Method for predicting vehicle emission information, electronic device and vehicle

Abstract

Embodiments of the present application provide a vehicle emission information prediction method, an electronic device and a vehicle. Specifically, the method comprises obtaining first information related to original emission under a first working condition; obtaining second information related to tail gas emission under the first working condition based on the first information related to the original emission and the chemical reaction kinetics parameters of the aftertreatment model; wherein the chemical reaction kinetics parameters of the aftertreatment model are obtained by optimizing a preset algorithm, and the preset algorithm takes the minimization of the error between the tail gas emission information actually measured under a second working condition and the tail gas emission information predicted by the aftertreatment model as the optimization target, and iteratively solves the chemical reaction kinetics parameters. The embodiments of the present application can improve the accuracy and robustness of the prediction results of the aftertreatment model.

Description

Technical Field

[0001] This application relates to the field of automotive exhaust aftertreatment technology, and in particular to a method for predicting vehicle emission information, an electronic device, and a vehicle. Background Technology

[0002] The three-way catalytic converter aftertreatment system is a key component of automotive aftertreatment, involving complex heat transfer, mass transfer, and chemical reactions. An accurate and reasonable aftertreatment model helps reduce emissions calibration on actual vehicles and in-vehicle testing resources, shortens calibration cycles, and expands the boundaries of emissions verification.

[0003] In related technologies, parameter calibration of post-processing models is generally carried out by manual parameter tuning. However, due to the complexity of the system, the large number of calibration parameters, and the mutual influence and constraints of the parameters, especially for chemical reaction kinetic parameters, it is difficult to ensure the robustness of the post-processing model for exhaust emission calculation under various engine operating conditions through manual parameter tuning. In addition, manual parameter adjustment is time-consuming and labor-intensive, and it is difficult to guarantee the global optimality of manually adjusted parameters for overall prediction accuracy. Summary of the Invention

[0004] This application provides a method, electronic device, and vehicle for predicting vehicle emission information, aiming to solve the problem of multiple calibration parameters and mutual influence among parameters in the post-processing model parameter tuning process, which makes it difficult to guarantee the global optimum. This is beneficial to improving the accuracy and robustness of the prediction results of the post-processing model.

[0005] In a first aspect, embodiments of this application provide a method for predicting vehicle emission information, including: Obtain initial information related to raw emissions under the first operating condition; Based on the chemical reaction kinetic parameters of the post-processing model, and using the first information related to the original emissions, a second information related to exhaust emissions under the first operating condition is obtained. The chemical reaction kinetic parameters of the post-processing model are obtained by optimization using a preset algorithm. The preset algorithm aims to minimize the error between the measured exhaust gas emission information and the exhaust gas emission information predicted by the post-processing model under the second operating condition, and iteratively solves to obtain the chemical reaction kinetic parameters.

[0006] In this embodiment, based on the original emissions, the optimized post-processing model can quickly and accurately predict the final exhaust emissions under any operating conditions. The automatic search for optimal chemical reaction kinetic parameters using a preset algorithm avoids errors caused by manual parameter tuning, improves the efficiency of parameter determination, and enhances the accuracy and robustness of the post-processing model's prediction results.

[0007] In one possible implementation, the pre-defined algorithm includes a particle swarm optimization algorithm, and the determination of the chemical reaction kinetic parameters of the post-processing model includes: Acquire the calibration condition data, which includes third information related to the original emissions and fourth information related to the exhaust emissions under the second operating condition. The third information and the fourth information are the measured exhaust emission information. Get the initialized particle swarm; Based on the data of the operating conditions to be calibrated, the initial particle swarm is subjected to fitness iteration to obtain the particle parameter with the smallest fitness value under the condition that the iteration meets the preset conditions, and this parameter is taken as the global optimal parameter. The fitness value represents the error between the exhaust emission-related prediction information obtained by the post-processing model under the current particle parameters and the fourth information. The prediction information is based on the third information related to the original emissions under the second operating condition. The chemical reaction kinetic parameters in the post-processing model are determined based on the globally optimal parameters.

[0008] In this embodiment, the chemical reaction kinetic parameters of the post-processing model are automatically optimized and calibrated using the particle swarm optimization algorithm. This can minimize the prediction error of the post-processing model based on the measured exhaust gas emission information, improve the accuracy of the post-processing model, reduce the error and cost of manual calibration, improve the efficiency of parameter calibration and global optimization capability, and thus improve the accuracy and robustness of the prediction results of the post-processing model.

[0009] In one possible implementation, obtaining the initialized particle swarm includes: Initialize the particle swarm to obtain the first generation particle swarm. The dimension of the particle swarm is the same as the number of chemical reaction kinetic parameters to be calibrated in the post-processing model. The particle parameters of the particles in the particle swarm are candidate solutions for the chemical reaction kinetic parameters. The fitness of each particle in the first generation particle swarm is calculated to obtain the fitness value of each particle in the first generation particle swarm. Based on the fitness values ​​of each particle in the initial particle swarm, the parameter of the particle with the smallest fitness value in the initial particle swarm is obtained as the historical global optimal parameter.

[0010] In the embodiments of this application, the quality of particle parameters is quantified by fitness calculation, which can provide a reference for algorithm iteration. The determination of the historical optimal particle parameters and historical global optimal parameters of the first generation particle swarm can serve as the direction of optimization, so as to gradually converge to the global optimum.

[0011] In one possible implementation, the fitness of the initialized particle swarm is iterated, including repeatedly performing the following steps until the iteration satisfies a preset condition: For each particle in the particle swarm of the previous iteration, determine whether the number of iterations in which the particle parameters of the particle have not been updated has reached a preset threshold, and update the particle parameters of the particle using the corresponding particle swarm algorithm based on the determination result, so as to obtain the particle swarm of the current iteration. Based on the calibration data, the fitness of the particle swarm in the current iteration is evaluated to obtain the fitness value of each particle in the particle swarm in the current iteration. The particle swarm for the current iteration is updated based on the fitness value of each particle.

[0012] In this embodiment, by determining whether the number of iterations in which particle parameters have not been updated for a continuous period has reached a preset threshold, and updating the particle parameters accordingly based on this determination, particles that have not been optimized for a long time have a certain possibility of jumping out of their current region to explore other regions in the search space. This helps avoid the problem of particle swarm optimization getting trapped in local optima, increases the possibility of finding the global optimum, and improves the algorithm's search capability and optimization effect. During the iteration process, the fitness of the particle swarm in each iteration is evaluated based on the calibration data, and the particle swarm is updated according to the evaluation results. This allows the algorithm to optimize with the goal of minimizing the error between measured and predicted information, which helps enhance the stability and reliability of the algorithm and makes the post-processing model corresponding to the optimization result more applicable.

[0013] In one possible implementation, the particle parameters include the particle position. For each particle in the particle swarm from the previous iteration, it is determined whether the number of consecutive iterations in which the particle parameters of the particle have not been updated has reached a preset threshold. Based on the determination result, the particle parameters of the particle are updated using a corresponding particle swarm optimization algorithm, including: If the number of iterations in which the particle parameters have not been updated consecutively does not reach a preset threshold, the particle's velocity and position are updated using the first particle swarm algorithm. The update of velocity and position is related to inertia factor, cognitive factor and social factor. Inertia factor is used to control the degree of continuity of the particle's velocity from the previous iteration. Cognitive factor is used to characterize the degree of influence of the particle's historical best position on the current update. Social factor is used to characterize the degree of influence of the historical global best position on the current update. If the number of iterations in which the particle parameters have not been updated for a consecutive period reaches a preset threshold, then the particle position is updated by randomly selecting one of at least two preset position calculation methods through the second particle swarm algorithm.

