gas emissions: In the existing technology, the shipping industry mainly uses fossil fuels that are more polluting, resulting in prominent carbon emission issues. In order to achieve a significant reduction in greenhouse gas emissions, a feasible method must be found to limit the greenhouse gas intensity of fuels used by ships to a sustainable level. Therefore, the primary goal of this patent is to solve the problem of carbon emissions from ships and to promote shipowners to adopt renewable and low-carbon fuels through technical means to reduce the overall greenhouse gas emissions of maritime transportation.

[0006] 2) Technical issues in optimizing ship energy consumption: Existing ship energy consumption prediction methods are limited by simple statistical models and cannot fully utilize the ship's own characteristics and complex operating environment data. In order to improve the accuracy and reliability of ship energy consumption prediction, an innovative method is needed to combine the basic properties of the ship, meteorological data and navigation data to build a more accurate energy consumption prediction model. Therefore, this patent aims to solve the technical problems of ship energy consumption prediction by introducing a prediction model based on artificial neural network (ANN) to more comprehensively consider the ship's characteristics and operating conditions to improve the accuracy of the prediction.

[0007] 3) Technical issues of minimizing carbon fines: In view of the shortcomings of existing technologies in optimizing ship energy consumption, a system and method are needed to minimize carbon fines while meeting the ship's energy consumption range. To achieve this goal, this patent adopts the quasi-Newton method to find the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality that can minimize fines through optimization algorithms. This technical solution helps shipping companies to minimize the cost of carbon fines while controlling carbon emissions, providing a more economical and environmentally friendly option for ship operations. Summary of the invention

[0008] In order to solve the technical problems that the existing technology cannot effectively predict ship energy consumption, solve the optimal energy consumption combination of ships, minimize carbon penalties, and assist ship decision-making, the present invention proposes a system and method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution. By setting an artificial neural network (ANN) ship energy consumption prediction model in combination with basic ship attributes, meteorological data and navigation data, the optimal energy consumption combination is predicted and solved on the basis of meeting the latest requirements of the EU and IMO for emission reduction in the shipping field, so as to minimize carbon penalties and realize intelligent decision-making on carbon emissions of shipping companies.

[0009] The specific plan is as follows:

[0010] A method to minimize carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution.

[0011] S1: Ship data collection: collect ship AIS dynamic data, MRV message data, ship fuel consumption message data and meteorological data; obtain ship load data based on MRV message; obtain ship operation data based on ship AIS dynamic data and MRV message data;

[0012] S2: Constructing an artificial neural network model: constructing an ANN model based on the ship data described in S1, taking ship operation data, ship load data and meteorological data as input features of the ANN model; setting the output layer to predict the energy consumption of the ship; and using historical ship data to train the ANN model;

[0013] S3: Use the trained ANN model to predict energy consumption and set energy consumption range:

[0014] S31: Energy consumption prediction: Using the ANN model trained in S2, simulate the ship operation in the next year and calculate the total energy consumption of traditional ship fuel and liquefied natural gas (LNG);

[0015] S32: Energy consumption range setting: Set the total energy consumption range as follows: Based on the total energy consumption predicted for the European segment in one year, set the fluctuation range to between 80% and 120% of the total energy consumption, and the fluctuation range is adjustable;

[0016] S4: Optimize ship energy consumption based on the quasi-Newton method: Take the total energy consumption, the quality of traditional ship fuel and the quality of LNG as the input variables of the objective function, and minimize the carbon penalty as the output value of the objective function. Use the quasi-Newton method optimization algorithm to solve the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality that can achieve the minimum penalty.

[0017] Preferably, S1 collects ship AIS dynamic data, ship fuel consumption message data and meteorological data:

[0018] The ship AIS dynamic data includes: the ship's position, speed, heading and ship's operating status;

[0019] The ship fuel consumption message data includes: fuel consumption, fuel quality, and establishment of a historical fuel consumption database;

[0020] The meteorological data are meteorological data related to the ship's operating environment, including: wind speed and sea conditions.

[0021] Preferably, the energy consumption of the ship predicted in S2 includes: energy consumption of traditional ship fuel and liquefied natural gas.

