Digital twinborn model construction method and system of new energy heavy truck battery power system

Abstract

The invention discloses a digital twinborn model construction method and system for a new energy heavy truck battery power system, and the method comprises the steps: constructing a multi-scale mapping model of the battery power system, which comprises hierarchical models from a single battery to a module, a whole pack, a controller and a drive motor; performing multi-physical field coupling modeling of an electrochemical field, a temperature field and a stress field on the constructed multi-scale mapping model to obtain a coupled multi-scale mapping model; based on real-time monitoring data, a data driving model is adopted to carry out parameter updating on the coupled multi-scale mapping model, and a preliminary digital twinborn model is generated; and performing dynamic time sequence modeling on sensor data in the preliminary digital twinborn model through a graph neural network-long short-term memory neural network structure to obtain a final digital twinborn model. According to the method, the complex characteristics of the new energy heavy truck battery power system in actual operation can be accurately reflected, and the model prediction precision and the system control performance are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of digital twin technology, and in particular to a method and system for constructing a digital twin model of a battery power system of a new energy heavy-duty truck. Background Art

[0002] In the field of digital twin modeling of new energy heavy-duty truck battery power systems, traditional methods are mainly based on single physical field models and offline parameter calibration. The typical approach is: first, an electrochemical model (such as a P2D model) or a thermal model (such as a lumped parameter model) is established for the battery cell, and then the parameter values ​​under limited operating conditions are obtained through bench experiments, and they are fixed and applied to system-level simulations. For example, some studies only construct a temperature field model of the battery thermal management system and simulate the coolant flow path through fluid mechanics equations, but ignore the reaction of temperature changes on the battery internal resistance and electrochemical processes. This method has significant limitations in practical applications: on the one hand, a single physical field model cannot reflect the coupling effect of the electrochemical-thermal-stress field in the battery system. For example, the thermal stress caused by the temperature gradient will cause the electrode structure to deform, thereby affecting the battery internal resistance and charge and discharge efficiency; on the other hand, the static parameters of the offline calibration cannot adapt to dynamic changes such as battery aging and operating condition fluctuations, resulting in a significant increase in model prediction error with running time.

[0003] Therefore, the existing digital twin model cannot accurately reflect the complex characteristics of the battery power system of new energy heavy-duty trucks in actual operation, resulting in insufficient model prediction accuracy and system control performance. Summary of the Invention

[0004] The embodiments of the present invention provide a method and system for constructing a digital twin model of a new energy heavy-duty truck battery power system, which can accurately reflect the complex characteristics of the new energy heavy-duty truck battery power system in actual operation, and effectively improve the model prediction accuracy and system control performance.

[0005] An embodiment of the present invention provides a method for constructing a digital twin model of a new energy heavy-duty truck battery power system, comprising:

[0006] Construct a multi-scale image model of the battery power system, including hierarchical models from battery cells to modules, entire packs, controllers, and drive motors;

[0007] Conduct multi-physics field coupling modeling of electrochemical field, temperature field, and stress field on the constructed multi-scale imaging model to obtain a coupled multi-scale imaging model;

[0008] Based on real-time monitoring data, a data-driven model is used to update the parameters of the coupled multi-scale image model to generate a preliminary digital twin model;

[0009] The sensor data in the preliminary digital twin model is dynamically modeled using the graph neural network-long short-term memory neural network structure to obtain the final digital twin model.

[0010] As an improvement to the above solution, the multi-scale image model of the battery power system is constructed, including a hierarchical model from battery cells to modules, entire packs, controllers, and drive motors, and includes the following sub-steps:

[0011] Based on the energy conservation equation, a cell-scale electrochemical-thermal model is constructed on top of the P2D model, resulting in a model structure that integrates electrochemical and thermal characteristics at the cell scale.

[0012] Based on the basic model at the cell scale, the thermal stress and thermal expansion caused by temperature gradients during battery operation are considered, and a thermal-mechanical coupling model at the cell scale is constructed to form a comprehensive model that includes characteristics at both the cell and cell scales.

[0013] Based on the comprehensive model, starting from the scale of the power battery module and the entire pack, industrial modeling software is used to build a 3D CAD model corresponding to the actual shape of the physical battery. This accurately presents the key structural parameters, the constraints and positioning relationships between cells and modules, and obtains a 3D model containing multi-scale structural information.

[0014] Pixyz Studio is used to perform surface subdivision, topology correction, extraction, and repair operations on the three-dimensional model containing multi-scale structural information. Then, Hertz theory and the extended finite element method are combined to establish an overall digital physical model, clarify the semantic information annotation objects, and complete the construction of the multi-scale imaging model of the battery power system.

[0015] As an improvement to the above solution, the multi-physics field coupling modeling of electrochemical field, temperature field, and stress field is performed on the constructed multi-scale imaging model to obtain the coupled multi-scale imaging model, including the following sub-steps:

[0016] For the completed multi-scale imaging model of the battery power system, the parameter information related to the electrochemical field, temperature field, and stress field is comprehensively sorted out to obtain the parameter sets of each physical field at different scales;

[0017] Based on the parameter set, analyze the transfer path and method of each physical field parameter between different levels of models, clarify the law of parameter transfer, and obtain the specific rule description of parameter transfer;

[0018] Based on the obtained parameter transfer rule description, the interaction mechanism between different physical fields and different hierarchical models is studied, the influencing factors of the interaction are determined, and a list of key factors affecting the interaction is obtained;

[0019] According to the parameter set, the description of the parameter transfer rules and the list of key factors affecting the interaction, the multi-physics field coupling integration is performed on the multi-scale imaging model to obtain the coupled multi-scale imaging model.

[0020] As an improvement to the above solution, based on real-time monitoring data, a data-driven model is used to update the parameters of the coupled multi-scale image model to generate a preliminary digital twin model, which includes the following sub-steps:

[0021] Through sensors pre-deployed in the battery power system, real-time monitoring data during system operation is collected to obtain a data set containing the real-time status of the battery power system;

[0022] A data-driven approach is used to extract information from a dataset containing the real-time status of the battery power system, filtering out valid information related to the multi-scale image model parameters to form a valid information set.

[0023] Comparing and matching the information in the valid information set with the parameters of the coupled multi-scale imaging model, determining the parameters that need to be updated and the update values, and obtaining a parameter update list;

[0024] According to the parameter update list, the parameters of the coupled multi-scale image model are updated to generate a preliminary digital twin model.

[0025] As an improvement to the above solution, the method of dynamically modeling the sensor data in the preliminary digital twin model through the graph neural network-long short-term memory neural network structure to obtain the final digital twin model includes the following sub-steps:

[0026] The sensor data in the preliminary digital twin model is abstracted into a network topology diagram of the sensor data time series accumulation process according to the time series and spatial relationship, and a topological structure reflecting the spatiotemporal characteristics of the sensor data is obtained;

[0027] A graph neural network-long short-term memory neural network structure is used to process the topological structure reflecting the spatiotemporal characteristics of sensor data, integrating the structural information and time information to form intermediate data that integrates spatiotemporal characteristics.