[0014] In this embodiment, if the number of iterations in which the particle parameters have not been updated consecutively does not reach a preset threshold, it indicates that the particle is still in a normal search. Therefore, the particle parameters can be updated using the first particle swarm algorithm without additional interference, thereby improving the efficiency of optimization. If the number of iterations in which the particle parameters have not been updated consecutively reaches a preset threshold, it indicates that the particle is likely to get stuck in a local optimum. The improved particle swarm algorithm can ensure the diversity of particle update rules and increase the particle's exploration performance.

[0015] In one possible implementation, the fitness of the particle swarm in the current iteration is evaluated based on the calibration condition data, including performing the following steps for each particle: Using the post-processing model, under the particle parameters of the current iteration, the fifth information related to exhaust emissions under the second operating condition is predicted based on the third information related to the original emissions under the second operating condition. Obtain the first weighting coefficient for the second operating condition and the second weighting coefficients for at least two emissions included in the fourth information; Based on the first and second weighting coefficients, the weighted error between the fourth and fifth information is calculated, and the weighted error is used as the fitness value of the particle.

[0016] In this embodiment, fitness evaluation is performed based on the data of the operating conditions to be calibrated (such as the second operating condition), enabling the particle swarm optimization algorithm to optimize the post-processing model for actual operating conditions. Exhaust gas emission characteristics vary significantly under different operating conditions. The fitness evaluation provided in this application ensures that the post-processing model accurately predicts exhaust gas emissions under various operating conditions, which is beneficial for improving the adaptability of the post-processing model to actual operating conditions. Specifically, when calculating the fitness value, weighting coefficients for different emissions and different operating conditions are considered, which allows for a more comprehensive assessment of the difference between the post-processing model's prediction results and the measured results. This helps reduce model bias caused by a single emission or a single operating condition, thereby improving the overall prediction accuracy of the post-processing model.

[0017] In one possible implementation, the particle swarm for the current iteration is updated based on the fitness value of each particle, including: For each particle, if the fitness value corresponding to the current particle parameter is less than the fitness value corresponding to the historical best particle parameter, then the historical best particle parameter of the particle is updated to the current particle parameter; otherwise, the particle parameter of the particle is not updated. From the current particle parameters of each particle in the particle swarm of the current iteration, obtain the target particle parameter with the smallest fitness value. If the fitness value corresponding to the target particle parameter is less than the fitness value corresponding to the historical global best parameter, then update the historical global best parameter to the target particle parameter; otherwise, do not update the historical global best parameter.

[0018] In this embodiment, by continuously updating the historical best particle parameters of individuals and the historical global best parameters, the particle swarm can quickly focus on the corresponding search region, and the algorithm can converge to the region close to the optimal solution more quickly, improving optimization efficiency. Furthermore, updating parameters by comparing fitness values ​​helps ensure that the algorithm maintains the same optimization objective during iteration, thus improving the algorithm's stability and robustness.

[0019] In one possible implementation, after the particle parameters are updated, the following is also included: If the particle parameters of a particle have a dimension value that exceeds the preset boundary interval after the update, then the dimension value that exceeds the preset boundary interval will be updated based on the preset boundary interval. The preset boundary interval is determined based on the chemical reaction kinetic parameters to be calibrated in the post-processing model.

[0020] In this embodiment, considering that if the particle position exceeds the preset boundary interval, using these invalid values ​​for fitness calculation and other operations may lead to incorrect calculation results or failure to obtain reasonable optimization results, for dimension values ​​that exceed the preset boundary interval, the dimension values ​​are reconfigured within the preset boundary interval. This can effectively ensure that the algorithm searches and calculates in an effective solution space, improve the efficiency and reliability of the algorithm's optimization, and thus improve the accuracy of the prediction results of the post-processing model.

[0021] In one possible implementation, the second operating condition includes the operating condition corresponding to the scan point operating point where the engine speed is in a preset speed range and the torque is in a preset torque range. The data for the calibration condition includes at least one of the following related to the second operating condition: engine speed, engine torque, exhaust mass flow rate, exhaust temperature, and air-fuel ratio. Third information related to primary emissions includes at least one of the following: hydrocarbon primary emission concentration, nitrogen oxide primary emission concentration, and carbon monoxide primary emission concentration; The fourth piece of information related to exhaust emissions includes at least one of hydrocarbon exhaust emission concentration, nitrogen oxide exhaust emission concentration, and carbon monoxide exhaust emission concentration.

[0022] In this embodiment of the application, by covering the entire operating range of the engine under the second operating condition, it is possible to ensure that the input data is complete and consistent with the actual emission situation. Furthermore, by optimizing and calibrating the parameters based on the emission concentration, the applicability and accuracy of the after-processing model can be improved.

[0023] Secondly, embodiments of this application provide a vehicle emission information prediction device, comprising: The acquisition module is used to acquire the first information related to the original emissions under the first operating condition; The prediction module is used to predict, based on the first information related to the original emissions, the chemical reaction kinetic parameters of the post-processing model, and obtain the second information related to the exhaust emissions under the first operating condition. The chemical reaction kinetic parameters of the post-processing model are obtained by optimization using a preset algorithm. The preset algorithm aims to minimize the error between the measured exhaust gas emission information and the exhaust gas emission information predicted by the post-processing model under the second operating condition, and iteratively solves to obtain the chemical reaction kinetic parameters.

[0024] Thirdly, embodiments of this application propose an electronic device, including a processor and a memory, wherein: the memory is used to store computer programs; and the processor is used to execute the programs stored in the memory to implement the method of the first aspect.

[0025] Fourthly, embodiments of this application propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method of the first aspect.

[0026] Fifthly, embodiments of this application provide a computer program product comprising a computer program that, when executed by a processor, implements the method of the first aspect.

[0027] Sixthly, embodiments of this application propose a vehicle including electronic equipment as described in the third aspect. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating a method for predicting vehicle emission information according to an embodiment of this application.

[0029] Figure 2 This is a flowchart illustrating a method for predicting vehicle emission information according to another embodiment of this application.

[0030] Figure 3 This is a flowchart illustrating a method for predicting vehicle emission information according to another embodiment of this application.

[0031] Figure 4 This is a flowchart illustrating a method for predicting vehicle emission information according to another embodiment of this application.

[0032] Figure 5 This is a structural block diagram of a vehicle emission information prediction device according to an embodiment of this application.

[0033] Figure 6 This is a structural diagram of the electronic device provided in the embodiments of this application. Detailed Implementation

[0034] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the following detailed description is provided in conjunction with embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0035] Terminology Explanation: Raw emissions: also known as in-engine raw emissions, refer to the waste pollutants and their component concentrations emitted directly from the exhaust port after fuel combustion in the internal combustion engine cylinders, without undergoing aftertreatment. Raw emissions include carbon monoxide (CO), hydrocarbons (HC), and nitrogen oxides (NOx). And particulate matter, etc. Original emissions and...

[0036] Exhaust emissions: also known as tail gas emissions, refer to the gases that are ultimately released into the atmosphere after the original exhaust gases have been purified by after-treatment devices (such as catalytic oxidation and reduction). Exhaust emissions are the final result of a vehicle's emissions.

[0037] Aftertreatment model: Used to describe the flow, heat transfer, component transport, and catalytic conversion processes of exhaust gas within an aftertreatment device (such as a three-way catalytic converter, TWC). It is a mathematical simulation model based on fluid mechanics, heat and mass transfer, and surface catalytic reaction mechanisms. The aftertreatment model can predict pollutant conversion efficiency, air-fuel ratio characteristics, oxygen storage capacity, and outlet emission concentration under different operating conditions.

