Whole vehicle thermal management model reduction method, device, medium and equipment

Abstract

Embodiments of the present application relate to the technical field of vehicles, and disclose a whole vehicle thermal management model reduction method, device, medium and equipment. The embodiments of the present application generate training data based on a target working condition simulation logic closed loop through a calibrated and checked thermal management model, and then build and optimize a thermal management reduction model by applying a preset modeling algorithm and a preset optimization algorithm, so as to at least reflect the hysteresis effect of temperature and improve the thermal management accuracy, thereby replacing the conventional physical model and being applied to a vehicle to realize the reduction operation of the whole vehicle thermal management model. The model operation amount can be greatly reduced while ensuring the simulation accuracy, and the optimization control of the controlled object model is completed.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and in particular to a method, apparatus, medium, and equipment for reducing the order of a vehicle thermal management model. Background Technology

[0002] The performance of the thermal management system is a key factor affecting the vehicle's energy consumption and comfort. Due to the high degree of coupling, strong nonlinearity, and demanding dynamic response requirements of the thermal management system, it is difficult to perform precise control. Furthermore, the high complexity and computational burden of the thermal management system model prevent its integration into embedded controllers for optimization calculations. Therefore, current thermal management control efforts primarily rely on experience and parameter calibration optimization using PID control, which is highly dependent on the environment, has long calibration cycles, high costs, and makes it difficult to achieve optimal results.

[0003] Specifically, Patent Document 1 (CN105270186A) discloses a reduced-order battery thermodynamic modeling method for control, which can be used to reduce the order of the thermal management model of a tractor's power battery, enabling rapid temperature prediction. The resulting reduced-order model requires less execution time while achieving similar accuracy to the full-order model, and can obtain the characteristic temperature range for each battery cell. However, this method is based on steady-state data for mathematical solutions and does not reflect the hysteresis effect of temperature changes in the thermal management system, thus it can only be applied to stable vehicle driving conditions.

[0004] Patent document 2 (CN118780343A) discloses a method for training a thermal management agent. The reduced-order model is a neural network model constructed based on the mapping relationship between parameters of the thermal management system; the vehicle's thermal management agent is trained based on the reduced-order model; the agent is a reinforcement learning system that learns based on observations to output the optimal decision. However, this method ignores key signals such as vehicle speed and thermostat opening, which play an important role in temperature changes, thus significantly affecting the model training accuracy.

[0005] Patent document 3 (CN118052145A) discloses a method for predicting the key performance of electric drive assemblies using multi-field coupling modeling and data-driven approaches. It obtains a high-precision model of the assembly by establishing a joint simulation model and calibrating model parameters. Then, by using a reduced-order thermal network model of key nodes, a prediction model for the electric drive assembly with higher prediction accuracy and lower computational cost is obtained. However, this method only reduces the order of specific nodes and cannot predict the overall system response, thus limiting its applicability.

[0006] Reference 1 (Research on Thermal Model Reduction Method and Thermal Management Strategy for Lithium-ion Battery Packs) provides a method for reducing the order of the finite element model of a battery pack. Based on the POD-Galerkin reduction principle, the flow field and temperature field of the air-cooled battery system are reduced in order. These two fields are coupled through a heat exchange system to form a reduced-order thermal model of the air-cooled battery system, thus reducing the computation time of the battery pack finite element model. However, this method is suitable for steady-state battery pack simulation and cannot predict the dynamic temperature response of the system. Summary of the Invention

[0007] The purpose of this invention is to provide a method, apparatus, medium, and equipment for reducing the order of a vehicle thermal management model, which can at least realize the reduction operation of the vehicle thermal management model, and can significantly reduce the amount of model calculation while ensuring simulation accuracy, and complete the optimization control of the controlled object model.

[0008] To address the aforementioned technical problems, in a first aspect, the present invention provides a method for reducing the order of a vehicle thermal management model, comprising at least: The initial input data required for vehicle thermal management simulation is obtained to build a vehicle thermal management model, and then the system calibration operation is performed on the vehicle thermal management model to obtain a calibrated thermal management model. The simulation result data is output through the calibrated thermal management model, and the data correction operation is performed on the simulation result data to obtain the corrected result data; To accelerate the convergence of the model gradient descent, normalization is performed on each of the corrected result data to obtain normalized data, and training data segmentation is performed on the normalized data to obtain a training set, a test set, and a validation set with a preset ratio. A thermal management reduction model is constructed based on a preset modeling algorithm, and a corresponding preset optimization algorithm is selected to optimize the thermal management reduction model in combination with the application background of vehicle thermal management, so as to at least make the optimized reduction model reflect the temperature hysteresis effect. After training the optimized reduced-order model based on the training set and thermal management characteristics, the trained reduced-order model is tested and verified based on the test set and the validation set to determine whether the trained reduced-order model meets the preset accuracy requirements. When the trained reduced-order model does not meet the preset accuracy requirement, the training data segmentation and model training, testing and verification operations are re-executed based on the characteristics of the test and verification data and the analysis results until the trained reduced-order model meets the preset accuracy requirement. After the trained reduced-order model meets the preset accuracy requirement, the trained reduced-order model is compiled, and the compiled reduced-order model is output to the controller integration at least.

