A digital twin driven emission prediction and optimization method for marine diesel engines

Abstract

The application discloses a kind of digital twin driven marine diesel engine emission prediction and optimization method, comprising the following steps: (1) build comprehensive performance mechanism model, to obtain twin data by the characterization and verification of comprehensive performance mechanism model;(2) design calibration module based on deep learning algorithm, with real data and twin data as input, output calibrated data by the training of model;(3) build combustion stage mechanism module, optimize intake duct structure by the simulation of oil-gas mixing characteristics and combustion efficiency by combustion mechanism model, improve mixing uniformity;(4) build data-driven model, fit the relationship between turbulent kinetic energy index and emission;(5) the above-mentioned comprehensive performance mechanism model, calibration module, combustion stage mechanism model and data-driven model fusion technology system as fitness function, find optimal injection control parameter based on particle swarm optimization algorithm, formulate optimization control strategy suitable for different working conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to digital twin technology, and in particular to a digital twin driven marine diesel engine emission prediction and optimization method. BACKGROUND

[0002] Under the objective factors of strict international emission standards and rising fuel costs, energy saving and emission reduction has become an important guideline for the engine industry. With the continuous improvement of China's energy saving and emission reduction indicators and the increasingly stringent trend of emission regulations, the demand for low-emission and high-efficiency new internal combustion engines has increased dramatically. In the era of new energy revolution, new energy technology is in a period of rapid development, but there are still some problems to be solved in energy storage, use, endurance, power, stability and reliability. In the next few decades, diesel engines will continue to shine in the field of engineering machinery and military equipment until the arrival of mature new energy power devices. During this period, marine diesel engines need to constantly update technology to meet the requirements of energy saving and emission reduction, low carbon and environmental protection under the background of the times.

[0003] With the enhancement of global environmental awareness and the increasingly stringent emission standards, the shipping industry is facing dual pressures of reducing emissions and improving fuel efficiency. As the main power source of ships, the emission problem of marine diesel engines has become a research hotspot. Traditional emission control methods, such as hardware upgrades and emission aftertreatment technology, have been effective, but often have high costs and are difficult to monitor and adjust emissions in real time. In the field of computer data regression analysis, artificial neural networks can reasonably predict different data after good training and show superior performance. However, general deep learning methods usually require a long time to train and a large amount of historical data. To address this issue, digital twin technology, as a new digital means, builds a virtual model of physical entities to achieve real-time monitoring, simulation and optimization of complex systems. Applying digital twin technology to marine diesel engines can not only accurately predict engine emissions, but also find ways to optimize operating parameters through simulation and data analysis, thereby minimizing emissions. SUMMARY

[0004] The present application aims at the problems existing in the prior art, and provides a digital twin driven marine diesel engine emission prediction and optimization method: first, a comprehensive performance mechanism model is constructed based on Simulink to describe the working principle and internal operation mechanism of the diesel engine, and the operating parameters of the diesel engine are obtained, such as in-cylinder pressure, in-cylinder temperature and CO. The model adopts the idea of decomposition and combination, and decomposes the diesel engine system into a plurality of independent and interrelated subsystems, so as to be more explanatory. Secondly, a calibration module is constructed, and a fusion technology system of calibrating the comprehensive performance mechanism model based on deep learning is proposed. The module has the characteristics of non-invasive data monitoring, and the comprehensive performance mechanism model is calibrated through the deep learning algorithm, the generalization ability of the model is improved, and a series of problems of the traditional sensor monitoring method are solved. Then, a combustion stage mechanism model is constructed to solve the problem that the comprehensive mechanism model cannot reflect the heat conduction of the diesel engine combustion process. The model deeply explores the influence of in-cylinder temperature distribution and oil-gas mixing characteristics on combustion efficiency, and improves the uniformity and flow characteristics of oil-gas mixing by optimizing the structure of the intake pipe. On this basis, a data-driven model is established, and the relationship between the turbulent kinetic energy index and the emission is further fitted by using the deep learning algorithm, so as to understand the internal relationship between the high-fidelity deduction of the combustion stage and the emission. Finally, from the perspective of optimization control strategy, the target of reducing pollutant emission is taken, and based on the particle swarm optimization algorithm, the control strategy that makes the emission lowest is found. These strategies take the comprehensive performance mechanism model, the calibration module, the combustion stage mechanism model and the data-driven model fusion technology model constructed as the fitness function to optimize the injection control parameters, so as to realize the goal of more sufficient combustion and lower pollutant emission.

