CN-120449674-B - Digital twin method and system of battery power system applied to new energy heavy truck

CN120449674BCN 120449674 BCN120449674 BCN 120449674BCN-120449674-B

Abstract

The invention discloses a digital twin method and a digital twin system of a battery power system applied to a new energy heavy truck, wherein the method comprises the steps of obtaining basic data; the method comprises the steps of obtaining a multi-scale mapping model of a battery power system based on basic data, adding a data driving model, extracting information from real-time monitoring data of the battery power system, supplementing the information to the multi-scale mapping model, constructing a digital twin model of the battery power system, abstracting sensor data of the battery power system into a network topology diagram through a graph neural network-long-short-term memory neural network structure, carrying out time-space relation multi-scale modeling, utilizing the constructed digital twin model of the battery power system to perceive running state information of the battery power system, and utilizing a cross-mode deep semantic matching mechanism to construct a depth semantic matching fusion model of incomplete multi-mode multi-physical fields aiming at multi-mode data contained in the perceived running state information.

Inventors

GUO HONGFEI
HE ZHIHUI
ZHANG RUI
LI JIANQING
CHAO BAO

Assignees

内蒙古工业大学

Dates

Publication Date: 20260512
Application Date: 20250428

Claims (8)

1. The digital twin method of the battery power system applied to the new energy heavy truck is characterized by comprising the following steps of: Acquiring digital twin basic data of a battery power system applied to a new energy heavy truck, wherein the basic data comprises full-element parameter information under the multi-physical field coupling state of an electrochemical field, a temperature field and a stress field of the battery power system of the new energy heavy truck, and multi-scale data from a battery monomer to a module, a whole package, a controller and a driving motor; Based on the basic data, an electrochemical-thermal coupling model is built from an electric core of the battery power system to a module scale, a thermal-mechanical coupling model is built in a single scale in consideration of thermal stress and thermal expansion behaviors, a geometric model is built and optimized from the module and the whole package scale of the battery power system, and meanwhile, a dynamic model of the battery power system is built, so that the omnibearing abstraction of the battery power system is realized, and a multi-scale mapping model of the battery power system is obtained; adding a data driving model, extracting information from the real-time monitoring data of the battery power system, supplementing the extracted information into the multi-scale mapping model, and constructing a digital twin model of the battery power system; abstracting sensor data of the battery power system into a network topological graph through a graph neural network-long-short-term memory neural network structure to perform space-time relationship multi-scale modeling, and sensing running state information of the battery power system by using the constructed digital twin model of the battery power system; utilizing a cross-mode depth semantic matching mechanism, constructing a shared feature subspace among modes aiming at the multi-mode data contained in the perceived running state information, and designing a invariant graph regularization factor to establish a depth semantic matching fusion model of incomplete multi-mode multi-physical fields, so as to reduce semantic deviation; the adding of the data driving model, the information extraction of the real-time monitoring data of the battery power system, the supplementation of the extracted information into the multi-scale mapping model, the construction of the digital twin model of the battery power system, comprises the following steps: Fusing the model system, namely fusing the multi-scale mapping model obtained in the multi-scale modeling step with the data-driven model to form a model system before fusion, wherein the multi-scale mapping model is fused with the data-driven model Data driven model The fusion is carried out by the following modes: where +.indicates the feature stitching operation, The model is driven by the data, Monitoring data in real time; The real-time data processing comprises extracting information from real-time monitoring data by a data driving method, supplementing the extracted information to a multiscale mapping model part in a model system before fusion to form a supplemented multiscale mapping model, wherein the real-time monitoring data comprises sensor data of a physical layer and simulation data of a virtual layer, and the real-time data processing comprises the steps of The following non-negative matrix factorization is applied: , wherein, As a whole of the modal data, As an incomplete mode data, the data is processed, In the form of a base matrix, the base matrix, And To share the feature matrix, extracting the features And Inputting a multiscale mapping model, and updating parameters through the following data-driven model supplementing mechanism: Wherein For learning rate, as per element multiplication, the ReLU activation function ensures non-negativity; And constructing a multi-scale digital twin model of the battery power system based on the supplemented multi-scale mapping model, and realizing data-driven model fusion.