[0009] Optionally, during the data correction operation on the simulation result data, for the instantaneous data loss caused by the simulation convergence problem, at least interpolation method and estimation and supplementation based on the physical relationship of vehicle thermal management are used to estimate the instantaneous data loss; and at least the Z-score method based on the 3σ principle and the preset physical constraint method are used to identify data anomalies, and the data anomalies are deleted or empirically corrected.

[0010] Optionally, the normalized data is determined at least by the following means: X norm =( X - X min ) / ( X max - X min ); In the above formula, X norm This refers to the normalized data; X This represents the corrected result data; X min This represents the minimum value of the input variable; X max This indicates the maximum value of the input variable.

[0011] Optionally, the preset modeling algorithm is at least a nonlinear autoregressive neural network algorithm; The mathematical expression of the thermal management order reduction model is at least: y ( t )= F [ y ( t -1), y ( t -2),…, y ( t - n y ), u ( t -1), u ( t -2),…, u ( t - n u )]+ L ( t ); In the above formula, y () indicates the output of the vehicle thermal management system at a certain moment; F [] denotes the nonlinear mapping function implemented by the feedforward neural network; u() indicates the input of the vehicle thermal management system at a certain moment; n y Indicates the output delay order; n u Indicates the input delay order; L ( t ) represents the model error.

[0012] Optionally, when the application background of the vehicle thermal management is to train the water temperature response of the thermal management system based on operating condition data, the preset optimization algorithm is at least the scaling conjugate gradient algorithm; The weight formula of the preset optimization algorithm is at least: w k+1 = w k + α k p k ; In the above formula, w k Indicates the weight at the current time. w k+1 This indicates the weight of the next time step from the current time step. α k Indicates the step size; p k Indicates the conjugate direction; p k At least by the current gradient g k Conjugate direction with the previous moment p k-1 Joint decision: p k =- g k + β k p k-1 ; in, β k This represents the conjugate parameter.

[0013] Optionally, when the application background of the vehicle thermal management is a scenario of fitting accessory power in the thermal management system, the preset optimization algorithm is at least the Levenberg-Marquardt algorithm; The weight formula of the preset optimization algorithm is at least: Δ w =-( J T J +μI ) -1 J T e ; In the above formula, Δ w This indicates the magnitude of the weight change between adjacent time points; J Indicates error e relative weight w Jacobian matrix; μ Represents the damping parameter, when μ When the inequality approaches 0, the Levenberg-Marquardt algorithm degenerates into the Gauss-Newton method. μ As the value approaches infinity, the Levenberg-Marquardt algorithm degenerates into gradient descent. I Represents the identity matrix.

[0014] Optionally, when the application background of the vehicle thermal management is a model training scenario where the training data of the thermal management system is limited, the preset optimization algorithm is at least a Bayesian regularization algorithm; The weight formula of the preset optimization algorithm is at least: E BR ( w )= αE W + βE D ; In the above formula, E BR ( w ) represents the overall objective function after Bayesian regularization; α Indicates the weighted penalty term E W hyperparameters; β Indicates the error term E D Hyperparameters.

[0015] Based on the same concept, in a second aspect, the present invention also provides a vehicle thermal management model reduction device for performing the vehicle thermal management model reduction method described in any one of the first aspects; The vehicle thermal management model reduction device includes at least: The model building and calibration module is used to obtain the initial input data required for vehicle thermal management simulation, so as to build a vehicle thermal management model, and then perform system calibration operation on the vehicle thermal management model to obtain a calibrated thermal management model; The result output and correction module is used to output simulation result data through the calibrated thermal management model, and to perform data correction operations on the simulation result data to obtain corrected result data. The normalization module is used to perform normalization processing on each of the corrected result data to obtain normalized data in order to accelerate the convergence of the model gradient descent, and to perform training data splitting operation on the basis of the normalized data to obtain a training set, a test set and a validation set with a preset ratio. The model optimization module is used to construct a thermal management reduced-order model based on a preset modeling algorithm, and select a corresponding preset optimization algorithm to optimize the thermal management reduced-order model in combination with the application background of vehicle thermal management, so as to at least make the optimized reduced-order model reflect the temperature hysteresis effect. The testing and verification module is used to perform testing and verification on the trained reduced-order model based on the test set and the verification set after training the optimized reduced-order model according to the training set and thermal management characteristics, so as to determine whether the trained reduced-order model meets the preset accuracy requirements. The re-execution module is used to re-execute the training data segmentation and model training, testing and verification operations based on the characteristics of the test and verification data and the analysis results when the trained reduced-order model does not meet the preset accuracy requirements, until the trained reduced-order model meets the preset accuracy requirements. The compilation output module is used to perform a compilation operation on the trained reduced-order model after the trained reduced-order model meets the preset accuracy requirement, and to output the compiled reduced-order model to the controller integration at least.

[0016] Based on the same concept, in a third aspect, the present invention also provides an electronic device, including a memory and a processor, the memory storing a computer program executable on the processor, wherein the processor, when executing the program, implements the steps in the vehicle thermal management model reduction method of any one of the first aspects.

[0017] Based on the same concept, in a fourth aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps in the vehicle thermal management model reduction method described in any one of the first aspects.