[0005] The technical scheme of the present application is as follows:

[0006] (1) A comprehensive performance mechanism model is constructed to describe the working principle and internal operation mechanism of the diesel engine, and the operating parameters of the diesel engine are obtained, including in-cylinder pressure, in-cylinder temperature, CO and the like in the combustion stage. The present application adopts the idea of decomposition and combination, and decomposes the diesel engine system into a series of independent and interrelated subsystems, including a compressor submodule, a intercooler submodule, an intake pipe submodule, a cylinder submodule, an exhaust pipe submodule, a turbine submodule, a supercharger rotor submodule and an injection pump submodule. The present application considers the interaction and influence between different subsystems, and is more explanatory.

[0007] (2) Construct a calibration module, and propose a fusion technology system based on deep learning optimization of the calibration comprehensive performance mechanism model, which also has the characteristics of non-invasive data monitoring. The mechanism model can be explained, but due to the fixed formula, the generalization ability is insufficient, and it cannot cope with complex working conditions, so the present application proposes a calibration method, which uses a deep learning algorithm to calibrate the comprehensive performance mechanism model, improves the generalization ability of the comprehensive performance mechanism model, and solves the problems of high cost, insufficient economy and short service life of the traditional sensor monitoring method.

[0008] (3) Construct a combustion stage mechanism model, and the main factor of high pollutant emission is insufficient combustion. In order to solve the problem that the comprehensive mechanism model cannot reflect the heat conduction of the diesel engine combustion process, the present application combines the combustion stage mechanism model, including the intake manifold, the resonance cavity, the intake manifold, the oil injector and the like, and deeply explores the influence of the in-cylinder temperature distribution and the oil-gas mixing characteristics on the combustion efficiency. First, the output of the calibrated mechanism model, including: injection pressure, inlet and outlet pressure, in-cylinder temperature, is used as the inlet and outlet conditions, wall conditions and initial conditions of the combustion stage mechanism model. Secondly, the turbulent kinetic energy is used as an index of the uniformity of oil-gas mixing, and the structure of the intake pipe is optimized, aiming at improving the uniformity and flow characteristics of oil-gas mixing.

[0009] (4) Establish a data-driven model. In order to further understand the internal relationship between high-fidelity deduction of the combustion stage and emission, based on the above combustion process mechanism model, the relationship between the turbulent kinetic energy index and the emission is further fitted using a deep learning algorithm.

[0010] (5) Finally, from the perspective of optimization control strategy, the optimal strategy is found to make the combustion more sufficient and the pollutant emission lower. The above-mentioned comprehensive performance mechanism model, calibration module, combustion stage mechanism model and data-driven model fusion technology model are used as the fitness function, and the least emission is taken as the target, an artificial intelligence optimization algorithm is designed, the optimal value of the injection control parameters such as main injection timing, pre-injection timing and injection quantity under different conditions is found, the control strategy is optimized, and the control strategy with the lowest emission is found.

[0011] Further, the pre-processing step of the historical data in step (2) is as follows:

[0012] (1-1) The historical data is cleaned, and the abnormal and repeated data is deleted, and the missing data is filled by using the difference method. The interpolation formula is as follows:

[0013] l t =(l t+1 +l t-1 ) / 2

[0014] In the formula, l t represents the missing value at time t; l t+1is the normal eigenvalue at time t+1; l t-1 is the normal eigenvalue at time t-1. That is, the mean value of the time point before and after the missing value is taken as the interpolation.

[0015] (1-2) Use variance filtering method to remove engine operating parameters with variance of 0.

[0016] Further, the parameter setting of Bilstm in step (3) is: the number of BiLstm model layers is 3, dropout=0.5, hiden size=400, input size=64. The setting variables for the CNN model are: channel=3, stride=1, padding=1, and the activation function is mean square error.

[0017] (2-1) First, for each time step t, the input gate decides how much new information should be added to the cell state, and the output of the input gate i t is calculated by the following formula:

[0018] i t =σ(w i [h t-1 ,x t ])+b i

[0019] At the same time, use the tanh activation function to add new candidate memory cells:

[0020] c' t =tanh(w c [h t-1 ,x t ]+b c )

[0021] Where, σ is the Sigmoid activation function, w i , w c are the weights of i t , c t respectively; b i , b c are the bias terms of i t , c t , respectively, x t is the input of the input sequence at time step t, that is, the input, h t-1 is the hidden state of the previous time step t-1.