[0028] Using the designed hierarchical feature aggregation method, different feature aggregators are learned on the intermediate data of fused spatiotemporal features in neighborhoods of different depths to obtain aggregated data containing multi-scale features;

[0029] Perform dynamic time series modeling on sensor data, integrate aggregated data containing multi-scale features into the model, and obtain the final digital twin model.

[0030] Another embodiment of the present invention provides a digital twin model construction system for a new energy heavy-duty truck battery power system, including:

[0031] A construction module for building a multi-scale image model of the battery power system, including hierarchical models from battery cells to modules, entire packs, controllers, and drive motors;

[0032] The first modeling module is used to perform multi-physics field coupling modeling of electrochemical field, temperature field, and stress field on the constructed multi-scale imaging model to obtain a coupled multi-scale imaging model;

[0033] The update module is used to update the parameters of the coupled multi-scale image model based on real-time monitoring data using a data-driven model to generate a preliminary digital twin model;

[0034] The second modeling module is used to perform dynamic time series modeling of the sensor data in the preliminary digital twin model through the graph neural network-long short-term memory neural network structure to obtain the final digital twin model.

[0035] As an improvement to the above solution, the building blocks are specifically used for:

[0036] Based on the energy conservation equation, a cell-scale electrochemical-thermal model is constructed on top of the P2D model, resulting in a model structure that integrates electrochemical and thermal characteristics at the cell scale.

[0037] Based on the basic model at the cell scale, the thermal stress and thermal expansion caused by temperature gradients during battery operation are considered, and a thermal-mechanical coupling model at the cell scale is constructed to form a comprehensive model that includes characteristics at both the cell and cell scales.

[0038] Based on the comprehensive model, starting from the scale of the power battery module and the entire pack, industrial modeling software is used to build a 3D CAD model corresponding to the actual shape of the physical battery. This accurately presents the key structural parameters, the constraints and positioning relationships between cells and modules, and obtains a 3D model containing multi-scale structural information.

[0039] Pixyz Studio is used to perform surface subdivision, topology correction, extraction, and repair operations on the three-dimensional model containing multi-scale structural information. Then, Hertz theory and the extended finite element method are combined to establish an overall digital physical model, clarify the semantic information annotation objects, and complete the construction of the multi-scale imaging model of the battery power system.

[0040] As an improvement to the above solution, the first modeling module is specifically configured to:

[0041] For the completed multi-scale imaging model of the battery power system, the parameter information related to the electrochemical field, temperature field, and stress field is comprehensively sorted out to obtain the parameter sets of each physical field at different scales;

[0042] Based on the parameter set, analyze the transfer path and method of each physical field parameter between different levels of models, clarify the law of parameter transfer, and obtain the specific rule description of parameter transfer;

[0043] Based on the obtained parameter transfer rule description, the interaction mechanism between different physical fields and different hierarchical models is studied, the influencing factors of the interaction are determined, and a list of key factors affecting the interaction is obtained;

[0044] According to the parameter set, the description of the parameter transfer rules and the list of key factors affecting the interaction, the multi-physics field coupling integration is performed on the multi-scale imaging model to obtain the coupled multi-scale imaging model.

[0045] As an improvement to the above solution, the update module is specifically configured to:

[0046] Through sensors pre-deployed in the battery power system, real-time monitoring data during system operation is collected to obtain a data set containing the real-time status of the battery power system;

[0047] A data-driven approach is used to extract information from a dataset containing the real-time status of the battery power system, filtering out valid information related to the multi-scale image model parameters to form a valid information set.

[0048] Comparing and matching the information in the valid information set with the parameters of the coupled multi-scale imaging model, determining the parameters that need to be updated and the update values, and obtaining a parameter update list;

[0049] According to the parameter update list, the parameters of the coupled multi-scale image model are updated to generate a preliminary digital twin model.

[0050] As an improvement to the above solution, the second modeling module is specifically configured to:

[0051] The sensor data in the preliminary digital twin model is abstracted into a network topology diagram of the sensor data time series accumulation process according to the time series and spatial relationship, and a topological structure reflecting the spatiotemporal characteristics of the sensor data is obtained;

[0052] A graph neural network-long short-term memory neural network structure is used to process the topological structure reflecting the spatiotemporal characteristics of sensor data, integrating the structural information and time information to form intermediate data that integrates spatiotemporal characteristics.

[0053] Using the designed hierarchical feature aggregation method, different feature aggregators are learned on the intermediate data of fused spatiotemporal features in neighborhoods of different depths to obtain aggregated data containing multi-scale features;

[0054] Perform dynamic time series modeling on sensor data, integrate aggregated data containing multi-scale features into the model, and obtain the final digital twin model.

[0055] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0056] First, a multiscale image model is constructed, spanning the battery cell to the drive motor. This hierarchical architecture comprehensively covers all levels of the system, ensuring accurate representation of characteristics from the microscopic to the macroscopic level. Multi-physics coupled modeling of electrochemical, temperature, and stress fields is then performed, transforming the traditional single-physics modeling approach. This model reproduces the real-world interactions of various physical fields during system operation, enhancing the model's ability to simulate complex operating conditions. Real-time monitoring data is then used to update the parameters of the coupled multi-scale image model through a data-driven model, enabling the model to respond in real time to dynamic factors such as battery aging and operating conditions, maintaining synchronization with the actual system. Finally, a graph neural network (GNN)-long short-term memory (LSTM) neural network architecture is employed. The former mines spatial correlations in sensor data, while the latter captures time series features, enabling spatiotemporal dynamic analysis of the system's operating status. Compared to existing technologies, which suffer from isolated multi-scale modeling, insufficient physical field coupling, data-driven lag, and limitations in time series modeling, this embodiment of the present invention achieves high-precision mapping and dynamic prediction of the operating status of the battery powertrain system for new energy heavy-duty trucks through hierarchical multi-scale modeling, deep physical field coupling, dynamic data-driven modeling, and spatiotemporal feature mining. This provides a solid and reliable digital foundation for efficient management and optimization of the system throughout its lifecycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of a method for constructing a digital twin model of a new energy heavy-duty truck battery power system provided by one embodiment of the present invention;

[0058] Figure 2 It is a structural schematic diagram of a digital twin model construction system for a new energy heavy-duty truck battery power system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0060] See also Figure 1 , is a flow chart of a method for constructing a digital twin model of a new energy heavy-duty truck battery power system according to one embodiment of the present invention. The method for constructing a digital twin model of a new energy heavy-duty truck battery power system comprises the following steps:

[0061] S10: Build a multi-scale image model of the battery power system, including hierarchical models from battery cells to modules, entire packs, controllers, and drive motors;

[0062] S11, performing multi-physics field coupling modeling of electrochemical field, temperature field, and stress field on the constructed multi-scale imaging model to obtain a coupled multi-scale imaging model;

[0063] S12, based on real-time monitoring data, uses a data-driven model to update the parameters of the coupled multi-scale image model to generate a preliminary digital twin model;

[0064] S13, dynamically model the sensor data in the preliminary digital twin model through the graph neural network-long short-term memory neural network structure to obtain the final digital twin model.