[0018] The technical solution provided by this invention firstly acquires the initial input data required for vehicle thermal management simulation to build a vehicle thermal management model, and then performs a system calibration operation on the vehicle thermal management model to obtain a calibrated thermal management model. Further, it outputs simulation result data through the calibrated thermal management model and performs data correction operations on the simulation result data to obtain corrected result data. Further, to accelerate the convergence of model gradient descent, it performs normalization processing on each corrected result data to obtain normalized data, and performs training data segmentation operations on the normalized data to obtain training sets, test sets, and validation sets with preset ratios. Further, it constructs a thermal management order reduction model based on a preset modeling algorithm, and selects appropriate algorithms based on the application background of vehicle thermal management. The preset optimization algorithm optimizes the thermal management reduction model to at least ensure that the optimized reduction model reflects the temperature hysteresis effect. Further, after training the optimized reduction model based on the training set and thermal management characteristics, the trained reduction model is tested and validated based on the test set and validation set to determine whether it meets the preset accuracy requirements. Further, when the trained reduction model does not meet the preset accuracy requirements, the training data segmentation and model training, testing, and validation operations are re-executed based on the characteristics of the test and validation data and the analysis results, until the trained reduction model meets the preset accuracy requirements. Further, after the trained reduction model meets the preset accuracy requirements, a compilation operation is performed on the trained reduction model, and the compiled reduction model is output to at least the controller integration.

[0019] Therefore, the embodiments of the present invention, through a calibrated and verified thermal management model, training data (i.e., the aforementioned training set) generated by simulation of target operating conditions, and then applying preset modeling algorithms and preset optimization algorithms to build and optimize a thermal management reduction model, can at least reflect the hysteresis effect of temperature and improve the accuracy of thermal management. This replaces the conventional physical model and is applied to vehicles, realizing the reduction of the overall vehicle thermal management model. While ensuring simulation accuracy, it can significantly reduce the amount of model computation and complete the optimization control of the controlled object model. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method for reducing the order of a vehicle thermal management model provided in an embodiment of the present invention; Figure 2 This is a comparison chart of predicted value curve and actual value curve provided by an embodiment of the present invention; Figure 3 This is a schematic diagram of a vehicle thermal management model reduction device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0023] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0024] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.

[0025] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0026] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0027] It should be noted that any symbols and / or numbers present in the specification that are not marked in the accompanying drawings are not reference numerals.

[0028] Figure 1 This is a flowchart of a vehicle thermal management model reduction method provided by an embodiment of the present invention. This embodiment is applicable to at least any vehicle thermal management system optimization scenario. The vehicle thermal management model reduction method can be, but is not limited to, executed by the vehicle thermal management model reduction device in this embodiment of the present invention. This execution entity can be implemented in software and / or hardware. Figure 1 As shown, the method for reducing the order of the vehicle thermal management model includes at least the following steps: S1. Obtain the initial input data required for vehicle thermal management simulation to build a vehicle thermal management model, and then perform system calibration operation on the vehicle thermal management model to obtain a calibrated thermal management model.

[0029] The initial input data can include geometric parameters, physical property parameters, operating condition parameters, and component characteristic data of the vehicle's thermal management system. Specifically, geometric parameters can include pipe length and diameter, radiator size, and component layout orientation; physical property parameters can include the thermal properties of the coolant / refrigerant and the thermal conductivity of the material; operating condition parameters can include ambient temperature, solar radiation, and road conditions; and component characteristic data can include the PQ curve of the pump / fan, the valve opening-flow characteristics, and the accuracy of sensors.

[0030] For example, the process of building a vehicle thermal management model can involve integrating the vehicle assembly and thermal management-related component models such as pumps, valves, heat exchangers, pipelines, and controllers according to their actual physical connections to establish a complete model reflecting the system's energy flow, mass flow, and signal transmission. Furthermore, it is ensured that all original data input to the model (i.e., the aforementioned initial input data) has a reliable source and that the error is within an acceptable range. For non-critical or less sensitive parameters, empirical or standard values ​​can be used for initialization.

[0031] In addition, system calibration can involve completing subsystem integration and performing model stability calibration tests, rationality calibration tests, and accuracy calibration tests to ensure that the model accuracy meets calibration requirements after subsystem integration. Simultaneously, it ensures that the model can solve stably under various set operating conditions without issues such as numerical divergence, oscillations, or convergence failures; and it qualitatively judges the correctness of the model's behavior from the perspective of physical laws and engineering common sense.

[0032] S2. Output simulation result data by calibrating the thermal management model, and perform data correction operation on the simulation result data to obtain corrected result data.

[0033] Data correction operations can include data cleaning processes, such as handling missing values, duplicate values, and performing data consistency checks.

[0034] In one specific implementation, optionally, during the data correction operation on the simulation result data, for the instantaneous data missing caused by the simulation convergence problem, at least interpolation method and estimation and supplementation based on the physical relationship of vehicle thermal management are used to estimate the instantaneous data missing; and at least the Z-score method based on the 3σ principle and the preset physical constraint method are used to identify data anomalies, and the data anomalies are deleted or empirically corrected.

[0035] S3. In order to accelerate the convergence of the model gradient descent, normalization is performed on each corrected result data to obtain normalized data. Then, training data splitting is performed on the normalized data to obtain training set, test set and validation set with preset ratio.

[0036] In order to ensure that all features are in a similar numerical range and to accelerate gradient descent convergence, this embodiment performs data normalization processing.