[0022] Preferably, the ANN model constructed in S2 consists of an input layer, four hidden layers and an output layer;

[0023] The input layer contains 7 neurons, corresponding to 7 input features: meteorological data includes: wind speed and wave height; ship operation data includes: main engine speed, speed and sailing distance; ship load data includes: empty and full load;

[0024] In the four hidden layers, the number of neurons is 6, 4, 4, and 4 respectively. The neurons in each layer process the input data through the activation function and gradually extract useful features;

[0025] One neuron in the output layer will output the predicted energy consumption of the ship, i.e., MRV daily fuel consumption (tons / day), based on the high-level features extracted from the hidden layer.

[0026] Preferably, the objective function described in S4 is:

[0027] M2=491*(M-M1) / 405

[0028]

[0029] Constraints:

[0030] min≤M≤max

[0031] 0≤M1≤M

[0032] 0≤M2≤M

[0033] M1+M2=M

[0034] Where M is the total energy consumption, the benchmark value is given by the ANN model, M1 is the mass of traditional fuel, M2 is the mass of LNG, α is the carbon emission factor of the whole life cycle, required is the fuel intensity limit value announced by IMO for that year, GHGIE actual is the average annual GHG intensity of the energy used on board during the relevant reporting period.

[0035] min is the minimum value of the total energy consumption, max is the maximum value of the total energy consumption, and M is the total energy consumption, and the benchmark value is given by the ANN model; the total energy consumption is required to set upper and lower limits of 0≤M1≤M; 0≤M2≤M on the benchmark value given by the ANN model.

[0036] Preferably, the α is a carbon emission factor for the entire life cycle, in units of gCO2e / gFuel, which represents the greenhouse gas emissions per unit weight of fuel for the entire life cycle, and the factor is related to the fuel type and consumption device. For LNG, the escape rates of the four different fuel consumption devices are different, resulting in multiple different carbon emission factors α for the entire life cycle of LNG. In the calculation, the corresponding carbon emission factor α for the entire life cycle can be used according to the fuel consumption device actually selected for the LNG dual-fuel ship.

[0037] Preferably, the steps of the quasi-Newton optimization algorithm in S4 are:

[0038] S41: Initialization: Select the initial point x0 = (M, M1, M2) and calculate the gradient of the objective function

[0039] S42: Select an initial matrix: initialize a positive definite matrix B0;

[0040] S43: Iteration: For the kth iteration, calculate the search direction p k : Determine the step size α through one-dimensional search k ; Update parameter: x k+1 =xk +α k p k ;

[0041] x k+1 is the combination of (M, M1, M2) after the K+1th iteration; x k is the combination of (M, M1, M2) after the Kth iteration; α k is a one-dimensional search to determine the step size; B k is the positive definite pair matrix after the Kth iteration;

[0042] S44: Update matrix: Calculate s k =x k+1 -x k and Then update the matrix B k :

[0043]

[0044] B k+1 is the positive definite pair matrix after the K+1th iteration; s k is x k+1 and x k The vector difference of y k It is the difference between the gradient value of the objective function after the k+1th iteration and the gradient value of the objective function after the kth iteration;

[0045] S45: Convergence judgment: judging whether the stop condition is met: the stop condition includes: the gradient is less than a set threshold or the parameter change is less than a set threshold;

[0046] S46: If the stop condition is met, output the current solution x k+1 As the optimal solution; otherwise, return to S43.

[0047] A carbon emission penalty minimization system for LNG dual-fuel ships in the European section based on optimization solution, including:

[0048] Ship data collection module: used to collect ship AIS dynamic data, ship fuel consumption message data and meteorological data;

[0049] ANN artificial neural network model module: including: setting input features and output layer ANN model submodule: including: ANN model input features constructed by basic ship attributes, ship AIS dynamic data, ship fuel consumption message data and meteorological data; using the predicted ship energy consumption as the ANN model output layer; and training submodule for training the ANN model using historical ship data;

[0050] Energy consumption prediction and energy consumption range setting module: used to use the trained ANN model to predict ship energy consumption and calculate the total energy consumption of traditional ship fuel and liquefied natural gas (LNG). It also includes an energy consumption range setting submodule based on setting the total energy consumption range:

[0051] Ship energy consumption optimization module: used to optimize ship energy consumption: including: an objective function constructed based on the quasi-Newton optimization algorithm, with the total energy consumption, traditional ship fuel quality and LNG quality as the input variables of the objective function, and minimizing carbon penalties as the output value of the objective function, to solve the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality that can achieve the minimum penalty.

