Vehicle control method and device, electronic equipment and computer readable storage medium

Abstract

The application relates to a vehicle control method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring multi-source sensing data of a vehicle, and determining target health indexes of key components of the vehicle based on the multi-source sensing data; determining target residual life of the key components based on the target health indexes of the key components; determining a global vulnerability index of the vehicle based on the target health indexes of the key components and the target residual life of the key components; establishing a multi-objective optimization function with the global vulnerability index as a constraint condition, and generating a control instruction based on an optimal solution of the multi-objective optimization function, wherein the multi-objective optimization function takes comfort, energy efficiency and component damage value of the vehicle as optimization targets; and controlling the vehicle based on the control instruction. In this way, the vehicle can be controlled in linkage according to the residual life of the key components of the vehicle, and the excessive wear of the key components of the vehicle and the safety risk of the vehicle can be effectively reduced.

Description

Technical Field

[0001] This application relates to the field of vehicle intelligent control technology, specifically to a vehicle control method, device, electronic device, and computer-readable storage medium. Background Technology

[0002] During vehicle use, key components (such as power batteries, motors, and electronic control systems) tend to age or wear out gradually over time, causing their remaining lifespan to change in real time. However, current technologies do not link vehicle control strategies with the remaining lifespan of key components, making it impossible to implement coordinated vehicle control based on these remaining lifespans. This leads to excessive wear and tear on key components and even safety risks. Therefore, how to implement coordinated vehicle control based on the remaining lifespan of key components has become an urgent technical problem to be solved. Summary of the Invention

[0003] This application provides a vehicle control method, device, electronic device, and computer-readable storage medium to solve the problem in related technologies that the vehicle cannot be controlled in conjunction with the remaining lifespan of key vehicle components, resulting in excessive wear and tear on key vehicle components and even safety risks.

[0004] In a first aspect, this application provides a vehicle control method, the method comprising: Acquire multi-source sensor data of the vehicle, and determine the target health index of each key component of the vehicle based on the multi-source sensor data; Based on the target health index of each key component, determine the target remaining life of each key component; Based on the target health index and target remaining life of each key component, the global vulnerability index of the vehicle is determined, wherein the global vulnerability index is used to characterize the health status of the most vulnerable component among the key components. Using the global vulnerability index as a constraint, a multi-objective optimization function is established, and control commands are generated based on the optimal solution of the multi-objective optimization function. The multi-objective optimization function aims to optimize the vehicle's comfort, energy efficiency, and component damage. The vehicle is controlled based on the control commands.

[0005] Optionally, determining the target health index of each key component of the vehicle based on the multi-source sensor data includes: The multi-source sensing data is input into the physical model corresponding to each key component to predict the initial health index of each key component. The physical model corresponding to each key component is constructed based on the physical properties of each key component. The multi-source sensor data is input into the residual learning network corresponding to each key component to predict the residual value of each key component. The residual learning network corresponding to each key component is used to learn the nonlinear time-varying residual between the health index predicted by the physical model corresponding to each key component and the actual health index of each key component. Calculate the difference between the initial health index of each key component and the residual value of each key component, and determine the difference as the target health index of each key component.

[0006] Optionally, determining the target remaining lifespan of each key component based on its target health index includes: The target health index of each key component is input into the physical mechanism prediction model corresponding to each key component to predict the first remaining life of each key component. The physical mechanism prediction model corresponding to each key component is constructed based on the physical degradation mechanism of each key component. The target health index of each key component is input into the data-driven model corresponding to each key component to predict the second remaining life of each key component. The data-driven model corresponding to each key component is pre-trained based on the historical operating data of each key component. Feature vectors of each key component are extracted from the multi-source sensing data, and the feature vectors of each key component are input into a pre-trained limit gradient boosting tree model to predict the working condition matching factor of each key component. By utilizing the operating condition matching factor of each key component, the first weight value of the first remaining life of each key component and the second weight value of the second remaining life of each key component are dynamically adjusted. By using the adjusted first weight value and the adjusted second weight value, the first remaining life and the second remaining life of each key component are fused to determine the target remaining life of each key component.

[0007] Optionally, the target remaining life of each key component is expressed using the following formula: + ; ; ; in, This indicates the target remaining lifespan of the critical components. This indicates the first remaining lifespan of the target critical component. This indicates the second remaining lifespan of the target critical component. This represents the adjusted first weight value. This represents the adjusted second weight value. This represents the operating condition matching factor of the target key component. This represents the prediction standard deviation of the physical mechanism prediction model corresponding to the target key component. This represents the prediction standard deviation of the data-driven model corresponding to the target key component.

[0008] Optionally, determining the global vulnerability index of the vehicle based on the target health index and target remaining life of each key component includes: Based on the target health index of each key component, determine the minimum health index value; Calculate the ratio of the target remaining life of each key component to the rated design life of each key component, and determine the minimum relative remaining life. The global vulnerability index is determined based on the minimum health index and the minimum relative remaining lifespan.

[0009] Optionally, the global vulnerability index is expressed using the following formula: ; in, This represents the global vulnerability index. This represents the preset baseline vulnerability index. This represents the preset vulnerability growth sensitivity coefficient. This represents the minimum value of the health index. This represents the target health index of the i-th critical component. This represents the minimum relative remaining lifetime. This represents the target remaining lifespan of the i-th critical component. This represents the rated design life of the i-th critical component, where i ranges from 1 to n.

[0010] Optionally, the multi-objective optimization function is expressed by the following formula: ; in, This refers to finding an optimal set of control variables within the control cycle. , so that the objective function Minimize the value of . This indicates the preset comfort weight. This indicates the preset energy efficiency weight. This represents the preset damage weight. Represents the comfort function. Represents the energy efficiency function. Represents the damage function, This indicates the upper limit of the allowable output power of the vehicle's motor. Indicates the driving mode of the vehicle, This indicates the upper limit of the allowed charging current for the vehicle. This indicates the state of charge window of the vehicle's power battery. This indicates the preset control cycle duration.

[0011] Optionally, generating control commands based on the optimal solution of the multi-objective optimization function includes: Based on the optimal solution of the multi-objective optimization function, the upper limit of the allowable output power of the vehicle's motor, the driving mode of the vehicle, the upper limit of the allowable charging current of the vehicle, and the state of charge window of the vehicle's power battery are determined. Based on the upper limit of output power, the driving mode, the upper limit of charging current, and the state of charge window, a control strategy for controlling the vehicle is determined, and a control command corresponding to the control strategy is generated. The control strategy includes at least one of a power output strategy, a charging and discharging strategy, and a driving mode.