[0037] In another specific implementation, the normalized data may optionally be determined at least by the following means: X norm =( X - X min ) / ( X max - X min ); In the above formula, X norm This represents normalized data; X This indicates the corrected data. X min This represents the minimum value of the input variable (for example, it could refer to the minimum value of all corrected result data). X max This indicates the maximum value of the input variable (for example, it could refer to the maximum value of all corrected result data).

[0038] For example, the training data segmentation operation can be as follows: considering that the amount of training data for the thermal management system model simulation is about 100,000 seconds, the ratio of the training set, validation set, and test set can be set to 60 / 20 / 20. This can ensure that the model has sufficient data for training, improve the model's generalization ability, and prevent the model from overfitting, so that the model can simulate real vehicle scenarios after evaluation.

[0039] S4. Construct a thermal management downgrade model based on a preset modeling algorithm, and select a corresponding preset optimization algorithm to optimize the thermal management downgrade model in combination with the application background of vehicle thermal management, so as to at least make the optimized downgrade model reflect the temperature hysteresis effect.

[0040] There are multiple options for choosing between the preset modeling algorithm and the preset optimization algorithm.

[0041] In yet another specific implementation, the preset modeling algorithm is optionally at least a nonlinear autoregressive neural network algorithm; The mathematical expression for the thermal management order reduction model is at least: y ( t )= F [ y ( t -1), y ( t -2),…, y ( t - n y ), u ( t -1), u ( t -2),…, u ( t - n u )]+ L ( t ); In the above formula, y () indicates the output of the vehicle's thermal management system at a certain moment (e.g., temperature, thermal management accessory power, etc.); F [] denotes the nonlinear mapping function implemented by the feedforward neural network; u () indicates the input of the vehicle thermal management system at a certain moment (such as engine speed, torque, vehicle speed, thermostat opening, fan speed, water pump speed, etc.). n y Indicates the output delay order; n u This indicates the input delay order (the above delay order can, for example, enable different delay settings for temperature and power). L ( t ) represents the model error.

[0042] In another specific implementation, optionally, when the application background of vehicle thermal management is to train the water temperature response of the thermal management system based on operating condition data, the preset optimization algorithm is at least the scaling conjugate gradient algorithm (the scaling conjugate gradient algorithm is suitable for long sequence prediction. When the predicted operating condition is a long-term, continuous and complete operating condition and the system has strong dynamic characteristics, the scaling conjugate gradient algorithm can show robust convergence and is not easy to get trapped in local minima). The weight formula for the preset optimization algorithm is at least: w k+1 = w k+ α k p k ; In the above formula, w k Indicates the weight at the current time. w k+1 This indicates the weight of the next time step from the current time step. α k Indicates the step size; p k Indicates the conjugate direction; p k At least by the current gradient g k Conjugate direction with the previous moment p k-1 Joint decision: p k =- g k + β k p k-1 ; in, β k This represents the conjugate parameter.

[0043] In another specific implementation, optionally, when the application background of vehicle thermal management is the fitting scenario of accessory power in thermal management system, the preset optimization algorithm is at least the Levenberg-Marquardt algorithm (the Levenberg-Marquardt algorithm is suitable for prediction with high real-time requirements, but its memory consumption is high). The weight formula for the preset optimization algorithm is at least: Δ w =-( J T J + μI ) -1 J T e ; In the above formula, Δ w This indicates the magnitude of the weight change between adjacent time points; J Indicates error e relative weight w Jacobian matrix; μ Represents the damping parameter, when μ When the time constant approaches 0, the Levenberg-Marquardt algorithm degenerates into the Gauss-Newton method. μAs the value approaches infinity, the Levenberg-Marquardt algorithm degenerates into gradient descent. I Represents the identity matrix.

[0044] In another specific implementation, optionally, when the application background of vehicle thermal management is a model training scenario where the training data of the thermal management system is limited, the preset optimization algorithm is at least the Bayesian regularization algorithm (the Bayesian regularization algorithm is suitable for model prediction with a small amount of data). The weight formula for the preset optimization algorithm is at least: E BR ( w )= αE W + βE D ; In the above formula, E BR ( w ) represents the overall objective function after Bayesian regularization; α Indicates the weighted penalty term E W hyperparameters; β Indicates the error term E D Hyperparameters.

[0045] In yet another specific implementation, E W (Sum of squares of weights) and E D The sum of squared data errors can be calculated in at least the following ways: ; In the above formula, This represents the model's predicted value. y i This represents the actual value.

[0046] ; In the above formula, w j Indicates the weight.

[0047] It is known that, α and βDuring training, the model can be continuously re-estimated based on the Bayesian evidence framework to automatically balance model fit and complexity. It is understood that the actual training of thermal management system models can be based on computational resource allocation and time and accuracy requirements, with corresponding algorithm selection (i.e., not limited to the nonlinear autoregressive neural network algorithm, scaled conjugate gradient algorithm, Levenberg-Marquardt algorithm, Bayesian regularization algorithm, etc., involved in this application), thereby achieving an optimal balance between prediction accuracy, training efficiency, computational resource consumption, and model generalization ability.

[0048] S5. After training the optimized reduced-order model based on the training set and thermal management characteristics, test and verify the trained reduced-order model based on the test set and validation set to determine whether the trained reduced-order model meets the preset accuracy requirements.

