New energy truck carbon emission correction method and system based on multi-objective collaborative optimization

Abstract

The application relates to the technical field of carbon emission correction, and discloses a new energy truck carbon emission correction method and system based on multi-target collaborative optimization, which comprises the following steps: simulating a carbon emission process according to historical operation data of a new energy truck, and constructing a first carbon emission model; clustering truck working condition data in the historical operation data, constructing constraints for each category in combination with calculation accuracy, response speed and data dependency, and solving a static reference point set corresponding to the working condition through multi-target optimization; setting corresponding dynamic adjustment factors to obtain a corresponding dynamic reference point set; solving correction parameters corresponding to each preset carbon emission correction model through multi-target optimization, and constructing a correction parameter library; matching target correction parameter groups in the correction parameter library, and correcting the carbon emission of the new energy truck in real time; the application can adapt to real-time working conditions and driving modes, quickly match corresponding correction parameters, and improve the effectiveness and correction efficiency of the carbon emission correction process.

Description

Technical Field

[0001] This application relates to the field of carbon emission correction technology, and more specifically to a method and system for carbon emission correction of new energy freight vehicles based on multi-objective collaborative optimization. Background Technology

[0002] Currently, accurate and real-time measurement and correction of carbon emissions from new energy trucks has become the data foundation for carbon emission regulation. However, carbon emissions from new energy trucks are affected by multi-source emissions from the power system, complex operating conditions, and differences in driving modes. Traditional carbon emission measurement methods often rely on single data modeling and static reference benchmarks, which are difficult to adapt to actual operating scenarios, resulting in large errors and poor real-time performance in carbon emission data.

[0003] Existing technologies suffer from the following problems: carbon emission model construction methods are simplistic, often based on modifications of traditional fuel vehicle models or only calculating emissions from electricity consumption, failing to incorporate the composite power structure of new energy trucks into the modeling, thus failing to comprehensively cover the contributions of multiple emission sources and resulting in significant errors in carbon emission data; operating condition processing and reference point setting methods are simplistic, merely classifying operating condition data without considering actual needs such as response speed and data dependency, making it difficult to adapt to complex application scenarios; carbon emission correction using static empirical values ​​results in unstable correction effects, failing to meet the precise correction requirements in complex scenarios; To address at least one of the above problems, this application proposes a carbon emission correction method and system for new energy trucks based on multi-objective collaborative optimization. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this application is to provide a method and system for correcting carbon emissions from new energy freight vehicles based on multi-objective collaborative optimization, which can effectively solve the problems in the background technology. The specific technical solution of this application is as follows:

[0005] A method for correcting carbon emissions from new energy freight vehicles based on multi-objective collaborative optimization includes:

[0006] Based on historical operating data of new energy freight vehicles, the carbon emission process is simulated to construct the first carbon emission model;

[0007] Clustering of truck operating condition data from historical operation data, and constructing constraints for each cluster based on computational accuracy, response speed and data dependency, the static reference point set for the corresponding operating condition is obtained through multi-objective optimization;

[0008] Analyze the characteristics of truck driving modes under each working condition, set corresponding dynamic adjustment factors, and dynamically adjust the static reference point set to obtain the corresponding dynamic reference point set.

[0009] Combining the first carbon emission model, the dynamic reference point set, and operating condition data, the correction parameters corresponding to each preset carbon emission correction model are solved through multi-objective optimization. The solved correction parameters are then correlated to construct a correction parameter library.

[0010] Based on the real-time operating data of new energy trucks, target correction parameter groups are matched in the correction parameter library to correct the carbon emissions of new energy trucks in real time.

[0011] Specifically, the process of simulating carbon emission processes and constructing a first carbon emission model based on historical operating data of new energy freight vehicles includes:

[0012] Based on the power system structure of new energy trucks, the carbon emission process is broken down into engine emissions, electricity consumption emissions, and auxiliary emissions.

[0013] For each layer, a corresponding carbon emission sub-model is constructed. Based on the dynamic allocation strategy during truck operation, the carbon emission sub-models are weighted and fused to construct the first carbon emission model.

[0014] Specifically, the process involves clustering truck operating condition data from historical operational data, constructing constraints for each cluster based on computational accuracy, response speed, and data dependency, and then using multi-objective optimization to solve for the corresponding set of static reference points for each operating condition, including:

[0015] The truck operating condition data in the historical operation data is clustered to obtain multiple clusters, and each cluster corresponds to one operating condition;

[0016] Analyze the data distribution characteristics in each cluster, determine the boundary requirements for computational accuracy, response speed and data dependence under the corresponding working conditions, and construct constraints.

[0017] Based on the aforementioned constraints, the optimal solution set for the corresponding working condition is obtained through multi-objective optimization.

[0018] Inflection points in different directions are selected from the optimal solution set to identify the set of static reference points corresponding to the working conditions.

[0019] Specifically, inflection points in different directions are selected from the optimal solution set, and a set of static reference points corresponding to the working conditions is identified, including:

[0020] Calculate the degree of improvement of each solution and its neighboring solutions in different objective dimensions in the optimal solution set, and calculate the curvature characteristics of each solution;

[0021] The optimal solution whose curvature feature is greater than a preset curvature threshold is selected as a candidate inflection point, and a candidate inflection point set is obtained.

[0022] In the candidate inflection point set, the candidate inflection points are clustered according to different target dimensions, and the intersection of multiple clustering results is selected to obtain the static reference point set for the corresponding working condition.

[0023] Specifically, the analysis of truck driving mode characteristics for each working condition, setting corresponding dynamic adjustment factors, and dynamically adjusting the static reference point set to obtain the corresponding dynamic reference point set includes:

[0024] The truck driving mode for each working condition is analyzed, and the corresponding features are extracted through a preset feature extraction model to determine the adjustment direction and adjustment weight, thereby obtaining the dynamic adjustment factor.

[0025] The coordinates of each reference point in the static reference point set are dynamically weighted and adjusted according to the dynamic adjustment factor, and the range of coordinate adjustment is limited by constraints to obtain the corresponding dynamic reference point set.

[0026] Specifically, the analysis of truck driving modes under each working condition involves extracting corresponding features using a preset feature extraction model, determining the adjustment direction and adjustment weight, and obtaining dynamic adjustment factors, including:

[0027] The truck driving mode under each working condition is analyzed, and corresponding features are extracted through a preset feature extraction model. Feature vectors are constructed, the degree of speed change of the driving mode is analyzed, and the corresponding aggressiveness value is calculated.

[0028] The aggressiveness value is compared with a preset aggressiveness threshold to determine the adjustment direction. If the aggressiveness value is greater than the preset aggressiveness threshold, the adjustment direction is set to the response speed target. If the aggressiveness value is less than or equal to the preset aggressiveness threshold, the adjustment direction is set to the calculation accuracy target.

[0029] The integrity of data in each working condition is analyzed by calculating information entropy, the data quality score is calculated, and the adjustment weight of the data dependency target is determined.

[0030] Combining the adjustment direction and adjustment weight, a corresponding dynamic adjustment factor is generated through a preset vector generation model.

[0031] Specifically, combining the first carbon emission model, the dynamic reference point set, and operating condition data, the correction parameters for each preset carbon emission correction model are solved through multi-objective optimization. The solved correction parameters are then correlated to construct a correction parameter library, including:

[0032] Based on the distribution of each reference point in the dynamic reference point set, the space is divided into multiple subspaces, and each subspace corresponds to an optimization objective combination;

[0033] In each subspace, target weights for computational accuracy, response speed, and data dependence are set respectively. Based on the difference between the results of the first carbon emission model and the results of the preset carbon emission correction model, a mapping relationship between the correction parameters and the optimization objectives is established.

[0034] Based on the mapping relationship and target weights, the correction parameters corresponding to each preset carbon emission correction model are solved through multi-objective optimization. The solved correction parameters are then correlated to construct a correction parameter library.

[0035] Specifically, the step of matching a target correction parameter group in the correction parameter library based on the real-time operating data of the new energy trucks to perform real-time correction of the carbon emissions of the new energy trucks includes:

[0036] Based on the real-time operating data of new energy trucks, the corresponding operating condition features are extracted and an operating condition feature vector is constructed.