[0012] Optionally, before determining the target health index of each key component of the vehicle based on the multi-source sensor data, the method further includes: The system receives a remaining lifespan reference value sent by a cloud server. The remaining lifespan reference value is obtained by the cloud server after performing information aggregation based on its own constructed graph neural network. Each node in the graph neural network is used to represent different vehicles, and each edge in the graph neural network is used to characterize the similarity weight between two adjacent nodes. The remaining lifespan reference value is used as a prior knowledge anchor point for online transfer learning of the local model on the vehicle.

[0013] Secondly, this application also provides a vehicle control device, the device comprising: The acquisition and determination module is used to acquire multi-source sensor data of the vehicle and determine the target health index of each key component of the vehicle based on the multi-source sensor data. The first determining module is used to determine the target remaining life of each key component based on the target health index of each key component. The second determining module is used to determine the global vulnerability index of the vehicle based on the target health index and the target remaining life of each key component, wherein the global vulnerability index is used to characterize the health status of the most vulnerable component among the key components. The generation module is used to establish a multi-objective optimization function with the global vulnerability index as a constraint, and to generate control commands based on the optimal solution of the multi-objective optimization function, wherein the multi-objective optimization function is optimized with the vehicle's comfort, energy efficiency and component damage values ​​as optimization objectives. The control module is used to control the vehicle based on the control commands.

[0014] Thirdly, this application also provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the vehicle control method described in the first aspect.

[0015] Fourthly, this application also provides a computer-readable storage medium storing computer-executable instructions for performing the vehicle control method described in the first aspect.

[0016] The beneficial effects of this application are: (1) This application determines the global vulnerability index of the vehicle by using the target health index and target remaining life of each key component, and establishes a multi-objective optimization function with the global vulnerability index as a constraint. Then, a set of optimal solutions that meet the constraint conditions are obtained by using the multi-objective optimization function to generate vehicle control commands, thereby realizing the linkage between the vehicle control strategy and the remaining life of the vehicle's key components. It can perform linkage control of the vehicle based on the remaining life of the vehicle's key components, effectively reducing the excessive wear and tear of the vehicle's key components and the vehicle's safety risks.

[0017] (2) Since the multi-objective optimization function in this application is based on the vehicle's comfort, energy efficiency and component damage values, it can improve the vehicle's comfort and energy efficiency as much as possible while reducing the damage to key vehicle components. Attached Figure Description

[0018] Figure 1 A schematic flowchart of a vehicle control method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a vehicle control device provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] The embodiments of this application will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be understood that the preferred embodiments are only for illustrating this application and are not intended to limit the scope of protection of this application.

[0020] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0021] For ease of description, spatial relative terms may be used in the text to describe the relative position or movement of one element or feature relative to another element or feature, as shown in the figure. These relative terms include, for example, "inside," "outside," "middle," "outer," "below," "below," "above," "front," "back," etc. Such spatial relative terms are intended to include different orientations of the device in use or operation, other than those depicted in the figure. For example, if the device in the figure undergoes a positional flip, orientation change, or change of motion, these directional indications will change accordingly. For instance, an element described as "below other elements or features" or "below other elements or features" will subsequently be oriented "above other elements or features" or "above other elements or features." Therefore, the example term "below" can include both upper and lower orientations. The device may be otherwise oriented (rotated 90 degrees or in other directions), and the spatial relative descriptors used in the text will be interpreted accordingly.

[0022] To address the problem in related technologies that the inability to perform coordinated vehicle control based on the remaining lifespan of key vehicle components, leading to excessive wear and tear on these components and even safety risks, this application provides a vehicle control method, device, electronic device, and computer-readable storage medium capable of performing coordinated vehicle control based on the remaining lifespan of key vehicle components.

[0023] See Figure 1 , Figure 1 This is a schematic flowchart illustrating a vehicle control method provided in an embodiment of this application. Figure 1 As shown, the vehicle control method may include the following steps: Step S102: Acquire multi-source sensor data of the vehicle, and determine the target health index of each key component of the vehicle based on the multi-source sensor data.

[0024] Specifically, the key components of the aforementioned vehicle may include, but are not limited to, components such as the power battery, motor, and electronic control system. The aforementioned multi-source sensing data refers to data collected by various sensors on the vehicle, which may include data related to the key components of the vehicle, vehicle environmental data, and driver behavior data. As an optional implementation, the aforementioned multi-source sensing data may include the following data: (1) Data related to power batteries, such as battery pack voltage, individual cell voltage, current, temperature (cell and module), state of charge (SOC), state of health (SOH), number of cycles, charge and discharge rate, etc.

[0025] (2) Motor-related data, such as motor speed, current, voltage, temperature, vibration signal, magnetic flux intensity, load torque, etc.

[0026] (3) Data related to the electrical control system, such as control signal status, switch status, current, voltage, power consumption, temperature, frequency fluctuations, etc.

[0027] (4) Vehicle environmental data and driver driving behavior data, such as vehicle speed, accelerator pedal opening, brake pedal status, steering angle, road slope, traffic flow, and external temperature and humidity.

[0028] It should be noted that after acquiring the multi-source sensor data of the vehicle, it can be preprocessed, such as time synchronization alignment and normalization, to facilitate the determination of the target health index of each key component of the vehicle based on the preprocessed multi-source sensor data.

[0029] When determining the target health index of each key component based on multi-source sensor data, the target health index of each key component can be directly predicted based on the physical model corresponding to each key component. Alternatively, after predicting the health index of each key component based on the physical model corresponding to each key component, the health index predicted by the physical model corresponding to each key component can be corrected based on the residual value predicted by the residual learning network corresponding to each key component, thereby obtaining the target health index of each key component. The embodiments of this application do not make specific limitations.

[0030] Step S104: Determine the target remaining life of each key component based on the target health index of each key component.

[0031] Specifically, the target remaining lifetime of each key component can be directly predicted based on the corresponding physical mechanism prediction model, or it can be directly predicted based on the corresponding data-driven model. Alternatively, two remaining lifetimes can be predicted separately based on the corresponding physical mechanism prediction model and the corresponding data-driven model, and then these two remaining lifetimes can be fused to obtain the target remaining lifetime of each key component. This application does not impose specific limitations on these methods. The physical mechanism prediction model and the data-driven model will be described in detail in subsequent embodiments and will not be repeated here.