2. The digital twin method of a battery power system applied to a new energy heavy truck according to claim 1, wherein the constructing an electrochemical-thermal coupling model from a cell to a module scale of the battery power system based on the basic data, constructing a thermal-mechanical coupling model in a single scale by considering thermal stress and thermal expansion behavior, constructing a geometric model and optimizing from the module and the whole package scale of the battery power system, and simultaneously constructing a dynamics model of the battery power system, realizing omnibearing abstraction of the battery power system, and obtaining a multi-scale mapping model of the battery power system comprises: Modeling from a battery cell to a module, namely establishing an electrochemical-thermal model on the battery cell scale according to an energy conservation equation and on the basis of a P2D model to obtain an electrochemical-thermal model of the battery cell scale; modeling the module and the whole package scale, namely constructing a three-dimensional CAD model of the power battery by using an electrochemical-thermal model of the battery cell scale and a thermal-force coupling model of the single scale as bases and adopting industrial modeling software to accurately express key structural parameters, constraint and positioning relations among the single and the module to obtain the three-dimensional CAD model; and (3) dynamic modeling, namely taking the optimized geometric model as input, and combining mechanical geometric shapes, material parameters, stress analysis and speed load information of the power battery pack, the motor, the inverter and the drive control unit to build a dynamic model containing the mechanical geometric shapes, the material parameters, the stress analysis and the speed load information.
3. The digital twin method of a battery power system applied to a new energy heavy truck according to claim 1, wherein the abstracting sensor data of the battery power system into a network topology map through a graph neural network-long-short-term memory neural network structure to perform space-time relationship multi-scale modeling, and sensing the running state information of the battery power system by using the constructed digital twin model of the battery power system comprises: Abstracting sensor data into a network topology graph of a time sequence accumulation process of the sensor data, defining connection among nodes as sequence events, and forming the network topology graph of the sensor data; Constructing a time sequence dynamic diagram structure of battery-motor-brake characteristic information fusion based on a network topological diagram of the sensor data to obtain a time sequence dynamic diagram containing battery-motor-brake characteristics; and (3) characteristic learning processing, namely integrating the structural information and the time information of the time sequence dynamic graph by adopting a graph neural network-long-short-term memory neural network structure, designing a layered characteristic aggregation method, and learning a characteristic aggregator in different depth neighborhood to realize information flow from higher depth to nodes so as to obtain a state sensing result.
4. The digital twin method of battery power system applied to new energy re-card according to claim 1, wherein the utilizing a cross-mode deep semantic matching mechanism, for the multi-mode data included in the perceived running state information, constructs a shared feature subspace between modes, and designs a invariant graph regularization factor to build a depth semantic matching fusion model of incomplete multi-mode multi-physical fields, and reduces semantic deviation, comprising: Modeling in a cross-mode manner, namely establishing an incomplete cross-mode depth semantic matching fusion model by adopting a cross-mode depth semantic matching mechanism and through multi-layer nonlinear correlation among multi-level and multi-scale mode data to obtain an initial fusion model; Constructing a shared feature subspace among modes based on the initial fusion model, and learning the sharing of incomplete cross-mode data to form the shared feature subspace; regularization processing, namely designing invariant graph regularization factors in the shared feature subspace, ensuring local similarity characteristics of all mode data, and obtaining a regularized shared subspace; And establishing a new objective function based on the regularized shared subspace to form a complete multi-mode data fusion model so as to reduce semantic deviation.
5. The utility model provides a battery power system digital twin system for new forms of energy heavy truck which characterized in that includes: The acquisition module is used for acquiring digital twin basic data of the battery power system applied to the new energy heavy truck, wherein the basic data comprise full-element parameter information of the battery power system of the new energy heavy truck in a multi-physical field coupling state of an electrochemical field, a temperature field and a stress field and multi-scale data from a battery monomer to a module, a whole package, a controller and a driving motor; the first construction module is used for constructing an electrochemical-thermal coupling model from an electric core of the battery power system to a module scale based on the basic data, constructing a thermal-mechanical coupling model by considering thermal stress and thermal expansion behaviors in a single scale, constructing a geometric model and optimizing the model from the module and the whole package scale of the battery power system, and simultaneously constructing a dynamic model of the battery power system, so that the omnibearing abstraction of the battery power system is realized, and a multi-scale mapping model of the battery power system is obtained; the second construction module is used for adding a data driving model, extracting information from the real-time monitoring data of the battery power system, supplementing the extracted information into the multi-scale mapping model, and constructing a digital twin model of the battery power system; the third