[0065] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0066] First, a multiscale image model is constructed, spanning the battery cell to the drive motor. This hierarchical architecture comprehensively covers all levels of the system, ensuring accurate representation of characteristics from the microscopic to the macroscopic level. Multi-physics coupled modeling of electrochemical, temperature, and stress fields is then performed, transforming the traditional single-physics modeling approach. This model reproduces the real-world interactions of various physical fields during system operation, enhancing the model's ability to simulate complex operating conditions. Real-time monitoring data is then used to update the parameters of the coupled multi-scale image model through a data-driven model, enabling the model to respond in real time to dynamic factors such as battery aging and operating conditions, maintaining synchronization with the actual system. Finally, a graph neural network (GNN)-long short-term memory (LSTM) neural network architecture is employed. The former mines spatial correlations in sensor data, while the latter captures time series features, enabling spatiotemporal dynamic analysis of the system's operating status. Compared to existing technologies, which suffer from isolated multi-scale modeling, insufficient physical field coupling, data-driven lag, and limitations in time series modeling, this embodiment of the present invention achieves high-precision mapping and dynamic prediction of the operating status of the battery powertrain system for new energy heavy-duty trucks through hierarchical multi-scale modeling, deep physical field coupling, dynamic data-driven modeling, and spatiotemporal feature mining. This provides a solid and reliable digital foundation for efficient management and optimization of the system throughout its lifecycle.

[0067] As one example, the multi-scale image model of the battery power system is constructed, including a hierarchical model from battery cells to modules, entire packs, controllers, and drive motors, and includes the following sub-steps:

[0068] Based on the energy conservation equation, a cell-scale electrochemical-thermal model is constructed on top of the P2D model, resulting in a model structure that integrates electrochemical and thermal characteristics at the cell scale.

[0069] Based on the basic model at the cell scale, the thermal stress and thermal expansion caused by temperature gradients during battery operation are considered, and a thermal-mechanical coupling model at the cell scale is constructed to form a comprehensive model that includes characteristics at both the cell and cell scales.

[0070] Based on the comprehensive model, starting from the scale of the power battery module and the entire pack, industrial modeling software is used to build a 3D CAD model corresponding to the actual shape of the physical battery. This accurately presents the key structural parameters, the constraints and positioning relationships between cells and modules, and obtains a 3D model containing multi-scale structural information.

[0071] Pixyz Studio is used to perform surface subdivision, topology correction, extraction, and repair operations on the three-dimensional model containing multi-scale structural information. Then, Hertz theory and the extended finite element method are combined to establish an overall digital physical model, clarify the semantic information annotation objects, and complete the construction of the multi-scale imaging model of the battery power system.

[0072] In this example, a high-precision multi-scale image model of the battery power system is constructed through a hierarchical modeling strategy that integrates multiple physical properties from micro to macro. First, based on the energy conservation equation, a cell-scale electrochemical-thermal model is constructed on the P2D model. This couples the electrochemical reactions within the battery with the heat transfer process, accurately depicting the core characteristics of the cell at the microscopic level. Next, based on the cell model, thermal stress and thermal expansion factors caused by temperature gradients are introduced to construct a coupled thermal-mechanical model at the cell scale. This extends the physical characteristics from microscopic cells to individual cells, enabling the model to reflect more complex physical phenomena. Then, based on the integrated model, industrial modeling software is used to construct 3D CAD models at the power battery module and full pack scales. This accurately represents the structural parameters of the battery entity and the constraints between components, completing structural modeling from the cell to the system level. Finally, the 3D model is optimized using Pixyz Studio, and a holistic digital physical model is established by combining Hertzian theory and the extended finite element method. This model is clearly annotated and achieves a deep integration of multi-scale structural and physical properties. In summary, this embodiment effectively solves the problems of multi-scale characteristic fragmentation and insufficient physical field coupling in traditional modeling through a layer-by-layer progressive modeling approach. It not only achieves accurate modeling of the full-level structure and physical characteristics from battery cells to the entire pack, but also ensures that the model can truly reflect the complex state of the battery power system in actual operation, providing a solid and reliable foundation for subsequent multi-physical field coupling modeling, parameter updating and dynamic timing analysis, and significantly improving the mapping accuracy and simulation capabilities of the digital twin model for the battery power system.

[0073] As an example, the working process of this embodiment is as follows:

[0074] First, when constructing a cell-scale electrochemical-thermal model, we build on the P2D (pseudo-two-dimensional) model based on the energy conservation equation. During the cell charge and discharge process, we consider the heat generated by the electrochemical reaction and the conduction and convection of heat within the cell. We introduce an improved heat transfer algorithm formula: Among them, Q total Represents the total heat generated by the battery cell; n is the number of microelements divided in the battery cell; α i is the convective heat transfer coefficient between the ith element and the environment, which can be obtained by experimentally measuring the heat transfer between the battery cell and the surrounding air under different materials and surface conditions; T i is the temperature of the ith element, which is collected in real time by high-precision temperature sensors pre-placed at different positions of the battery cell; T amb is the ambient temperature, measured by an ambient temperature sensor; β is a coefficient related to the cell material and structure, calculated by analyzing the cell material properties and internal structural parameters; I is the current flowing through the cell, monitored in real time by the battery management system; and R is the cell's internal resistance, which is calculated by combining factory parameters with an aging model. By combining the electrochemical heat generation and heat transfer processes of the cell, we develop a model structure that integrates electrochemical and thermal characteristics at the cell scale.

[0075] Next, based on the above-mentioned basic model at the cell scale, a thermal-mechanical coupling model at the cell scale is constructed. The thermal stress and thermal expansion factors caused by the temperature gradient during battery operation are taken into account. For the calculation of thermal stress, an improved stress calculation model is used: thermal =γ·ΔT·E·(1-v). Among them, σ thermal is the thermal stress; γ is the thermal expansion coefficient of the cell material, determined through material characterization experiments; ΔT is the temperature difference between different locations within the cell, calculated from the temperatures of each microelement in the previously constructed cell-scale model; E is the elastic modulus of the material, which can be modified based on the material manual and actual production process; and v is the Poisson's ratio of the material, obtained through experimental measurements. This formula is used to calculate thermal stress and combine it with the cell-scale model, taking into account the structural changes caused by thermal expansion of the battery cell, to form a comprehensive model that incorporates both cell- and cell-scale characteristics.

[0076] Then, based on the comprehensive model, starting from the scale of the power battery module and the entire pack, industrial modeling software (such as CATIA) is used to create a 3D CAD model corresponding to the actual shape of the physical battery. During the modeling process, based on the battery design drawings and actual production process parameters, the size, shape, and relative positions of the battery module and the entire pack are precisely set, presenting key structural parameters such as the arrangement of single cells, the connection structure between modules, and the location of fixings. At the same time, the software's constraint and positioning functions are used to accurately simulate the constraint and positioning relationships between cells and modules, obtaining a 3D model containing multi-scale structural information.