[0038] Chemical reaction kinetic parameters are a set of physical properties and reaction characteristics that describe the reaction rate, reaction pathway, and reaction extent on a catalytic surface. These include reaction rate constants, activation energy, pre-exponential factors, adsorption / desorption coefficients, and oxygen storage and release rate coefficients. These parameters determine the accuracy of post-treatment models in calculating pollutant conversion processes.

[0039] Particle Swarm Optimization (PSO) is a heuristic global optimization algorithm based on swarm intelligence. It iteratively searches for the optimal solution in the solution space by simulating the foraging behavior of bird flocks. The algorithm uses particle positions to represent the parameters to be optimized, and a fitness function to evaluate the quality of solutions. Individual and global extrema guide the updating of particle speed and position, ultimately converging to the globally optimal solution. In this embodiment, PSO can be used to calibrate chemical reaction kinetic parameters in a post-processing model.

[0040] The three-way catalytic converter aftertreatment system is a key component of automotive aftertreatment, involving complex heat transfer, mass transfer, and chemical reactions. An accurate and reasonable aftertreatment model helps reduce emissions calibration on actual vehicles and in-vehicle testing resources, shortens calibration cycles, and expands the boundaries of emissions verification.

[0041] In related technologies, parameter calibration for post-processing models is typically performed manually, such as through manual parameter tuning. However, due to the complexity of the post-processing model system and the coupling of its parameters, especially for chemical reaction kinetic parameters, manual parameter tuning struggles to guarantee robustness in emission prediction performance across various operating conditions, thus limiting the application of post-processing models. Therefore, proper parameter calibration for post-processing models, particularly for chemical reaction kinetic parameters, is crucial for improving the prediction accuracy and robustness of the model.

[0042] In view of this, this application proposes a method for predicting vehicle emission information. In this method, a post-processing model predicts second information related to exhaust emissions under a first operating condition based on first information related to the original emissions. The chemical reaction kinetic parameters of the post-processing model are optimized using a preset algorithm. The preset algorithm uses the error between the measured exhaust emissions under the second operating condition and the exhaust emissions predicted by the post-processing model as the optimization objective, and iteratively solves to obtain the chemical reaction kinetic parameters. This application automatically finds and optimizes the chemical reaction kinetic parameters through a preset algorithm, avoiding errors caused by manual parameter tuning, improving the efficiency of parameter calibration, and enhancing the accuracy and robustness of the prediction results from the post-processing model.

[0043] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for predicting vehicle emissions information according to an embodiment of this application. Figure 1 The method shown can be executed by an electronic device, which may include a server or a terminal. The terminal may be a smartphone, tablet, laptop, desktop computer, smart TV, smart home device, in-vehicle terminal, vehicle after-processing controller, vehicle engine controller, etc., without specific limitations. In this embodiment, the server may be a standalone physical server, a server cluster consisting of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0044] Specifically, such as Figure 1 As shown, the method provided in this application embodiment may include S110 to S120: S110, Obtain first information related to the original emissions under the first operating condition.

[0045] Optionally, the first operating condition can be the vehicle's operating condition under a specific speed, load, temperature, and air-fuel ratio. The first information related to raw emissions can be the exhaust information of the engine under the first operating condition before after-treatment, which may include raw exhaust flow rate, temperature, air-fuel ratio, and raw CO / HC ratio. Concentration, etc.

[0046] Optionally, the first information may be the vehicle's raw emissions information under the first operating condition, collected through engine bench testing, on-board sensors, or simulation models.

[0047] S120. Based on the first information related to the original emissions, the chemical reaction kinetic parameters of the post-processing model are used to predict the second information related to the exhaust emissions under the first operating condition. The chemical reaction kinetic parameters of the post-processing model are obtained by optimization using a preset algorithm. The preset algorithm aims to minimize the error between the measured exhaust emissions information and the exhaust emissions information predicted by the post-processing model under the second operating condition, and iteratively solves to obtain the chemical reaction kinetic parameters.

[0048] Optionally, the second piece of information related to exhaust emissions may refer to the information on pollutants that are finally released into the atmosphere after the original emissions have been purified by the after-treatment device. This may include the predicted exhaust emission concentration and conversion efficiency output by the after-treatment model.

[0049] Optionally, the first information can be input into the post-processing model. The post-processing model processes the oxidation, reduction and conversion of pollutants in the catalyst based on the internal chemical reaction kinetic parameters, and outputs second information related to exhaust emissions, such as the purified exhaust gas concentration, total emissions, and conversion efficiency.

[0050] Optionally, the preset algorithm can be a particle swarm optimization algorithm, an intelligent optimization algorithm, etc., which can be used to optimize and calibrate chemical reaction kinetic parameters.

[0051] Optionally, the measured exhaust emission information can be the actual exhaust emission concentration information obtained by instruments during whole vehicle or bench testing, which can be used as the true value for calibration reference. The predicted exhaust emission information is the exhaust emission concentration information predicted and output by the after-processing model, which is the calculated value of the after-processing model. The error between the measured exhaust emission information and the exhaust emission information predicted by the after-processing model can be expressed in the form of mean square error (MSE), absolute error, etc., which is the objective function of the preset algorithm optimization.

[0052] Optionally, the parameters to be calibrated in the post-processing model (such as chemical reaction kinetic parameters) are obtained by optimization using a preset algorithm. During the optimization process, the measured exhaust gas emission data under multiple operating conditions can be used as a benchmark, and the optimization objective is to minimize the error between the measured value and the model prediction value. The preset algorithm is used to iteratively update the parameters until the error converges, so as to obtain the optimal parameters that enable the post-processing model to match the measured data with high accuracy.

[0053] In this embodiment, based on the original emissions, the optimized post-processing model can quickly and accurately predict the final exhaust emissions under any operating conditions. The automatic search for optimal chemical reaction kinetic parameters using a preset algorithm avoids errors caused by manual parameter tuning, improves the efficiency of parameter determination, and enhances the accuracy and robustness of the post-processing model's prediction results.

[0054] In one feasible embodiment, the preset algorithm includes a particle swarm optimization algorithm. The following embodiments use a particle swarm optimization algorithm as an example for illustration. Optionally, such as... Figure 2 As shown, the determination of chemical reaction kinetic parameters for the post-processing model includes S210 to S240: S210. Obtain the calibration condition data, which includes third information related to the original emissions and fourth information related to the exhaust emissions under the second operating condition. The third information and the fourth information are the measured exhaust emission information.

[0055] Optionally, the calibration data (such as engine calibration data) can be organized in tabular form. For example, each row in the table represents a steady-state condition obtained from the engine bench, and each column in the list represents the corresponding characteristic exhaust parameters (such as operating data, original exhaust data, and exhaust emission data) under each condition. During the iterative parameter update process using the particle swarm optimization algorithm, the acquired calibration data can be input into the MATLAB (Matrix Laboratory) workspace in tabular form. Each row in the table is provided as a time series of a fixed duration (such as 50 seconds) to the post-processing model for input simulation, and the exhaust emission result at the end of the calculation is used as the prediction result of the post-processing model for accuracy comparison.