[0049] Specifically, step S5 may include: specifying the hidden layers and delay step size of the neural network, and defining and testing training parameters based on the characteristics of the thermal management system.

[0050] More specifically, a neural network model with input time step and hidden layer number can be obtained through model training. The neural network input layer mainly includes engine speed, torque, vehicle speed, water pump speed, fan speed, thermostat opening, ambient temperature, compressor speed, etc.; the hidden layer mainly includes model input parameters - algorithm, specified training parameters, etc.; the output layer mainly includes engine water temperature, fan power, water pump power, etc., providing the controlled object for subsequent optimization control.

[0051] For example, closed-loop multi-step prediction can be used to verify the long-term predictive ability and robustness of the model. Specifically, the model can use its own prediction output... As input for the next step, only real external inputs will be used. u true ( t The formula is as follows: ; In the above formula, [] represents the mapping function after training is complete.

[0052] Based on the above process, the predicted value vs. the actual value curve for the entire test sequence is plotted according to the prediction results. Figure 2 This is a comparison chart of predicted value curves and actual value curves provided in an embodiment of the present invention. See also... Figure 2 The black curve represents the predicted value curve, and the gray curve represents the actual value curve, which can intuitively assess the degree of fit and divergence trend of the predicted trajectory.

[0053] S6. When the trained reduced-order model does not meet the preset accuracy requirements, the training data segmentation and model training, testing and verification operations are re-executed based on the characteristics of the test and verification data and the analysis results until the trained reduced-order model meets the preset accuracy requirements.

[0054] S7. After the trained reduced-order model meets the preset accuracy requirements, perform a compilation operation on the trained reduced-order model and output the compiled reduced-order model to the controller integration at least.

[0055] The preset accuracy requirements can be adapted to the actual vehicle application, and this invention does not limit them.

[0056] The technical solution provided in this embodiment first obtains the initial input data required for vehicle thermal management simulation to build a vehicle thermal management model, and then performs a system calibration operation on the vehicle thermal management model to obtain a calibrated thermal management model. Further, it outputs simulation result data through the calibrated thermal management model and performs data correction operations on the simulation result data to obtain corrected result data. Further, to accelerate the convergence of model gradient descent, it performs normalization processing on each corrected result data to obtain normalized data, and performs training data segmentation operations on the normalized data to obtain training sets, test sets, and validation sets with preset ratios. Further, it constructs a thermal management order reduction model based on a preset modeling algorithm, and selects appropriate algorithms based on the application background of vehicle thermal management. A preset optimization algorithm is used to optimize the thermal management reduction model to at least ensure that the optimized reduction model reflects the temperature hysteresis effect. Further, after training the optimized reduction model based on the training set and thermal management characteristics, the trained reduction model is tested and validated using the test and validation sets to determine if it meets the preset accuracy requirements. Further, if the trained reduction model does not meet the preset accuracy requirements, the training data segmentation and model training, testing, and validation operations are re-executed based on the characteristics of the test and validation data and the analysis results, until the trained reduction model meets the preset accuracy requirements. Further, after the trained reduction model meets the preset accuracy requirements, a compilation operation is performed on the trained reduction model, and the compiled reduction model is output to the controller integration at least once.

[0057] Therefore, this embodiment, through a comprehensive analysis of the influencing factors of the thermal management system and the application of a nonlinear autoregressive algorithm, combined with training using full simulation data, can at least achieve accurate prediction of engine coolant temperature and power consumption of thermal management system accessories, while significantly improving computational efficiency and enabling the deployment of the reduced-order model at the controller end. The reduced-order model can reflect the cumulative and transient response of the thermal management system temperature, providing an accurate, fast, and controllable model of the controlled object for subsequent thermal management system control optimization, thereby achieving overall vehicle energy-saving effects.

[0058] Figure 3This is a schematic diagram of a vehicle thermal management model reduction device provided in an embodiment of the present invention. This embodiment is applicable to at least any vehicle thermal management system optimization scenario. The vehicle thermal management model reduction device can be implemented in software and / or hardware. Figure 3 As shown, the vehicle thermal management model reduction device is used to execute the vehicle thermal management model reduction method of any of the foregoing embodiments or implementations.

[0059] The vehicle thermal management model downgrade device includes at least the following: The model building and calibration module 110 is used to obtain the initial input data required for vehicle thermal management simulation, so as to build a vehicle thermal management model, and then perform system calibration operation on the vehicle thermal management model to obtain a calibrated thermal management model. The result output and correction module 120 is used to output simulation result data through the calibration thermal management model, and to perform data correction operations on the simulation result data to obtain corrected result data. The normalization module 130 is used to perform normalization processing on each corrected result data in order to accelerate the convergence of the model gradient descent, obtain normalized data, and perform training data splitting operation on the basis of normalized data to obtain training set, test set and validation set with preset ratio; The model optimization module 140 is used to construct a thermal management reduced-order model based on a preset modeling algorithm, and select a corresponding preset optimization algorithm to optimize the thermal management reduced-order model in combination with the application background of vehicle thermal management, so as to at least make the optimized reduced-order model reflect the temperature hysteresis effect. The test and verification module 150 is used to perform tests and verifications on the trained reduced-order model based on the test set and verification set after training the optimized reduced-order model according to the training set and thermal management characteristics, so as to determine whether the trained reduced-order model meets the preset accuracy requirements. The re-execution module 160 is used to re-execute the training data segmentation and model training, testing and verification operations based on the characteristics of the test and verification data and the analysis results when the trained reduced model does not meet the preset accuracy requirements, until the trained reduced model meets the preset accuracy requirements. The compilation output module 170 is used to perform a compilation operation on the trained reduced-order model after the trained reduced-order model meets the preset accuracy requirements, and to output the compiled reduced-order model to the controller integration at least.