[0052] Beneficial effects:

[0053] The present invention proposes a method and system for minimizing carbon emission fines for LNG dual-fuel ships in the European section based on optimization solution. By establishing an energy consumption prediction model based on artificial neural network (ANN), the basic properties, dynamic data and meteorological data of the ship are fully considered, so that the energy consumption prediction is more accurate. This accurate prediction can guide shipping companies to adopt more environmentally friendly renewable and low-carbon fuels, thereby significantly reducing greenhouse gas emissions from ships, and helping to meet international and EU environmental protection requirements. First, the present invention significantly improves energy efficiency: through optimization algorithms such as the quasi-Newton method, the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality is achieved, making the utilization of ship energy more efficient. The optimized energy configuration not only reduces operating costs, but also complies with environmental protection regulations, and promotes the development of the shipping industry in a more energy-saving and sustainable direction. Secondly, the minimization of carbon penalty costs: by optimizing the energy consumption configuration, the present invention achieves the goal of minimizing carbon penalties. Shipping companies can minimize the expenditure of carbon penalties on the premise of meeting environmental protection regulations, improve economic benefits, and provide shipowners with a more flexible and economically reasonable operating strategy. Third, technological innovation and intelligent decision support: This invention introduces advanced technologies such as artificial neural network (ANN) to innovatively solve the problem of ship energy consumption prediction and optimal configuration. Ship owners can use this intelligent system to make more intelligent and environmentally friendly operational decisions based on accurate predictions and optimal decisions.

[0054] In summary, this patent optimizes the technical problems existing in the prior art, and at the same time brings significant beneficial effects in reducing greenhouse gas emissions, improving energy efficiency and minimizing carbon penalty costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 Flowchart of a method for minimizing carbon emission penalties for LNG dual-fuel ships on the European voyage based on optimization solution.

[0056] Figure 2 A structural diagram of a system for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution.

[0057] Figure 3 ANN energy consumption prediction model test results in the embodiment are shown in the figure. DETAILED DESCRIPTION

[0058] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0059] like Figure 1 As shown in Figure 1, a method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution is presented.

[0060] S1: Ship data collection: collect ship AIS dynamic data, MRV message data, ship fuel consumption message data and meteorological data; obtain ship load data based on MRV message; obtain ship operation data based on ship AIS dynamic data and MRV message data;

[0061] S2: Constructing an artificial neural network model: constructing an ANN model based on the ship data described in S1, taking ship operation data, ship load data and meteorological data as input features of the ANN model; setting the output layer to predict the energy consumption of the ship; and using historical ship data to train the ANN model;

[0062] S3: Use the trained ANN model to predict energy consumption and set energy consumption range:

[0063] S31: Energy consumption prediction: Using the ANN model trained in S2, simulate the ship operation in the next year and calculate the total energy consumption of traditional ship fuel and liquefied natural gas (LNG);

[0064] S32: Energy consumption range setting: Set the total energy consumption range as follows: Based on the total energy consumption predicted for the European segment in one year, set the fluctuation range to between 80% and 120% of the total energy consumption, and the fluctuation range is adjustable;

[0065] S4: Optimize ship energy consumption based on the quasi-Newton method: Take the total energy consumption, the quality of traditional ship fuel and the quality of LNG as the input variables of the objective function, and minimize the carbon penalty as the output value of the objective function. Use the quasi-Newton method optimization algorithm to solve the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality that can achieve the minimum penalty.

[0066] The data sources used include AIS data provided by China Ocean Shipping Group (COSCO), MRV data, and meteorological data provided by the Shanghai Meteorological Bureau. These three data sets are merged into a data set for dual-energy ship energy consumption modeling. The data set was collected from January 17, 2021 to September 10, 2023. The LNG dual-fuel ship-Yuan Ruiyang will be taken as an example.

[0067] (1) Yuanruiyang Energy Consumption Prediction Model Based on ANN

[0068] The extracted data set is divided into input variables and output variables for predictive analysis. The input variables include wind force, wave size, sailing speed, daily main engine speed, sailing distance, full load, no load, and the state between full load and no load, a total of 7 characteristic variables. Among them, full load and no load are obtained by the ship's draft (half load). The seven input variables can be divided into four categories, representing the ship's load state, engine operating state, ship speed, and meteorological conditions. Subsequently, the daily fuel consumption of the ship is used as the output variable, as shown in Table 1.

