Hybrid power diesel engine high cold start process energy optimization method based on multi-objective genetic algorithm
By optimizing the energy distribution of hybrid diesel engines using a multi-objective genetic algorithm and a combustion parameter prediction model, the problem of high starting power consumption in cold environments was solved, enabling rapid, low-power cold starts and improving starting efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-26
- Publication Date
- 2026-03-10
AI Technical Summary
In cold environments, the starting process of a hybrid diesel engine involves numerous and interdependent control parameters, making it difficult to find the optimal energy distribution strategy through experiments. This results in high starting power consumption and difficulty in accurate simulation.
A combustion parameter prediction model was established by combining a multi-objective genetic algorithm with a long short-term memory neural network and a double Weber function linearization method. The preheating of the medium and the power consumption of the motor were optimized. The power consumption function relationship was fitted by experimental data to achieve the optimization of energy distribution.
In cold environments, the hybrid diesel engine achieves rapid, low-energy-consumption cold start, reducing start-up time by 32.6% and energy consumption by 51.7%, thus improving the efficiency and energy utilization of the start-up process.
Smart Images
Figure CN121635162A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of diesel engine starting control, and in particular to an energy optimization method for the cold-start process of a hybrid diesel engine based on a multi-objective genetic algorithm. This method primarily addresses the challenges of accurately simulating significant inter-cycle differences, numerous and coupled control parameters, and the difficulty in finding the optimal energy allocation strategy through experimentation during the cold start process of a diesel engine in extremely cold environments. It ensures that the diesel engine can achieve rapid, low-power cold start and operation under these conditions. Background Technology
[0002] The starting auxiliary measures used in the cold start of hybrid diesel engines, such as medium preheating and motor drive, all require energy from the power battery. However, in extremely cold environments, the available battery capacity decreases, making the rational allocation of battery energy crucial for improving starting performance. However, the starting process involves numerous and interdependent control parameters, making it difficult to find the optimal control strategy through experimentation. Establishing a cold start simulation model suitable for extremely cold environments to obtain the power consumption patterns of each auxiliary measure and its impact on starting characteristics is particularly important for optimizing the cold start control strategy. Current cold start models developed domestically and internationally use fixed Weber parameters for simulation, but the differences between cold start cycles are significant, making it impossible to accurately simulate the starting process in extremely cold conditions. Furthermore, current research lacks a multi-component coupled power consumption model for the starting process, making it impossible to calculate starting power consumption.
[0003] Therefore, in view of the characteristics of starting in high cold weather, it is urgent to establish an energy optimization method for the starting process based on the simulation model of the starting performance of hybrid diesel engines in high cold weather, so as to realize the rapid and low power consumption starting of diesel engines under different ambient temperatures. Summary of the Invention
[0004] To address the shortcomings of the aforementioned technologies, this invention proposes an energy optimization method for the high-cold-weather start-up process of a hybrid diesel engine based on a multi-objective genetic algorithm. This method reduces the start-up time and power consumption of the hybrid diesel engine in high-cold-weather conditions, relying only on partial experimental data.
[0005] To solve the above technical problems, the present invention includes the following steps: Step 1: Collect data on cyclic fuel injection quantity, combustion, and power consumption during the start-up process of the hybrid diesel engine under different ambient temperatures and medium preheating temperatures. Step 2: Identify Weber parameters based on double Weber function linearization combined with the Levenberg-Marquardt algorithm, and correlate model parameters with the actual heat release process. Identify parameters separately for premixed combustion and diffusion combustion stages to reduce computational workload. Divide the measured integral value of the cyclic heat release rate by the equivalent fuel quantity to obtain the combustion fuel quantity fraction, quantifying the difference between the cyclic fuel injection quantity and the actual combustion fuel quantity caused by fuel adhesion to the walls. Step 3: Establish a combustion parameter prediction model using experimental data and a long short-term memory neural network. Use Pearson correlation analysis to select input parameters and train the neural network, selecting the final compression pressure. Compression end temperature Coolant temperature Diesel engine speed Fuel injection timing and the amount of fuel injected in circulation As input parameters of the long short-term memory neural network, the Weber parameters and fuel fraction are predicted. Combined with the calculation of heat dissipation loss during the starting process and system dynamics calculation, the starting speed curve is predicted. Step four, the function relationship between medium preheating power consumption and motor power consumption is obtained by fitting power consumption test data. Step five, with coolant preheating temperature, lubricating oil preheating temperature, intake air preheating temperature and motor torque compensation time as optimization parameters, and starting time and power consumption as optimization objectives, the optimal energy distribution during the high-cold start-up process of the hybrid diesel engine under different operating conditions is achieved by using a multi-objective genetic algorithm.