[0032] Step S106: Based on the target health index and target remaining life of each key component, determine the global vulnerability index of the vehicle, wherein the global vulnerability index is used to characterize the health status of the most vulnerable component among the key components.

[0033] Specifically, the aforementioned global vulnerability index is used to characterize the health status of the most vulnerable component among all critical components. The lower the value of the global vulnerability index, the worse the health status and the shorter the lifespan of the most vulnerable component of the vehicle, which means that the vehicle's tolerance for "damage" is lower, and therefore the control constraints should be more stringent; the higher the value of the global vulnerability index, the better the health status and the longer the lifespan of the most vulnerable component of the vehicle, which means that the vehicle's tolerance for "damage" is higher, and therefore the control constraints should be more lenient.

[0034] Step S108: Establish a multi-objective optimization function with the global vulnerability index as a constraint, and generate control commands based on the optimal solution of the multi-objective optimization function. The multi-objective optimization function optimizes vehicle comfort, energy efficiency and component damage values.

[0035] Specifically, the aforementioned multi-objective optimization function is an optimization function that uses the global vulnerability index as a constraint and aims to optimize vehicle comfort, energy efficiency, and component damage. In other words, the purpose of this multi-objective optimization function is to maximize vehicle comfort and energy efficiency while satisfying the constraint of the global vulnerability index.

[0036] After solving the multi-objective optimization function, control commands for the vehicle can be generated based on the optimal solution. These control commands include, but are not limited to, control commands related to power output, charging / discharging, and driving modes.

[0037] Step S110: Control the vehicle based on control commands.

[0038] Specifically, after generating control commands, the control commands can be sent to relevant components for vehicle control.

[0039] In this way, the global vulnerability index of the vehicle can be determined by the target health index and target remaining life of each key component. A multi-objective optimization function constrained by this global vulnerability index can then be established. A set of optimal solutions satisfying the constraints can then be obtained through this multi-objective optimization function, thereby generating vehicle control commands. This achieves linkage between the vehicle control strategy and the remaining life of key vehicle components, enabling coordinated vehicle control based on the remaining life of key components, effectively reducing excessive wear and tear on key components and minimizing vehicle safety risks. Furthermore, since the multi-objective optimization function in this application optimizes vehicle comfort, energy efficiency, and component damage, it can maximize vehicle comfort and energy efficiency while reducing damage to key components.

[0040] In an optional embodiment, step S102, determining the target health index of each key component of the vehicle based on multi-source sensor data, includes: Multi-source sensor data is input into the physical model corresponding to each key component to predict the initial health index of each key component. The physical model corresponding to each key component is constructed based on the physical properties of each key component. Multi-source sensor data is input into the residual learning network corresponding to each key component to predict the residual value of each key component. The residual learning network corresponding to each key component is used to learn the nonlinear time-varying residual between the health index predicted by the physical model corresponding to each key component and the actual health index of each key component. Calculate the difference between the initial health index of each key component and the residual value of each key component, and determine the difference as the target health index of each key component.

[0041] Specifically, the physical models corresponding to the aforementioned key components are constructed based on the physical properties of each key component. Taking a power battery as an example, this application uses the Arrhenius semi-empirical attenuation formula to construct the physical model corresponding to the power battery. The input data for this physical model can be the historical stress sequence of the power battery under operating conditions, specifically including the current I( ), temperature T ( ) and State of Charge (SOC) ), etc., to calculate the current cumulative capacity loss rate of the power battery. The specific calculation formula is as follows: d ; in, I(represents the current cumulative capacity loss rate of the power battery) T represents the current of the power battery. The State of Charge (SOC) indicates the temperature of the power battery. This indicates the state of charge of the power battery. The fitting coefficients are related to the material. Here, R is the activation energy, and R is the gas constant. To be related to the state of charge Related correction functions.

[0042] Next, the physical model can be based on the current cumulative capacity loss rate. The initial health index of the power battery is calculated and output. The initial health index of the power battery can be calculated using the following formula: ; in, This indicates the initial health index of the power battery. This indicates the current cumulative capacity loss rate of the power battery. This indicates the initial rated capacity of the power battery. This indicates the lifespan end-of-life threshold of the power battery. As an optional implementation, this lifespan end-of-life threshold... It can be the initial rated capacity 80%.

[0043] The above formula can be used to map the absolute physical quantities calculated from the electrochemical mechanism of a power battery into a dimensionless initial health index.

[0044] It should be noted that although the above steps use a power battery as an example for formula derivation, the framework of "physical model derivation + residual compensation" has universal applicability. For the motor, a thermal aging model based on coil insulation life can be used; for the electronic control system, a Coffin-Manson fatigue model based on power cycling can be used. Each key component executes the same process in parallel, outputting its corresponding initial health index. , , This is for use in subsequent calculations of the global vulnerability index.

[0045] The aforementioned residual learning network can be implemented based on a Long Short-Term Memory (LSTM) network, which is mainly used to learn the health index predicted by the physical model corresponding to each key component. The true health index of each key component The nonlinear time-varying residual e(t) between them. Therefore, during the training phase of the residual learning network, the labels of each key component can be marked with the true health index. The health index predicted by the working condition data sample and the corresponding physical model It is used as training data for training. Here, the data used for supervised training is... Data tags are derived from periodically calibrated measured capacity (for batteries) or insulation resistance test values ​​(for motors) during offline accelerated aging tests on benches in the vehicle development phase. In the initial stages of actual vehicle deployment, calibrated data from similar models in a cloud-based historical database can be used as cold-start replacement tags. During vehicle operation, approximate values ​​can be generated online through periodic SOC-OCV (Open Circuit Voltage) curve correction under specific steady-state conditions (such as after prolonged inactivity). To achieve continuous self-learning.

[0046] It should be emphasized that, for the power battery, drive motor, and electronic control system, this application can deploy three residual learning networks with identical structures but independent model parameters and input features, respectively. , and Each residual learning network performs the same residual fitting task: Here This can include sequential data such as voltage, current, temperature, and SOC. Each residual learning network is trained and predicts independently, outputting the residual values ​​of each key component. , and The root cause of this residual value lies in the influence of complex dynamic operating conditions (such as transient high current discharge and irregular charging and discharging behavior) that are difficult for physical models to cover, as well as individual differences caused by inconsistencies in cell manufacturing.

[0047] When determining the target health index of each key component of a vehicle based on multi-source sensor data, the multi-source sensor data can be input into the physical model corresponding to each key component to predict the initial health index of each key component. The multi-source sensor data can also be input into the residual learning network corresponding to each key component to predict the residual value of each key component. Then, the difference between the initial health index of each key component and the residual value of each key component can be calculated, and the difference can be determined as the target health index of each key component.