[0037] Based on the operating condition feature vector, the target correction parameter group is matched in the correction parameter library to correct the carbon emissions of new energy trucks in real time.

[0038] Specifically, based on the aforementioned operating condition feature vector, a target correction parameter group is matched in the correction parameter library to perform real-time correction of the carbon emissions of new energy freight vehicles, including:

[0039] Analyze the correction parameter information corresponding to different working conditions in the correction parameter library, and generate the corresponding working condition fingerprint vector;

[0040] Calculate the similarity between the working condition feature vector and the working condition fingerprint vector, and select correction parameters with similarity greater than a preset similarity threshold as a set of candidate parameters;

[0041] Based on the real-time operation strategy mode of new energy freight vehicles, the utility value of different parameter combinations in the candidate parameter set is calculated, and the parameter combination with the highest utility value is selected as the target correction parameter group to correct the carbon emissions of new energy freight vehicles in real time.

[0042] A carbon emission correction system for new energy freight vehicles based on multi-objective collaborative optimization is used to implement the aforementioned carbon emission correction method for new energy freight vehicles based on multi-objective collaborative optimization, including:

[0043] The carbon emission model building module simulates the carbon emission process based on historical operating data of new energy trucks and builds the first carbon emission model.

[0044] The static reference point filtering module clusters the truck operating condition data in the historical operation data. For each category, it constructs constraints based on calculation accuracy, response speed, and data dependency. Through multi-objective optimization, it solves for the set of static reference points for the corresponding operating condition.

[0045] The dynamic reference point adjustment module analyzes the characteristics of truck driving modes under each working condition, sets corresponding dynamic adjustment factors, and dynamically adjusts the static reference point set to obtain the corresponding dynamic reference point set.

[0046] The correction parameter library construction module combines the first carbon emission model, the dynamic reference point set, and the operating condition data to solve the correction parameters corresponding to each preset carbon emission correction model through multi-objective optimization, and then constructs the correction parameter library by associating the solved correction parameters.

[0047] The carbon emission correction module matches the target correction parameter group in the correction parameter library based on the real-time operating data of the new energy trucks, and corrects the carbon emissions of the new energy trucks in real time.

[0048] The beneficial effects of this application are as follows: By constructing a carbon emission model through hierarchical modeling and fusion, the accuracy of the carbon emission simulation process can be improved; a set of static reference points is obtained through working condition clustering and multi-objective optimization, and initial balance points of accuracy, speed, and data dependence are selected for different working conditions; dynamic adjustment factors are calculated based on driving mode characteristics, and the static reference points are adjusted in real time with weighted adjustments; an optimization subspace is divided based on the dynamic reference points, and a corresponding correction parameter library is constructed. Real-time parameter matching is performed by combining feature vector matching and utility value calculation, and the matched parameters are used to correct carbon emissions in real time. By determining static reference points through working condition clustering combined with constraints, and calculating corresponding dynamic adjustment factors based on the driving characteristics of each driving mode, the static reference points are dynamically adjusted to adapt to real-time working conditions and driving modes, quickly matching the corresponding correction parameters, and improving the effectiveness and efficiency of the carbon emission correction process. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the carbon emission correction method for new energy freight vehicles based on multi-objective collaborative optimization in Embodiment 1 of this application.

[0050] Figure 2 This is a schematic diagram of the set of static reference points in Embodiment 1 of this application;

[0051] Figure 3 This is a flowchart illustrating the dynamic adjustment factor calculation process in Embodiment 1 of this application;

[0052] Figure 4 This is a schematic diagram of the structure of the new energy truck carbon emission correction system based on multi-objective collaborative optimization in Embodiment 1 of this application. Detailed Implementation

[0053] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. In the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or advantageous than other embodiments or design options. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0054] Hereinafter, the terms "first," "second," and other generic terms are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0055] Example 1:

[0056] refer to Figure 1 The image shows a specific implementation of the carbon emission correction method for new energy freight vehicles based on multi-objective collaborative optimization, as described in this application, including:

[0057] S101. Based on the historical operating data of new energy freight vehicles, simulate the carbon emission process and construct the first carbon emission model;

[0058] S102. Cluster the truck operating condition data in the historical operation data, and construct constraints for each class by combining calculation accuracy, response speed and data dependency. Solve the static reference point set for the corresponding operating condition through multi-objective optimization.

[0059] S103. Analyze the characteristics of the truck driving mode for each working condition, set the corresponding dynamic adjustment factor, and dynamically adjust the static reference point set to obtain the corresponding dynamic reference point set.

[0060] S104. Combining the first carbon emission model, the dynamic reference point set, and the operating condition data, the correction parameters corresponding to each preset carbon emission correction model are solved through multi-objective optimization. The solved correction parameters are then associated to construct a correction parameter library.

[0061] S105. Based on the real-time operating data of the new energy trucks, match the target correction parameter group in the correction parameter library to correct the carbon emissions of the new energy trucks in real time.

[0062] In the transportation sector, the promotion of new energy freight vehicles is a key measure to reduce carbon emissions; however, the accuracy of monitoring and correcting carbon emissions from new energy freight vehicles cannot meet the needs of practical applications. Most existing correction methods rely on simple average emission factors or static operating condition mapping tables, ignoring the dynamic switching characteristics of the hybrid power source in new energy freight vehicles and the impact of driver behavior on the emission process. Traditional single-objective optimization strategies struggle to meet the multiple requirements of real-time computation, result accuracy, and data robustness in practical applications, resulting in low correction accuracy for the carbon emission correction process of new energy vehicles and an inability to provide accurate carbon emission data support.

[0063] In this embodiment, based on historical operating data of new energy trucks, the carbon emission process is decomposed into three levels: emissions from the direct combustion of fuel by the engine, indirect emissions from the consumption of grid electricity, and emissions from the operation of auxiliary systems such as air conditioning. For each level, a corresponding sub-model is constructed to simulate the carbon emission process. Dynamic weights are assigned to each sub-model according to the energy allocation strategies of different power sources during actual truck operation. The sub-models are then weighted and fused to construct the first carbon emission model. Through hierarchical modeling and dynamic fusion, errors caused by simplifying the new energy truck as a single emission source can be reduced. The first carbon emission model allows for the analysis of the complex carbon emission process of new energy trucks under multi-mode switching, providing accurate data support for the carbon emission process, improving the realism and reliability of the carbon emission simulation process under complex operating conditions, and enhancing the accuracy of the carbon emission correction process.

[0064] Specifically, the truck operating condition data from historical operational data is clustered. For each cluster, constraints are constructed based on computational accuracy, response speed, and data dependency. These three objectives are used as optimization directions to construct a multi-objective optimization problem. The static reference point set for the corresponding operating condition is obtained through multi-objective optimization. By combining clustering and multi-objective optimization, a set of static reference points that achieve the best balance between computational accuracy, response speed, and data dependency is selected for each typical operating condition. This avoids the limitation of using a single parameter to address all scenarios and provides high-quality data references for dynamic correction and adjustment processes.

[0065] Furthermore, the driving mode characteristics of trucks under each operating condition are analyzed, and the aggressiveness of speed changes under the corresponding operating condition is calculated to quantify the driver's aggressive or moderate driving tendency. Corresponding dynamic adjustment factors are set, and the static reference point set is dynamically adjusted to obtain the corresponding dynamic reference point set. By calculating the dynamic adjustment factors, the reference points can be dynamically adjusted according to the real-time driving mode. Under the premise of ensuring optimal overall performance, the corresponding needs can be dynamically prioritized, thereby significantly improving the practicality and robustness of the carbon emission correction process in real and variable driving environments.

[0066] Specifically, by combining the primary carbon emission model, the dynamic reference point set, and operating condition data, the correction parameters for each preset carbon emission correction model are solved through multi-objective optimization. These correction parameters are then correlated with the corresponding operating condition characteristics and dynamic reference point information to construct a correction parameter library. This library provides a rich selection of parameters for real-time carbon emission correction, allowing for rapid matching of appropriate parameters from the library based on real-time conditions to correct carbon emissions. The optimized parameter combinations can then be quickly invoked, improving the efficiency and effectiveness of the carbon emission correction process.