[0056] Optionally, the second operating condition includes the operating conditions corresponding to the scanned operating points where the engine speed is within a preset speed range and the torque is within a preset torque range. The preset speed range includes the range set from the lowest speed to the highest speed. The preset torque range includes the range set from the lowest torque to the external characteristic torque. The lowest torque can be the minimum torque at which the engine can stably operate at the current speed, also known as the torque near idle or the minimum load torque. The external characteristic torque can be the maximum torque that the engine can output at the current speed, also known as the full load torque or the maximum torque. The second operating condition can be an operating condition in a steady-state operating condition dataset formed by scanning points within the preset speed range and the preset torque range. Optionally, in this embodiment, the second operating condition does not specifically refer to a particular operating condition, but can represent a certain operating condition in the steady-state operating condition dataset. Optionally, the second operating condition and the first operating condition can be the same or different operating conditions, without limitation.

[0057] Optionally, the information related to the second operating condition in the data to be calibrated includes engine speed. Engine torque Exhaust mass flow rate Exhaust temperature air-fuel ratio At least one of the following. Third information related to primary emissions includes hydrocarbon primary emission concentrations. Nitrogen oxide emission concentration and carbon monoxide emission concentration At least one of the following. Fourth information related to exhaust emissions includes hydrocarbon exhaust emission concentrations. Nitrogen oxide emissions concentration and carbon monoxide exhaust emission concentration At least one of them.

[0058] In this embodiment of the application, by covering the entire operating range of the engine under the second operating condition, it is possible to ensure that the input data is complete and consistent with the actual emission situation. Furthermore, by optimizing and calibrating the parameters based on the emission concentration, the applicability and accuracy of the after-processing model can be improved.

[0059] S220, Obtain the initialized particle swarm.

[0060] Optionally, a particle swarm is a set of candidate solutions, where each candidate solution (e.g., a particle) represents a set of possible values ​​for the chemical reaction kinetic parameters to be calibrated in the post-processing model. Initializing the particle swarm can be understood as adapting the parameter generation to generate an initial set of candidate solutions. The dimension of the particle swarm refers to the number of parameters contained in each particle. In this embodiment, the dimension of the particle swarm is the same as the number of chemical reaction kinetic parameters to be calibrated in the post-processing model. Particle parameters include the specific numerical values ​​constituting the particle, representing candidate solutions for the chemical reaction kinetic parameters. For example, if there are two parameters to be calibrated, the particle parameters of a particle include the specific numerical values ​​of both parameters.

[0061] Optional, such as Figure 3 As shown, the initial particle swarm obtained in S220 includes S221 to S223: S221. Initialize the particle swarm to obtain the initial particle swarm. The dimension of the particle swarm is the same as the number of chemical reaction kinetic parameters to be calibrated in the post-processing model. The particle parameters of the particles in the particle swarm are candidate solutions for the chemical reaction kinetic parameters.

[0062] Optionally, particle swarm optimization is performed by... It consists of a two-dimensional matrix, where This indicates the population size. A larger population size results in higher search efficiency and a greater probability of locating the global optimum, but also a greater computational load. The individual dimension of the particle is represented by the same number of parameters to be calibrated in the post-processing model. This allows for optimization of the included chemical reaction kinetic parameters. Each particle can be represented as a vector. in This indicates the number of chemical reactions, including redox reactions of pollutants and oxygen storage and release reactions. This represents the pre-exponential factor of the reaction rate. This represents the activation energy of the reaction rate. The initialization of the population involves generating random numbers, ensuring that the values ​​of each dimension of each individual are randomly distributed within the upper and lower limits of a baseline value. These upper and lower limits, also known as preset boundary intervals, can be determined based on the value range of the chemical reaction kinetic parameters to be calibrated. During initialization, the particle parameters (such as position) of all particles can be randomly generated within these preset boundary intervals. The setting of the preset boundary intervals can refer to relevant technologies, and this application does not limit this aspect.

[0063] Optionally, the initial particle swarm can be regarded as the first randomly generated particle swarm, serving as the starting point for algorithm iteration.

[0064] S222. Calculate the fitness of each particle in the first generation particle swarm to obtain the fitness value of each particle in the first generation particle swarm.

[0065] Optionally, fitness calculation can involve substituting particle parameters into a post-processing model to calculate the error between the predicted and measured exhaust emissions. Since the optimization objective is to minimize the error between the measured and predicted exhaust emissions, in this embodiment, a smaller fitness value indicates better parameters.

[0066] Optional, fitness for each individual particle Perform calculations and evaluations, among which The index of a particle is represented by the individual index, and the individual fitness is calculated by the following formula (1): Formula (1) in, This represents the fitness weighting coefficient for each steady-state condition, and their sum is 1. Indicates the working condition index. This represents the weighting coefficient for the three emissions, which sums to 1. , , This represents the exhaust gas emission concentration output by the post-processing model. The operating condition weighting coefficient and the emission weighting coefficient are used to adjust the proportion of a certain operating condition and emission in the optimization objective, which can transform a multi-objective optimization problem into a single-objective optimization problem.

[0067] S223. Based on the fitness values ​​of each particle in the first generation particle swarm, obtain the parameter of the particle with the smallest fitness value in the first generation particle swarm as the historical global optimal parameter.

[0068] Optionally, the particle parameters of the particle with the smallest fitness value (smallest error) in the initial particle swarm can be determined as the global optimal parameters for the current stage, that is, the current optimal combination of dynamic parameters.

[0069] Optionally, since the comparison object of the first-generation particle swarm is itself, the fitness value of each particle in the first-generation particle swarm is also the best individual of each particle in each generation. (e.g., the best particle parameters in history). Based on this, the best individual among the best individuals in history (e.g., the particle parameters of the particle with the smallest fitness value) can be selected as the best global individual in history. (Such as historical global optimal parameters). Among them, the smaller the fitness value, the better the evaluation, and correspondingly, the smaller the overall prediction error of the post-processing model.

[0070] In this embodiment, the initial particle swarm is initialized by random generation, which ensures that the optimization space covers a reasonable range and that the values ​​of each dimension of the individual particles are randomly distributed within the upper and lower limits of the baseline value. Quantifying the quality of particle parameters through fitness calculation provides a reference for algorithm iteration. Furthermore, determining the historical optimal particle parameters and historical globally optimal parameters of the initial particle swarm can serve as the direction for optimization, gradually leading to convergence towards the global optimum.

[0071] S230. Based on the data of the operating conditions to be calibrated, the initial particle swarm is subjected to fitness iteration to obtain the particle parameter with the smallest fitness value under the condition that the iteration meets the preset conditions, and this parameter is used as the global optimal parameter. The fitness value represents the error between the exhaust emission-related prediction information obtained by the post-processing model under the current particle parameters and the fourth information. The prediction information is obtained based on the third information related to the original emissions under the second operating condition.

[0072] Optionally, after fitness evaluation, the initialized particle swarm (such as the initial particle swarm) yields historically optimal particle parameters and historically globally optimal parameters. Based on this, fitness iteration can be performed on the particle swarm, and if the iterations meet preset conditions, the particle parameters with the smallest fitness value can be obtained as the globally optimal parameters. These preset conditions may include a set threshold for the number of iterations, or the error between predicted and measured information being less than a set threshold.

[0073] S240. Determine the chemical reaction kinetic parameters in the post-processing model based on the globally optimal parameters.

[0074] Optionally, at the end of the iteration, the globally optimal parameters of the particle swarm obtained in the last iteration can be used as chemical reaction kinetic parameters in the post-processing model.

[0075] In this embodiment, the chemical reaction kinetic parameters of the post-processing model are automatically optimized and calibrated using the particle swarm optimization algorithm. This can minimize the prediction error of the post-processing model based on the measured exhaust gas emission information, improve the accuracy of the post-processing model, reduce the error and cost of manual calibration, improve the efficiency of parameter calibration and global optimization capability, and thus improve the accuracy and robustness of the prediction results of the post-processing model.