[0069] Table 1

[0070]

[0071] After data cleaning, data fusion, data clustering and encoding of the original three segments of data, point-by-point fusion data similar to AIS data was obtained. Some of the data are shown in Table 2.

[0072] Table 2

[0073]

[0074] Use Distance, Wind_val, Wave_val, Stream_val, Average_Engine_Speed_Daily, No_load, and Full_load as the input variables of the model, and Daily_Fuel_Mrv as the output variable of the model to build the ANN model.

[0075] In order to increase the amount of training data for the model, the data of sister ships built with the same drawings as Yuan Ruiyang are selected to form the training data of the ANN energy consumption prediction model. Yuan Ruiyang is an LNG dual-fuel ship put into use at the end of February 2022, and began to record LNG usage on November 7, 2022. Therefore, the data before April 1, 2022 is used as the training data, and the data from April 7, 2022 to November 7, 2022 is used as the test set. The results are as follows Figure 3 shown.

[0076] The figure shows the comparison between the energy consumption prediction and actual energy consumption of a sister ship of Yuanruiyang. The predicted value in red is more consistent with the actual value in black, and the cumulative error is about 1.44%, which shows that the energy consumption prediction model based on ANN can accurately predict the daily energy consumption of ships. It can effectively predict the total dual-fuel energy consumption of Yuanruiyang after it starts using LNG.

[0077] (2) Optimize the usage ratio of LNG and traditional ship fuel based on the quasi-Newton method to minimize the penalty.

[0078] The energy consumption forecast results of Yuanruiyang from November 7, 2022 to September 10, 2023 based on ANN are shown in Table 3:

[0079] Table 3

[0080] Date Daily_Fuel_MRV_Actual Daily_Emi_Lng_Actual Predict_Energy_Consumption 1 2022-11-07 27.324876 5 31.634056 2 2022-11-08 26.124876 5.1 30.309494 3 2022-11-09 24.331095 4.7 26.215759 4 2022-11-10 41.237313 5.4 43.558018 ... ... ... ... ... 215 2023-09-07 33.824876 5.1 35.494778 216 2023-09-08 31.131095 5.1 33.045116 217 2023-09-09 27.924876 5.2 30.339142 218 2023-09-10 11.246095 5.6 13.349802

[0081] The first two columns are the daily consumption of traditional fuel and LNG after the actual LNG usage record, and the last column is the daily energy consumption predicted by the ANN model, which is based on the same standard as traditional fuel. The mass of LNG can be converted from the total energy consumption and the mass of traditional fuel. The formula is as follows

[0082] M_Lng=491(M-M_Mrv) / 405

[0083] Where M is the mass of total energy consumption, M_Mrv is the energy consumption of traditional ships, and M_lng is the consumption of Lng. According to the Predict_Energy_Consumption in the above table, the total predicted energy consumption for this time period is calculated. Since it is a predicted value, in order to improve the reference of the model, a certain fluctuation range is set, namely M_min and M_max.

[0084]

[0085] Then, M is between M_min and M_max

[0086] Set the minimum penalty as the objective function of the model

[0087]

[0088] in

[0089]

[0090] The constraints are as follows

[0091] M_min <M<M_max

[0092] M1≤M

[0093] M2≤M

[0094] M1+M2=M

[0095] sM i ≥0,i=1,2

[0096] Where M is the total energy consumption, the benchmark value is given by the ANN model, M1 is the mass of traditional fuel, M2 is the mass of LNG, α is the carbon emission factor of the whole life cycle, required is the fuel intensity limit value announced by IMO for that year, GHGIE actual is the average annual GHG intensity of the energy used on board during the relevant reporting period.