[0067] Specifically, based on the real-time operating data of new energy freight vehicles, target correction parameter sets are matched in the correction parameter library to correct the carbon emissions of the new energy freight vehicles in real time. By quickly matching the corresponding parameter sets in the correction parameter library, the most suitable correction parameters can be matched by combining the real-time operating status and corresponding working conditions of the new energy freight vehicles, so as to obtain accurate carbon emission data that meets the operating targets and improve the correction effect and accuracy of the carbon emission process.

[0068] This application improves the accuracy of carbon emission simulation by constructing a carbon emission model through hierarchical modeling and fusion. It obtains a set of static reference points through operating condition clustering and multi-objective optimization, selecting initial balance points for accuracy, speed, and data dependence for different operating conditions. Dynamic adjustment factors are calculated based on driving mode characteristics, and the static reference points are adjusted in real-time with weighted adjustments. An optimization subspace is partitioned based on the dynamic reference points, and a corresponding correction parameter library is constructed. Real-time parameter matching is performed using feature vector matching and utility value calculation, and the matched parameters are used to correct carbon emissions in real-time. By determining static reference points through operating condition clustering combined with constraints, and calculating corresponding dynamic adjustment factors based on driving characteristics of each driving mode, the static reference points are dynamically adjusted to adapt to real-time operating conditions and driving modes, quickly matching corresponding correction parameters, and improving the effectiveness and efficiency of the carbon emission correction process.

[0069] Furthermore, based on historical operating data of new energy freight vehicles, a carbon emission process is simulated to construct the first carbon emission model, including:

[0070] S201. Based on the power system structure of new energy trucks, the carbon emission process is decomposed into engine emissions, electricity consumption emissions, and auxiliary emissions.

[0071] S202. Construct a corresponding carbon emission sub-model for each layer. Based on the dynamic allocation strategy during truck operation, perform weighted fusion of the carbon emission sub-models to construct the first carbon emission model.

[0072] In this embodiment, based on the power system structure of the new energy truck, the power system topology of the new energy truck is analyzed to identify all components or energy flow paths that directly or indirectly generate carbon emissions. The carbon emission process of the entire vehicle is divided into three levels: engine emissions, which directly produce greenhouse gases such as carbon dioxide during the combustion of fossil fuels; electricity consumption emissions, which generate carbon emissions at the production end when the vehicle obtains electrical energy from the grid and uses it to drive the motor; and auxiliary carbon emissions, which are generated by the energy consumed by auxiliary systems to ensure the normal operation of the vehicle rather than being directly used for driving. By structuring the carbon emission process hierarchically, the problem of significant errors caused by treating the truck as a whole and using a single emission model for rough estimation can be avoided. By adopting corresponding modeling rules for each level, the accuracy of the carbon emission analysis process can be improved, providing an accurate data foundation for carbon emission correction.

[0073] Specifically, a corresponding carbon emission sub-model is constructed for each layer. For engine emissions, an engine emission sub-model is constructed based on the engine's instantaneous fuel consumption rate and the corresponding carbon emission factor. For electricity consumption emissions, an electricity consumption emission sub-model is constructed based on the instantaneous electrical power obtained by the vehicle from the power grid and the power grid carbon emission factor reflecting the average carbon emission level of the regional power grid. For auxiliary emissions, an auxiliary emission sub-model is constructed based on the rated power or measured power of the auxiliary system and its duty cycle or start-stop state. The weights of each sub-model are dynamically determined according to the energy dynamic allocation strategy adopted by the vehicle controller in real-time operation. For example, in pure electric mode, the weight of the engine emission sub-model is zero, and the weight of the electricity consumption emission sub-model is close to 1. The engine emission sub-model, electricity consumption emission sub-model, and auxiliary emission sub-model are weighted and fused according to their corresponding weights to obtain the first carbon emission model.

[0074] It should be noted that by constructing sub-models in layers and integrating them to form the first carbon emission model, the changes in energy flow paths of vehicles under different operating conditions and control strategies can be accurately captured. The real-time contribution of each emission source can be calculated dynamically, continuously and accurately, improving the model calculation accuracy of the carbon emission process and providing accurate data support for the real-time correction and optimization of carbon emissions.

[0075] Furthermore, the truck operating condition data from historical operation data is clustered. For each cluster, constraints are constructed based on computational accuracy, response speed, and data dependency. A set of static reference points for the corresponding operating condition is then obtained through multi-objective optimization, including:

[0076] S301. Cluster the truck operating condition data in the historical operation data to obtain multiple clusters, each cluster corresponding to one operating condition;

[0077] S302. Analyze the data distribution characteristics in each cluster, determine the boundary requirements for calculation accuracy, response speed and data dependence under the corresponding working conditions, and construct constraints.

[0078] S303. Based on the constraints, the optimal solution set for the corresponding working condition is obtained through multi-objective optimization.

[0079] S304. Select inflection points in different directions from the optimal solution set and identify the set of static reference points corresponding to the working conditions.

[0080] In this embodiment, based on truck operating condition data from historical operational data, feature parameters that characterize the essence of the operating conditions are extracted and feature vectors are constructed. These feature parameters include, but are not limited to, average vehicle speed, standard deviation of vehicle speed, and average acceleration. Based on these feature vectors, AP clustering is used to obtain multiple clusters, each corresponding to a typical truck operating condition. By transforming a large amount of raw operational data into a set of corresponding typical operating conditions through clustering, the amount of data in the carbon emission analysis process can be reduced, computational efficiency improved, and in-depth analysis and parameter optimization can be performed on representative operating condition patterns, enhancing the effectiveness and computational efficiency of the analysis process.

[0081] Specifically, the data distribution characteristics within each cluster are analyzed to determine the boundary requirements for computational accuracy, response speed, and data dependency under corresponding operating conditions. For example, for high-speed cruising conditions with high average vehicle speed and gradual speed changes, a higher lower limit can be set for the computational accuracy of carbon emission calculations due to the stable state. This lower limit is calculated based on the operating condition information. Simultaneously, due to the high vehicle speed, the system is required to respond quickly to state changes, necessitating a strict maximum delay limit for the response speed. For urban congestion conditions with high data noise and drastic feature changes, a minimum data completeness threshold needs to be set. This minimum data completeness threshold is a minimum standard used to quantitatively evaluate the completeness of operating condition data within a specific time window; its specific value is set based on engineering experience and reliability requirements. For example, for urban congestion conditions with high data noise and drastic feature changes, this threshold can be set to 90%. By calculating the variance of speed and acceleration, and the statistical data on missing data rates within the clusters, and combining the calculated data information with the carbon emission correction accuracy requirements, corresponding conditional thresholds are set to construct constraints. By constructing constraints, the optimal solution can be found within a specific feasible domain for each operating condition, avoiding the problem of some operating conditions having excessive performance while others have substandard performance due to differences in operating conditions. This improves the applicability of the optimization results to each typical operating scenario and enhances the computational performance of the carbon emission process.

[0082] For example, the conditional thresholds include, but are not limited to, the lower limit of calculation accuracy, the upper limit of response speed, and the data completeness threshold. For high-speed cruise conditions, the upper limit of the allowable relative error of the carbon emission calculation results can be set to 2%. For high-speed cruise conditions, in order to ensure that the system can quickly follow the state changes, the maximum processing delay from data input to correction result output can be set to 100 milliseconds. For urban congestion conditions, in order to ensure basic reliability in an environment where data is susceptible to interference, the minimum data completeness threshold can be set to 85%.

[0083] Specifically, based on the constructed constraints, the optimal solution set for the corresponding operating conditions is obtained through multi-objective optimization. For each specific cluster of operating conditions, the constructed constraints are used as constraints for the optimization problem, and the multi-objective optimization method is used to solve the optimization problem. Specifically, a sorting genetic algorithm is used to initialize a random population of candidate solutions, each solution representing a specific set of system parameter configurations. During the iteration process, the algorithm continuously generates new solutions by simulating selection, crossover, and mutation operations in biological evolution, and performs non-dominated sorting of all solutions according to Pareto dominance. At the same time, crowding is calculated to maintain the diversity of solutions in the objective space. After the algorithm converges, a set of Pareto optimal solutions is output. Through automatic matching search, a comprehensive and evenly distributed set of Pareto optimal solutions is found for each type of operating condition, providing a large amount of data support for the carbon emission correction process and improving the efficiency and effectiveness of carbon emission correction.