[0077] Finally, Pixyz Studio was used to perform surface subdivision, topology correction, extraction, and repair on the 3D model containing multi-scale structural information. On this basis, the overall digital physical model was established by combining Hertz theory and the extended finite element method. During the modeling process, the contact mechanics between battery cells was analyzed based on Hertz theory using the formula: Among them, F contact is the contact force between cells; k is the contact stiffness coefficient, which is related to material properties and contact surface geometry and is calculated through analysis of battery cell materials and surface treatment processes; δ is the contact deformation between cells, calculated using finite element analysis software. The extended finite element method is used to simulate and analyze potential battery defects such as cracks. After model construction is complete, semantic information is clearly annotated, such as key battery components and parameter measurement points, to complete the construction of a multi-scale image model of the battery power system.

[0078] As one example, the multi-physics field coupling modeling of electrochemical field, temperature field, and stress field is performed on the constructed multi-scale imaging model to obtain the coupled multi-scale imaging model, including the following sub-steps:

[0079] For the completed multi-scale imaging model of the battery power system, the parameter information related to the electrochemical field, temperature field, and stress field is comprehensively sorted out to obtain the parameter sets of each physical field at different scales;

[0080] Based on the parameter set, analyze the transfer path and method of each physical field parameter between different levels of models, clarify the law of parameter transfer, and obtain the specific rule description of parameter transfer;

[0081] Based on the obtained parameter transfer rule description, the interaction mechanism between different physical fields and different hierarchical models is studied, the influencing factors of the interaction are determined, and a list of key factors affecting the interaction is obtained;

[0082] According to the parameter set, the description of the parameter transfer rules and the list of key factors affecting the interaction, the multi-physics field coupling integration is performed on the multi-scale imaging model to obtain the coupled multi-scale imaging model.

[0083] In this embodiment, a deep coupling of multiple physical fields in a multi-scale image model of a battery power system is achieved through systematic parameter combing, transfer rule mining, and interaction analysis. First, the relevant parameters of the electrochemical field, temperature field, and stress field in the multi-scale image model are comprehensively combed, and parameter sets at different scales are constructed to provide a data foundation for subsequent coupling. Next, the transfer paths and methods of each physical field parameter between hierarchical models are deeply analyzed, the parameter transfer rules are clarified, and parameter transfer rule descriptions are established to solve the problem of isolated and unrelated parameters in traditional models. Then, the interaction mechanism between different physical fields and hierarchical models is studied, and the key factors affecting the interaction are determined. A list of key factors is formed to provide a direction for coupled modeling. Finally, the parameter set, transfer rules, and key factors are integrated into the multi-scale image model to complete the multi-physics field coupled modeling. In summary, this embodiment overcomes the defects of multi-physical field separation and insufficient coupling in traditional modeling through systematic research on physical field parameters, transfer rules and interaction mechanisms, realizes collaborative modeling of electrochemical fields, temperature fields and stress fields at different scales, and significantly improves the simulation accuracy of the digital twin model for the complex physical phenomena of the battery power system, providing a more realistic and reliable model basis for system performance analysis and optimization, and enhancing the model's adaptability and predictive ability to actual working conditions.

[0084] As an example, the working process of this embodiment is as follows:

[0085] 1. For the constructed multi-scale imaging model, a comprehensive review of the parameters of each physical field is first conducted. For the electrochemical field, at the cell scale, the exchange current density and lithium ion diffusion coefficient of the electrode material are measured using an electrochemical workstation; at the module and whole package scale, multi-point voltage monitoring combined with an equivalent circuit model is used to calculate the internal resistance distribution and SOC (state of charge) distribution. For the temperature field, thermocouples and infrared thermal imagers distributed at key positions of the battery are used to obtain temperature distribution data, and thermal conductivity and specific heat capacity are calculated using thermal imaging analysis software. For the stress field, optical fiber sensors are implanted during the cell manufacturing process to monitor stress changes during charging and discharging in real time. Combined with the elastic modulus and Poisson's ratio measured by material mechanics experiments, a set of stress field parameters is constructed. By establishing a parameter mapping table, the parameters of each physical field are classified and stored according to the four scale levels of cell, monomer, module, and whole package to form a complete parameter matrix.

[0086] 2. Based on the parameter set obtained, the transmission paths and methods of each physical field parameter between different hierarchical models are analyzed. Taking the coupling of temperature field and electrochemical field as an example, temperature changes will significantly affect the electrochemical reaction rate and lithium ion diffusion coefficient. This paper improves the traditional Arrhenius equation, proposes a temperature correction factor ξ, and establishes a dynamic correlation model between temperature and electrochemical parameters: Among them, C Li is the lithium ion concentration, D eff is the effective diffusion coefficient, is the Laplace operator, is the temperature change rate. The temperature correction factor ξ is obtained by fitting experimental data and reflects the degree of influence of different temperature ranges on lithium-ion diffusion. This model accurately describes the transfer process of temperature field parameters to electrochemical field parameters. Similarly, for parameter transfer between other physical fields, similar methods are used to establish transfer rules, ultimately forming a complete description of parameter transfer rules.

[0087] 3. Based on the parameter transfer rules, we conduct in-depth research on the interaction mechanism between different physical fields and different hierarchical models. Taking the interaction between the temperature field and the stress field as an example, temperature changes will cause the battery material to expand and contract, generating thermal stress; and stress changes will affect the thermal conductivity of the material, forming a bidirectional coupling. This paper proposes a thermal stress calculation model: σ thermal =γ·ΔT·E·(1-v), where σ thermal is thermal stress, γ is the thermal expansion coefficient, ΔT is the temperature difference, E is the elastic modulus, and v is the Poisson's ratio. Through this model, the thermal stress caused by temperature changes can be accurately calculated. At the same time, the study found that stress changes will cause changes in the internal microstructure of the material, which in turn affects the thermal conductivity coefficient. A stress-thermal conduction coupling model was established: λ(σ)=λ0·(1+α·σ), where λ(σ) is the thermal conductivity coefficient under stress, λ0 is the initial thermal conductivity coefficient, α is the stress influence coefficient, and σ is the stress value. Through the above two models, the key influencing factors of the interaction between the temperature field and the stress field are clarified, forming a complete list of key factors affecting the interaction.

[0088] 4. Based on the parameter set, parameter transfer rules, and the list of key interaction factors, the multi-scale mapping model is coupled with multiple physical fields. This paper proposes a comprehensive coupling weight algorithm to unify the modeling of electrochemical fields, temperature fields, and stress fields: Among them, S total is the comprehensive state parameter after multi-physics field coupling, S i is the state parameter of each physical field (i=1, 2, 3 represent electrochemical field, temperature field, stress field, respectively), w i is the weight coefficient of each physical field, λ ijis the coupling coefficient between physical fields. Weight coefficient w i Dynamic adjustment according to the working state of the battery, for example, in the fast charging state, the electrochemical field weight increases; in the high temperature environment, the temperature field weight increases. ij The results are obtained by fitting experimental data and reflect the interaction strength between different physical fields. Through this comprehensive coupling model, the organic integration of the three physical fields is achieved, and a complete coupled multi-scale imaging model is obtained.