[0048] Taking power batteries as an example, their target health index can be calculated using the following formula: ; in, This indicates the target health index for power batteries. This indicates the initial health index of the power battery. This represents the residual value of the power battery.

[0049] In this way, a basic physical model can be established as the digital thread for key components such as power batteries and motors, and a residual learning network can be built simultaneously. The residual learning network can be used to specifically learn the nonlinear time-varying residual between the predicted value and the actual value of the physical model, thereby capturing the performance deviation caused by complex dynamic working conditions and individual differences. This makes the target health index of each key component in the final output integrate the determinism of physical mechanism and the adaptability of data-driven approach.

[0050] In an optional embodiment, step S104, determining the target remaining lifespan of each key component based on its target health index, includes: The target health index of each key component is input into the physical mechanism prediction model corresponding to each key component to predict the first remaining life of each key component. The physical mechanism prediction model corresponding to each key component is constructed based on the physical degradation mechanism of each key component. The target health index of each key component is input into the data-driven model corresponding to each key component to predict the second remaining life of each key component. The data-driven model corresponding to each key component is pre-trained based on the historical operating data of each key component. Feature vectors of each key component are extracted from multi-source sensor data, and the feature vectors of each key component are input into a pre-trained limit gradient boosting tree model to predict the working condition matching factor of each key component. By utilizing the operating condition matching factor of each key component, the first weight value of the first remaining life of each key component and the second weight value of the second remaining life of each key component are dynamically adjusted. By using the adjusted first weight value and the adjusted second weight value, the first remaining life and the second remaining life of each key component are fused to determine the target remaining life of each key component.

[0051] Specifically, the physical mechanism prediction models for each key component are constructed based on the physical degradation mechanisms of each key component. Taking a power battery as an example, its physical mechanism model can be based on the current moment... Target Health Index and its first derivative The equation for (attenuation rate) obtained by forward extrapolation is as follows: = ; in, Indicates the first remaining lifespan of the power battery. Indicates the current moment of the power battery The target health index, This represents the preset end-of-life threshold of the power battery (e.g., 0.6). Solving this equation yields the first remaining lifespan of the power battery. This refers to the remaining lifespan predicted by the physical mechanism prediction model corresponding to the power battery.

[0052] The data-driven models for each key component are pre-trained based on historical operating data of each component, and can be implemented using a Transformer-based sequence prediction architecture. Specifically, this is illustrated using a power battery as an example, whose input data... The specific components are as follows: Target Health Index: ; Current characteristics: charging and discharging current I Current fluctuation rate; Voltage characteristics: Individual unit / module voltage V; Temperature characteristics: cell temperature T, temperature rise rate; State of charge (SOC) and SOC working window; Driving conditions: vehicle speed v acceleration a Braking frequency; The data-driven model uses an encoder to extract features from a historical sequence of length W, and a decoder generates predicted health index values ​​for the next P steps using an autoregressive approach. After prediction, it can automatically detect the relationship between the predicted curve and the failure threshold line. The intersection point, the difference between the current time and the intersection point time is the second remaining lifespan of the power battery. If the intersection point falls outside the prediction window, linear extrapolation or direct output of a health index prediction value greater than P steps is used, and the confidence interval is widened accordingly.

[0053] The aforementioned operating condition matching factor is used to characterize the degree of intrinsic fit between the current real-time operating conditions of each key component and the applicable boundary of the physical mechanism model. Taking a power battery as an example, when obtaining the operating condition matching factor of the power battery, a parameter of length [length missing] can be constructed. A 60-second sliding window is used to extract the following 7-dimensional real-time feature vectors within the window. : (Average cell temperature). (Standard deviation of temperature); (Standard deviation of current) (Mean rate of change of current) (Peak magnification); (Standard deviation of vehicle speed) (Absolute value of average acceleration); Then, a pre-trained extreme gradient boosting (XGBoost) model is used to analyze this feature vector. Make predictions and output the predicted probability values. After Platt scaling calibration, the result is obtained. The physical meaning of this value is the probability that the prediction result of the physical mechanism prediction model is reliable under the current operating conditions. During actual vehicle operation, the onboard controller updates the sliding window data at a frequency of 1Hz, calculates the above 7-dimensional features in real time, and inputs them into the trained XGBoost model; the output is the value at the current moment. value.

[0054] When determining the target remaining life of each key component based on its target health index, the target health index can be input into the corresponding physical mechanism prediction model to predict the first remaining life of each key component. Then, the target health index can be input into the corresponding data-driven model to predict the second remaining life of each key component. Next, feature vectors of each key component are extracted from multi-source sensor data and input into a pre-trained extreme gradient boosting tree model to predict the operating condition matching factor of each key component. Then, the operating condition matching factor can be used to dynamically adjust the first weight value of the first remaining life and the second weight value of the second remaining life of each key component. Finally, the adjusted first and second weight values ​​are used to fuse the first and second remaining lifespans of each key component to determine the target remaining life of each key component.

[0055] In this way, the fusion weights of the two predictions from the physical mechanism prediction model and the data-driven model can be dynamically adjusted based on the operating condition matching factor. When the vehicle is under stable operating conditions, the fusion weight corresponding to the physical mechanism prediction model is increased; when the vehicle is under operating conditions that the physical mechanism prediction model cannot cover, such as severe driving or extreme environments, the fusion weight corresponding to the data-driven model is smoothly enhanced. This mechanism solves the problem of a single model being prone to inaccuracy under global operating conditions, and outputs a truly reliable dynamic remaining lifetime.

[0056] In an optional embodiment, the target remaining life of each critical component is expressed by the following formula: + ; ; ; in, This indicates the target remaining lifespan of the critical components. Indicates the first remaining lifespan of the target critical component. This indicates the second remaining lifespan of the target's critical components. This represents the adjusted first weight value. This represents the adjusted second weight value. This represents the condition matching factor of the target key components. This represents the prediction standard deviation of the physical mechanism prediction model corresponding to the target key component. This represents the prediction standard deviation of the data-driven model corresponding to the target key component.

[0057] Specifically, according to the above formula, the first weight value Second weight value Subject to working condition matching factor This dynamic adjustment mechanism is key to accurate predictions across all operating conditions. When the operating condition match is high, the vehicle relies on the physical mechanism prediction model to ensure the stability and interpretability of the predictions. When the operating condition deteriorates, it seamlessly switches to relying on the data-driven model to ensure the robustness of the predictions. This dynamic adjustment mechanism enables the vehicle to perceive its own operating conditions and intelligently balance the interpretability and stability of the physical mechanism prediction model with the flexibility of the data-driven model.