[0084] Specifically, inflection points in different directions are selected from the set of optimal solutions, and a set of static reference points for the corresponding operating conditions is identified. A small number of representative key reference points are selected from a large set of Pareto optimal solutions. These selected static reference points reflect the critical states of switching between different optimization directions, which can reduce the complexity of real-time matching and decision-making, improve the quality of static reference points, and enhance the effectiveness of the carbon emission correction process.

[0085] Furthermore, inflection points in different directions are selected from the optimal solution set to identify the corresponding set of static reference points for the working conditions, including:

[0086] S401. Calculate the degree of improvement of each solution and its neighboring solutions in different target dimensions in the optimal solution set, and calculate the curvature characteristics of each solution;

[0087] S402. Select the optimal solution whose curvature feature is greater than the preset curvature threshold as candidate inflection points to obtain a set of candidate inflection points;

[0088] S403. In the candidate inflection point set, the candidate inflection points are clustered according to different target dimensions, and the intersection of multiple clustering results is selected to obtain the static reference point set corresponding to the working condition.

[0089] In this embodiment, the improvement degree of each solution and its neighboring solutions in different target dimensions is calculated in the optimal solution set, and the curvature feature of each solution is calculated. Based on three dimensions—computational accuracy, response speed, and data dependency—topological structure analysis is performed on the optimal solution set. Within the optimal solution set, the Euclidean distance between solutions is calculated, and the nearest solution is selected to determine the neighboring solutions of each solution. The improvement degree of each solution and its neighboring solutions in each target dimension is analyzed. For example, for computational accuracy and response speed, the change in the computational accuracy target value when the computational response speed target value changes by one unit is calculated, and the rate of change of this change among neighboring solutions is calculated. The rates of change between computational accuracy and response speed, computational accuracy and data dependency, and response speed and data dependency are calculated separately. The average of these rates of change is then used to calculate the curvature feature of each solution. By calculating the local curvature feature of each solution, key points on the Pareto front can be quickly identified, providing accurate data support for the rapid selection of key reference points, improving the speed and quality of static reference point selection, and thus improving the quality of carbon emission correction.

[0090] Specifically, optimal solutions with curvature features greater than a preset curvature threshold are selected as candidate inflection points, resulting in a candidate inflection point set. Based on the statistical distribution characteristics of the curvature features of all solutions, the top 20% quantiles of the curvature feature value distribution are selected as the curvature threshold. The curvature features of all solutions are compared with the curvature threshold, and solutions with curvature feature values ​​greater than the threshold are selected as candidate inflection points. The selected solutions indicate a dramatic change in direction at their location on the Pareto front, reflecting significant turning points or boundary points between different optimization tendencies. The candidate inflection points are combined to construct the candidate inflection point set. Threshold screening ensures that the selected candidate inflection points are representative solutions in terms of the objective trade-off relationship, quickly and effectively filtering out solutions located in the approximately linear trade-off region and those with weak representativeness, reducing the amount of data processed subsequently. At the same time, it ensures that each point in the final determined static reference point set corresponds to a clear performance trade-off strategy, improving the efficiency and effectiveness of the carbon emission correction process.

[0091] like Figure 2As shown, candidate inflection points are clustered according to different objective dimensions within the candidate inflection point set. The intersection of multiple clustering results is then selected to obtain the static reference point set for the corresponding operating condition. For computational accuracy, response speed, and data dependency, AP clustering is performed on candidate inflection points using the value of a single optimization objective as a feature. Inflection points with similar values ​​on the corresponding objective are grouped together. Clustering is then performed sequentially for the three objective dimensions of computational accuracy, response speed, and data dependency, resulting in three distinct representative inflection point subsets. The intersection of these three subsets is then calculated. Inflection points in the intersection represent points that are identified as representative from different optimization objective perspectives. These inflection points are then determined as the static reference point set for the corresponding operating condition. By selecting the intersection through multiple clustering, the comprehensiveness and representativeness of the static reference point set can be improved. This effectively avoids the problem of static reference points being overly concentrated on a single optimization dimension, ensuring that the final reference point set can cover every optimization direction. The selected static reference point set is relatively small in number and can completely represent all key performance trade-off patterns under the corresponding operating condition, thus improving the efficiency and quality of the carbon emission correction process.

[0092] Furthermore, the characteristics of truck driving modes under each working condition are analyzed, corresponding dynamic adjustment factors are set, and the static reference point set is dynamically adjusted to obtain the corresponding dynamic reference point set, including:

[0093] S501. Analyze the truck driving mode for each working condition, extract the corresponding features through the preset feature extraction model, determine the adjustment direction and adjustment weight, and obtain the dynamic adjustment factor.

[0094] S502. The coordinates of each reference point in the static reference point set are dynamically weighted and adjusted according to the dynamic adjustment factor, and the range of coordinate adjustment is limited by the constraint conditions to obtain the corresponding dynamic reference point set.

[0095] In this embodiment, the static reference point reflects the category reference information of the working condition. However, different driving behaviors under the same working condition will have a significant impact on the real-time requirements of carbon emission calculation. By analyzing the truck driving mode of each working condition, the corresponding features are extracted through a preset feature extraction model to determine the adjustment direction and adjustment weight, and the corresponding dynamic adjustment factor is obtained. By calculating the dynamic adjustment factor, the driving behavior mode is transformed into a specific adjustment instruction. By combining the static working condition and the driver's dynamic driving mode, precise dynamic adjustment can be performed, which can improve the accuracy and environmental adaptability of the carbon emission correction process.

[0096] Specifically, the input data for the feature extraction model is a multi-dimensional time-series signal, including a standardized vehicle speed sequence, a longitudinal acceleration sequence, and a torque request percentage sequence of the motor / engine. The model can adopt a simple three-layer one-dimensional convolutional neural network structure. The first CNN layer uses multiple filters to convolve in the time dimension to extract local temporal patterns. The second layer is a pooling layer to reduce the time dimension. The third layer is a fully connected layer that maps the features into a fixed-length feature vector. Finally, a scalar value, i.e., the aggression level value, is output through a sigmoid activation function. The training data comes from a large number of labeled historical driving data segments. These segments are labeled by experienced engineers or by clustering algorithms combined with the frequency of vehicle deceleration and acceleration events, classifying driving behavior into aggressive (label 1) and stable (label 0).

[0097] Furthermore, the coordinates of each reference point in the static reference point set are dynamically weighted and adjusted according to the dynamic adjustment factor, and the adjustment range is limited by constraints to obtain the corresponding dynamic reference point set. For each static reference point, a vector is constructed based on its data in terms of computational accuracy, response speed, and data dependency. This vector is then dynamically weighted according to the adjustment direction and weight of the dynamic adjustment factor. For example, if the adjustment direction points to response speed, the coordinate values ​​of that reference point in the response speed dimension are weighted and enhanced. During the dynamic adjustment process, to prevent over-adjustment leading to system performance loss of control, the new coordinates of each dynamic reference point during and after the adjustment are checked against the constraints to ensure they still meet the requirements. For cases exceeding the boundaries, the corresponding coordinates are pulled back to the nearest feasible boundary point, and the dynamic reference point set is output. Through the dynamic weighting and constraint checking process, it can be ensured that the reference points can adapt to different operating conditions and corresponding driving modes, improving the environmental adaptability of the carbon emission correction process and enhancing the accuracy of carbon emission correction.

[0098] like Figure 3 As shown, the truck driving mode under each working condition is analyzed. Corresponding features are extracted using a pre-set feature extraction model to determine the adjustment direction and weight, resulting in dynamic adjustment factors, including:

[0099] S601. Analyze the truck driving mode for each working condition, extract the corresponding features through the preset feature extraction model, construct the feature vector, analyze the speed change of the driving mode, and calculate the corresponding aggressiveness value.

[0100] S602. Compare the aggressiveness value with the preset aggressiveness threshold to determine the adjustment direction. If the aggressiveness value is greater than the preset aggressiveness threshold, the adjustment direction is set to the response speed target. If the aggressiveness value is less than or equal to the preset aggressiveness threshold, the adjustment direction is set to the calculation accuracy target.