[0089] As one example, based on real-time monitoring data, a data-driven model is used to update the parameters of the coupled multi-scale image model to generate a preliminary digital twin model, which includes the following sub-steps:

[0090] Through sensors pre-deployed in the battery power system, real-time monitoring data during system operation is collected to obtain a data set containing the real-time status of the battery power system;

[0091] A data-driven approach is used to extract information from a dataset containing the real-time status of the battery power system, filtering out valid information related to the multi-scale image model parameters to form a valid information set.

[0092] Comparing and matching the information in the valid information set with the parameters of the coupled multi-scale imaging model, determining the parameters that need to be updated and the update values, and obtaining a parameter update list;

[0093] According to the parameter update list, the parameters of the coupled multi-scale image model are updated to generate a preliminary digital twin model.

[0094] In this embodiment, a dynamic parameter update mechanism driven by real-time data is constructed to ensure the synchronization of the digital twin model and the physical system. First, real-time monitoring data is collected from sensors pre-deployed in the battery power system to obtain the original operating status information of the system, providing a data foundation for model updates. Second, a data-driven approach is used to deeply mine the original dataset, filtering out valid information related to the multi-scale image model parameters, eliminating redundant data, and improving data utilization efficiency. Then, the valid information is compared and matched with the model parameters to accurately locate the parameters and their values ​​that need to be updated, forming a parameter update list and clarifying the update direction. Finally, the parameters of the coupled multi-scale image model are adjusted based on the update list, so that the model can reflect the changes in the operating status of the battery power system in real time, and a preliminary digital twin model is generated. In summary, this implementation breaks the limitations of static setting of traditional model parameters through real-time data collection, effective information extraction, precise parameter matching and dynamic updating, solves the problem of disconnection between the model and the actual system operating status, and significantly improves the dynamic adaptability and accuracy of the digital twin model, enabling the model to respond to dynamic factors such as battery aging and operating condition changes in a timely manner, providing more reliable digital model support for real-time monitoring, fault diagnosis and optimization control of battery power systems.

[0095] As an example, the working process of this embodiment is as follows:

[0096] Step 1: Real-time monitoring data collection: Build a comprehensive and sophisticated sensor monitoring network in the new energy heavy-duty truck battery power system. At the cell level, high-precision micro-thermocouple sensors are installed at the positive and negative tabs of each cell, with a measurement accuracy of ±0.1°C, which can accurately capture subtle changes in the local temperature of the cell during the charge and discharge process; on both sides of the cell diaphragm, a special micro-solid electrolyte lithium ion concentration sensor is embedded. This sensor can collect lithium ion concentration data in real time, with a fast response speed and a measurement range of 0.1-2 mol / m 3. At the module level, high-precision Hall current sensors (accuracy ±0.2%) are used to monitor the current distribution of each module. At the same time, strain gauges are attached to the key stress-bearing parts of the module to measure the mechanical stress changes inside the module caused by battery expansion and contraction. At the battery pack level, pressure sensors are reasonably arranged at key positions such as the top, bottom and sides of the box to monitor the air pressure fluctuations inside the whole pack; and a high-resolution infrared thermal imager is installed above the whole pack to perform real-time scanning of the temperature field distribution of the whole pack with a resolution of up to 640×480. All sensors are connected to the data acquisition unit via the CAN bus. The data acquisition unit will adopt a dynamic sampling strategy based on the parameter change characteristics: for parameters that change more slowly, such as voltage and temperature, sampling is performed at a frequency of 1Hz; for parameters that change rapidly, such as current and stress, high-frequency sampling is performed at a frequency of 100Hz. The collected data contains information such as timestamp, sensor ID, measurement value, and data quality identification, and ultimately forms a multidimensional data set D covering the real-time status of the battery power system.

[0097] Step 2, data-driven effective information extraction: When filtering out effective information related to the multi-scale imaging model parameters from the original data set D, an improved feature selection method is used. First, the principal component analysis (PCA) algorithm is used to reduce the dimensionality of the original data. The PCA algorithm maps the original data to a new low-dimensional space through linear transformation, removes redundant information in the data, reduces the data dimension while retaining the main feature information. After completing the dimensionality reduction, the improved recursive feature elimination (RFE) algorithm is introduced and combined with the model parameter sensitivity index for feature screening. The parameter sensitivity formula is defined as:

[0098]

[0099] Among them, S i represents the sensitivity of the i-th parameter, which is used to measure the influence of the parameter on the model output; M represents the output of the multi-scale image model, such as the battery voltage prediction value, temperature prediction value, etc.; P i is the i-th parameter in the model; n is the total number of model parameters. The sensitivity of each data feature and model parameter is calculated by this formula, and the features with high sensitivity are retained first, and these features are integrated to form an effective information set I eff This improved algorithm significantly improves the accuracy and pertinence of information screening by quantifying the impact of features on model output, and can more accurately find data features that are valuable for updating model parameters.

[0100] Step 3, parameter comparison and matching: Set the valid information set I effCompare with the coupled multi-scale image model parameters to determine the parameters that need to be updated and their update values. Here, an improved dynamic threshold matching algorithm is used to construct the parameter deviation evaluation formula as follows:

[0101]

[0102] Among them, Δ k It represents the deviation ratio of the kth parameter, which is used to intuitively reflect the degree of deviation between the measured value and the model predicted value; is the measured value of the kth parameter, which is collected by the sensor in step 1 and obtained after data processing; is the model prediction value of the kth parameter, which is calculated by the multi-scale imaging model based on the current parameter state; The reference value of the parameter is usually the design nominal value or the stable value obtained through historical data statistics. Set the dynamic threshold θ k =θ0×(1+β×SOH+γ×SOC), where θ0 is the basic threshold, which is the initial judgment standard set based on experience and experiments; β and γ are adjustment coefficients, which are optimized and determined through a large amount of experimental data and actual operation conditions; SOH represents the battery health state, reflecting the degree of battery aging; SOC is the state of charge, indicating the current remaining capacity of the battery. k >θ k When , it is determined that the parameter needs to be updated, and then a parameter update list L is generated. The list records in detail the name, current value, target update value, priority identifier and other information of the parameter to be updated, providing a clear basis for subsequent model parameter updates.

[0103] Step 4: Multi-scale image model parameter update: Based on the parameter update list L, a hierarchical iterative update strategy is used to modify the parameters of the coupled multi-scale image model. For independent parameters, such as cell internal resistance, which are less affected by other parameters, the updated value can be directly used to replace the original parameter to complete the parameter update. For coupled parameters, such as electrochemical-thermal-stress coupled parameters, a collaborative update algorithm is used:

[0104]

[0105] in, Represents the updated i-th parameter value; is the original value of the parameter; ΔP i is the parameter update amount calculated in step 3; iis the update weight coefficient, determined by the parameter's level and coupling relationship. Generally speaking, cell-level parameters have a higher weight due to their more direct and critical impact on battery performance, while package-level parameters have a lower weight. During the update process, through continuous iterative calculations and continuous adjustment of model parameters, we ensure that the error between the model output and the measured data gradually converges to within a pre-set threshold, ultimately generating a preliminary digital twin model that accurately reflects the real-time status of the battery power system.