[0006] Furthermore, in step two of this invention, the parameter identification method based on double Weber function linearization combined with the Levenberg-Marquardt algorithm associates the model parameters with the actual exothermic process and identifies them separately in two stages: premixed combustion and diffusion combustion, in order to reduce the computational workload. The linearization formula for the dual Weber function, considering only diffusion combustion, is as follows:
[0007] The linearized formula for the dual Weiber function, considering only premixed combustion, is as follows:
[0008] In the formula: It is the combustion fraction; Premixed combustion fraction; The premixed combustion index; The diffusion flammability index; For Weber parameters; Crankshaft rotation angle, unit: ; The time at which premixed combustion begins, in units of ; The duration of premixed combustion is expressed in units of... ; The time of start of diffusion combustion, in units of ; The duration of diffusion combustion, in units of .
[0009] Furthermore, in step two of this invention, the formula for calculating the fuel fraction is:
[0010] In the formula: It represents the fuel fraction. This represents the actual amount of fuel burned, in units of... ; This refers to the cyclic fuel injection quantity, in units of... ; This refers to the lower calorific value of diesel fuel, in units of... ; The heat release rate is expressed in units of... .
[0011] Furthermore, in step three of this invention, the formula for calculating the Pearson correlation analysis is as follows:
[0012] In the formula, The correlation coefficient; For input parameters; The average value of the input parameters; For output parameters; This is the average value of the output parameters; The number of cycles; Using the data obtained from the experiment as a sample, the correlation coefficient between the input parameters and the combustion parameters and fuel fraction obtained from the experiment was calculated, and the final compression pressure was selected. Compression end temperature Coolant temperature Diesel engine speed Fuel injection timing and the amount of fuel injected in circulation As input parameters for long short-term memory neural networks.
[0013] Furthermore, in step three of this invention, the training process of the long short-term memory neural network is as follows: during the training process, the experimental data is randomly divided into a training set, a test set, and a validation set in a ratio of 8:1:1, and the training framework is TensorFlow.
[0014] Furthermore, in step three of this invention, the formula for calculating heat loss during the high-temperature start-up process is as follows:
[0015] In the formula: The in-cylinder heat dissipation rate, in units of ; The cylinder heat dissipation area, in units of ; This is the heat transfer coefficient, in units of... ; Engine speed, unit: ; Cylinder diameter, unit: ; The average piston speed, in units of ; Pressure, unit: ; Temperature, unit: ; , , , It is a constant.
[0016] Furthermore, in step three of this invention, the system dynamics calculation formula for the diesel engine speed during the high-cold-weather start-up process is:
[0017] In the formula: The crankshaft angular velocity, in units of ; Motor torque, unit: ; Frictional resistance torque, in units of ; The torque of the gas inside the cylinder, in units of ; The moment of inertia of the rotating component referred to the crankshaft, in units of . ; Mass of reciprocating moving parts, unit: ; Crank radius, in units of ; is the conversion factor.
[0018] Furthermore, in step four of this invention, the calculation formulas for the power consumption of medium preheating and motor power consumption based on the fitted experimental data are as follows:
[0019] In the formula: , , , These are the power consumption for intake air preheating, coolant preheating, lubricating oil preheating, and motor power consumption, respectively, in units of... ; This refers to the specific heat capacity of the medium at constant pressure, in units of... ; Medium quality, unit: ; The difference between the preheating temperature and the ambient temperature, in units of... ; The starting time of motor operation, in units of ; The time when the motor finishes running, in units of ; Motor speed, unit: ; Motor torque, unit: ; The motor efficiency is given by the efficiency characteristic graph, in units of... .