[0101] S603. Analyze the integrity of data in each working condition through information entropy calculation, calculate the data quality score, and determine the adjustment weight of the data dependency target.

[0102] S604. Combining the adjustment direction and adjustment weight, a corresponding dynamic adjustment factor is generated through a preset vector generation model.

[0103] In this embodiment, the truck driving mode for each working condition is analyzed. Corresponding features are extracted using a preset feature extraction model to construct a feature vector. The degree of speed variation in the driving mode is analyzed, and the corresponding aggression value is calculated. For each truck driving data condition, corresponding features are extracted using a preset feature extraction model. This model includes, but is not limited to, a neural network model pre-trained using a large amount of historical truck driving data. The extracted features include, but are not limited to, the average acceleration, the standard deviation of acceleration, the number of rapid acceleration events, and the number of rapid deceleration events. These features are combined to construct a feature vector. Based on the feature vector, the intensity of driving behavior and the degree of speed variation in the driving mode are analyzed. The model calculates and outputs the corresponding aggression value. By extracting and analyzing the corresponding driving modes, accurate data references are provided for calculating dynamic adjustment factors, improving the accuracy of driving behavior analysis.

[0104] Specifically, the aggressiveness level is compared with a preset aggressiveness threshold to determine the adjustment direction. If the aggressiveness level is greater than the preset threshold, the adjustment direction is set to the response speed target; if the aggressiveness level is less than or equal to the preset threshold, the adjustment direction is set to the calculation accuracy target. An aggressiveness threshold is set by analyzing the clustering boundaries between safe, economical, and aggressive driving in historical driving data. The aggressiveness level is compared with this threshold. If the aggressiveness level is greater than the threshold, the current mode is determined to be aggressive driving, with rapid changes in vehicle status requiring the system to quickly track and respond. In this case, the adjustment direction prioritizes ensuring response speed, and dynamic adjustments tend to improve system response speed. If the aggressiveness level is less than or equal to the threshold, the current mode is determined to be stable driving, with lower real-time requirements and higher requirements for carbon emission calculation accuracy. In this case, the adjustment direction prioritizes ensuring calculation accuracy. By linking driving behavior perception with the system's core optimization strategies, speed can be prioritized when rapid response is needed, and accuracy can be prioritized when the state is stable. This ensures that the dynamic adjustment process always aligns with the corresponding application requirements, improving the accuracy of carbon emission correction.

[0105] For example, by performing cluster analysis on normalized features representing driving style from a large amount of historical driving data (including but not limited to acceleration standard deviation and frequency of rapid acceleration and deceleration), a numerical boundary is identified between the centers of the economical and safe driving cluster and the aggressive driving cluster. For instance, the calculated normalized aggression values ​​range from 0 to 1, with the aggression values ​​of the economical and safe driving cluster mainly distributed in the 0-0.6 range, while the values ​​of the aggressive driving cluster are mainly distributed in the 0.7-1.0 range. The midpoint of the boundary between the two clusters is taken as a threshold, and a preset aggression threshold of 0.65 is set. When the real-time calculated aggression value is greater than 0.65, it is determined to be aggressive driving, and the direction adjustment is set to prioritize response speed; conversely, the direction adjustment is set to prioritize calculation accuracy.

[0106] Specifically, the integrity of data within each operating condition is analyzed through information entropy calculation, a data quality score is calculated, and the adjustment weights for data dependency targets are determined. The data is divided into several time units, and each unit is checked for invalid conditions such as data loss, abnormal value ranges, or persistent instability. The corresponding information entropy is calculated; the more complete the data and the more consistent the changes with expectations, the higher the entropy value; the more missing or abnormal data there is, the lower the entropy value. The calculated information entropy value is normalized to between 0 and 1 to obtain a data quality score; the higher the score, the better the data quality. The adjustment weights for data dependency targets are set based on the data quality score, calculated as the reciprocal of the data quality score. By calculating information entropy and dynamically adjusting the weights for data dependency, accurate calculations can be performed by fully utilizing reliable data sources, and a lower data dependency operating mode can be switched when data quality deteriorates, improving adaptability to in-vehicle environments with data interference.

[0107] Specifically, by combining the adjustment direction and adjustment weights, a pre-defined vector generation model generates corresponding dynamic adjustment factors. The vector generation model selects the appropriate adjustment vector template based on the input adjustment direction and weights. If the adjustment direction is biased towards speed, the base vector has a positive weight in the response speed dimension and a negative weight in the calculation accuracy dimension. The adjustment weight values ​​of the data dependency target are integrated into the base vector, and the corresponding weight values ​​are added to the data dependency dimension of the vector to obtain the dynamic adjustment factor. By constructing dynamic adjustment factors, a comprehensive adjustment strategy reflecting real-time driving characteristics and data environment status can be achieved, providing accurate data references for the dynamic adjustment process and improving adjustment efficiency and effectiveness.

[0108] Specifically, the vector generation model input includes adjustment direction identifiers and data dependency adjustment weights, and the model output is a dynamic adjustment factor. In this embodiment, the dynamic adjustment factor can be defined as a triple, including a tendency coefficient for offsetting the three optimization objectives of computational accuracy, response speed, and data dependency. The correlation between input and output is based on the adjustment direction identifier to set a basic tendency, and the data dependency adjustment weights are incorporated. The corresponding tendency coefficients are calculated by weighting to obtain the dynamic adjustment factor, which clearly indicates the positive and negative directions and relative magnitudes of the adjustments in each objective dimension.

[0109] Furthermore, combining the first carbon emission model, the dynamic reference point set, and operating condition data, the correction parameters corresponding to each preset carbon emission correction model are solved through multi-objective optimization. The solved correction parameters are then correlated to construct a correction parameter library, including:

[0110] S701. Based on the distribution of each reference point in the dynamic reference point set, the space is divided into multiple subspaces, and each subspace corresponds to an optimization objective combination;

[0111] S702. Set target weights for computational accuracy, response speed and data dependence in each subspace, and establish a mapping relationship between correction parameters and optimization objectives by combining the differences between the results of the first carbon emission model and the results of the preset carbon emission correction model.

[0112] S703. Based on the mapping relationship and target weight, the correction parameters corresponding to each preset carbon emission correction model are solved through multi-objective optimization. The solved correction parameters are associated to construct a correction parameter library.

[0113] In this embodiment, based on the distribution of each reference point in the dynamic reference point set, the space is divided into multiple subspaces, each corresponding to an optimization objective combination. Each dynamic reference point is used as a generation point, and the entire objective space is divided into multiple units. Each unit contains all spatial locations that are closest to the dynamic reference points within that unit, resulting in multiple subspaces. In each subspace, the performance tendency represented by the corresponding dynamic reference point is taken as the core feature, corresponding to a specific optimization objective combination. By dividing the space, the global multi-objective optimization problem is decomposed into multiple local subproblems with clear objective orientation and limited search space. This avoids searching in the global space, improves the efficiency and specificity of the optimization process, and enhances the efficiency and accuracy of parameter solving.

[0114] Specifically, target weights for computational accuracy, response speed, and data dependency are set in each subspace. A mapping relationship between correction parameters and optimization objectives is established by combining the differences between the results of the first carbon emission model and the results of the preset carbon emission correction model. For each subspace, corresponding target weights are assigned to the three optimization objectives based on the core characteristics of that subspace. The output of the first carbon emission model under the given operating conditions is used as a high-precision benchmark value, and the output of the preset carbon emission correction model under the same data is used as the value to be corrected. The difference between the benchmark value and the value to be corrected is calculated, and a mapping relationship between correction parameters and optimization objectives is constructed based on this difference. This difference is represented as a function of the correction model parameters. The difference is associated with the computational accuracy objective, the computational complexity of the correction model is associated with the response speed objective, and the sensitivity of the correction model to the input data dimension is associated with the data dependency objective. By adjusting the underlying model parameters, the impact on system performance can be accurately assessed, improving the accuracy of the solved correction parameters while simultaneously improving the overall system performance.