[0106] As one example, the method of dynamically modeling the sensor data in the preliminary digital twin model using the graph neural network-long short-term memory neural network structure to obtain the final digital twin model includes the following sub-steps:

[0107] The sensor data in the preliminary digital twin model is abstracted into a network topology diagram of the sensor data time series accumulation process according to the time series and spatial relationship, and a topological structure reflecting the spatiotemporal characteristics of the sensor data is obtained;

[0108] A graph neural network-long short-term memory neural network structure is used to process the topological structure reflecting the spatiotemporal characteristics of sensor data, integrating the structural information and time information to form intermediate data that integrates spatiotemporal characteristics.

[0109] Using the designed hierarchical feature aggregation method, different feature aggregators are learned on the intermediate data of fused spatiotemporal features in neighborhoods of different depths to obtain aggregated data containing multi-scale features;

[0110] Perform dynamic time series modeling on sensor data, integrate aggregated data containing multi-scale features into the model, and obtain the final digital twin model.

[0111] In this embodiment, real-time operating data is first collected by sensors deployed in the battery power system to obtain original information on the system's true state and build a data foundation. Next, a data-driven approach is used to deeply process the original data set, filtering out valid information related to the multi-scale image model parameters from the massive data, removing redundancy, and improving data quality. Then, the valid information is accurately compared with the coupled multi-scale image model parameters to identify parameter differences, determine the parameters that need to be updated and their corresponding values, and form a parameter update list. Finally, the model parameters are adjusted based on the list to make the model fit the system's operating status in real time, generating a preliminary digital twin model. In summary, this embodiment effectively solves the problem of traditional model parameters being fixed and difficult to reflect dynamic changes in the system. Through real-time data collection and precise parameter updates, it ensures that the digital twin model can dynamically adjust to the operating status of the battery power system, significantly improving the matching and synchronization between the model and the actual system, providing dynamic and accurate model support for the monitoring, analysis, and optimization of battery power systems based on digital twins, and enhancing the reliability of system prediction and decision-making.

[0112] As an example, the working process of this embodiment is as follows:

[0113] Step 1, sensor data topology map construction: In the preliminary digital twin model, the spatiotemporal features of the sensor data of the battery power system are abstracted. Taking the battery module as an example, the battery cell temperature sensor, voltage sensor, current sensor and other devices are defined as network nodes according to their physical location and connection relationship; the physical connection lines between sensors (such as CAN bus connection, module internal circuit connection) or data interaction logical relationship are used as network edges. For time series data, 1 minute is used as the time window, and the data collected by the sensor in each window is used as the node attribute. For example, the node v i Contains the average temperature T of the i-th cell in the current time window i , voltage V i , charge and discharge current I i Equal attributes, forming attribute vector X i =[T i , V i , I i In this way, the sensor data is constructed as a directed weighted network topology graph with attributes. in is the node set, ε is the edge set, and X is the attribute matrix of all nodes. This topological graph fully reflects the spatial distribution (node ​​location and connection relationship) and time series characteristics (data aggregation within the window) of sensor data.

[0114] Step 2: Spatiotemporal feature fusion processing: An improved graph neural network-long short-term memory neural network (GNN-LSTM) fusion structure is used for data processing. First, the topological graph $\mathcal{G}$ is input into the graph neural network layer (GNNLayer), and the node features are updated using an improved message passing algorithm:

[0115]

[0116] in, is the hidden state of node i in the lth layer GNN; is the set of neighbor nodes of node i; e ij is the weight of the edge between nodes i and j (set according to the physical connection strength or data correlation); W (l) and b (l) is the weight matrix and bias vector of the lth layer; σ is the activation function (such as ReLU). This formula strengthens the extraction of spatial structure features by weighted aggregation of neighbor node information.

[0117] Subsequently, the node feature sequence output by GNN is input into the long short-term memory neural network (LSTM) layer. In order to adapt to the multi-scale characteristics of sensor data, the forget gate, input gate and output gate of LSTM are improved, and the spatiotemporal weight coefficient α is introduced. t :

[0118] f t =σ(W f ·[h t-1 , x t ]+b f )·α t

[0119] i t =σ(W i ·[h t-1 , x t ]+b i )·(1-α t )

[0120] o t =σ(W o ·[h t-1 , x t ]+b o )·α t

[0121] Among them, f t 、i t 、o t are the outputs of the forget gate, input gate, and output gate respectively; h t-1 is the hidden state at the previous moment; x t Input for the current moment; Wf 、W i 、W o and b f 、b i 、b o are weight and bias parameters; α t is the spatiotemporal weight coefficient, which is dynamically adjusted according to the fluctuation of sensor data in the current time window (calculated through historical data regression). After the joint processing of GNN and LSTM, the intermediate data Z that integrates spatiotemporal features is generated.

[0122] Step 3, hierarchical feature aggregation: Design a hierarchical feature aggregation method to achieve multi-scale feature extraction. Divide the intermediate data Z into three categories according to the node level: cell level, module level, and whole package level. In each level, an improved aggregator is used Perform feature learning:

[0123]

[0124] in, is the output of the aggregator of node i at the kth layer; is the set of nodes in the neighborhood of node i in the kth layer; z j is the eigenvector of node j. Taking the cell level as an example, Using a weighted average aggregator:

[0125]

[0126] Among them, w ij is the correlation weight between nodes i and j, calculated based on the thermal conductivity and electrical connection tightness between cells. Attention and pooling aggregators are used at the module and package levels, respectively, to highlight key node features. Through this three-layer aggregation process, aggregated data Y is generated, encompassing cell microscale, module mesoscale, and package macroscale features.

[0127] Step 4, dynamic time series modeling and final model generation: Based on the aggregated data Y, a dynamic time series prediction model is constructed. An improved version of the autoregressive integrated moving average model (ARIMA) is used, and the multi-scale feature weighting coefficient ω is introduced. k :

[0128]

[0129] in, is the predicted value at time t; ω k is the weight coefficient of the k-th scale feature (optimized by cross-validation); φ i，k and θ i，k are the autoregressive and moving average coefficients respectively; ∈ tis a white noise sequence; p and q are the model orders. This formula enhances the model's adaptability to complex dynamic changes by integrating multi-scale features. The prediction results are fed back into the digital twin model to update the model parameters, ultimately generating a final digital twin model that accurately reflects the dynamic behavior of the battery powertrain.

[0130] It should be noted that the details of the relevant algorithm technologies that are not described in detail above can be referred to the existing relevant algorithm technologies and will not be repeated here.

[0131] See also Figure 2 , is a structural diagram of a digital twin model construction system for a new energy heavy-duty truck battery power system provided by one embodiment of the present invention. The digital twin model construction system for a new energy heavy-duty truck battery power system includes:

[0132] Construction module 10 is used to build a multi-scale image model of the battery power system, including hierarchical models from battery cells to modules, entire packs, controllers, and drive motors;

[0133] The first modeling module 11 is used to perform multi-physics field coupling modeling of electrochemical field, temperature field, and stress field on the constructed multi-scale imaging model to obtain a coupled multi-scale imaging model;

[0134] An updating module 12 is used to update the parameters of the coupled multi-scale image model based on real-time monitoring data using a data-driven model to generate a preliminary digital twin model;

[0135] The second modeling module 13 is used to perform dynamic time series modeling on the sensor data in the preliminary digital twin model through the graph neural network-long short-term memory neural network structure to obtain the final digital twin model.