[0020] Furthermore, in step five of this invention, the preheating temperature range of the coolant is 243-313K, the preheating temperature range of the lubricating oil is 243-293K, the preheating temperature range of the intake air is 263-323K, and the motor torque compensation time is 0-15s.
[0021] Compared with existing technologies, the present invention, employing the above technical solutions, has the following advantages: The cold start process combustion parameter prediction model established in this invention considers the influence of fuel adhesion to the walls, achieving cyclic-level prediction of Weber parameters and fuel fraction, greatly reducing testing requirements. The energy optimization method proposed in this invention enables rapid, low-power cold start of hybrid diesel engines in cold environments, providing support for efficient optimization of the start-up process. Attached Figure Description
[0022] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a comparison between the experimental and simulated starting speed values of a hybrid diesel engine in an embodiment of the present invention; Figure 3 This is a comparison between the experimental and simulated starting power consumption values of a hybrid diesel engine in an embodiment of the present invention; Figure 4 This invention relates to the energy optimization method used to improve power consumption during startup. Figure 5 This invention relates to an energy optimization method for improving the start-up time during the startup process. Detailed Implementation
[0023] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings. These embodiments are based on the technical solutions of the present invention and provide detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments. Any process scheme that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experiments on the basis of the prior art is within the scope of protection defined by the claims of the present invention.
[0024] Embodiments Specific implementation examples Figures 1 to 5 As shown in the figure, firstly, data on the cyclic fuel injection quantity, combustion, and power consumption during the start-up process of the hybrid diesel engine were collected under different ambient temperatures and medium preheating temperatures.
[0025] Subsequently, based on the linearization of the double Weber function combined with the Levenberg-Marquardt algorithm, Weber parameters were identified, and the model parameters were correlated with the actual heat release process. The parameters were identified separately for the premixed combustion and diffusion combustion stages to reduce computational workload. The fuel quantity fraction was obtained by dividing the measured integral value of the cycle heat release rate by the cycle fuel quantity to quantify the difference between the cycle fuel injection quantity and the actual fuel quantity caused by fuel adhesion to the combustion walls.
[0026] The linearization formula for the dual Weber function, considering only diffusion combustion, is as follows:
[0027] The linearized formula for the dual Weiber function, considering only premixed combustion, is as follows:
[0028] In the formula: It is the combustion fraction; Premixed combustion fraction; The premixed combustion index; The diffusion flammability index; For Weber parameters; Crankshaft rotation angle, unit: ; The time at which premixed combustion begins, in units of ; The duration of premixed combustion is expressed in units of... ; The time of start of diffusion combustion, in units of ; The duration of diffusion combustion, in units of .
[0029] Fuel fraction Defined as:
[0030] In the formula: It represents the fuel fraction. This represents the actual amount of fuel burned, in units of... ; This refers to the cyclic fuel injection quantity, in units of... ; This refers to the lower calorific value of diesel fuel, in units of... ; The heat release rate is expressed in units of... .
[0031] Subsequently, using the experimental data as a sample, the Pearson correlation coefficients of the input parameters and the experimental combustion parameters and fuel fraction were calculated, and the compression final pressure was selected. Compression end temperature Coolant temperature Diesel engine speed Fuel injection timing and the amount of fuel injected in circulation The data were used as input parameters for a Long Short-Term Memory (LSTM) neural network. A combustion parameter prediction model was established using an LSM network, and the network was trained. During training, the experimental data was randomly divided into training, testing, and validation sets in an 8:1:1 ratio. The training framework used was TensorFlow. The established model can predict the changes in Weber parameters and fuel fraction with the cycle. Combined with calculations of heat loss during startup and system dynamics, it predicts the startup speed curve.
[0032] Among them, Pearson correlation analysis is required for the parameter selection of the long short-term memory neural network combustion prediction model:
[0033] In the formula, The correlation coefficient; For input parameters; The average value of the input parameters; For output parameters; This is the average value of the output parameters; The number is the repeating number.
[0034] The formula for calculating heat loss during high-altitude, cold-weather startup is:
[0035] In the formula: The in-cylinder heat dissipation rate, in units of ; The cylinder heat dissipation area, in units of ; This is the heat transfer coefficient, in units of... ; Engine speed, unit: ; Cylinder diameter, unit: ; The average piston speed, in units of ; Pressure, unit: ; Temperature, unit: ; , , , It is a constant.