[0115] Specifically, based on mapping relationships and target weights, multi-objective optimization is used to solve for the correction parameters corresponding to each preset carbon emission correction model. These corrective parameters are then correlated to construct a correction parameter library. For each subspace and its corresponding preset carbon emission correction model, a multi-objective optimization problem is determined based on the mapping relationship and target weights. This problem is solved using a multi-objective evolutionary algorithm, quickly converging to one or a few optimal solutions under the conditions of that subspace. The correction parameters for this correction model under these conditions are then obtained, and the correction parameters corresponding to the optimal solutions are combined to construct the correction parameter library. By constructing the correction parameter library, the optimal correction parameters for the current state can be quickly obtained from the library, reducing computation time, improving system response speed, and ensuring that correction performance remains at a high level, thus improving the overall system correction performance.

[0116] Furthermore, based on the real-time operating data of the new energy trucks, a target correction parameter group is matched in the correction parameter library to perform real-time correction of the carbon emissions of the new energy trucks, including:

[0117] S801. Based on the real-time operating data of new energy freight vehicles, extract the corresponding operating condition features and construct an operating condition feature vector.

[0118] S802. According to the operating condition feature vector, match the target correction parameter group in the correction parameter library to correct the carbon emissions of new energy trucks in real time.

[0119] In this embodiment, based on the real-time operating data of the new energy truck, corresponding operating condition features are extracted to construct an operating condition feature vector. These features include, but are not limited to, average vehicle speed, standard deviation of vehicle speed, maximum acceleration, and average acceleration. The extracted operating condition features are combined sequentially to obtain the operating condition feature vector. By constructing the operating condition feature vector, the dynamically changing and complex vehicle operating state is transformed into vector data, improving the computational and correction efficiency of the correction process.

[0120] Specifically, based on the aforementioned operating condition feature vector, a target correction parameter group is matched in the correction parameter library to perform real-time correction of carbon emissions from new energy trucks. By performing vector matching in the correction parameter library, complex optimization calculations can be avoided in resource-constrained on-board environments. Correction parameters that are adapted to the current state and have undergone global optimization can be directly matched, improving the accuracy and real-time performance of carbon emission correction results. This approach can adapt to different driving styles and operating strategies, thereby enhancing the accuracy and efficiency of the carbon emission correction process.

[0121] Furthermore, based on the aforementioned operating condition feature vector, a target correction parameter group is matched in the correction parameter library to perform real-time correction of the carbon emissions of new energy freight vehicles, including:

[0122] S901. Analyze the correction parameter information corresponding to different working conditions in the correction parameter library, and generate the corresponding working condition fingerprint vector.

[0123] S902. Calculate the similarity between the working condition feature vector and the working condition fingerprint vector, and select correction parameters with similarity greater than the preset similarity threshold as a set of candidate parameters.

[0124] S903. Based on the real-time operation strategy mode of new energy freight vehicles, calculate the utility value of different parameter combinations in the candidate parameter set, select the parameter combination with the highest utility value as the target correction parameter group, and make real-time corrections to the carbon emissions of new energy freight vehicles.

[0125] In this embodiment, the correction parameter information corresponding to different working conditions is analyzed in the correction parameter library to generate corresponding working condition fingerprint vectors. Each correction parameter information corresponding to different working conditions in the correction parameter library is standardized and normalized. Principal component analysis is then used for feature dimensionality reduction to extract the corresponding feature components, generating the corresponding working condition fingerprint vector. By generating working condition fingerprint vectors, data references are provided for the real-time and rapid parameter matching process, improving the efficiency and effectiveness of parameter matching.

[0126] Specifically, the similarity between the operating condition feature vector and the operating condition fingerprint vector is calculated, and correction parameters with similarity greater than a preset similarity threshold are selected as candidate parameters. The operating condition feature vector undergoes the same normalization and dimensionality reduction processing to ensure its dimension matches that of the operating condition fingerprint vector. The cosine similarity between the two vectors is then calculated. A similarity threshold is set based on the accuracy requirements of the carbon emission correction process, and correction parameters with cosine similarity greater than this threshold are selected to construct the candidate parameter set. Through similarity comparison, correction parameters highly correlated with the current real-time operating conditions can be quickly selected, and the dimensionality of these parameters can be rapidly reduced, improving the efficiency and effectiveness of the parameter matching process.

[0127] For example, the preset similarity threshold can be set based on the success rate statistics of historical matching experiments. Analysis of a large number of known operating conditions reveals that when the cosine similarity is greater than 0.85, the overall compliance rate of the matched parameter set in terms of correction accuracy and response performance under that operating condition exceeds 95%. Therefore, the preset similarity threshold is set to 0.85. During real-time matching, only when the cosine similarity between the real-time operating condition feature vector and a certain operating condition fingerprint vector is greater than 0.85 will the correction parameters corresponding to that fingerprint vector be included in the candidate parameter set. This threshold ensures matching accuracy while avoiding the inclusion of low-quality parameters with excessively low similarity as candidates, thus ensuring the high reliability of the subsequent decision-making basis.

[0128] Specifically, based on the real-time operation strategy mode of new energy freight vehicles, the utility values ​​of different parameter combinations in the candidate parameter set are calculated. The parameter combination with the highest utility value is selected as the target correction parameter group to correct the carbon emissions of new energy freight vehicles in real time. The current real-time operation strategy mode of the vehicle is obtained, and each strategy mode corresponds to a utility function. The performance attributes associated with the candidate parameter combinations are analyzed through the utility function, and the computational accuracy, response speed, and data dependency emphasized in their respective subspaces are mapped to corresponding utility values. For example, in the economic mode, since accurate carbon emission data is the foundation of energy-saving management, the utility function is set with a very high weight for computational accuracy; in the performance mode, which requires a rapid response to the driver's power requests, the utility function is set with a very high weight for response speed. For each parameter combination in the candidate parameter set, the corresponding utility value is calculated based on the corresponding performance attribute label and the utility function corresponding to the current strategy mode. The parameter combination with the highest utility value is selected as the target correction parameter group. Parameter combination selection through utility functions ensures that the carbon emission correction strategy is synergistic with the overall operational goals of the vehicle, improving the efficiency and effectiveness of the carbon emission correction process.

[0129] like Figure 4 As shown, a carbon emission correction system for new energy freight vehicles based on multi-objective collaborative optimization is used to implement a carbon emission correction method for new energy freight vehicles based on multi-objective collaborative optimization, including:

[0130] The carbon emission model building module simulates the carbon emission process based on historical operating data of new energy trucks and builds the first carbon emission model.

[0131] The static reference point filtering module clusters the truck operating condition data in the historical operation data. For each category, it constructs constraints based on calculation accuracy, response speed, and data dependency. Through multi-objective optimization, it solves for the set of static reference points for the corresponding operating condition.

[0132] The dynamic reference point adjustment module analyzes the characteristics of truck driving modes under each working condition, sets corresponding dynamic adjustment factors, and dynamically adjusts the static reference point set to obtain the corresponding dynamic reference point set.

[0133] The correction parameter library construction module combines the first carbon emission model, the dynamic reference point set, and the operating condition data to solve the correction parameters corresponding to each preset carbon emission correction model through multi-objective optimization, and then constructs the correction parameter library by associating the solved correction parameters.

[0134] The carbon emission correction module matches the target correction parameter group in the correction parameter library based on the real-time operating data of the new energy trucks, and corrects the carbon emissions of the new energy trucks in real time.

[0135] In this embodiment, the carbon emission model construction module deconstructs the carbon emission process into layers based on the physical structure of the new energy truck's power system, including engine emissions, electricity consumption emissions, and auxiliary emissions. A corresponding model is built for each layer, and the models are dynamically weighted and fused according to real-time energy allocation to construct the first carbon emission model. Through layered modeling and model fusion, the error of simplified models that treat vehicles as a single emission source can be reduced, improving the accuracy of the carbon emission simulation process and providing accurate data support for the carbon emission correction process. The static reference point selection module performs cluster analysis on historical operating condition data. For each typical operating condition, constraints are constructed based on three objectives: computational accuracy, response speed, and data dependency. The optimal solution set is obtained through a multi-objective optimization algorithm, and a set of static reference points corresponding to critical points in different optimization directions is selected through curvature analysis. This decomposes the global optimization problem into optimization processes for specific operating conditions, providing a high-quality initial parameter benchmark for the carbon emission optimization process and improving the correction quality.