[0060] Optionally, during the data correction operation on the simulation results data, for the instantaneous data missing caused by the simulation convergence problem, at least interpolation method and estimation and supplementation based on the physical relationship of vehicle thermal management are used to estimate the instantaneous data missing; and at least the Z-score method based on the 3σ principle and the preset physical constraint method are used to identify data outliers, and the data outliers are deleted or empirically corrected.

[0061] Optionally, the normalized data shall be determined at least in the following ways: X norm =( X - X min ) / ( X max - X min ); In the above formula, X norm This represents normalized data; X This indicates the corrected data. X min This represents the minimum value of the input variable; X max This indicates the maximum value of the input variable.

[0062] Optionally, the preset modeling algorithm is at least a nonlinear autoregressive neural network algorithm; The mathematical expression for the thermal management order reduction model is at least: y ( t )= F [ y ( t -1), y ( t -2),…, y ( t - n y ), u ( t -1), u ( t -2),…, u ( t - n u )]+ L ( t ); In the above formula, y () indicates the output of the vehicle thermal management system at a certain moment; F [] denotes the nonlinear mapping function implemented by the feedforward neural network; u () indicates the input of the vehicle thermal management system at a certain moment; n y Indicates the output delay order; n u Indicates the input delay order; L ( t ) represents the model error.

[0063] Optionally, when the application background of vehicle thermal management is to train the water temperature response of the thermal management system based on operating condition data, the preset optimization algorithm is at least the scaling conjugate gradient algorithm; The weight formula for the preset optimization algorithm is at least: w k+1 = w k + α k p k ; In the above formula, w k Indicates the weight at the current time. w k+1 This indicates the weight of the next time step from the current time step. α k Indicates the step size; p k Indicates the conjugate direction; p k At least by the current gradient g k Conjugate direction with the previous moment p k-1 Joint decision: p k =- g k + β k p k-1 ; in, β k This represents the conjugate parameter.

[0064] Optionally, when the application background of vehicle thermal management is the fitting scenario of accessory power in thermal management system, the preset optimization algorithm is at least the Levenberg-Marquardt algorithm; The weight formula for the preset optimization algorithm is at least: Δ w =-( J T J + μI ) -1 J T e ; In the above formula, Δ w This indicates the magnitude of the weight change between adjacent time points; J Indicates error e relative weight w Jacobian matrix; μ Represents the damping parameter, when μ When the time constant approaches 0, the Levenberg-Marquardt algorithm degenerates into the Gauss-Newton method. μ As the value approaches infinity, the Levenberg-Marquardt algorithm degenerates into gradient descent. I Represents the identity matrix.

[0065] Optionally, when the application background of vehicle thermal management is a model training scenario where the training data of the thermal management system is limited, the preset optimization algorithm is at least the Bayesian regularization algorithm; The weight formula for the preset optimization algorithm is at least: E BR ( w )= αE W + βE D ; In the above formula, E BR ( w ) represents the overall objective function after Bayesian regularization; α Indicates the weighted penalty term E W hyperparameters; β Indicates the error term E D Hyperparameters.

[0066] The technical solution provided in this embodiment firstly obtains the initial input data required for vehicle thermal management simulation through the model building and calibration module to build a vehicle thermal management model, and then performs system calibration on the vehicle thermal management model to obtain a calibrated thermal management model. Further, the result output and correction module outputs simulation result data through the calibrated thermal management model and performs data correction on the simulation result data to obtain corrected result data. Further, to accelerate the convergence of model gradient descent, the normalization module performs normalization processing on each corrected result data to obtain normalized data, and performs training data segmentation operation on the normalized data to obtain training set, test set, and validation set with preset ratios. Further, the model optimization module constructs a thermal management reduced-order model based on a preset modeling algorithm, and combines it with vehicle thermal management applications. The background involves selecting a pre-defined optimization algorithm to optimize the thermal management reduction model, ensuring that the optimized model at least reflects the temperature hysteresis effect. Further, after training the optimized reduction model based on the training set and thermal management characteristics, a testing and verification module performs tests and verifications on the trained reduction model using the test and verification sets to determine if it meets the preset accuracy requirements. Further, if the trained reduction model does not meet the preset accuracy requirements, a re-execution module re-executes the training data segmentation, model training, testing, and verification operations based on the characteristics of the test and verification data and the analysis results, until the trained reduction model meets the preset accuracy requirements. Finally, after the trained reduction model meets the preset accuracy requirements, a compilation and output module compiles the trained reduction model and outputs the compiled reduction model, at least to the controller integration.