construction module is used for abstracting sensor data of the battery power system into a network topological graph through a graph neural network-long-short-term memory neural network structure to carry out space-time relationship multi-scale modeling, and sensing the running state information of the battery power system by utilizing the constructed digital twin model of the battery power system; The fourth construction module is used for constructing a shared feature subspace among modes according to the multi-mode data contained in the perceived running state information by utilizing a cross-mode depth semantic matching mechanism, and designing a invariant graph regularization factor so as to establish a depth semantic matching fusion model of incomplete multi-mode multi-physical fields and reduce semantic deviation; the second construction module is specifically configured to: Fusing the model system, namely fusing the multi-scale mapping model obtained in the multi-scale modeling step with the data-driven model to form a model system before fusion, wherein the multi-scale mapping model is fused with the data-driven model Data driven model The fusion is carried out by the following modes: where +.indicates the feature stitching operation, The model is driven by the data, Monitoring data in real time; The real-time data processing comprises extracting information from real-time monitoring data by a data driving method, supplementing the extracted information to a multiscale mapping model part in a model system before fusion to form a supplemented multiscale mapping model, wherein the real-time monitoring data comprises sensor data of a physical layer and simulation data of a virtual layer, and the real-time data processing comprises the steps of The following non-negative matrix factorization is applied: , wherein, As a whole of the modal data, As an incomplete mode data, the data is processed, In the form of a base matrix, the base matrix, And To share the feature matrix, extracting the features And Inputting a multiscale mapping model, and updating parameters through the following data-driven model supplementing mechanism: Wherein For learning rate, as per element multiplication, the ReLU activation function ensures non-negativity; And constructing a multi-scale digital twin model of the battery power system based on the supplemented multi-scale mapping model, and realizing data-driven model fusion.
6. The digital twin system of a battery power system applied to a new energy heavy truck as defined in claim 5, wherein the first construction module is specifically configured to: Modeling from a battery cell to a module, namely establishing an electrochemical-thermal model on the battery cell scale according to an energy conservation equation and on the basis of a P2D model to obtain an electrochemical-thermal model of the battery cell scale; modeling the module and the whole package scale, namely constructing a three-dimensional CAD model of the power battery by using an electrochemical-thermal model of the battery cell scale and a thermal-force coupling model of the single scale as bases and adopting industrial modeling software to accurately express key structural parameters, constraint and positioning relations among the single and the module to obtain the three-dimensional CAD model; and (3) dynamic modeling, namely taking the optimized geometric model as input, and combining mechanical geometric shapes, material parameters, stress analysis and speed load information of the power battery pack, the motor, the inverter and the drive control unit to build a dynamic model containing the mechanical geometric shapes, the material parameters, the stress analysis and the speed load information.
7. The digital twin system of a battery power system applied to a new energy heavy truck as defined in claim 5, wherein the third construction module is specifically configured to: Abstracting sensor data into a network topology graph of a time sequence accumulation process of the sensor data, defining connection among nodes as sequence events, and forming the network topology graph of the sensor data; Constructing a time sequence dynamic diagram structure of battery-motor-brake characteristic information fusion based on a network topological diagram of the sensor data to obtain a time sequence dynamic diagram containing battery-motor-brake characteristics; and (3) characteristic learning processing, namely integrating the structural information and the time information of the time sequence dynamic graph by adopting a graph neural network-long-short-term memory neural network structure, designing a layered characteristic aggregation method, and learning a characteristic aggregator in different depth neighborhood to realize information flow from higher depth to nodes so as to obtain a state sensing result.
8. The digital twin system of a battery power system applied to a new energy heavy truck as defined in claim 5, wherein the fourth building module is specifically configured to: Modeling in a cross-mode manner, namely establishing an incomplete cross-mode depth semantic matching fusion model by adopting a cross-mode depth semantic matching mechanism and through multi-layer nonlinear correlation among multi-level and multi-scale mode data to obtain an initial fusion model; Constructing a shared feature subspace among modes based on the initial fusion model, and learning the sharing of incomplete cross-mode data to form the shared feature subspace; regularization processing, namely designing invariant graph regularization factors in the shared feature subspace, ensuring local similarity characteristics of all mode data, and obtaining a regularized shared subspace; And establishing a new objective function based on the regularized shared subspace to form a complete multi-mode data fusion model so as to reduce semantic deviation.