[0136] Specifically, the dynamic reference point adjustment module analyzes the driving behavior characteristics of trucks in real time and, combined with data quality assessment results, generates dynamic adjustment factors. These factors are then used to transform the coordinate space of the static reference point set accordingly. By incorporating each driver's driving mode, the static optimization benchmark is dynamically evolved, improving the system's adaptability to complex operating environments. The correction parameter library construction module divides the dynamic reference point set into optimization subspaces. Within each subspace, target weights are set based on performance preferences. A mapping relationship is used to associate the output differences between the first carbon emission model and the preset correction model with the optimization objective. Multi-objective optimization is used to solve for the optimal parameters for each correction model and construct a related database. Based on the global optimization process, an optimal parameter database is built, allowing for real-time matching of corresponding correction parameters, improving correction effectiveness and efficiency.

[0137] Specifically, the carbon emission correction module continuously extracts the operating condition feature vectors from real-time operating data. By performing similarity matching and utility value decision-making based on operating strategies in the correction parameter library, it quickly matches the optimal target correction parameter set and performs real-time calibration of the original carbon emission calculation. Through efficient retrieval and intelligent correction, it ensures low latency, high accuracy, and strategy synergy in carbon emission correction, improves correction effect and efficiency, and achieves reliable automatic carbon emission correction in complex environments.

[0138] Example 2:

[0139] This embodiment takes the carbon emission correction process of a plug-in hybrid electric new energy truck in a typical transportation task as an example to fully demonstrate the entire chain of technology implementation from data preparation and offline model building to online real-time correction.

[0140] This embodiment is based on historical data accumulated by the vehicle during long-term operation. This data is collected by the onboard terminal and uploaded to a cloud server for offline model training and optimization. The historical data mainly includes vehicle CAN bus signals, such as vehicle speed, engine speed, instantaneous fuel flow rate, motor current and voltage, battery SOC value, and air conditioning compressor status, with a sampling frequency of 1Hz. The system has collected over three months of travel data for the truck, covering various road conditions including urban areas, suburbs, and highways, totaling over 2000 hours.

[0141] Specifically, a first carbon emission model is constructed using historical data. Based on the powertrain structure, carbon emissions are decomposed into three levels: engine emissions, electricity consumption emissions, and auxiliary system emissions. For example, the engine emission sub-model takes the engine fuel consumption rate (in L / h) as input and calculates the engine's real-time carbon emission rate by querying the carbon emission coefficient of that diesel model (e.g., 2.68 kg CO2 / L). The electricity consumption emission sub-model takes the DC-side power of the battery (in kW) as input and multiplies it by the average carbon emission factor of the power grid (e.g., 0.6 kg CO2 / kWh, or 0.000167 kg CO2 / kJ) to obtain the indirect carbon emission rate. The auxiliary emission sub-model estimates the average power of high-power accessories such as air conditioners based on their duty cycles and converts this into carbon emissions. By analyzing the energy flow distribution ratio under different driving modes (pure electric, hybrid, and engine direct drive) in historical data, a dynamic strategy for weighted fusion of each sub-model was determined. Finally, a first carbon emission model that can output the comprehensive carbon emission rate based on the real-time power distribution status was constructed. The overall estimation error of this model on historical data was assessed to be less than 5%.

[0142] After obtaining a high-precision first carbon emission model, historical operating conditions were clustered to extract typical patterns. A large number of 600-second segments were extracted from the historical data, and 12 feature parameters, including average vehicle speed, standard deviation of speed, average acceleration, and idling time ratio, were calculated for each segment to form a feature vector. For example, the characteristics of a typical urban congestion segment might be: average vehicle speed 18 km / h, standard deviation of speed 12 km / h, and average acceleration 0.15 m / s². 2 The idling time ratio was 40%. The AP clustering algorithm was used to analyze these feature vectors, automatically identifying five representative operating condition clusters, corresponding to severe congestion, general urban roads, smooth suburban traffic, highway cruising, and high-speed heavy-load uphill driving. For the highway cruising operating condition cluster, its data characteristics were further analyzed. Vehicles in this cluster are in a stable state, and the coefficient of variation of their carbon emission values ​​is small. Therefore, a high lower limit for computational accuracy was set; for example, the relative root mean square error between the corrected carbon emission value and the baseline value of the first model should not exceed 3%. Simultaneously, considering high-speed driving safety, a strict upper limit for response speed was set, requiring the delay from new data input to the completion of the corrected output to not exceed 80 milliseconds. Furthermore, the data in this operating condition is usually relatively complete, so the data completeness threshold was set to 90%. Using the above boundary requirements as constraints, under this operating condition, with the objectives of minimizing computational accuracy error, minimizing response speed delay, and maximizing the robustness of data dependency to missing data, the NSGA-II multi-objective optimization algorithm was used to solve the problem, obtaining a set of Pareto optimal solutions containing hundreds of non-dominated solutions.

[0143] To extract the most critical representative points from a large number of optimal solutions, a geometric analysis of the Pareto front was performed. The local curvature of each solution on two-dimensional projection planes such as precision-velocity and precision-dependency was calculated. For example, the curvature eigenvalue of a certain solution on the precision-velocity plane was calculated to be 0.42. Based on the statistical distribution of curvature values ​​across all solutions, a curvature threshold of 0.35 was set. Solutions with curvature greater than 0.35 were selected as candidate inflection points, resulting in 15 points. These 15 points were then subjected to K-means clustering (K=3) based on their calculated precision, response speed, and data dependency values. Points that were consistently assigned to clusters near their respective cluster centers across all three clustering results were selected, ultimately identifying four of the most representative static reference points. These four points clearly characterize four different optimization tendencies under this condition: Point A (high precision, medium velocity, medium dependency), Point B (medium precision, high velocity, medium dependency), Point C (very high precision, low velocity, low dependency), and Point D (medium precision, medium velocity, high robustness).

[0144] The core task of the offline phase is to construct a correction parameter library. For the highway cruise scenario and its four dynamic reference points, the optimization space is divided into four sub-regions. Within each sub-region, target weights are set according to its core tendency. For example, for a sub-region biased towards high accuracy, the weights are set as (accuracy: 0.7, speed: 0.2, robustness: 0.1). The preset carbon emission correction model uses a lightweight gradient boosting tree model. Within this sub-region, using the output of the first carbon emission model as the training objective and readily available real-time vehicle operating parameters (such as vehicle speed, acceleration, motor power, and engine load rate) as input features, a multi-objective evolutionary algorithm is used to adjust the hyperparameters of the gradient boosting tree model (such as the number of trees, depth, and learning rate) to find the optimal parameter combination under given target weights. The parameter set, along with the corresponding scenario identifier, dynamic reference point coordinates, and target weights, is stored in the correction parameter library. This process is repeated for all sub-regions of all scenarios, ultimately forming a large, indexed correction parameter knowledge base.

[0145] During the online real-time correction phase, the vehicle is cruising at approximately 90 km / h on a highway. The onboard computing unit continuously collects real-time data from the past 60 seconds. The feature extraction module quickly calculates the feature vector for the current segment, for example: average vehicle speed 88 km / h, speed standard deviation 5 km / h, and average acceleration 0.05 m / s². 2The idling time is 1%. Simultaneously, the driving behavior analysis model analyzes this data segment, outputting an aggression level value of 0.3 (less than the threshold of 0.65), classifying it as smooth driving; therefore, adjusting the direction is prioritized for accuracy. The data quality assessment module calculates the information entropy of the current key signal, resulting in a data quality score of 0.95 (above the threshold of 0.8), thus assigning a relatively low adjustment weight of 0.2 to the data dependency target. Combining direction and weight, a dynamic adjustment factor is generated.

[0146] The matching module calculates the similarity between the current real-time feature vector and the operating condition fingerprint in the correction parameter library, achieving a cosine similarity of 0.92 with the highway cruise operating condition fingerprint. Given that the vehicle is currently in economy mode, the utility function prioritizes accuracy; therefore, from the matched candidate parameters, a parameter set corresponding to high accuracy, medium speed, and medium dependency tendency was selected. The onboard system immediately loads this parameter set into the gradient boosting tree correction model. When new instantaneous data (e.g., vehicle speed 89 km / h, acceleration 0.02 m / s²) is encountered... 2 When inputting a motor power of 30kW and an engine load rate of 0%, the correction model outputs a corrected carbon emission rate value within milliseconds. For example, the original simple model calculates a value of 245.6 g / km, while the final result after correction by this system is 238.2 g / km. This result is not only closer to the high-precision benchmark value of the first carbon emission model, but the entire processing also meets real-time requirements and fully utilizes data information while maintaining excellent data quality. Through the above process, the system achieves accurate, efficient, and adaptive real-time correction of carbon emissions from new energy trucks.