[0022] (2-2) The forget gate decides which information in the new report state should be forgotten or retained. The output of the forget gate f t is calculated by the following formula:

[0023] ft =σ(w f [h t-1 ,x t ]+b f )

[0024] (2-3) Next, use the output i of the input gate t and the output f of the forget gate t To update cell state c t :

[0025] c t =f t c t-1 +i t c' t

[0026] (2-4) Output gate according to h t-1 x t Calculate output information o t The tanh activation function combines the output gate information o t Get the current hidden layer state

[0027] o t =σ(w o [h t-1 ,x t ]+b o )

[0028] h t =o t *tanh(c t )

[0029] In the formula, w o b represents the output gate weights; o This is the output gate bias term.

[0030] Hidden layer state h t Transmitted to the next time step, while cell state c t This information is also passed to the next step to be retained as long-term memory. At this point, the neurons inside the LSTM have completed one computational cycle.

[0031] Specifically, the formula for calculating the mean squared error loss function is as follows:

[0032]

[0033] In the formula, N is the number of samples, y is the true value, and a is the predicted value. The formula for calculating the predicted value a is:

[0034] a=f(z=f(w*x+b0)

[0035] In the formula, x is the input, w is the weight, b is the bias term, and f is the activation function.

[0036] Further, the step (3) Attention hidden layer state value:

[0037] e i =tanh(Wh i +b)

[0038]

[0039] In the formula: W and b are weight and threshold value respectively; h i is the hidden unit state value output by the CNN, the score e i of each hidden unit state value is obtained, the score of the attention mechanism is normalized, finally e i and a i are weighted and summed to obtain the hidden layer state value c t .

[0040] Further, in the step (4), when calculating the turbulent flow model, the standard k-ε model is selected, and when solving the velocity-pressure correlation model, the SIMPLE algorithm is selected.

[0041] Further, in the step (5), the particle optimization process is:

[0042]

[0043] In the formula: i is 1, 2...N; N is the total number of particles in the population; t is the iteration number of particles; omega is the weight; v is the speed of each particle in the population; r1, r2 are random numbers; x i is the current position of the i particle; is the current particle individual optimal position of the i particle; and is the optimal position of the particle population; c1, c2 are learning factors.

[0044] The beneficial effects of the present application are:

[0045] Compared with the prior art, the present application has the following advantages: the present application discloses a digital twin driven marine diesel engine emission prediction and optimization method. The digital twin system is composed of a one-dimensional physical model and a calibration module, a three-dimensional physical model, and a comprehensive diesel engine digital twin system is constructed. This system can not only accurately simulate the emission of diesel engine under different working conditions, but also improve the accuracy and reliability of the model through the non-invasive method of the calibration module, solve the problem of low accuracy of one-dimensional physical model and heat conduction. At the same time, based on the high-fidelity data produced by digital twin, the emission prediction and multi-angle emission optimization are realized, including the optimization of intake pipe structure and injection timing, which improves the combustion efficiency and emission control ability. In addition, the real-time big data of diesel engine is used for online model updating, so that the digital twin model can adapt to various complex working conditions in time, and the accuracy and practicality of emission prediction are improved. In summary, the present application brings a comprehensive and accurate solution to the field of diesel engine emission prediction and optimization, and has important technical and economic value. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 is a calibration module structure diagram of the present application;

[0047] Figure 2 is a NOx emission experiment diagram before and after optimization of each working condition of the present application X

[0048] Figure 3 is a flow chart of the present application. DETAILED DESCRIPTION

[0049] In order to more clearly understand the purposes, features and advantages of the present application, the present application will be further described below in combination with the drawings and examples. In the following description, many specific details are set forth in order to fully understand the present application, however, the present application can also be implemented in other ways different from those described herein, therefore, the present application is not limited to the specific examples disclosed below.