Description

Digital twin method and system of battery power system applied to new energy heavy truck Technical Field The invention relates to the technical field of data processing, in particular to a digital twin method and a digital twin system of a battery power system applied to a new energy heavy truck. Background In the field of new energy heavy-duty battery power systems, the prior art generally adopts a rule-based energy management strategy and a simplified battery model for system control. For example, conventional approaches construct a single-body-level electrochemical model by merely monitoring apparent parameters of battery voltage, temperature, etc., and distributing power of the power source based on fixed thresholds. The scheme realizes basic energy management function, but has the obvious technical defects that on one hand, the model can not accurately reflect the actual state of the battery under the complex working condition due to the dynamic influence of the coupling effect of multiple physical fields (such as an electrochemical field, a temperature field and a stress field) on the battery performance, and on the other hand, a model updating mechanism driven by real-time data is lacking, so that the scheme is difficult to adapt to the nonlinear characteristic of a new energy heavy truck in the operation of multiple working conditions. The defects cause the problems of low energy utilization efficiency, lag system response and the like in the prior art when the dynamic environment changes are solved, and the requirements of the new energy heavy truck on high energy efficiency and high reliability cannot be met. Disclosure of Invention The embodiment of the invention provides a digital twin method and a digital twin system of a battery power system applied to a new energy heavy truck, which can be used for solving the problem of the prior art. The embodiment of the invention provides a digital twin method of a battery power system applied to a new energy heavy truck, which comprises the following steps: Acquiring digital twin basic data of a battery power system applied to a new energy heavy truck, wherein the basic data comprises full-element parameter information under the multi-physical field coupling state of an electrochemical field, a temperature field and a stress field of the battery power system of the new energy heavy truck, and multi-scale data from a battery monomer to a module, a whole package, a controller and a driving motor; Based on the basic data, an electrochemical-thermal coupling model is built from an electric core of the battery power system to a module scale, a thermal-mechanical coupling model is built in a single scale in consideration of thermal stress and thermal expansion behaviors, a geometric model is built and optimized from the module and the whole package scale of the battery power system, and meanwhile, a dynamic model of the battery power system is built, so that the omnibearing abstraction of the battery power system is realized, and a multi-scale mapping model of the battery power system is obtained; adding a data driving model, extracting information from the real-time monitoring data of the battery power system, supplementing the extracted information into the multi-scale mapping model, and constructing a digital twin model of the battery power system; abstracting sensor data of the battery power system into a network topological graph through a graph neural network-long-short-term memory neural network structure to perform space-time relationship multi-scale modeling, and sensing running state information of the battery power system by using the constructed digital twin model of the battery power system; And constructing a shared feature subspace among modes according to the multi-mode data contained in the perceived running state information by utilizing a cross-mode depth semantic matching mechanism, and designing a invariant graph regularization factor to establish a depth semantic matching fusion model of incomplete multi-mode multi-physical fields, so as to reduce semantic deviation. As an improvement of the above solution, the building an electrochemical-thermal coupling model from the cell to the module scale of the battery power system based on the basic data, building a thermal-force coupling model in consideration of thermal stress and thermal expansion behavior in a single scale, building a geometric model from the module and the whole package scale of the battery power system and optimizing, and simultaneously building a dynamics model of the battery power system, realizing omnibearing abstraction of the battery power system, and obtaining a multi-scale mapping model of the battery power system, including: Modeling from a battery cell to a module, namely establishing an electrochemical-thermal model on the battery cell scale according to an energy conservation equation and on the basis of a P2D model to obtain an electrochemical-thermal model of th