[0076] In one feasible embodiment, such as Figure 3 As shown, in S230, the fitness of the initialized particle swarm is iterated, including repeatedly executing S231 to S233 below until the iteration meets the preset conditions: S231. For each particle in the particle swarm of the previous iteration, determine whether the number of iterations in which the particle parameters of the particle have not been updated has reached a preset threshold, and update the particle parameters of the particle using the corresponding particle swarm algorithm based on the judgment result, so as to obtain the particle swarm of the current iteration.

[0077] Optionally, the particle parameter update process involves updating particle parameters based on fitness values. For example, for a specific particle, the fitness value of that particle in the previous iteration's particle swarm is compared with the fitness value corresponding to the particle's historical best parameters. Based on the comparison result, it is determined whether to update the particle parameters. It is understood that there is a certain probability that the particle parameters will be updated in the previous iteration, and there is also a certain probability that they will not be updated. Correspondingly, the globally optimal parameters of the particle swarm in the previous iteration may also be updated or not. Based on this, embodiments of this application also count the update status of particle parameters and configure corresponding particle swarm optimization algorithms to update particle parameters based on the count.

[0078] For example, for particle A, if the particle parameters are not updated in the first iteration, the counter is incremented by 1; if the particle parameters are not updated in the second iteration, the counter is incremented by 2; if the particle parameters are not updated in the third iteration, the counter is incremented by 3; if the particle parameters are updated in the fourth iteration, the counter is reset to zero; if the particle parameters are not updated in the fifth iteration, the counter is incremented by 1... and so on.

[0079] Optionally, in this embodiment, for each particle in the particle swarm, it is determined whether the number G of consecutive unoptimized (unupdated particle parameters) iterations of that particle is less than a set threshold. The particle parameters are updated using the corresponding algorithm based on the judgment results.

[0080] S232. Based on the data of the working conditions to be calibrated, evaluate the fitness of the particle swarm in the current iteration and obtain the fitness value of each particle in the particle swarm in the current iteration.

[0081] Optionally, in each iteration, the fitness is calculated for each particle. The fitness value can be calculated by referring to formula (1) in the above embodiment to obtain the fitness value of each particle in the particle swarm in the current iteration.

[0082] S233. Update the particle swarm for the current iteration based on the fitness value of each particle.

[0083] Optionally, the decision to update particle parameters and global optimal parameters in the current iteration can be based on the fitness value. For example, before the iteration begins, the historically optimal particle parameters for each particle and the historically optimal parameters for the particle swarm are determined. Based on this, for each particle, its fitness value can be compared with the fitness value corresponding to the historically optimal particle parameters, and the comparison result determines whether to update the historically optimal particle parameters for that particle. For the particle swarm, the minimum fitness value in the current iteration can be compared with the fitness value corresponding to the historically optimal global parameters, and the comparison result determines whether to update the historically optimal global parameters.

[0084] In this embodiment, by determining whether the number of iterations in which particle parameters have not been updated for a continuous period has reached a preset threshold, and updating the particle parameters accordingly based on this determination, particles that have not been optimized for a long time have a certain possibility of jumping out of their current region to explore other regions in the search space. This helps avoid the problem of particle swarm optimization getting trapped in local optima, increases the possibility of finding the global optimum, and improves the algorithm's search capability and optimization effect. During the iteration process, the fitness of the particle swarm in each iteration is evaluated based on the calibration data, and the particle swarm is updated according to the evaluation results. This allows the algorithm to optimize with the goal of minimizing the error between measured and predicted information, which helps enhance the stability and reliability of the algorithm and makes the post-processing model corresponding to the optimization result more applicable.

[0085] In one feasible embodiment, the particle parameters include particle position. For example... Figure 3 As shown, in step S231, for each particle in the particle swarm from the previous iteration, it is determined whether the number of consecutive iterations in which the particle parameters of the particle have not been updated has reached a preset threshold, and the particle parameters of the particle are updated using the corresponding particle swarm algorithm based on the determination result, including steps A1 to A2: Step A1: If the number of iterations in which the particle parameters have not been updated consecutively does not reach a preset threshold, the particle's velocity and position are updated using the first particle swarm algorithm. The update of velocity and position is related to the inertia factor, cognitive factor, and social factor. The inertia factor is used to control the degree of continuity of the particle's velocity from the previous iteration. The cognitive factor is used to characterize the influence of the particle's historical optimal position on the current update. The social factor is used to characterize the influence of the historical global optimal position on the current update.

[0086] Optionally, if the number of iterations in which the particle parameters have not been updated for a consecutive period of time has not reached a preset threshold, it indicates that the particle is still in the normal search process. Therefore, the particle parameters can be updated by the first particle swarm algorithm without additional interference, thereby improving the efficiency of optimization.

[0087] Optionally, the next generation velocity and position vectors are updated according to the following formulas (2) and (3): Formula (2) Formula (3) in, The inertia factor represents the retention effect of the previous generation's velocity vector (such as the degree to which the particle maintains the velocity of the previous iteration). The cognitive factor representing a particle indicates the impact of its historical best position (such as the best particle parameters in the past) on the current update. The social factor representing a particle represents the impact of the globally optimal particle position (such as historical globally optimal parameters) on the current update. To refine the search in later iterations, the inertia factor, cognitive factor, and social factor can be correlated with the iteration number using the following formula: Formula (4) Formula (5) Formula (6) in, Indicates the maximum number of iterations. , , , , , These represent the maximum and minimum values ​​of each factor, respectively.

[0088] Step A2: If the number of iterations in which the particle parameters have not been updated for a consecutive period reaches a preset threshold, then the particle position is updated by randomly selecting one of at least two preset position calculation methods through the second particle swarm algorithm.

[0089] Optionally, if the number of iterations in which the particle parameters have not been updated for a consecutive period reaches a preset threshold, it indicates that the particle is likely to get stuck in a local optimum. Based on this, embodiments of this application provide an improved particle swarm algorithm (such as the second particle swarm algorithm) to update the particle parameters.

[0090] For example, the particle in the next iteration is selected with equal probability by generating random numbers, and its position is updated using a sine or cosine formula.

[0091] The sine formula is shown in formula (7): Formula (7) The cosine formula is expressed as shown in formula (8): Formula (8) in, , , Represents a random number between 0 and 1.

[0092] Optionally, other functions can be configured to update the particle's position, depending on the specific circumstances.

[0093] The embodiments of this application can ensure the diversity of particle update rules and increase the exploration performance of particles through an improved particle swarm algorithm.

[0094] In one feasible embodiment, such as Figure 3 As shown, after the particle parameters are updated, step A3 is also included: if the particle parameters of the particle have a dimension value that exceeds the preset boundary interval, then the dimension value that exceeds the preset boundary interval is updated based on the preset boundary interval.

[0095] The preset boundary interval is determined based on the chemical reaction kinetic parameters to be calibrated in the post-processing model.

[0096] Optionally, if after the particle position is updated, the dimension value [A1,A2...An,Ea1,Ea2...Ean] exceeds the upper and lower limits (such as the preset boundary interval), then for the dimension value that exceeds the boundary, the dimension value is randomly assigned to a value within the boundary by generating a random number.

[0097] In this embodiment, considering that if the particle position exceeds the preset boundary interval, using these invalid values ​​for fitness calculation and other operations may lead to incorrect calculation results or failure to obtain reasonable optimization results, for dimension values ​​exceeding the preset boundary interval, random numbers are generated to reconfigure the dimension values ​​within the preset boundary interval. This effectively ensures that the algorithm searches and calculates in an effective solution space, improves the efficiency and reliability of the algorithm's optimization, and thus improves the accuracy of the post-processing model's prediction results.