[0050] The present application discloses a digital twin driven marine diesel engine emission prediction and optimization method, comprising:

[0051] (1) Construct a comprehensive performance mechanism model to describe the working principle and internal operation mechanism of the diesel engine, and obtain the operating parameters of the diesel engine, including in-cylinder pressure, in-cylinder temperature, CO, etc. during the combustion stage. The present application adopts the idea of decomposition and combination, and decomposes the diesel engine system into a series of independent and interrelated subsystems, including compressor sub-module, intercooler sub-module, intake pipe sub-module, cylinder sub-module, exhaust pipe sub-module, turbine sub-module, supercharger rotor sub-module and fuel injection pump sub-module. The present application considers the interaction and influence between different subsystems, and is more explanatory. ​

[0052] (1-1) Fuel Injection Pump Submodule

[0053] The fuel injection pump sub-model obtains the single-cylinder cycle fuel injection quantity based on the diesel engine speed and the fuel injection pump rack position output by the governor. The relationship between the single-cylinder cycle fuel injection quantity and the fuel injection pump rack position and diesel engine speed is obtained based on the speed characteristic curve and load characteristic curve of the fuel injection pump.

[0054] g c =f(F r ,n s )

[0055] Where g c It is the fuel injection pump's single-cylinder circulating fuel supply volume (kg / cyl).

[0056] (1-2) Diesel engine cylinder module

[0057] The cylinder charging efficiency is only a function of the diesel engine speed; therefore, its change with speed can be approximated by a semi-empirical formula:

[0058]

[0059] Where n s This refers to the diesel engine speed, measured in r / min. It is the diesel engine speed used to determine the optimal timing, in r / min; η v It refers to the cylinder charging efficiency; It is the optimal timing for cylinder charging efficiency; B is a constant, which is 0.014 for a four-stroke diesel engine.

[0060] For a four-stroke diesel engine, the airflow through the intake valve can be divided into two parts: intake volume and scavenging volume. We define the charging mass flow rate q1 (kg / s), the scavenging mass flow rate q2 (kg / s), and the total air mass flow rate q... a The concept of (kg / s):

[0061] q a =q1+q2

[0062] The inflation flow rate can be obtained using the following formula:

[0063]

[0064] Where p in Intake manifold air pressure, Pa; T in is the average air temperature inside the intake manifold, K; Ri is the intake-side gas constant, taken as 286.846 J / kg·K; V is the diesel engine cylinder displacement, m. 3 S is the stroke coefficient, which is taken as 2 for a four-stroke diesel engine. The scavenging air mass flow rate can be obtained by the following formula:

[0065]

[0066] When p out / p in When >0.98, When p out / p in When ≤0.98,

[0067] S q It is the average value of the flow cross section during scavenging, m 3 ;k i The isentropic index of the intake gas is taken as 1.4; the average exhaust temperature of the engine is calculated according to the first law of thermodynamics of in-cylinder processes.

[0068] q f H u -L w -N e =(q a +q f C Pe T out -q a C Pi T in

[0069] Where q f The amount of fuel injected into the cylinder per unit time, expressed in kg / s; H u It refers to the lower calorific value of fuel oil, measured in J / kg; L w It is the amount of heat removed by cooling water per unit time, measured in J / s; N. e This refers to the effective power of a diesel engine, measured in watts (W). Pe It is the specific heat capacity at constant pressure of the intake working fluid, measured in J / (kg·K); C Pi It is the specific heat capacity at constant pressure of the exhaust working fluid, measured in J / (kg·K); T out It is the average exhaust temperature, in Kelvin (K).

[0070] Diesel engine effective power:

[0071] N e =n e q f H u

[0072] The heat removed by the cooling water is:

[0073] L w =δ w q f H u

[0074] Where ne is the diesel engine estimated effective efficiency; δ w is the cooling loss percentage. The above three equations can be combined to get:

[0075]

[0076] where T cool is the outlet temperature of the intercooler, in K. For diesel engines, the value of varies usually between 0.94 and 1.0, so it can be considered approximately equal to 1, and the above equation can be simplified to:

[0077]

[0078] The engine effective thermal efficiency n e is mainly affected by q a / q f , and the cooling loss δ w is mainly affected by q a / q f and the engine speed n s . Let the engine exhaust temperature be K T =(H u / C Pe )(1-n e -δ w ) = f(q a / q f , n s ), and finally the average engine exhaust temperature is:

[0079]

[0080] (1-3) Exhaust pipe sub-module

[0081] The volume of the exhaust manifold and each exhaust branch pipe is considered as one volume, with the volume being V e (m 3 ), the mass of the exhaust gas in the volume being Q e (kg), the mass flow rate through the diesel engine exhaust valve being q Exh (kg / s), which is equal to the sum of the total air mass flow rate q a (kg / s), the fuel mass flow rate q f (kg / s), and the mass flow rate through the supercharger turbine q tur (kg / s). At non-steady state, the difference between q Exh and q t causes the mass Q e in the intake pipe volume to change, i.e.:

[0082]

[0083] The average exhaust temperature from the cylinder into the exhaust pipe is T Exh (K), and then these air is uniformly mixed in the exhaust pipe, and flows out of the exhaust pipe into the turbine, and the temperature of the gas flowing into the turbine (i.e. the average temperature of the air in the exhaust pipe) is T tur (K), and from the energy equation and the assumption of uniform mixing, we have:

[0084]

[0085] From the above three equations, the temperature T Aout (K) flowing out of the exhaust pipe is:

[0086]

[0087] The change in volume and temperature causes a change in pressure, and according to the thermodynamic state equation, the exhaust pressure is:

[0088]

[0089] where T Aout is the average temperature flowing out of the exhaust pipe, in K; p Exh is the exhaust pressure of the exhaust pipe, in Pa; R e is the gas constant on the exhaust side, taken as 286.354 J / kg·K; k e is the gas isentropic index on the exhaust side, taken as 1.33.

[0090] The generation of NOx in a diesel engine is mainly through the reaction of nitrogen and oxygen in the air at high temperature, mainly including thermal NO, fast NO and fuel NO. Thermal NO is the main generation mechanism, which can be calculated by the Zeldovich mechanism:

[0091]

[0092] The generation of CO is mainly formed by the partial oxidation of hydrocarbons when the combustion is incomplete. Its generation is related to the type of fuel, air-fuel ratio, combustion temperature. The generation and consumption of CO can be represented by the following equilibrium equation:

[0093]

[0094] (2) Calibration module construction, a fusion technology system based on deep learning optimization calibration comprehensive performance mechanism model is proposed, which also has the characteristics of non-invasive data monitoring. The mechanism model can be explained, but due to the fixed formula, the generalization ability is insufficient, and it cannot cope with complex working conditions, so the present application proposes a calibration method, which uses deep learning algorithm to calibrate the comprehensive performance mechanism model, improves the generalization ability of the comprehensive performance mechanism model, and solves the problems of high cost, insufficient economy and short service life of the traditional sensor monitoring method.

[0095] (2-1) Use the history data to clean up, delete the abnormal and repeated data, and use the difference method to fill in the missing data. The interpolation formula is as follows:

[0096] l t =(l t+1 +l t-1 ) / 2

[0097] In the formula, l t represents the missing value at time t; l t+1 is the normal characteristic value at t+1; l t-1 is the normal characteristic value at t-1. That is, the mean value of the time point before and after the missing value is taken as the interpolation value.

[0098] (2-2) Use variance filtering method to remove engine operating parameters with variance of 0.

[0099] (2-3) Use CNN-Bilstm-Attention as calibration module, the neurons of LSTM long short-term memory neural network can extract and filter the data in the model, so as to improve the effect of the model. The internal calculation process of its neurons is as follows: first, for each time step t, the input gate decides how much new information should be added to the cell state, and the output of the input gate i t is calculated by the following formula:

[0100] i t =σ(w i [h t-1 ,x t ])+b i

[0101] At the same time, a new candidate memory cell is added by using tanh activation function:

[0102] c' t =tanh(w c [h t-1 ,x t ]+b c )

[0103] Where, sigma is the Sigmoid activation function, wi w c i t c' t weights; b i b c i t c' t The bias term, x t It is the input of the input sequence at time step t, that is, the input h. t-1 It is the hidden state of the previous time step t-1.

[0104] The forget gate determines which information in a new state should be forgotten or retained. The output f of the forget gate... t It is calculated using the following formula:

[0105] f t =σ(w f [h t-1 ,x t ]+b f )

[0106] Next, use the output i of the input gate t and the output f of the forget gate t To update cell state c t :

[0107] c t =f t c t-1 +i t c' t

[0108] The output gate depends on h t-1 x t Calculate output information o t The tanh activation function combines the output gate information o t Get the current hidden layer state

[0109] o t =σ(w o [h t-1 ,x t ]+b o )

[0110] h t =o t *tanh(c t )

[0111] In the formula, w o b represents the output gate weights; o This is the output gate bias term.

[0112] Hidden layer state h tis passed to the next time step, while the cell state C t is also passed to the next step to maintain long-term memory. At this point, the neuron of the LSTM neural network has completed a calculation process according to the input.