[0058] As another alternative implementation, when the vehicle detects that the historical prediction accuracies of two independent prediction models are on the same order of magnitude (i.e., ,in When the threshold is a very small number, the fusion formula can be simplified to an engineering approximation as follows: This simplified form facilitates rapid deployment at computationally limited edge environments. However, it should be emphasized that the core of this application's claim lies in introducing the inherent accuracy of the model. As a fusion anchor, the above simplification is merely an engineering dimensionality reduction application of the original solution under specific boundaries and does not constitute a limitation on the scope of protection of this patent.

[0059] In an optional embodiment, step S106, determining the vehicle's global vulnerability index based on the target health index and target remaining life of each key component, includes: Based on the target health index of each key component, determine the minimum health index value; Calculate the ratio of the target remaining life of each key component to the rated design life of each key component, and determine the minimum relative remaining life. The global vulnerability index is determined based on the minimum health index and the minimum relative remaining lifespan.

[0060] Specifically, when determining the global vulnerability index of a vehicle, the minimum health index can be determined based on the target health index of each key component. At the same time, the ratio of the target remaining life of each key component to its rated design life can be calculated to determine the minimum relative remaining life. Then, the global vulnerability index can be determined based on the minimum health index and the minimum relative remaining life.

[0061] The higher the global vulnerability index, the stricter the operational constraints imposed on the vehicle due to component aging should be; the lower the global vulnerability index, the looser the operational constraints imposed on the vehicle due to component aging should be. Therefore, it can serve as a bridge connecting the remaining lifespan of each key component with vehicle control, realizing the linkage between vehicle control strategy and the remaining lifespan of key vehicle components.

[0062] In an optional embodiment, the global vulnerability index is expressed using the following formula: ; in, This represents the global vulnerability index. This represents the preset baseline vulnerability index. This represents the preset vulnerability growth sensitivity coefficient. This represents the minimum value of the health index. This represents the target health index of the i-th critical component. This represents the minimum relative remaining lifespan. This represents the target remaining lifespan of the i-th critical component. This represents the rated design life of the i-th critical component, where i ranges from 1 to n.

[0063] As an optional implementation method, the benchmark vulnerability index This is dimensionless data, and its value range can be 0.1. 0.5; Vulnerability growth sensitivity coefficient The data is dimensionless, ranging from 1 to 5, and is calibrated by accelerated aging tests on the whole vehicle; the target health index of the i-th key component. It is dimensionless data, and its value range can be [0,1].

[0064] Using the above formula, the global vulnerability index can be accurately and quickly determined based on the minimum health index and the minimum relative remaining lifespan. This makes it easier to use the global vulnerability index to solve the optimal solution of the multi-objective optimization function and generate vehicle control commands.

[0065] In an optional embodiment, the multi-objective optimization function is expressed by the following formula: ; in, This refers to finding an optimal set of control variables within the control cycle. , so that the objective function Minimize the value of . This indicates the preset comfort weight. This indicates the preset energy efficiency weight. This represents the preset damage weight. Represents the comfort function. Represents the energy efficiency function. Represents the damage function, This indicates the maximum allowable output power of the vehicle's motor. Indicates the vehicle's driving mode, Indicates the upper limit of the charging current allowed for the vehicle. A window indicating the state of charge (SBC) of the vehicle's battery. This indicates the preset control cycle duration.

[0066] Specifically, , , , and This represents the solution value for the multi-objective optimization function. Comfort function. For use based on and Predict the rate of change of longitudinal acceleration of the whole vehicle, with a value range of [0,1]; energy efficiency function For use based on and Obtain the combined efficiency MAP of the battery and motor, with a value range of [0,1]; damage function For use based on Mode and The equivalent aging coefficient calculated by the current thermal effect is obtained, with a value range of [0,1]. , , and The specific values ​​are determined by the vehicle's current driving mode (Economy / Comfort / Sport) and are calibrated by the vehicle manufacturer.

[0067] The core hard constraint condition of this multi-objective optimization function is: This constraint ensures that the component damage value during this control cycle does not exceed the vehicle's global vulnerability index. .when When the power output increases due to component aging, the feasible range shrinks, and the vehicle automatically reduces power output and narrows the SOC window to ensure that the vehicle operates within a safe envelope.

[0068] Furthermore, confidence level softening corrections can be introduced: ,in, The preset relaxation coefficient (ranging from 0 to 0.5) is calibrated at the vehicle factory. This setting enables Model Predictive Control (MPC) to intelligently balance "aggressive protection" and "comfortable driving" based on the reliability of the current prediction.

[0069] In this way, by constructing a multi-objective optimization function, the abstract predicted remaining lifespan of components can be concretized into executable vehicle-side control commands such as power output limits and charging power curves. This "prediction-control" closed-loop linkage mechanism can imperceptibly slow down component aging and reduce performance degradation perceptible to the user during vehicle operation. Furthermore, since this multi-objective optimization function optimizes vehicle comfort, energy efficiency, and component damage, it can maximize vehicle comfort and energy efficiency while minimizing damage to key vehicle components.

[0070] In an optional embodiment, step S108, generating control commands based on the optimal solution of the multi-objective optimization function, includes: Based on the optimal solution of the multi-objective optimization function, the upper limit of the allowable output power of the vehicle's motor, the vehicle's driving mode, the upper limit of the allowable charging current of the vehicle, and the state of charge window of the vehicle's power battery are determined. Based on the upper limit of output power, driving mode, upper limit of charging current and state of charge window, a control strategy for controlling the vehicle is determined and a control command corresponding to the control strategy is generated. The control strategy includes at least one of power output strategy, charging and discharging strategy and driving mode.

[0071] Specifically, when generating control commands based on the optimal solution of a multi-objective optimization function, the upper limit of the allowable output power of the vehicle's motor, the vehicle's driving mode, the upper limit of the allowable charging current, and the state of charge window of the vehicle's power battery can be determined based on the optimal solution of the multi-objective optimization function. Then, based on the upper limit of output power, driving mode, upper limit of charging current, and state of charge window, the control strategy for controlling the vehicle (such as power output strategy, charging and discharging strategy, and driving mode) can be determined, and the corresponding control commands can be generated.