[0098] In one feasible embodiment, such as Figure 3 As shown, in S232, based on the calibration condition data, the fitness of the particle swarm in the current iteration is evaluated, including performing the following steps B1 to B3 for each particle: Step B1: Using the post-processing model, under the particle parameters of the current iteration, predict the fifth information related to exhaust emissions under the second operating condition based on the third information related to the original emissions under the second operating condition.

[0099] Step B2: Obtain the first weighting coefficient for the second operating condition and the second weighting coefficients corresponding to at least two emissions included in the fourth information.

[0100] Step B3: Based on the first weighting coefficient and the second weighting coefficient, calculate the weighted error between the fourth and fifth information, and use the weighted error as the fitness value of the particle.

[0101] Optionally, the information used to characterize the second operating condition includes at least one of engine speed, engine torque, exhaust mass flow rate, exhaust temperature, and air-fuel ratio. The third information related to the raw emissions may include at least one of hydrocarbon raw emission concentration, nitrogen oxide raw emission concentration, and carbon monoxide raw emission concentration. Based on this, particle parameters of the current particles can be obtained from the particle swarm of the current iteration as chemical reaction kinetic parameters to be calibrated in the post-processing model. Then, the post-processing model predicts the fifth information related to exhaust emissions under the second operating condition based on the third information. The fifth information may include at least one of hydrocarbon exhaust emission concentration, nitrogen oxide exhaust emission concentration, and carbon monoxide exhaust emission concentration.

[0102] Optionally, as shown in formula (1) above, each operating condition has a corresponding fitness weighting system, and each emission concentration has a corresponding weighting coefficient. Based on this, the fitness value of the particles can be calculated using formula (1). The first weighting coefficient and the second weighting coefficient can be set according to the actual situation and requirements. For example, if a certain second operating condition is a condition that occurs frequently in actual operation and has a significant impact on exhaust emissions, then a higher first weighting coefficient can be assigned to it.

[0103] Optionally, in the error calculation, the error between the fourth piece of information (such as measured exhaust emission information) and the fifth piece of information (such as predicted exhaust emission information output by the post-processing model) can be calculated. For each emission, the difference between its actual value and predicted value can be calculated, such as absolute error or relative error. Then, the calculated error can be weighted according to the first and second weighting coefficients obtained in step B2 to obtain the final weighted error. In this embodiment, the calculated weighted error is used as the fitness value of the particle. The fitness value can reflect the degree of agreement between the post-processing model prediction result and the measured result under the current particle parameters. The smaller the fitness value, the higher the accuracy of the post-processing model prediction result and the better the current particle parameters.

[0104] In this embodiment, fitness evaluation is performed based on the calibration data (second operating condition), enabling the particle swarm optimization algorithm to optimize the post-processing model for actual operating conditions. Exhaust gas emission characteristics vary significantly under different operating conditions. The fitness evaluation provided in this application ensures that the post-processing model accurately predicts exhaust gas emissions under various operating conditions, thus improving its adaptability to actual operating conditions. Furthermore, by considering weighting coefficients for different emissions and operating conditions when calculating the fitness value, the difference between the post-processing model's predictions and actual results can be more comprehensively assessed. This helps reduce model bias caused by a single emission or operating condition, thereby improving the overall prediction accuracy of the post-processing model.

[0105] In one feasible embodiment, such as Figure 3 As shown, S233 updates the particle swarm for the current iteration based on the fitness value of each particle, including steps C1 to C2: Step C1: For each particle, if the fitness value corresponding to the current particle parameter of the particle is less than the fitness value corresponding to the historical best particle parameter, then update the historical best particle parameter of the particle to the current particle parameter; otherwise, do not update the particle parameter of the particle.

[0106] Step C2: From the current particle parameters of each particle in the particle swarm of the current iteration, obtain the target particle parameter with the smallest fitness value. If the fitness value corresponding to the target particle parameter is less than the fitness value corresponding to the historical global optimal parameter, then update the historical global optimal parameter to the target particle parameter; otherwise, do not update the historical global optimal parameter.

[0107] In one example, the particle swarm can be traversed, and each particle can be processed: the fitness value of the particle is compared with the fitness value corresponding to the historical best particle parameters (which could be the fitness value corresponding to the particle parameters of any previous iteration of the particle swarm or the particle parameters of the initial particle swarm). If the fitness value of the particle in the current iteration is less than the fitness value corresponding to the historical best particle parameters, then the historical best particle parameters are updated to the current particle parameters of the particle. Furthermore, if the fitness value of the particle is less than the fitness value corresponding to the global historical best parameters, then the global historical best parameters are simultaneously updated to the current particle parameters of the particle.

[0108] For example, the fitness of the particle after its position is updated is evaluated, and if the fitness value is... Less than The fitness value will then be Updated to At the same time, if Less than ,Will Updated to ,like Greater than fitness value, then , It remains unchanged.

[0109] In another example, individual updates and global updates can be handled separately. For individual updates, the particle swarm in the current iteration is traversed, and each particle is processed. If the fitness value of a particle in the current iteration is less than the fitness value corresponding to the historical best particle parameter, then the historical best particle parameter is updated to the current particle parameter; otherwise, no update is performed. For global updates, based on the current particle parameters of all particles in the particle swarm in the current iteration, the target particle parameter corresponding to the particle with the smallest fitness value is determined. If the fitness value corresponding to the target particle parameter is less than the fitness value corresponding to the historical global best parameter, then the historical global best parameter is updated to the target particle parameter; otherwise, no update is performed.

[0110] In this embodiment, by continuously updating the historical best particle parameters of individuals and the historical best parameters of the global system, the particle swarm can quickly focus on the corresponding search region, enabling the algorithm to converge to a region close to the optimal solution more quickly and improving optimization efficiency. Furthermore, updating parameters by comparing fitness values ​​helps ensure that the algorithm maintains the same optimization objective during iteration, thus improving the algorithm's stability and robustness.

[0111] The following is combined Figure 4 The process for determining the chemical reaction kinetic parameters in the post-processing model of the method provided in the embodiments of this application is described.

[0112] Specifically, such as Figure 4 As shown, the determination of chemical reaction kinetic parameters in the post-processing model includes the following steps 1 to 12: Step 1: Organize the engine operating condition data to be calibrated into the corresponding file by row. That is, each row represents a certain steady-state operating condition obtained from the engine bench, and the columns represent the characteristic exhaust parameters corresponding to each operating condition.

[0113] Step 2: Read the files compiled in Step 1 into the MATLAB workspace in tabular form. Each row is provided as a time series of a fixed duration for the post-processing model as input for simulation. At the same time, the exhaust emission data is used as the output accuracy comparison of the post-processing model.

[0114] Step 3: Initialize the initial position and velocity of the particle swarm. The number of particles represents the overall population size, and the particle dimension is consistent with the number of parameters to be calibrated.

[0115] Step 4: Evaluate the fitness of the initialized particle swarm. The fitness is the weighted result of the post-processing model's prediction accuracy of exhaust emissions for the three types of emissions under all engine operating conditions in Step 1, given the current particle parameters.

[0116] Step 5: Set the initial population to the best population in history, and set the individual with the best fitness in the initial population (such as the historical best particle parameters) to the global best individual (such as the historical global best parameters).