[0097] Taking requied = 89.3 and alpha = 3.753556444 as an example, the quasi-Newton method is used to solve the problem. The solution results are shown in Table 4 below:

[0098] Table 4

[0099] Total energy consumption Quality of HFO LNG quality fine Optimization results 9242.49128475 8353.65839634 1077.572711627925 0.198 Actual Results 9197.450549725912 7775.248920398009 1724.2 396211.7565924728

[0100] It can be seen from the table that the total energy consumption in the optimization result is not much different from the total energy consumption in the actual result, but the use ratio of HFO is increased and the use ratio of LNG is reduced, which reduces the fine from US$396211.7565924728 to US$0.198, greatly reducing the shipowner's expenses and helping shipping companies save operating costs.

[0101] α is the carbon emission factor of the whole life cycle, with the unit of gCO2e / gFuel, which indicates the greenhouse gas emission of the whole life cycle of the unit weight of fuel. This factor is related to the type of fuel and the consumption device. For LNG, the escape rates of the four different fuel consumption devices are different, resulting in multiple different carbon emission factors α for the whole life cycle of LNG. In the calculation, the corresponding carbon emission factor α for the whole life cycle can be used according to the fuel consumption device actually selected for the LNG dual-fuel ship. Table 5 shows the reference of α values.

[0102] Table 5

[0103] LNG Otto(dual fuel medium speed) 4.37986382 LNG Otto(dual fuel slow speed) 4.06882274 LNG Diesel(dual fuel slow speed) 3.73556444 LBSI 4.26877772 No Escape 3.6611

[0104] like Figure 2 As shown, a carbon emission penalty minimization system for LNG dual-fuel ships in the European section based on optimization solution includes:

[0105] Ship data collection module: used to collect ship AIS dynamic data, ship fuel consumption message data and meteorological data;

[0106] ANN artificial neural network model module: including: setting input features and output layer ANN model submodule: including: ANN model input features constructed by basic ship attributes, ship AIS dynamic data, ship fuel consumption message data and meteorological data; using the predicted ship energy consumption as the ANN model output layer; and training submodule for training the ANN model using historical ship data;

[0107] Energy consumption prediction and energy consumption range setting module: used to use the trained ANN model to predict ship energy consumption and calculate the total energy consumption of traditional ship fuel and liquefied natural gas (LNG). It also includes an energy consumption range setting submodule based on setting the total energy consumption range:

[0108] Ship energy consumption optimization module: used to optimize ship energy consumption: including: an objective function constructed based on the quasi-Newton optimization algorithm, with the total energy consumption, traditional ship fuel quality and LNG quality as the input variables of the objective function, and minimizing carbon penalties as the output value of the objective function, to solve the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality that can achieve the minimum penalty.

[0109] Finally, it should be noted that the above is only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred arrangement scheme, a person skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution, characterized in that: S1: Ship data collection: collect ship AIS dynamic data, MRV message data, ship fuel consumption message data and meteorological data; obtain ship load data based on MRV message; obtain ship operation data based on ship AIS dynamic data and MRV message data; S2: Building an ANN artificial neural network model: Building an ANN model based on the ship data described in S1, using ship operation data, ship load data and meteorological data as input features of the ANN model; Set the output layer to predict the energy consumption of the ship; and Use historical ship data to train the ANN model; S3: Use the trained ANN model to predict energy consumption and set energy consumption range: S31: Energy consumption prediction: Using the ANN model trained in S2, simulate the ship operation in the next year and calculate the total energy consumption of traditional ship fuel and liquefied natural gas (LNG); S32: Energy consumption range setting: Set the total energy consumption range as follows: Based on the total energy consumption predicted for the European segment in one year, set the fluctuation range to between 80% and 120% of the total energy consumption, and the fluctuation range is adjustable; S4: Optimize ship energy consumption based on the quasi-Newton method: Take the total energy consumption, the quality of traditional ship fuel and the quality of LNG as the input variables of the objective function, and minimize the carbon penalty as the output value of the objective function. Use the quasi-Newton method optimization algorithm to solve the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality that can achieve the minimum penalty.

2. A method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution according to claim 1, characterized in that: S1 collects ship AIS dynamic data, ship fuel consumption message data and meteorological data: The ship AIS dynamic data includes: the ship's position, speed, heading and ship's operating status; The ship fuel consumption message data includes: fuel consumption, fuel quality, and establishment of a historical fuel consumption database; The meteorological data are meteorological data related to the ship's operating environment, including: wind speed and sea conditions.

3. A method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution according to claim 1, characterized in that: The energy consumption of ships predicted in S2 includes the energy consumption of traditional ship fuel and liquefied natural gas.