[0136] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0137] First, a multiscale image model is constructed, spanning the battery cell to the drive motor. This hierarchical architecture comprehensively covers all levels of the system, ensuring accurate representation of characteristics from the microscopic to the macroscopic level. Multi-physics coupled modeling of electrochemical, temperature, and stress fields is then performed, transforming the traditional single-physics modeling approach. This model reproduces the real-world interactions of various physical fields during system operation, enhancing the model's ability to simulate complex operating conditions. Real-time monitoring data is then used to update the parameters of the coupled multi-scale image model through a data-driven model, enabling the model to respond in real time to dynamic factors such as battery aging and operating conditions, maintaining synchronization with the actual system. Finally, a graph neural network (GNN)-long short-term memory (LSTM) neural network architecture is employed. The former mines spatial correlations in sensor data, while the latter captures time series features, enabling spatiotemporal dynamic analysis of the system's operating status. Compared to existing technologies, which suffer from isolated multi-scale modeling, insufficient physical field coupling, data-driven lag, and limitations in time series modeling, this embodiment of the present invention achieves high-precision mapping and dynamic prediction of the operating status of the battery powertrain system for new energy heavy-duty trucks through hierarchical multi-scale modeling, deep physical field coupling, dynamic data-driven modeling, and spatiotemporal feature mining. This provides a solid and reliable digital foundation for efficient management and optimization of the system throughout its lifecycle.

[0138] As one example, the building blocks are specifically used to:

[0139] Based on the energy conservation equation, a cell-scale electrochemical-thermal model is constructed on top of the P2D model, resulting in a model structure that integrates electrochemical and thermal characteristics at the cell scale.

[0140] Based on the basic model at the cell scale, the thermal stress and thermal expansion caused by temperature gradients during battery operation are considered, and a thermal-mechanical coupling model at the cell scale is constructed to form a comprehensive model that includes characteristics at both the cell and cell scales.

[0141] Based on the comprehensive model, starting from the scale of the power battery module and the entire pack, industrial modeling software is used to build a 3D CAD model corresponding to the actual shape of the physical battery. This accurately presents the key structural parameters, the constraints and positioning relationships between cells and modules, and obtains a 3D model containing multi-scale structural information.

[0142] Pixyz Studio is used to perform surface subdivision, topology correction, extraction, and repair operations on the three-dimensional model containing multi-scale structural information. Then, Hertz theory and the extended finite element method are combined to establish an overall digital physical model, clarify the semantic information annotation objects, and complete the construction of the multi-scale imaging model of the battery power system.

[0143] As one example, the first modeling module is specifically configured to:

[0144] For the completed multi-scale imaging model of the battery power system, the parameter information related to the electrochemical field, temperature field, and stress field is comprehensively sorted out to obtain the parameter sets of each physical field at different scales;

[0145] Based on the parameter set, analyze the transfer path and method of each physical field parameter between different levels of models, clarify the law of parameter transfer, and obtain the specific rule description of parameter transfer;

[0146] Based on the obtained parameter transfer rule description, the interaction mechanism between different physical fields and different hierarchical models is studied, the influencing factors of the interaction are determined, and a list of key factors affecting the interaction is obtained;

[0147] According to the parameter set, the description of the parameter transfer rules and the list of key factors affecting the interaction, the multi-physics field coupling integration is performed on the multi-scale imaging model to obtain the coupled multi-scale imaging model.

[0148] As one example, the update module is specifically configured to:

[0149] Through sensors pre-deployed in the battery power system, real-time monitoring data during system operation is collected to obtain a data set containing the real-time status of the battery power system;

[0150] A data-driven approach is used to extract information from a dataset containing the real-time status of the battery power system, filtering out valid information related to the multi-scale image model parameters to form a valid information set.

[0151] Comparing and matching the information in the valid information set with the parameters of the coupled multi-scale imaging model, determining the parameters that need to be updated and the update values, and obtaining a parameter update list;

[0152] According to the parameter update list, the parameters of the coupled multi-scale image model are updated to generate a preliminary digital twin model.

[0153] As one example, the second modeling module is specifically configured to:

[0154] The sensor data in the preliminary digital twin model is abstracted into a network topology diagram of the sensor data time series accumulation process according to the time series and spatial relationship, and a topological structure reflecting the spatiotemporal characteristics of the sensor data is obtained;

[0155] A graph neural network-long short-term memory neural network structure is used to process the topological structure reflecting the spatiotemporal characteristics of sensor data, integrating the structural information and time information to form intermediate data that integrates spatiotemporal characteristics.

[0156] Using the designed hierarchical feature aggregation method, different feature aggregators are learned on the intermediate data of fused spatiotemporal features in neighborhoods of different depths to obtain aggregated data containing multi-scale features;

[0157] Perform dynamic time series modeling on sensor data, integrate aggregated data containing multi-scale features into the model, and obtain the final digital twin model.

[0158] It should be noted that the system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the system embodiment provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive work.

[0159] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for constructing a digital twin model of a new energy heavy-duty truck battery power system, characterized in that: include: Construct a multi-scale image model of the battery power system, including hierarchical models from battery cells to modules, entire packs, controllers, and drive motors; Conduct multi-physics field coupling modeling of electrochemical field, temperature field, and stress field on the constructed multi-scale imaging model to obtain a coupled multi-scale imaging model; Based on real-time monitoring data, a data-driven model is used to update the parameters of the coupled multi-scale image model to generate a preliminary digital twin model; The sensor data in the preliminary digital twin model is dynamically modeled using the graph neural network-long short-term memory neural network structure to obtain the final digital twin model.

2. The method for constructing a digital twin model of a new energy heavy truck battery power system according to claim 1, characterized in that: The multi-scale image model of the battery power system is constructed, including a hierarchical model from battery cells to modules, entire packs, controllers, and drive motors, and includes the following sub-steps: Based on the energy conservation equation, a cell-scale electrochemical-thermal model is constructed on top of the P2D model, resulting in a model structure that integrates electrochemical and thermal characteristics at the cell scale. Based on the basic model at the cell scale, the thermal stress and thermal expansion caused by temperature gradients during battery operation are considered, and a thermal-mechanical coupling model at the cell scale is constructed to form a comprehensive model that includes characteristics at both the cell and cell scales. Based on the comprehensive model, starting from the scale of the power battery module and the entire pack, industrial modeling software is used to build a 3D CAD model corresponding to the actual shape of the physical battery. This accurately presents the key structural parameters, the constraints and positioning relationships between cells and modules, and obtains a 3D model containing multi-scale structural information. Pixyz Studio is used to perform surface subdivision, topology correction, extraction, and repair operations on the three-dimensional model containing multi-scale structural information. Then, Hertz theory and the extended finite element method are combined to establish an overall digital physical model, clarify the semantic information annotation objects, and complete the construction of the multi-scale imaging model of the battery power system.