[0067] Therefore, this embodiment uses a calibrated and verified thermal management model and training data (i.e., the aforementioned training set) generated by simulation of the target working condition to generate a logical closed loop. Then, it applies a preset modeling algorithm and a preset optimization algorithm to build and optimize a thermal management reduction model, so as to at least reflect the hysteresis effect of temperature and improve the accuracy of thermal management. This replaces the conventional physical model and is applied to the vehicle, realizing the reduction of the thermal management model of the whole vehicle. While ensuring the simulation accuracy, it can significantly reduce the amount of model calculation and complete the optimization control of the controlled object model.

[0068] This embodiment provides an electronic device. Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. See also: Figure 4The electronic device 1000 includes a processor 1001 and a memory 1002. The memory 1002 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 1001, the steps in any of the above-mentioned vehicle thermal management model downgrade methods are performed. Through the above technical solution, the processor 1001 and the memory 1002 are interconnected and communicate with each other through a communication bus and / or other forms of connection mechanism (not shown). The memory 1002 stores a computer program executable by the processor. When the electronic device 1000 is running, the processor 1001 executes the computer program to execute the vehicle thermal management model reduction method in any optional implementation of the above embodiments, so as to at least achieve the following functions: obtaining the initial input data required for vehicle thermal management simulation to build a vehicle thermal management model, and then performing system calibration operation on the vehicle thermal management model to obtain a calibrated thermal management model; outputting simulation result data through the calibrated thermal management model, and performing data correction operation on the simulation result data to obtain corrected result data; in order to accelerate the convergence of model gradient descent, performing normalization processing on each corrected result data to obtain normalized data, and based on the normalized data... The training data is segmented to obtain training, testing, and validation sets with preset proportions. A thermal management reduction model is constructed based on a preset modeling algorithm, and a corresponding preset optimization algorithm is selected to optimize the thermal management reduction model in light of the vehicle's thermal management application background, so that the optimized reduction model at least reflects the temperature hysteresis effect. After training the optimized reduction model based on the training set and thermal management characteristics, the trained reduction model is tested and validated based on the testing and validation sets to determine whether it meets the preset accuracy requirements. If the trained reduction model does not meet the preset accuracy requirements, the training data segmentation and model training, testing, and validation operations are repeated based on the characteristics of the test and validation data and the analysis results until the trained reduction model meets the preset accuracy requirements. After the trained reduction model meets the preset accuracy requirements, a compilation operation is performed on the trained reduction model, and the compiled reduction model is output to the controller integration at least once.

[0069] This embodiment provides a computer-readable storage medium storing a computer program. When executed by a processor, the program implements the vehicle thermal management model reduction method provided in all embodiments of this application: acquiring initial input data required for vehicle thermal management simulation to build a vehicle thermal management model; then performing system calibration on the vehicle thermal management model to obtain a calibrated thermal management model; outputting simulation result data through the calibrated thermal management model, and performing data correction on the simulation result data to obtain corrected result data; in order to accelerate the convergence of model gradient descent, performing normalization processing on each corrected result data to obtain normalized data, and performing training data segmentation operation on the normalized data to obtain a training set, test set, and validation set with a preset ratio; and constructing a model based on a preset modeling algorithm. A thermal management downgrade model is built, and a corresponding preset optimization algorithm is selected to optimize the thermal management downgrade model based on the application background of vehicle thermal management, so that the optimized downgrade model can at least reflect the temperature hysteresis effect. After training the optimized downgrade model based on the training set and thermal management characteristics, the trained downgrade model is tested and validated based on the test set and validation set to determine whether the trained downgrade model meets the preset accuracy requirements. When the trained downgrade model does not meet the preset accuracy requirements, the training data segmentation and model training, testing and validation operations are re-executed based on the characteristics of the test and validation data and the analysis results, until the trained downgrade model meets the preset accuracy requirements. After the trained downgrade model meets the preset accuracy requirements, the trained downgrade model is compiled, and the compiled downgrade model is output to the controller integration at least.

[0070] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0071] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0072] The program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0073] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for reducing the order of a vehicle thermal management model, characterized in that, At least including: The initial input data required for vehicle thermal management simulation is obtained to build a vehicle thermal management model, and then the system calibration operation is performed on the vehicle thermal management model to obtain a calibrated thermal management model. The simulation result data is output through the calibrated thermal management model, and the data correction operation is performed on the simulation result data to obtain the corrected result data; To accelerate the convergence of the model gradient descent, normalization is performed on each of the corrected result data to obtain normalized data, and training data segmentation is performed on the normalized data to obtain a training set, a test set, and a validation set with a preset ratio. A thermal management reduction model is constructed based on a preset modeling algorithm, and a corresponding preset optimization algorithm is selected to optimize the thermal management reduction model in combination with the application background of vehicle thermal management, so as to at least make the optimized reduction model reflect the temperature hysteresis effect. After training the optimized reduced-order model based on the training set and thermal management characteristics, the trained reduced-order model is tested and verified based on the test set and the validation set to determine whether the trained reduced-order model meets the preset accuracy requirements. When the trained reduced-order model does not meet the preset accuracy requirement, the training data segmentation and model training, testing and verification operations are re-executed based on the characteristics of the test and verification data and the analysis results until the trained reduced-order model meets the preset accuracy requirement. After the trained reduced-order model meets the preset accuracy requirement, a compilation operation is performed on the trained reduced-order model, and the compiled reduced-order model is output to the controller integration at least once. The ratio of training set, validation set, and test set is 60 / 20 / 20. The preset modeling algorithm is at least a nonlinear autoregressive neural network algorithm; When the application background of the vehicle thermal management is to train the water temperature response of the thermal management system based on operating condition data, the preset optimization algorithm is at least the scaling conjugate gradient algorithm; When the application background of the vehicle thermal management is a scenario of fitting accessory power in the thermal management system, the preset optimization algorithm is at least the Levenberg-Marquardt algorithm; When the application background of the vehicle thermal management is a model training scenario where the training data of the thermal management system is limited, the preset optimization algorithm is at least a Bayesian regularization algorithm.