[0113] Specifically, the calculation formula of the mean square error loss function is:

[0114]

[0115] In the formula, N is the number of samples, y is the true value, and a is the predicted value. The calculation formula of the predicted value a is:

[0116] a = f(z) = f(w*x+b)

[0117] In the formula, x is the input, w is the weight, b is the bias term, and f is the activation function.

[0118] Further, the step (3) Attention hidden layer state value:

[0119] e i = tanh(Wh i +b)

[0120]

[0121] In the formula: W and b are weights and thresholds respectively; h i is the hidden unit state value output by the CNN, the score e i of each hidden unit state value is obtained, the score of the attention mechanism is normalized, and finally e i and a i are weighted and summed to obtain the hidden layer state value c t .

[0122] (3) Construct a combustion phase mechanism model, and the main factor of high pollutant emission is insufficient combustion. In order to solve the problem that the comprehensive mechanism model cannot reflect the heat conduction of the diesel engine combustion process, the combustion phase mechanism model is combined in the present application, including an air intake manifold, a resonance cavity, an air intake manifold, an oil injector and the like, and the influence of in-cylinder temperature distribution and oil-gas mixing characteristics on combustion efficiency is deeply explored. First, the output of the calibrated mechanism model, including: injection pressure, inlet and outlet pressure, in-cylinder temperature, is used as the inlet and outlet conditions, wall conditions and initial conditions of the combustion phase mechanism model. Secondly, the turbulent energy is used as an index of the uniformity of oil-gas mixing, and the structure of the air intake pipe is optimized, aiming at improving the uniformity and flow characteristics of oil-gas mixing.

[0123] (3-1) Establish a CFD numerical simulation model, including an air intake manifold, a resonance cavity, an air intake manifold, an oil injector and the like.

[0124] (3-2) Intake grid division, extract the fluid domain of the intake pipe three-dimensional model, and divide the fluid domain into tetrahedral grids. Set the front end of the intake manifold as the inlet of the fluid, and the end of the intake manifold as the outlet of the fluid. Set the end of the oil injector as the methanol injection inlet, and the rest of the surface as the wall surface of the intake pipe. Perform local encryption processing on the grid at the end of the oil injector.

[0125] (3-3) Use the data in step (3) as the inlet and outlet conditions, wall conditions, and initial conditions of the combustion phase mechanism model. This includes: in-cylinder temperature, inlet and outlet pressure, and injector pressure.

[0126] (3-4) Determination of initial conditions and boundary conditions. When performing computational fluid dynamics calculations, appropriate initial and boundary conditions must be established. Selecting the appropriate boundary conditions requires consideration of the fluid's physical properties, calculation range, and control equations. This allows the fluid calculation results to be more consistent with reality. In fluid simulation calculations, there are multiple inlet and outlet boundary conditions to choose from. If the fluid's pressure, flow rate, and mass are known, the inlet (or outlet) can be set to the corresponding pressure, velocity, and mass boundary conditions. During the simulation, the inlet uses mass and pressure boundaries, and the outlet uses pressure boundaries. At the same time, the wall is set to be smooth and frictionless, and the standard wall function method is used for solving.

[0127] (3-5) Structure optimization, analyze the flow characteristics in the intake pipe. When the oil injector is not injecting oil, all intake manifolds are pressure outlets with an initial pressure of atmospheric pressure. Analyze the oil-gas mixing characteristics, and use the maximum value of the intake manifold's turbulent kinetic energy as an indicator of the uniformity of the oil-gas mixture. Adjust the length of the intake manifold in the vertical direction.

[0128] (3-6) Optimization verification, after the above step (4-5) optimization is completed, re-analyze the mixing characteristics, view the maximum value of the intake manifold's turbulent kinetic energy and the turbulent kinetic energy cloud map, and verify the effectiveness of the structure optimization.

[0129] (4) Construct a data-driven model. To further understand the internal relationship between high-fidelity deduction and emissions in the combustion phase, based on the above combustion process mechanism model, further use deep learning algorithms to fit the relationship between the maximum value of turbulent kinetic energy and emissions.

[0130] A convolutional neural network (CNN) module is established to extract the maximum value of turbulent kinetic energy and the features of operating condition data. Through the CNN module, the maximum value of turbulent kinetic energy and the feature information related to the operating conditions can be automatically extracted from the input data. The relationship between the maximum value of turbulent kinetic energy and the operating condition information and the emission data is predicted. Then, the model is trained, and the backpropagation algorithm and optimizer are used to adjust the parameters of the model to minimize the loss function MAE.