3. The method for constructing a digital twin model of a new energy heavy truck battery power system according to claim 1, characterized in that: The multi-scale imaging model constructed is subjected to multi-physics field coupling modeling of electrochemical field, temperature field, and stress field to obtain a coupled multi-scale imaging model, including the following sub-steps: For the completed multi-scale imaging model of the battery power system, the parameter information related to the electrochemical field, temperature field, and stress field is comprehensively sorted out to obtain the parameter sets of each physical field at different scales; Based on the parameter set, analyze the transfer path and method of each physical field parameter between different levels of models, clarify the law of parameter transfer, and obtain the specific rule description of parameter transfer; Based on the obtained parameter transfer rule description, the interaction mechanism between different physical fields and different hierarchical models is studied, the influencing factors of the interaction are determined, and a list of key factors affecting the interaction is obtained; According to the parameter set, the description of the parameter transfer rules and the list of key factors affecting the interaction, the multi-physics field coupling integration is performed on the multi-scale imaging model to obtain the coupled multi-scale imaging model.

4. The method for constructing a digital twin model of a new energy heavy truck battery power system according to claim 1, characterized in that: The method uses a data-driven model to update the parameters of the coupled multi-scale image model based on real-time monitoring data to generate a preliminary digital twin model, including the following sub-steps: Through sensors pre-deployed in the battery power system, real-time monitoring data during system operation is collected to obtain a data set containing the real-time status of the battery power system; A data-driven approach is used to extract information from a dataset containing the real-time status of the battery power system, filtering out valid information related to the multi-scale image model parameters to form a valid information set. Comparing and matching the information in the valid information set with the parameters of the coupled multi-scale imaging model, determining the parameters that need to be updated and the update values, and obtaining a parameter update list; According to the parameter update list, the parameters of the coupled multi-scale image model are updated to generate a preliminary digital twin model.

5. The method for constructing a digital twin model of a new energy heavy truck battery power system according to claim 1, characterized in that: The method of dynamically modeling the sensor data in the preliminary digital twin model through the graph neural network-long short-term memory neural network structure to obtain the final digital twin model includes the following sub-steps: The sensor data in the preliminary digital twin model is abstracted into a network topology diagram of the sensor data time series accumulation process according to the time series and spatial relationship, and a topological structure reflecting the spatiotemporal characteristics of the sensor data is obtained; A graph neural network-long short-term memory neural network structure is used to process the topological structure reflecting the spatiotemporal characteristics of sensor data, integrating the structural information and time information to form intermediate data that integrates spatiotemporal characteristics. Using the designed hierarchical feature aggregation method, different feature aggregators are learned on the intermediate data of fused spatiotemporal features in neighborhoods of different depths to obtain aggregated data containing multi-scale features; Perform dynamic time series modeling on sensor data, integrate aggregated data containing multi-scale features into the model, and obtain the final digital twin model.

6. A digital twin model construction system for a new energy heavy-duty truck battery power system, characterized in that: include: A construction module for building a multi-scale image model of the battery power system, including hierarchical models from battery cells to modules, entire packs, controllers, and drive motors; The first modeling module is used to perform multi-physics field coupling modeling of electrochemical field, temperature field, and stress field on the constructed multi-scale imaging model to obtain a coupled multi-scale imaging model; The update module is used to update the parameters of the coupled multi-scale image model based on real-time monitoring data using a data-driven model to generate a preliminary digital twin model; The second modeling module is used to perform dynamic time series modeling of the sensor data in the preliminary digital twin model through the graph neural network-long short-term memory neural network structure to obtain the final digital twin model.

7. The digital twin model construction system for the new energy heavy truck battery power system according to claim 6, characterized in that: The building blocks are specifically used for: Based on the energy conservation equation, a cell-scale electrochemical-thermal model is constructed on top of the P2D model, resulting in a model structure that integrates electrochemical and thermal characteristics at the cell scale. Based on the basic model at the cell scale, the thermal stress and thermal expansion caused by temperature gradients during battery operation are considered, and a thermal-mechanical coupling model at the cell scale is constructed to form a comprehensive model that includes characteristics at both the cell and cell scales. Based on the comprehensive model, starting from the scale of the power battery module and the entire pack, industrial modeling software is used to build a 3D CAD model corresponding to the actual shape of the physical battery. This accurately presents the key structural parameters, the constraints and positioning relationships between cells and modules, and obtains a 3D model containing multi-scale structural information. Pixyz Studio is used to perform surface subdivision, topology correction, extraction, and repair operations on the three-dimensional model containing multi-scale structural information. Then, Hertz theory and the extended finite element method are combined to establish an overall digital physical model, clarify the semantic information annotation objects, and complete the construction of the multi-scale imaging model of the battery power system.

8. The digital twin model construction system for the new energy heavy truck battery power system according to claim 6, characterized in that: The first modeling module is specifically used for: For the completed multi-scale imaging model of the battery power system, the parameter information related to the electrochemical field, temperature field, and stress field is comprehensively sorted out to obtain the parameter sets of each physical field at different scales; Based on the parameter set, analyze the transfer path and method of each physical field parameter between different levels of models, clarify the law of parameter transfer, and obtain the specific rule description of parameter transfer; Based on the obtained parameter transfer rule description, the interaction mechanism between different physical fields and different hierarchical models is studied, the influencing factors of the interaction are determined, and a list of key factors affecting the interaction is obtained; According to the parameter set, the description of the parameter transfer rules and the list of key factors affecting the interaction, the multi-physics field coupling integration is performed on the multi-scale imaging model to obtain the coupled multi-scale imaging model.

9. The digital twin model construction system for the new energy heavy truck battery power system according to claim 6, characterized in that: The update module is specifically used for: Through sensors pre-deployed in the battery power system, real-time monitoring data during system operation is collected to obtain a data set containing the real-time status of the battery power system; A data-driven approach is used to extract information from a dataset containing the real-time status of the battery power system, filtering out valid information related to the multi-scale image model parameters to form a valid information set. Comparing and matching the information in the valid information set with the parameters of the coupled multi-scale imaging model, determining the parameters that need to be updated and the update values, and obtaining a parameter update list; According to the parameter update list, the parameters of the coupled multi-scale image model are updated to generate a preliminary digital twin model.

10. The digital twin model construction system for the new energy heavy truck battery power system according to claim 6, characterized in that: The second modeling module is specifically used for: The sensor data in the preliminary digital twin model is abstracted into a network topology diagram of the sensor data time series accumulation process according to the time series and spatial relationship, and a topological structure reflecting the spatiotemporal characteristics of the sensor data is obtained; A graph neural network-long short-term memory neural network structure is used to process the topological structure reflecting the spatiotemporal characteristics of sensor data, integrating the structural information and time information to form intermediate data that integrates spatiotemporal characteristics. Using the designed hierarchical feature aggregation method, different feature aggregators are learned on the intermediate data of fused spatiotemporal features in neighborhoods of different depths to obtain aggregated data containing multi-scale features; Perform dynamic time series modeling on sensor data, integrate aggregated data containing multi-scale features into the model, and obtain the final digital twin model.

Images