2. The method for reducing the order of a vehicle thermal management model according to claim 1, characterized in that, During the data correction operation on the simulation results data, for the instantaneous data loss caused by the simulation convergence problem, at least interpolation method and estimation and supplementation based on the physical relationship of vehicle thermal management are used to estimate the instantaneous data loss; and at least the Z-score method based on the 3σ principle and the preset physical constraint method are used to identify data anomalies, and the data anomalies are deleted or empirically corrected.

3. The method for reducing the order of a vehicle thermal management model according to claim 1, characterized in that, The normalized data is determined at least through the following methods: X norm =( X - X min ) / ( X max - X min )； In the above formula, X norm This refers to the normalized data; X This represents the corrected result data; X min This represents the minimum value of the input variable; X max This indicates the maximum value of the input variable.

4. The method for reducing the order of a vehicle thermal management model according to claim 1, characterized in that, The mathematical expression of the thermal management order reduction model is at least: y ( t )= F [ y ( t -1), y ( t -2),…, y ( t - n y ), u ( t -1), u ( t -2),…, u ( t - n u )]+ L ( t )； In the above formula, y () indicates the output of the vehicle thermal management system at a certain moment; F [] denotes the nonlinear mapping function implemented by the feedforward neural network; u () indicates the input of the vehicle thermal management system at a certain moment; n y Indicates the output delay order; n u Indicates the input delay order; L ( t ) represents the model error.

5. The method for reducing the order of a vehicle thermal management model according to claim 1, characterized in that, The weight formula of the preset optimization algorithm is at least: w k+1 = w k + α k p k ； In the above formula, w k Indicates the weight at the current moment; w k+1 This indicates the weight of the next time step from the current time step. α k Indicates the step size; p k Indicates the conjugate direction; p k At least by the current gradient g k Conjugate direction with the previous moment p k-1 Joint decision: p k =- g k + β k p k-1 ； in, β k This represents the conjugate parameter.

6. The method for reducing the order of a vehicle thermal management model according to claim 1, characterized in that, The weight formula of the preset optimization algorithm is at least: D w =-( J T J + μI ) -1 J T e ; In the above formula, Δ w This indicates the magnitude of the weight change between adjacent time points; J Indicates error e relative weight w The Jacobian matrix; μ Represents the damping parameter, when μ When the inequality approaches 0, the Levenberg-Marquardt algorithm degenerates into the Gauss-Newton method. μ As the value approaches infinity, the Levenberg-Marquardt algorithm degenerates into gradient descent. I Represents the identity matrix.

7. The method for reducing the order of a vehicle thermal management model according to claim 1, characterized in that, The weight formula of the preset optimization algorithm is at least: E BR ( w )= αE W + βE D ； In the above formula, E BR ( w ) represents the overall objective function after Bayesian regularization; α Indicates the weighted penalty term E W hyperparameters; β Indicates the error term E D Hyperparameters.

8. A vehicle thermal management model order reduction device, characterized in that, Used to perform the vehicle thermal management model reduction method according to any one of claims 1-7; The vehicle thermal management model reduction device includes at least: The model building and calibration module is used to obtain the initial input data required for vehicle thermal management simulation, so as to build a vehicle thermal management model, and then perform system calibration operation on the vehicle thermal management model to obtain a calibrated thermal management model; The result output and correction module is used to output simulation result data through the calibrated thermal management model, and to perform data correction operations on the simulation result data to obtain corrected result data. The normalization module is used to perform normalization processing on each of the corrected result data to obtain normalized data in order to accelerate the convergence of the model gradient descent, and to perform training data splitting operation on the basis of the normalized data to obtain a training set, a test set and a validation set with a preset ratio. The model optimization module is used to construct a thermal management reduced-order model based on a preset modeling algorithm, and select a corresponding preset optimization algorithm to optimize the thermal management reduced-order model in combination with the application background of vehicle thermal management, so as to at least make the optimized reduced-order model reflect the temperature hysteresis effect. The testing and verification module is used to perform testing and verification on the trained reduced-order model based on the test set and the verification set after training the optimized reduced-order model according to the training set and thermal management characteristics, so as to determine whether the trained reduced-order model meets the preset accuracy requirements. The re-execution module is used to re-execute the training data segmentation and model training, testing and verification operations based on the characteristics of the test and verification data and the analysis results when the trained reduced-order model does not meet the preset accuracy requirements, until the trained reduced-order model meets the preset accuracy requirements. The compilation output module is used to perform a compilation operation on the trained reduced-order model after the trained reduced-order model meets the preset accuracy requirement, and to output the compiled reduced-order model to the controller integration at least.

9. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the vehicle thermal management model reduction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the vehicle thermal management model downgrade method according to any one of claims 1 to 7.

Images