[0117] Step 6: Traverse each particle in the population and perform local optimum judgment for each particle. If the optimal position of a particle has not been updated after several generations of updates, start the improved algorithm to make the particle jump out of the local optimum; otherwise, use the classic particle swarm algorithm to update the particle position.

[0118] In steps 7 and 6, if an improved particle swarm algorithm (such as the second particle swarm algorithm) is started, the particle position is updated by generating random numbers using the cosine or sine formula.

[0119] If the first particle swarm algorithm is used in steps 8 and 6, then the particle velocity and position vectors are updated according to the principles of the first particle swarm algorithm.

[0120] Step 9: Determine the upper and lower limits of the updated particle positions. No processing is done within the boundaries, and particle positions are randomly assigned if the positions exceed the boundaries.

[0121] Step 10: Evaluate the fitness of the particles after updating their positions. If the fitness is better than its own best individual in previous generations, then update its own best individual in previous generations and start the improved particle swarm algorithm to reset the threshold to zero. Otherwise, keep its own best individual in previous generations unchanged, start the improved particle swarm algorithm to increment the threshold by 1, and compare the fitness values ​​of its own historical best individual with the global best individual. If the fitness of its own historical best individual is better than the fitness of the global best individual, then update the global best individual to its own historical best individual. Otherwise, keep the global historical best individual unchanged.

[0122] Step 11: After updating the positions of all individuals in the particle swarm once, the number of iterations is increased by 1. If the number of iterations is less than the upper limit of the number of iterations, return to step 6 and continue to update the position of each particle. If the number of iterations is greater than or equal to the upper limit of the number of iterations, the algorithm ends.

[0123] Step 12: Output the current global optimal individual as the chemical reaction kinetic parameters of the final post-processing model.

[0124] In this embodiment of the application, the particle swarm optimization algorithm is used to automatically find feasible calibrations that meet the data of each working condition by taking advantage of its high efficiency and parallelism. This can avoid the blindness of manual calibration, which is conducive to speed and efficiency, saving manpower costs and shortening the calibration cycle. Moreover, it can obtain the optimal calibration under the constraints (such as setting weighting coefficients), which can improve the prediction accuracy and robustness of the post-processing model.

[0125] Please see Figure 5 , Figure 5This is a structural block diagram of a vehicle emissions information prediction device according to an embodiment of this application. Figure 5 The device can be applied to electronic devices, such as Figure 5 The device may include an acquisition module 310 and a prediction module 320, wherein: the acquisition module 310 is used to acquire first information related to the original emissions under a first operating condition; the prediction module 320 is used to predict, based on the first information related to the original emissions, the second information related to the exhaust emissions under the first operating condition by using the chemical reaction kinetic parameters of the post-processing model; wherein the chemical reaction kinetic parameters of the post-processing model are obtained by optimization using a preset algorithm, the preset algorithm taking the minimization of the error between the measured exhaust emission information under the second operating condition and the exhaust emission information predicted by the post-processing model as the optimization objective, and iteratively solving to obtain the chemical reaction kinetic parameters.

[0126] In one possible implementation, the preset algorithm includes a particle swarm optimization algorithm, and the device further includes a determination module for determining the chemical reaction kinetic parameters of the post-processing model, specifically for: Acquire the calibration condition data, which includes third information related to the original emissions and fourth information related to the exhaust emissions under the second operating condition. The third information and the fourth information are the measured exhaust emission information. Get the initialized particle swarm; Based on the data of the operating conditions to be calibrated, the initial particle swarm is subjected to fitness iteration to obtain the particle parameter with the smallest fitness value under the condition that the iteration meets the preset conditions, and this parameter is taken as the global optimal parameter. The fitness value represents the error between the exhaust emission-related prediction information obtained by the post-processing model under the current particle parameters and the fourth information. The prediction information is based on the third information related to the original emissions under the second operating condition. The chemical reaction kinetic parameters in the post-processing model are determined based on the globally optimal parameters.

[0127] In one possible implementation, when the determining module is used to perform the acquisition of the initialized particle swarm, it is specifically used for: Initialize the particle swarm to obtain the first generation particle swarm. The dimension of the particle swarm is the same as the number of chemical reaction kinetic parameters to be calibrated in the post-processing model. The particle parameters of the particles in the particle swarm are candidate solutions for the chemical reaction kinetic parameters. The fitness of each particle in the first generation particle swarm is calculated to obtain the fitness value of each particle in the first generation particle swarm. Based on the fitness values ​​of each particle in the initial particle swarm, the parameter of the particle with the smallest fitness value in the initial particle swarm is obtained as the historical global optimal parameter.

[0128] In one possible implementation, when the determining module performs fitness iteration on the initialized particle swarm, it specifically performs the following operation repeatedly until the iteration satisfies a preset condition: For each particle in the particle swarm of the previous iteration, determine whether the number of iterations in which the particle parameters of the particle have not been updated has reached a preset threshold, and update the particle parameters of the particle using the corresponding particle swarm algorithm based on the determination result, so as to obtain the particle swarm of the current iteration. Based on the calibration data, the fitness of the particle swarm in the current iteration is evaluated to obtain the fitness value of each particle in the particle swarm in the current iteration. The particle swarm for the current iteration is updated based on the fitness value of each particle.

[0129] In one possible implementation, the particle parameters include particle position. The determination module is used to determine, for each particle in the particle swarm from the previous iteration, whether the number of consecutive iterations in which a particle's particle parameters have not been updated has reached a preset threshold. Based on the determination result, when updating the particle parameters of the particle using the corresponding particle swarm algorithm, it is specifically used for: If the number of iterations in which the particle parameters have not been updated consecutively does not reach a preset threshold, the particle's velocity and position are updated using the first particle swarm algorithm. The update of velocity and position is related to inertia factor, cognitive factor and social factor. Inertia factor is used to control the degree of continuity of the particle's velocity from the previous iteration. Cognitive factor is used to characterize the degree of influence of the particle's historical best position on the current update. Social factor is used to characterize the degree of influence of the historical global best position on the current update. If the number of iterations in which the particle parameters have not been updated for a consecutive period reaches a preset threshold, then the particle position is updated by randomly selecting one of at least two preset position calculation methods through the second particle swarm algorithm.

[0130] In one possible implementation, the determination module, when performing fitness evaluation of the particle swarm in the current iteration based on the calibration condition data, specifically performs the following operations for each particle: Using the post-processing model, under the particle parameters of the current iteration, the fifth information related to exhaust emissions under the second operating condition is predicted based on the third information related to the original emissions under the second operating condition. Obtain the first weighting coefficient for the second operating condition and the second weighting coefficients for at least two emissions included in the fourth information; Based on the first and second weighting coefficients, the weighted error between the fourth and fifth information is calculated, and the weighted error is used as the fitness value of the particle.

[0131] In one possible implementation, the determining module, when performing the particle swarm update based on the fitness values ​​of each particle in the current iteration, is specifically used for: For each particle, if the fitness value corresponding to the current particle parameter is less than the fitness value corresponding to the historical best particle parameter, then the historical best particle parameter of the particle is updated to the current particle parameter; otherwise, the particle parameter of the particle is not updated. From the current particle parameters of each particle in the particle swarm of the current iteration, obtain the target particle parameter with the smallest fitness value. If the fitness value corresponding to the target particle parameter is less than the fitness value corresponding to the historical global best parameter, then update the historical global best parameter to the target particle parameter; otherwise, do not update the historical global best parameter.

[0132] In one possible implementation, the determination module, after being used for particle parameter updates, is also used for: If the particle parameters of a particle have a dimension value that exceeds the preset boundary interval after the update, then the dimension value that exceeds the preset boundary interval will be updated based on the preset boundary interval. The preset boundary interval is determined based on the chemical reaction kinetic parameters to be calibrated in the post-processing model.