[0131]

[0132] (5) From the optimization control strategy, find the optimal strategy, so that the combustion is more sufficient, and the pollutant emission is lower, the above-constructed comprehensive performance mechanism model, the calibration module, the combustion stage mechanism model, the data-driven model fusion technology model are used as the fitness function, the least emission is taken as the target, the particle swarm optimization algorithm is used to find the optimal value of the main injection timing, the pre-injection timing, the injection amount and other injection control parameters under different conditions, the optimization control strategy is optimized, and the control strategy with the lowest emission is found.

[0133] Particle optimization process:

[0134]

[0135] In the formula: i is 1, 2...N; N is the total number of particles in the population; t is the iteration number of particles; omega is the weight; v is the speed of each particle in the population; r1, r2 are random numbers; x i is the current position of the i particle; is the current particle individual optimal position of the i particle; and is the optimal position of the particle population at present; c1, c2 are learning factors.

[0136] The remaining matters of the application are known technologies.

[0137] The above examples only illustrate the technical concept and characteristics of the application, and the purpose is to enable those skilled in the art to understand the content of the application and implement it, and cannot limit the protection scope of the application. Any equivalent changes or modifications made according to the spirit and essence of the application should be covered within the protection scope of the application.

Claims

1. A digital twin-driven method for predicting and optimizing emissions from marine diesel engines, characterized in that, Includes the following steps: (1) Construct a comprehensive performance mechanism model to describe the working principle and internal operation mechanism of the diesel engine, obtain the operating parameters of the diesel engine, and use the idea of ​​decomposition and combination to decompose the diesel engine system into a series of independent but interconnected subsystems, and consider the interaction and influence between different subsystems. (2) Construct a calibration module and propose a fusion technology system based on deep learning to optimize the calibration comprehensive performance mechanism model. Use deep learning algorithms to calibrate the comprehensive performance mechanism model, improve the generalization ability of the comprehensive performance mechanism model, and solve the problems of high cost, insufficient economy and short service life of traditional sensor monitoring methods. (3) Construct a combustion stage mechanism model, use the output of the calibrated comprehensive performance mechanism model as the inlet and outlet conditions, wall conditions and initial conditions of the combustion stage mechanism model, use turbulent kinetic energy as an index of the uniformity of oil-gas mixing, and optimize the structure of the intake pipe to improve the uniformity and flow characteristics of oil-gas mixing. (4) Establish a data-driven model. In order to further understand the intrinsic relationship between high-fidelity simulation of combustion stage and emissions, based on the combustion process mechanism model, use deep learning algorithms to fit the relationship between turbulent kinetic energy index and emissions. (5) From the perspective of optimizing control strategies, find the optimal strategy to make combustion more complete and pollutant emissions lower. Using the integrated performance mechanism model, calibration module, combustion stage mechanism model and data-driven model fusion technology system as the fitness function, with the goal of minimizing emissions, design an artificial intelligence optimization algorithm to optimize fuel injection control parameters and determine the optimal fuel injection control strategy under different operating conditions. The construction of the calibration module in step (2) includes the following steps: (2-1) Preprocess the output and historical data of the comprehensive performance mechanism model in step (2), including cleaning the data to remove abnormal and duplicate data and fill in missing data; use variance filtering to remove data with variance of 0, and divide the data into training set, test set and validation set; (2-2) Establish a CNN module, input the divided dataset into the convolutional neural network for feature extraction, extract data through the convolutional layer and process it with the activation function before passing it to the pooling layer, use the neurons in the pooling layer to perform secondary local feature extraction and simplification on the data from the first feature extraction, and perform hierarchical extraction on the input data, integrate the extracted local information, and lay the foundation for using BiLSTM to extract deep temporal features. (2-3) Establish a BiLSTM module to process and analyze data to capture long-term dependencies in time series data. Use a bidirectional long short-term memory network model to extract time series information from the data and analyze the patterns and regularities therein. (2-4) An Attention mechanism is introduced, and the information extracted by CNN is used as a new input variable to LSTM. The weights of the hidden layer of LSTM are calculated by using the Attention mechanism through the forward and backward propagation of deep learning. The gate weights and bias matrices of the hidden layer states at each time series point are propagated. The error between the predicted value and the actual value is adjusted and optimized. The gate weights and biases optimized by the Attention mechanism are updated through the error. The Attention mechanism is added to the hidden layer of BiLSTM to learn the temporal change pattern in depth, accurately capture key information and prevent data generalization, and finally obtain a suitable and stable deep learning model.