[0036] The system dynamics calculation formula for diesel engine speed during high-altitude cold start-up is:
[0037] In the formula: The crankshaft angular velocity, in units of ; Motor torque, unit: ; Frictional resistance torque, in units of ; The torque of the gas inside the cylinder, in units of ; The moment of inertia of the rotating component referred to the crankshaft, in units of . ; Mass of reciprocating moving parts, unit: ; Crank radius, in units of ; is the conversion factor.
[0038] Then, by fitting the power consumption test data, the functional relationship between the power consumption of the medium preheating and the power consumption of the motor was obtained:
[0039] In the formula: , , , These are the power consumption for intake air preheating, coolant preheating, lubricating oil preheating, and motor power consumption, respectively, in units of... ; This refers to the specific heat capacity of the medium at constant pressure, in units of... ; Medium quality, unit: ; The difference between the preheating temperature and the ambient temperature, in units of... ; The starting time of motor operation, in units of ; The time when the motor finishes running, in units of ; Motor speed, unit: ; Motor torque, unit: ; The motor efficiency is given by the efficiency characteristic graph, in units of... .
[0040] Finally, a multi-objective genetic algorithm was used to obtain the optimal energy management strategy for the high-cold start-up process of a hybrid diesel engine with start-up time and power consumption as objectives. The coolant preheating temperature range was 243-313K, the lubricating oil preheating temperature range was 243-293K, the intake air preheating temperature range was 263-323K, and the motor torque compensation time was 0-15s.
[0041] Taking the specific test results of a hybrid diesel engine as an example, after applying the energy optimization method of this invention, from Figure 2 and Figure 3 It can be seen that the combustion parameter prediction model established by experimental data and long short-term memory neural network can accurately predict the starting speed curve, and the power consumption function obtained by fitting power consumption test data can accurately calculate the medium preheating and motor power consumption during the starting process, greatly reducing the experimental requirements of the optimization process. Figure 4 and Figure 5 The results show that, with starting time and power consumption as objectives, the energy optimization of the high-altitude cold-start process of hybrid diesel engines using a multi-objective genetic algorithm can reduce power consumption by up to 51.7% while keeping the starting time unchanged, and shorten the starting time by up to 32.6% while keeping the power consumption unchanged. This enables the hybrid diesel engine to start quickly and with low power consumption in high-altitude cold environments, and can provide support for the efficient optimization of the starting process.
Claims
1. A hybrid diesel engine high-cold starting process energy optimization method based on a multi-objective genetic algorithm, characterized by The method comprises the following steps: Step one, collecting the cycle fuel injection amount, combustion and power consumption data during the starting process of the hybrid diesel engine under different environmental temperatures and medium preheating temperature working conditions; Step two, identifying the Weber parameters based on the linearization of the double Weber function combined with the Levenberg-Marquardt algorithm, and using the converted oil amount divided by the cycle oil amount to obtain the combustion oil amount fraction to quantify the difference between the cycle fuel injection amount and the actual combustion oil amount caused by fuel wall adhesion; Step three, establishing a combustion parameter prediction model through test data and a long short-term memory neural network, selecting input parameters using Pearson correlation analysis, training the neural network, predicting the Weber parameters and combustion oil amount fraction cycle by cycle, and combining the starting process heat loss calculation and system dynamics calculation to predict the starting speed curve; Step four, fitting the medium preheating power consumption and motor power consumption function relationship formula through power consumption test data; Step five, obtaining the optimal energy distribution strategy of the high-cold starting process of the hybrid diesel engine by using the multi-objective genetic algorithm, taking the coolant preheating temperature, lubricating oil preheating temperature, intake air preheating temperature and motor torque compensation time as optimization parameters, and taking the starting time and power consumption as optimization objectives.