[0072] This allows for linkage between vehicle control strategies and the remaining lifespan of key vehicle components, enabling coordinated vehicle control based on the remaining lifespan of these components, effectively reducing excessive wear and tear on key components and minimizing vehicle safety risks.

[0073] In an optional embodiment, before step S102 above, which determines the target health index of each key component of the vehicle based on multi-source sensor data, the method further includes: The system receives a remaining life reference value sent by a cloud server. This remaining life reference value is obtained by the cloud server through information aggregation based on its own graph neural network. Each node in the graph neural network represents a different vehicle, and each edge in the graph neural network represents the similarity weight between two adjacent nodes. Using the remaining lifespan reference value as a priori knowledge anchor, online transfer learning is performed on the local model on the vehicle.

[0074] Specifically, for individual vehicles with limited data, especially new vehicles or vehicles operating under simple conditions, local models face the challenge of a cold start, where data sparsity leads to slow convergence and low accuracy in prediction models. Furthermore, in fleet management scenarios, the operating environments and task assignments of different vehicles influence each other, making global optimization difficult through single-vehicle optimization alone. To address this, this application designs a global collaborative and personalized evolution mechanism on a cloud server, which is a core component of this application's vehicle-cloud integrated architecture.

[0075] Traditional cloud-based solutions often employ simple, centralized training with large datasets, treating all vehicle data as independent, identically distributed samples. The drawback of this approach is that it ignores the similarities between vehicles caused by usage environments, driving habits, and other factors—relationships that are precisely valuable prior knowledge for solving the cold start problem. This application introduces Graph Neural Networks (GNNs) for the first time to capture these relationships, and their specific construction method is as follows: Node definition: Each vehicle is modeled as a node in the graph, with its initial feature vector... It includes the vehicle's model, cumulative mileage, local health index sequence, driving behavior statistics, etc.

[0076] Edge definition: Edge connections between vehicles are not arbitrarily established, but based on multi-dimensional similarity calculations. Dimensions of similarity calculation include: geographical environment similarity (e.g., operating in cities within the same temperature zone), vehicle type and component configuration similarity, and driving condition distribution similarity (e.g., buses with high-frequency acceleration and deceleration vs. logistics vehicles with high-speed steady-state operation). The higher the similarity, the greater the edge weight.

[0077] Then, through graph convolution operations, the degradation pattern information of similar neighbor nodes can be aggregated for a given vehicle node. This means that a newly launched car can inherit verified degradation patterns from its neighbor nodes through a graph network as strong prior knowledge, thereby significantly accelerating the convergence of its local prediction model.

[0078] The final calculated reference value for remaining lifetime is obtained using the following formula: ; in, This indicates that the cloud server is for vehicles. Calculated reference value for remaining useful life. This represents a graph neural network-long short-term memory network joint model. Indicates vehicle Its own static and dynamic feature vectors, Indicates vehicle The feature vector of the external environment. Indicates vehicle The set of neighbor nodes in a graph network Indicates vehicle The edge weight between the user and neighbor j is a dimensionless value, ranging from [0,1]. =1. Indicates vehicle The hidden degradation pattern is represented by aggregating neighboring vehicles j.

[0079] The remaining lifetime reference value is not directly used as the final decision result locally, but rather as a prior knowledge anchor point to guide the online transfer learning process of the local model on the vehicle (i.e., the residual network and Transformer model mentioned above).

[0080] Building upon the generation of individual vehicle control strategies, the cloud server also undertakes the task of macro-level scheduling for the entire fleet. This application models this task as a constrained discrete global optimization problem. The optimization objective is to minimize the total expected value of the weighted fault risk of the entire fleet, and its objective function is as follows: ; Where A is the task matrix (such as driving route, load class) assigned to each vehicle by the cloud server. For vehicles Task priority weight (value range 0-1, set by the platform based on task urgency), and =1); Indicates vehicle At the current remaining life reference value The estimated failure probability is calculated using the following mapping function: , here The characteristic lifetime scaling factor calibrated for the cloud server (the value can be...) ), so that when Much larger hour, Approaching 0; when much smaller hour, Approaching 1.

[0081] Its global constraints are as follows: 1. Total energy consumption constraint: Ensure that the total power consumption of all vehicles performing tasks does not exceed the total power available from the charging station; The estimated power consumption for the vehicle to perform the mission; This represents the upper limit of total global energy consumption.

[0082] 2. Task completion rate constraint This ensures that the fleet's overall transportation volume meets the operator's minimum required percentage. and These represent the task completion amount and its target value, respectively.

[0083] The cloud server runs once a day when the departure plan is set, outputting recommended task types and departure priority rankings for each vehicle, rather than directly issuing specific accelerator / brake commands. After executing the strategy issued by the cloud server, the vehicles can send back actual effect data, such as energy consumption change rate, temperature rise change rate, and actual component degradation rate, to the cloud server.

[0084] At this point, the cloud server doesn't simply retrain the model with new data. Instead, it constructs a reinforcement learning-style policy evaluation-improvement loop: using feedback data as a reward signal to evaluate the effectiveness of the previous policy, and then fine-tuning the parameters of the graph network model and the policy generation network online. In this way, each policy execution contributes high-quality data labeled with "causal effects" to the model, driving the entire "prediction-policy-feedback-optimization" data flywheel to spin faster, ultimately achieving personalized lifespan and vehicle management that is "personalized for each individual and improves with use."

[0085] Therefore, the vehicle control method provided in this application has the following beneficial effects: 1. Accurate prediction based on uncertainty self-awareness: This application achieves Bayesian dynamic fusion of physical mechanism prediction model and data-driven model. Its core lies in dynamically adjusting the fusion weight of the prediction results of the two models according to the working condition matching factor, which solves the problem that a single model is prone to inaccuracy under complex working conditions.

[0086] 2. Deep Strategy Linkage Between Component Health Status and Vehicle Control: This application goes beyond mere early warning prompts. It establishes a multi-objective optimization function that translates "health index" into "vehicle control parameters," concretizing abstract component remaining lifespan predictions into executable vehicle-side control commands such as power output limits and charging power curves. This closed-loop "prediction-control" linkage mechanism can imperceptibly slow down component aging during vehicle operation, reducing user-perceptible performance degradation.

[0087] 3. Continuous Evolution of Personalized Strategies from a Global Perspective: This application innovatively introduces a cloud-based graph neural network model for "fleet-vehicle" collaboration. By using the massive operational and environmental data of vehicles in the fleet as global prior knowledge, it injects the data into the personalized prediction model of each vehicle. This effectively solves the "cold start" problem caused by the scarcity of data for new or individual vehicles, which leads to slow prediction startup and low accuracy. Furthermore, based on execution feedback, the strategy is continuously iterated and optimized to achieve "personalized" vehicle lifespan management.