[0147] The above description is merely a preferred embodiment of this application. The scope of protection of this application is not limited to the above embodiments. All technical solutions falling within the scope of this application's concept are within the scope of protection of this application. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of this application should also be considered within the scope of protection of this application.

Claims

1. A new energy truck carbon emission correction method based on multi-objective collaborative optimization, characterized in that, The application relates to a method for real-time correction of carbon emission of a new energy truck. According to historical operation data of the new energy truck, a first carbon emission model is constructed by simulating a carbon emission process; The truck working condition data in the historical operation data are clustered, and for each cluster, constraints are constructed by combining calculation accuracy, response speed and data dependency, and a static reference point set corresponding to the working condition is solved by multi-objective optimization. The driving mode characteristics of each working condition are analyzed, a corresponding dynamic adjustment factor is set, the static reference point set is dynamically adjusted, and a corresponding dynamic reference point set is obtained. Based on the distribution positions of the reference points in the dynamic reference point set, a space is divided into multiple subspaces, and each subspace corresponds to an optimization target combination. In each subspace, the target weights of calculation accuracy, response speed and data dependency are set, a mapping relationship between correction parameters and optimization targets is established by combining the differences between the first carbon emission model result and a preset carbon emission correction model result, correction parameters corresponding to each preset carbon emission correction model are solved by multi-objective optimization based on the mapping relationship and the target weights, the solved correction parameters are associated, and a correction parameter library is constructed. According to real-time operation data of the new energy truck, target correction parameter groups are matched in the correction parameter library, and carbon emission of the new energy truck is corrected in real time. The first carbon emission model is constructed by simulating a carbon emission process according to historical operation data of the new energy truck.

2. The multi-objective collaborative optimization based new energy truck carbon emission correction method according to claim 1, characterized in that, According to the structure of the power system of the new energy truck, the carbon emission process is divided into engine emission, power consumption emission and auxiliary emission. For each layer, a corresponding carbon emission submodel is constructed, the carbon emission submodels are weighted and fused according to the dynamic allocation strategy in the truck operation, and the first carbon emission model is constructed. The truck working condition data in the historical operation data are clustered, and for each cluster, constraints are constructed by combining calculation accuracy, response speed and data dependency, and a static reference point set corresponding to the working condition is solved by multi-objective optimization.

3. The multi-objective collaborative optimization based new energy truck carbon emission correction method according to claim 1, characterized in that, The truck working condition data in the historical operation data are clustered, and for each cluster, constraints are constructed by combining calculation accuracy, response speed and data dependency, and a static reference point set corresponding to the working condition is solved by multi-objective optimization. The truck working condition data in the historical operation data are clustered, and for each cluster, constraints are constructed by combining calculation accuracy, response speed and data dependency, and a static reference point set corresponding to the working condition is solved by multi-objective optimization. The truck working condition data in the historical operation data are clustered, and for each cluster, constraints are constructed by combining calculation accuracy, response speed and data dependency, and a static reference point set corresponding to the working condition is solved by multi-objective optimization. The truck working condition data in the historical operation data are clustered, and for each cluster, constraints are constructed by combining calculation accuracy, response speed and data dependency, and a static reference point set corresponding to the working condition is solved by multi-objective optimization. ​ 4. The multi-objective collaborative optimization based new energy truck carbon emission correction method according to claim 3, characterized in that, ​ ​ ​ ​ 5. The multi-objective collaborative optimization based new energy truck carbon emission correction method according to claim 4, characterized in that, The analysis of the driving mode characteristics of each working condition of the truck sets a corresponding dynamic adjustment factor, dynamically adjusts the static reference point set, and obtains a corresponding dynamic reference point set, including: The driving mode of each working condition of the truck is analyzed, the corresponding features are extracted through the preset feature extraction model, the adjustment direction and the adjustment weight are determined, and a dynamic adjustment factor is obtained. According to the dynamic adjustment factor, the coordinates of each reference point in the static reference point set are dynamically weighted and adjusted, and the coordinate adjustment range is limited through the constraint condition, to obtain a corresponding dynamic reference point set.

6. The multi-objective collaborative optimization based new energy truck carbon emission correction method according to claim 5, characterized in that, The analysis of the driving mode characteristics of each working condition of the truck sets a corresponding dynamic adjustment factor, dynamically adjusts the static reference point set, and obtains a corresponding dynamic reference point set, including: The driving mode of each working condition of the truck is analyzed, the corresponding features are extracted through the preset feature extraction model, the adjustment direction and the adjustment weight are determined, and a dynamic adjustment factor is obtained. The driving mode of each working condition of the truck is analyzed, the corresponding features are extracted through the preset feature extraction model, the adjustment direction and the adjustment weight are determined, and a dynamic adjustment factor is obtained. The driving mode of each working condition of the truck is analyzed, the corresponding features are extracted through the preset feature extraction model, the adjustment direction and the adjustment weight are determined, and a dynamic adjustment factor is obtained. The driving mode of each working condition of the truck is analyzed, the corresponding features are extracted through the preset feature extraction model, the adjustment direction and the adjustment weight are determined, and a dynamic adjustment factor is obtained.

7. The multi-objective collaborative optimization based new energy truck carbon emission correction method according to claim 1, characterized in that, The driving mode of each working condition of the truck is analyzed, the corresponding features are extracted through the preset feature extraction model, the adjustment direction and the adjustment weight are determined, and a dynamic adjustment factor is obtained. According to the real-time running data of the new energy truck, the corresponding working condition characteristics are extracted, and a working condition characteristic vector is constructed. According to the working condition characteristic vector, a target correction parameter group is matched in the correction parameter library, and the carbon emission of the new energy truck is corrected in real time.

8. The multi-objective collaborative optimization based new energy truck carbon emission correction method according to claim 7, characterized in that, According to the working condition characteristic vector, a target correction parameter group is matched in the correction parameter library, and the carbon emission of the new energy truck is corrected in real time. According to the working condition characteristic vector, a target correction parameter group is matched in the correction parameter library, and the carbon emission of the new energy truck is corrected in real time. The correction parameter information corresponding to different working conditions in the correction parameter library is analyzed, and a corresponding working condition fingerprint vector is generated. The similarity between the working condition characteristic vector and the working condition fingerprint vector is calculated, and the correction parameters with a similarity greater than a preset similarity threshold are selected as a candidate parameter set.

9. A new energy truck carbon emission correction system based on multi-objective collaborative optimization, characterized in that, According to the real-time running strategy mode of the new energy truck, the utility values of different parameter combinations in the candidate parameter set are calculated, the parameter combination with the highest utility value is selected as the target correction parameter group, and the carbon emission of the new energy truck is corrected in real time. A method for implementing the multi-objective collaborative optimization-based new energy truck carbon emission correction method according to any one of claims 1-8, comprising: A carbon emission model construction module simulates the carbon emission process according to the historical running data of the new energy truck, and constructs a first carbon emission model; A static reference point screening module clusters the truck working condition data in the historical running data, constructs constraints for each class in combination with the calculation accuracy, response speed and data dependency, and solves the static reference point set of the corresponding working condition through multi-objective optimization. A dynamic reference point adjustment module analyzes the truck driving mode characteristics of each working condition, sets a corresponding dynamic adjustment factor, dynamically adjusts the static reference point set, and obtains a corresponding dynamic reference point set; A correction parameter library construction module combines the first carbon emission model, the dynamic reference point set, and the working condition data, solves the correction parameters corresponding to each preset carbon emission correction model through multi-objective optimization, associates the solved correction parameters, and constructs a correction parameter library; A carbon emission correction module matches a target correction parameter group in the correction parameter library according to real-time running data of the new energy truck, and corrects the carbon emission of the new energy truck in real time.

Images