2. The method of claim 1, wherein the method is a method of energy optimization for high-cold starting process of a hybrid diesel engine based on a multi-objective genetic algorithm, characterized in that, The parameter identification method based on the linearization of the double Weber function combined with the Levenberg-Marquardt algorithm in step two is to associate the model parameters with the actual heat release process, and to identify the pre-mixed combustion and diffusion combustion stages respectively, so as to reduce the calculation workload; The linearization formula of the double Weber function when only considering diffusion combustion is: ； The linearization formula of the double Weber function when only considering pre-mixed combustion is: ； In the formulae: is the combustion fraction; is the premixed combustion fraction; is the premixed combustion index; is the diffusion combustion index; is the Weber number; is the crank angle in ; is the premixed combustion start in ; is the premixed combustion duration in ; is the diffusion combustion start in ; is the diffusion combustion duration in .
3. The method of claim 1, wherein the method is a method for energy optimization of a high-cold starting process of a hybrid diesel engine based on a multi-objective genetic algorithm, characterized in that, The calculation formula of the combustion oil amount fraction in step two is: ； wherein: is the fraction of the fuel quantity combusted; is the actual fuel quantity, in g; ; is the injection quantity of the cycle, in g; ; is the lower heating value of the diesel fuel, in MJ / kg; ; is the heat release rate, in MJ / (kg s); .
4. The method of claim 1, wherein the method is a method for energy optimization of a high-cold starting process of a hybrid diesel engine based on a multi-objective genetic algorithm. The calculation formula of the Pearson correlation analysis in step three is: ； wherein is a correlation coefficient; is an input parameter; is an input parameter average; is an output parameter; is an output parameter average; is a number of cycles; With the data obtained in the experiment as samples, the correlation coefficients of the input parameters and the combustion parameters and the fraction of the combustion oil amount obtained in the experiment are calculated, and the compression end pressure , the compression end temperature , the coolant temperature , the diesel engine speed , the injection timing , and the cycle injection amount are selected as the input parameters of the long short-term memory neural network.
5. The method of claim 1, wherein the method is a method for energy optimization of high-cold starting process of a hybrid diesel engine based on a multi-objective genetic algorithm. In step three, the process of training the long short-term memory neural network is as follows: in the training process, the test data is randomly divided into training set, test set and validation set according to the ratio of 8:1:1, and the training framework is TensorFlow.
6. The method of claim 1, wherein the method is a method of energy optimization for high-cold starting process of a hybrid diesel engine based on a multi-objective genetic algorithm. The calculation formula of the high-cold starting process heat loss in step three is: ； wherein: is the rate of heat dissipation in the cylinder, and is ; is the heat transfer coefficient, and is ; is the engine speed, and is ; is the cylinder diameter, and is ; is the piston average speed, and is ; is the pressure, and is ; is the temperature, and is ; is the temperature, and is ; , , , is a constant.
7. The multi-objective genetic algorithm based energy optimization method for high and cold start process of a hybrid diesel engine of claim 1, wherein The system dynamics calculation formula of the high-cold starting process diesel engine speed in step three is: ； In the formula: is the crankshaft angular velocity, in rad / s ; is the motor torque, in Nm ; is the friction torque, in Nm ; is the in-cylinder gas acting torque, in Nm ; is the rotational inertia of the rotating member converted to the crankshaft, in kg·m2 ; is the reciprocating member mass, in kg ; is the crank radius, in m ; is the conversion factor.
8. The method of claim 1, wherein the method is a method of energy optimization for high-cold starting process of a hybrid diesel engine based on a multi-objective genetic algorithm. The calculation formula of the medium preheating power consumption and motor power consumption based on test data fitting in step four is: ； In the formula: , , , are the electric power consumptions of the intake air preheating, the coolant preheating, the lubricating oil preheating and the motor respectively, and the unit is ; is the specific heat capacity of the medium at constant pressure, and the unit is ; is the mass of the medium, and the unit is ; is the difference between the preheating temperature and the ambient temperature, and the unit is ; is the starting time of the motor operation, and the unit is ; is the ending time of the motor operation, and the unit is ; is the motor speed, and the unit is ; is the motor torque, and the unit is ; is the motor efficiency given by the efficiency characteristic map, and the unit is . 9.The hybrid diesel engine high-cold-start process energy optimization method based on the multi-objective genetic algorithm of claim 1, characterized in that In step five, the coolant preheating temperature interval is 243-313K, the lubricating oil preheating temperature interval is 243-293K, the intake air preheating temperature interval is 263-323K, and the motor torque compensation time is 0-15s.