[0088] See Figure 2 , Figure 2 This is a schematic diagram of the structure of a vehicle control device provided in an embodiment of this application. Figure 2 As shown, the vehicle control device 200 includes: The acquisition and determination module 202 is used to acquire multi-source sensor data of the vehicle and determine the target health index of each key component of the vehicle based on the multi-source sensor data. The first determining module 204 is used to determine the target remaining life of each key component based on the target health index of each key component. The second determining module 206 is used to determine the global vulnerability index of the vehicle based on the target health index and the target remaining life of each key component. The global vulnerability index is used to characterize the health status of the most vulnerable component among the key components. The generation module 208 is used to establish a multi-objective optimization function with the global vulnerability index as a constraint, and generate control commands based on the optimal solution of the multi-objective optimization function. The multi-objective optimization function is optimized with the vehicle's comfort, energy efficiency and component damage values ​​as optimization objectives. The control module 210 is used to control the vehicle based on control commands.

[0089] Furthermore, the acquisition and determination module 202 includes: The first prediction submodule is used to input multi-source sensor data into the physical model corresponding to each key component and predict the initial health index of each key component. The physical model corresponding to each key component is constructed based on the physical properties of each key component. The second prediction submodule is used to input multi-source sensor data into the residual learning network corresponding to each key component to predict the residual value of each key component. The residual learning network corresponding to each key component is used to learn the nonlinear time-varying residual between the health index predicted by the physical model corresponding to each key component and the actual health index of each key component. The first determination submodule is used to calculate the difference between the initial health index of each key component and the residual value of each key component, and to determine the difference as the target health index of each key component.

[0090] Furthermore, the first determining module 204 includes: The third prediction submodule is used to input the target health index of each key component into the physical mechanism prediction model corresponding to each key component, and predict the first remaining life of each key component. The physical mechanism prediction model corresponding to each key component is constructed based on the physical degradation mechanism of each key component. The fourth prediction submodule is used to input the target health index of each key component into the data-driven model corresponding to each key component to predict the second remaining life of each key component. The data-driven model corresponding to each key component is pre-trained based on the historical operating data of each key component. The fifth prediction submodule is used to extract the feature vectors of each key component from multi-source sensor data, and input the feature vectors of each key component into the pre-trained limit gradient boosting tree model to predict the working condition matching factor of each key component. The adjustment submodule is used to dynamically adjust the first weight value of the first remaining life of each key component and the second weight value of the second remaining life of each key component by utilizing the working condition matching factor of each key component. The fusion submodule is used to fuse the first remaining lifetime and the second remaining lifetime of each key component using the adjusted first weight value and the adjusted second weight value, so as to determine the target remaining lifetime of each key component.

[0091] Furthermore, the target remaining life of each key component is expressed using the following formula: + ; ; ; in, This indicates the target remaining lifespan of the critical components. Indicates the first remaining lifespan of the target critical component. This indicates the second remaining lifespan of the target's critical components. This represents the adjusted first weight value. This represents the adjusted second weight value. This represents the condition matching factor of the target key components. This represents the prediction standard deviation of the physical mechanism prediction model corresponding to the target key component. This represents the prediction standard deviation of the data-driven model corresponding to the target key component.

[0092] Furthermore, the second determining module 206 includes: The second determining submodule is used to determine the minimum health index based on the target health index of each key component; The third determination submodule is used to calculate the target remaining life of each key component and the ratio of the target remaining life to the rated design life of each key component, and to determine the minimum relative remaining life. The fourth determination submodule is used to determine the global vulnerability index based on the minimum health index and the minimum relative remaining lifespan.

[0093] Furthermore, the global vulnerability index is expressed using the following formula: ; in, This represents a global vulnerability index. This represents the preset baseline vulnerability index. This represents the preset vulnerability growth sensitivity coefficient. This represents the minimum value of the health index. This represents the target health index of the i-th critical component. This represents the minimum relative remaining lifespan. This represents the target remaining lifespan of the i-th critical component. This represents the rated design life of the i-th critical component, where i ranges from 1 to n.

[0094] Furthermore, the multi-objective optimization function is expressed by the following formula: ; in, This refers to finding an optimal set of control variables within the control cycle. , so that the objective function Minimize the value of . This indicates the preset comfort weight. This indicates the preset energy efficiency weight. This represents the preset damage weight. Represents the comfort function. Represents the energy efficiency function. Represents the damage function, This indicates the maximum allowable output power of the vehicle's motor. Indicates the vehicle's driving mode, Indicates the upper limit of the charging current allowed for the vehicle. A window indicating the state of charge (SBC) of the vehicle's battery. This indicates the preset control cycle duration.

[0095] Furthermore, the generation module 208 includes: The fifth determination submodule is used to determine the upper limit of the allowable output power of the vehicle's motor, the vehicle's driving mode, the upper limit of the allowable charging current of the vehicle, and the state of charge window of the vehicle's power battery based on the optimal solution of the multi-objective optimization function. The generation submodule is used to determine the control strategy for controlling the vehicle based on the upper limit of output power, driving mode, upper limit of charging current and state of charge window, and generate control commands corresponding to the control strategy. The control strategy includes at least one of power output strategy, charging and discharging strategy and driving mode.

[0096] Furthermore, the vehicle control device 200 also includes: The receiving module is used to receive the remaining life reference value sent by the cloud server. The remaining life reference value is obtained by the cloud server after performing information aggregation based on the graph neural network it has built. Each node in the graph neural network is used to represent different vehicles, and each edge in the graph neural network is used to represent the similarity weight between two adjacent nodes. The learning module is used to perform online transfer learning on the local model on the vehicle, using the remaining life reference value as a prior knowledge anchor.

[0097] It should be noted that the vehicle control device 200 can implement the vehicle control method provided in any of the aforementioned method embodiments and achieve the same technical effect, which will not be elaborated here.

[0098] like Figure 3 As shown, this application embodiment also provides an electronic device, including a processor 311, a communication interface 312, a memory 313 and a communication bus 314, wherein the processor 311, the communication interface 312 and the memory 313 communicate with each other through the communication bus 314. Memory 313 is used to store computer programs; In one embodiment of this application, the processor 311, when executing the program stored in the memory 313, implements the vehicle control method provided in any of the foregoing method embodiments.

[0099] In addition, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the vehicle control method provided in any of the foregoing method embodiments.