[0133] In one possible implementation, the second operating condition includes the operating condition corresponding to the scan point operating point where the engine speed is in a preset speed range and the torque is in a preset torque range. The data for the calibration condition includes at least one of the following related to the second operating condition: engine speed, engine torque, exhaust mass flow rate, exhaust temperature, and air-fuel ratio. Third information related to primary emissions includes at least one of the following: hydrocarbon primary emission concentration, nitrogen oxide primary emission concentration, and carbon monoxide primary emission concentration; The fourth piece of information related to exhaust emissions includes at least one of hydrocarbon exhaust emission concentration, nitrogen oxide exhaust emission concentration, and carbon monoxide exhaust emission concentration.

[0134] The apparatus in this embodiment can be described with reference to the above method, and will not be repeated here.

[0135] This application also provides an electronic device, please refer to... Figure 6 , Figure 6 The electronic device 400 shown includes a processor 410 and a memory 420, wherein the memory 410 is used to store computer programs; and the processor 420 is used to execute the programs stored in the memory 410 to implement the methods described in any embodiment of this application.

[0136] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in any embodiment of this application.

[0137] This application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the methods described in any embodiment of this application.

[0138] This application also provides a vehicle that includes electronic equipment as described in the above embodiments.

[0139] In this application, "multiple" refers to two or more.

[0140] In this application, unless otherwise expressly defined, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0141] The terms “first,” “second,” “third,” “fourth,” etc., in this application (if any) are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0142] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0143] Unless otherwise specified, all steps in this application may be performed sequentially or randomly. For example, if a method includes steps A and B, it means that the method may include steps A and B performed sequentially, or it may include steps B and A performed sequentially. For example, if a method may also include step C, it means that step C may be added to the method in any order. For example, the method may include steps A, B, and C, or it may include steps A, C, and B, or it may include steps C, A, and B, etc.

[0144] The above are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for predicting vehicle emissions information, characterized in that, include: Obtain initial information related to raw emissions under the first operating condition; Based on the first information related to the original emissions, predictions are made using the chemical reaction kinetic parameters of the post-processing model to obtain the second information related to exhaust emissions under the first operating condition. The chemical reaction kinetic parameters of the post-processing model are obtained by optimization using a preset algorithm. The preset algorithm aims to minimize the error between the measured exhaust gas emission information under the second operating condition and the exhaust gas emission information predicted by the post-processing model, and iteratively solves to obtain the chemical reaction kinetic parameters.

2. The method according to claim 1, characterized in that, The preset algorithm includes a particle swarm optimization algorithm, and the determination of the chemical reaction kinetic parameters of the post-processing model includes: Acquire calibration condition data, which includes third information related to the original emissions and fourth information related to exhaust emissions under the second operating condition, wherein the third information and the fourth information are the measured exhaust emission information; Get the initialized particle swarm; Based on the calibration data, the initial particle swarm is subjected to fitness iteration to obtain the particle parameter with the smallest fitness value under the condition that the iteration meets the preset conditions, and this parameter is used as the global optimal parameter. The fitness value represents the error between the exhaust emission-related prediction information obtained by the post-processing model under the current particle parameters and the fourth information. The prediction information is obtained based on the third information related to the original emissions under the second condition. The chemical reaction kinetic parameters in the post-processing model are determined based on the globally optimal parameters.

3. The method according to claim 2, characterized in that, The process of obtaining the initialized particle swarm includes: Initialize the particle swarm to obtain the first generation particle swarm. The dimension of the particle swarm is the same as the number of chemical reaction kinetic parameters to be calibrated in the post-processing model. The particle parameters of the particles in the particle swarm are candidate solutions for the chemical reaction kinetic parameters. The fitness of each particle in the first generation particle swarm is calculated to obtain the fitness value of each particle in the first generation particle swarm. Based on the fitness values ​​of each particle in the initial particle swarm, the parameter of the particle with the smallest fitness value in the initial particle swarm is obtained as the historical global optimal parameter.

4. The method according to claim 2, characterized in that, The fitness iteration of the initialized particle swarm includes repeatedly performing the following steps until the iteration meets a preset condition: For each particle in the particle swarm of the previous iteration, determine whether the number of iterations in which the particle parameters of the particle have not been updated has reached a preset threshold, and update the particle parameters of the particle using the corresponding particle swarm algorithm based on the determination result, so as to obtain the particle swarm of the current iteration. Based on the calibration data, the fitness of the particle swarm in the current iteration is evaluated to obtain the fitness value of each particle in the particle swarm in the current iteration. The particle swarm for the current iteration is updated based on the fitness value of each particle.

5. The method according to claim 4, characterized in that, The particle parameters include particle position. For each particle in the particle swarm from the previous iteration, determining whether the number of consecutive iterations in which the particle parameters of a particle have not been updated has reached a preset threshold, and updating the particle parameters of the particle using a corresponding particle swarm optimization algorithm based on the determination result, includes: If the number of iterations in which the particle parameters have not been updated consecutively does not reach a preset threshold, the particle's velocity and position are updated using the first particle swarm algorithm. The updates of velocity and position are related to inertia factor, cognitive factor, and social factor. The inertia factor is used to control the degree to which the particle continues the velocity of the previous iteration. The cognitive factor is used to characterize the influence of the particle's historical optimal position on the current update. The social factor is used to characterize the influence of the historical global optimal position on the current update. If the number of iterations in which the particle parameters have not been updated for a consecutive period reaches a preset threshold, then the particle position is updated by randomly selecting one of at least two preset position calculation methods through the second particle swarm algorithm.

6. The method according to claim 4, characterized in that, The fitness evaluation of the particle swarm in the current iteration based on the calibration data includes performing the following steps for each particle: Using the post-processing model, under the particle parameters of the current iteration, the fifth information related to exhaust emissions under the second operating condition is predicted based on the third information related to the original emissions under the second operating condition. Obtain the first weighting coefficient for the second operating condition and the second weighting coefficients for at least two emissions included in the fourth information; Based on the first weighting coefficient and the second weighting coefficient, the weighted error between the fourth information and the fifth information is calculated, and the weighted error is used as the fitness value of the particle.

7. The method according to claim 4, characterized in that, The process of updating the particle swarm for the current iteration based on the fitness value of each particle includes: For each particle, if the fitness value corresponding to the current particle parameter is less than the fitness value corresponding to the historical best particle parameter, then the historical best particle parameter of the particle is updated to the current particle parameter; otherwise, the particle parameter of the particle is not updated. From the current particle parameters of each particle in the particle swarm of the current iteration, obtain the target particle parameter with the smallest fitness value. If the fitness value corresponding to the target particle parameter is less than the fitness value corresponding to the historical global optimal parameter, then update the historical global optimal parameter to the target particle parameter; otherwise, do not update the historical global optimal parameter.

8. The method according to claim 4, characterized in that, After the particle parameters are updated, the following is also included: If, after updating the particle parameters, a dimension value exceeds a preset boundary interval, then based on the preset boundary interval, the dimension value exceeding the preset boundary interval is updated. The preset boundary interval is determined based on the chemical reaction kinetic parameters to be calibrated in the post-processing model.

9. An electronic device, characterized in that, It includes a processor and a memory, wherein: the memory is used to store computer programs; and the processor is used to execute the programs stored in the memory to implement the method as claimed in any one of claims 1-8.

10. A vehicle, characterized in that, Including the electronic device as claimed in claim 9.

Images