4. A method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution according to claim 1, characterized in that: The network structure of the ANN model constructed in S2 is: The ANN model consists of an input layer, four hidden layers and an output layer; The input layer contains 7 neurons, corresponding to 7 input features: meteorological data includes: wind speed and wave height; ship operation data includes: main engine speed, speed and sailing distance; ship load data includes: empty and full load; In the four hidden layers, the number of neurons is 6, 4, 4, and 4 respectively. The neurons in each layer process the input data through the activation function and gradually extract useful features; One neuron in the output layer will output the predicted energy consumption of the ship, i.e., MRV daily fuel consumption (tons / day), based on the high-level features extracted from the hidden layer.

5. A method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution according to claim 1, characterized in that: The objective function described in S4 is: M2=491*(M-M1) / 405 Constraints: min≤M≤max 0≤M1≤M 0≤M2≤M M1+M2=M Where M is the total energy consumption, the benchmark value is given by the ANN model, M1 is the mass of traditional fuel, M2 is the mass of LNG, α is the carbon emission factor of the whole life cycle, required is the fuel intensity limit value announced by IMO for that year, GHGIE actual is the average annual GHG intensity of the energy used on board during the relevant reporting period. min is the minimum value of total energy consumption, max is the maximum value of total energy consumption, M is the total energy consumption, and the benchmark value is given by the ANN model; the total energy consumption is required to set the upper and lower limits of the benchmark value given by the ANN model within the range of 0≤M1≤M 0≤M2≤M.

6. A method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution according to claim 5, characterized in that: The α is the carbon emission factor for the entire life cycle and is selected according to the fuel consumption device actually selected for the LNG dual-fuel ship.

7. A method for minimizing carbon emission penalties for LNG dual-fuel ships in the European section based on optimization solution according to claim 1, characterized in that: The steps of the quasi-Newton optimization algorithm in S4 are: S41: Initialization: Select the initial point x0 = (M, M1, M2), calculate the gradient of the objective function S42: Select an initial matrix: initialize a positive definite matrix B0; S43: Iteration: For the kth iteration, calculate the search direction pk: Determine the step size α through one-dimensional search k ; Update parameter: x k+1 =x k +α k P k ; x k+1 is the combination of (M, M1, M2) after the K+1th iteration; x k is the combination of (M, M1, M2) after the Kth iteration; α k is a one-dimensional search to determine the step size; B k is the positive definite pair matrix after the Kth iteration; S44: Update matrix: Calculate s k =x k+1 -x k and Then update the matrix B k : B k+1 is the positive definite pair matrix after the K+1th iteration; s k is x k+1 and x k The vector difference of y k It is the difference between the gradient value of the objective function after the k+1th iteration and the gradient value of the objective function after the kth iteration; S45: Convergence judgment: judging whether the stop condition is met, wherein the stop condition includes: the gradient is less than a set threshold or the parameter change is less than a set threshold; S46: If the stop condition is met, output the current solution x k+1 As the optimal solution; otherwise, return to S43.

8. A carbon emission penalty minimization system for LNG dual-fuel ships in the European section based on optimization solution according to any one of claims 1-7, characterized in that: include: Ship data collection module: used to collect ship AIS dynamic data, ship fuel consumption message data and meteorological data; ANN artificial neural network model module: including: setting input features and output layer ANN model submodule: including: ANN model input features constructed by basic ship attributes, ship AIS dynamic data, ship fuel consumption message data and meteorological data; using the predicted ship energy consumption as the ANN model output layer; and training submodule for training the ANN model using historical ship data; Energy consumption prediction and energy consumption range setting module: used to use the trained ANN model to predict ship energy consumption and calculate the total energy consumption of traditional ship fuel and liquefied natural gas (LNG). It also includes an energy consumption range setting submodule based on setting the total energy consumption range: Ship energy consumption optimization module: used to optimize ship energy consumption: including: an objective function constructed based on the quasi-Newton optimization algorithm, with the total energy consumption, traditional ship fuel quality and LNG quality as the input variables of the objective function, and minimizing carbon penalties as the output value of the objective function, to solve the optimal combination of total energy consumption, traditional ship fuel quality and LNG quality that can achieve the minimum penalty.

Images