[0100] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0101] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0102] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0103] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A vehicle control method, characterized in that, The method includes: Acquire multi-source sensor data of the vehicle, and determine the target health index of each key component of the vehicle based on the multi-source sensor data; Based on the target health index of each key component, determine the target remaining life of each key component; Based on the target health index and target remaining life of each key component, the global vulnerability index of the vehicle is determined, wherein the global vulnerability index is used to characterize the health status of the most vulnerable component among the key components. Using the global vulnerability index as a constraint, a multi-objective optimization function is established, and control commands are generated based on the optimal solution of the multi-objective optimization function. The multi-objective optimization function aims to optimize the vehicle's comfort, energy efficiency, and component damage. The vehicle is controlled based on the control commands.

2. The method according to claim 1, characterized in that, The determination of the target health index of each key component of the vehicle based on the multi-source sensor data includes: The multi-source sensing data is input into the physical model corresponding to each key component to predict the initial health index of each key component. The physical model corresponding to each key component is constructed based on the physical properties of each key component. The multi-source sensor data is input into the residual learning network corresponding to each key component to predict the residual value of each key component. The residual learning network corresponding to each key component is used to learn the nonlinear time-varying residual between the health index predicted by the physical model corresponding to each key component and the actual health index of each key component. Calculate the difference between the initial health index of each key component and the residual value of each key component, and determine the difference as the target health index of each key component.

3. The method according to claim 1, characterized in that, The determination of the target remaining lifespan of each key component based on its target health index includes: The target health index of each key component is input into the physical mechanism prediction model corresponding to each key component to predict the first remaining life of each key component. The physical mechanism prediction model corresponding to each key component is constructed based on the physical degradation mechanism of each key component. The target health index of each key component is input into the data-driven model corresponding to each key component to predict the second remaining life of each key component. The data-driven model corresponding to each key component is pre-trained based on the historical operating data of each key component. Feature vectors of each key component are extracted from the multi-source sensing data, and the feature vectors of each key component are input into a pre-trained limit gradient boosting tree model to predict the working condition matching factor of each key component. By utilizing the operating condition matching factor of each key component, the first weight value of the first remaining life of each key component and the second weight value of the second remaining life of each key component are dynamically adjusted. By using the adjusted first weight value and the adjusted second weight value, the first remaining life and the second remaining life of each key component are fused to determine the target remaining life of each key component.

4. The method according to claim 3, characterized in that, The target remaining life of each key component is expressed by the following formula: + ； ； ； in, This indicates the target remaining lifespan of the critical components. This indicates the first remaining lifespan of the target critical component. This indicates the second remaining lifespan of the target critical component. This represents the adjusted first weight value. This represents the adjusted second weight value. This represents the operating condition matching factor of the target key component. This represents the prediction standard deviation of the physical mechanism prediction model corresponding to the target key component. This represents the prediction standard deviation of the data-driven model corresponding to the target key component.

5. The method according to claim 1, characterized in that, The determination of the vehicle's global vulnerability index based on the target health index and target remaining lifespan of each key component includes: Based on the target health index of each key component, determine the minimum health index value; Calculate the ratio of the target remaining life of each key component to the rated design life of each key component, and determine the minimum relative remaining life. The global vulnerability index is determined based on the minimum health index and the minimum relative remaining lifespan.

6. The method according to claim 5, characterized in that, The global vulnerability index is expressed by the following formula: ； in, This represents the global vulnerability index. This represents the preset baseline vulnerability index. This represents the preset vulnerability growth sensitivity coefficient. This represents the minimum value of the health index. This represents the target health index of the i-th critical component. This represents the minimum relative remaining lifetime. This represents the target remaining lifespan of the i-th critical component. This represents the rated design life of the i-th critical component, where i ranges from 1 to n.

7. The method according to claim 1, characterized in that, The multi-objective optimization function is expressed by the following formula: ； in, This refers to finding an optimal set of control variables within the control cycle. , so that the objective function Minimize the value of . This indicates the preset comfort weight. This indicates the preset energy efficiency weight. This represents the preset damage weight. Represents the comfort function. Represents the energy efficiency function. Represents the damage function, This indicates the upper limit of the allowable output power of the vehicle's motor. Indicates the driving mode of the vehicle, This indicates the upper limit of the allowed charging current for the vehicle. This indicates the state of charge window of the vehicle's power battery. This indicates the preset control cycle duration.

8. The method according to claim 7, characterized in that, The generation of control commands based on the optimal solution of the multi-objective optimization function includes: Based on the optimal solution of the multi-objective optimization function, the upper limit of the allowable output power of the vehicle's motor, the driving mode of the vehicle, the upper limit of the allowable charging current of the vehicle, and the state of charge window of the vehicle's power battery are determined. Based on the upper limit of output power, the driving mode, the upper limit of charging current, and the state of charge window, a control strategy for controlling the vehicle is determined, and a control command corresponding to the control strategy is generated. The control strategy includes at least one of a power output strategy, a charging and discharging strategy, and a driving mode.

9. The method according to claim 1, characterized in that, Before determining the target health index of each key component of the vehicle based on the multi-source sensor data, the method further includes: The system receives a remaining lifespan reference value sent by a cloud server. The remaining lifespan reference value is obtained by the cloud server after performing information aggregation based on its own constructed graph neural network. Each node in the graph neural network is used to represent different vehicles, and each edge in the graph neural network is used to characterize the similarity weight between two adjacent nodes. The remaining lifespan reference value is used as a prior knowledge anchor point for online transfer learning of the local model on the vehicle.

10. A vehicle control device, characterized in that, The device includes: The acquisition and determination module is used to acquire multi-source sensor data of the vehicle and determine the target health index of each key component of the vehicle based on the multi-source sensor data. The first determining module is used to determine the target remaining life of each key component based on the target health index of each key component. The second determining module is used to determine the global vulnerability index of the vehicle based on the target health index and the target remaining life of each key component, wherein the global vulnerability index is used to characterize the health status of the most vulnerable component among the key components. The generation module is used to establish a multi-objective optimization function with the global vulnerability index as a constraint, and to generate control commands based on the optimal solution of the multi-objective optimization function, wherein the multi-objective optimization function is optimized with the vehicle's comfort, energy efficiency and component damage values ​​as optimization objectives. The control module is used to control the vehicle based on the control commands.

11. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in a memory, implements the vehicle control method according to any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The device stores computer-executable instructions for performing the vehicle control method according to any one of claims 1-9.

Images