2. The method for predicting and optimizing emissions from marine diesel engines driven by digital twins according to claim 1, characterized in that, Construction of the comprehensive performance mechanism model in step (1): (1-1) Construct a comprehensive performance mechanism model, including compressor submodule, intercooler submodule, intake pipe submodule, cylinder submodule, exhaust pipe submodule, turbine submodule, turbocharger rotor submodule and fuel injection pump submodule; (1-2) Determine the output parameters of the comprehensive performance mechanism model, including the in-cylinder pressure, in-cylinder temperature and emission parameters such as CO, NO and NO2 during the combustion stage.

3. The method for predicting and optimizing emissions from marine diesel engines driven by digital twins according to claim 1, characterized in that, The construction of the combustion stage mechanism model in step (3) includes the following steps: (3-1) Establish a CFD numerical simulation model and define the structure of the intake manifold, resonant cavity, intake manifold and fuel injector; (3-2) Perform intake mesh generation, extract the fluid domain of the 3D model of the intake pipe, and perform tetrahedral mesh generation on the fluid domain; set the front end of the intake manifold as the fluid inlet and the end of the intake manifold as the fluid outlet; set the end of the fuel injector as the methanol injection inlet and the remaining surfaces as the intake pipe wall; perform local meshing on the end of the fuel injector. (3-3) Use the data in step (3) as the inlet and outlet conditions, wall conditions and initial conditions of the combustion stage mechanism model. The data includes the cylinder temperature, inlet and outlet pressure and injector pressure. (3-4) Determine the initialization conditions and boundary conditions, and establish the initial and boundary conditions based on computational fluid dynamics; when selecting boundary conditions, consider the physical properties of the fluid, the computational range, and the governing equations; set the inlet boundary as a mass and pressure boundary, and set the outlet boundary as a pressure boundary; set the wall surface as smooth and frictionless, and use the standard wall function method for solving; (3-5) Optimize the structure and analyze the flow characteristics in the intake manifold; when the injector is not injecting fuel, set all intake manifolds as pressure outlets and set the initial pressure to atmospheric pressure; analyze the fuel-air mixing characteristics and use the maximum value of the turbulent kinetic energy of the intake manifold as an indicator of the uniformity of fuel-air mixing; adjust the length of the intake manifold in the vertical direction to improve the mixing characteristics. (3-6) Optimize and verify the structure. After the structural optimization is completed, reanalyze the mixing characteristics. Observe the maximum value of turbulent kinetic energy and the turbulent kinetic energy cloud map of the intake manifold to verify the effectiveness of the structural optimization.

4. The method for predicting and optimizing emissions from marine diesel engines driven by digital twins according to claim 1, characterized in that, The step (3) of establishing the data-driven model includes the following steps: (4-1) The maximum value of turbulent kinetic energy and emission data under different working conditions are preprocessed, including cleaning and normalization, and divided into training set and test set; (4-2) Establish a convolutional neural network (CNN) module to extract the maximum value of turbulent kinetic energy and the features of the operating condition data; through the CNN module, automatically extract the maximum value of turbulent kinetic energy and the feature information related to the operating condition from the input data, and predict the relationship between the maximum value of turbulent kinetic energy and the operating condition information and emission data; (4-3) The model is trained by using the backpropagation algorithm and optimizer to adjust the model parameters in order to minimize the loss function.

5. The method for predicting and optimizing emissions from marine diesel engines driven by digital twins according to claim 1, characterized in that, The control strategy optimization in step (5) includes the following steps: (5-1) Determine the optimization parameters as fuel injection control parameters, including pre-injection timing, main injection timing, and fuel injection quantity; (5-2) Based on the particle swarm optimization algorithm, the data in step (5-1) is used as the initial particle swarm parameters, and the position and velocity of each particle are randomly initialized; the fusion system formed by steps (1), (2), (3), and (4) is used as the fitness function, and the minimum emission is used as the objective function to perform iterative optimization. (5-3) The pre-injection timing, main injection timing and injection quantity obtained by the fusion system consisting of steps (1), (2), (3) and (4) are verified to determine the corresponding emissions.

Images