Commercial vehicle battery life extension control method, system, device, and medium

Abstract

The application provides a commercial vehicle battery life extension control method, system, device and medium. The application collects multi-dimensional operation data of the commercial vehicle at the vehicle end and completes working condition scene clustering to obtain a scene label and a multi-dimensional working condition feature vector and upload to the cloud end; the cloud end extracts battery micro-aging characteristics in combination with working condition information relying on a pre-constructed battery digital twin model to deduce a battery health state decay rate; the cloud end takes the working condition characteristics and the battery decay rate as inputs, iteratively optimizes through a trained deep reinforcement learning model to generate a globally optimal control strategy that takes into account operation efficiency and battery life extension; the cloud end cuts the global strategy according to the real-time scene label reported by the vehicle end to generate a lightweight vehicle end control strategy and issues it to the vehicle end; the vehicle end dynamically adjusts battery charging and discharging and thermal management parameters relying on the strategy to realize battery loss suppression and health state optimization under complex working conditions and improve the control effect of the commercial vehicle battery life extension.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle battery management technology, and in particular to a method, system, device and medium for extending the life of commercial vehicle batteries. Background Technology

[0002] Currently, commercial vehicle battery management technology primarily employs a solution combining vehicle-side fixed control with cloud-based basic monitoring. Vehicle-side battery management systems generally utilize rule-driven and static MAP lookup table-based fixed control strategies. The vehicle manages the battery by pre-setting fixed charging cutoff voltage, thermal management start-stop temperature thresholds, and two-dimensional power limiting parameters based on state of charge and temperature. While this type of control logic is fixed and highly versatile, it cannot adapt to the diverse, complex, and extreme operating conditions of commercial vehicles, such as long-haul logistics, urban delivery, and heavy-duty mining operations. Under harsh conditions, this strategy can easily exacerbate internal battery side reactions and accelerate aging; under mild conditions, it can easily lead to excessive limitation of battery performance and reduced vehicle operating efficiency.

[0003] Cloud-based battery management platforms only provide basic data monitoring, relying on ampere-hour integrals and simplified equivalent circuit models to estimate the macroscopic battery health status. They also rely solely on fixed thresholds for over-temperature and over-voltage fault alarms. However, cloud-based data processing is limited in scope, failing to consider microscopic aging mechanisms such as electrode mechanical fatigue, localized lithium plating, and cell temperature gradient stress. This results in significant lag and large errors in battery health status estimation and lifespan prediction, failing to provide accurate and reliable data support for battery life extension management. Furthermore, existing battery management systems can only make fixed trade-offs between battery life, vehicle power performance, and charging efficiency, lacking multi-objective global dynamic optimization capabilities. They cannot achieve Pareto optimal decisions based on differentiated user operating needs and real-time operating conditions, making it difficult to maximize the total cost of ownership (TCO) of the vehicle. Therefore, an effective method for extending the lifespan of commercial vehicle batteries is urgently needed to address these issues. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed to overcome or at least partially solve the above problems.

[0005] To achieve the above and other related objectives, the present invention provides a method for controlling the life extension of commercial vehicle batteries, the method comprising: Multi-dimensional operating data of commercial vehicles are collected from the vehicle end, and the multi-dimensional operating data is clustered according to working conditions to obtain scene labels and multi-dimensional working condition feature vectors, which are then uploaded to the cloud. The cloud-based system extracts microscopic aging features under corresponding operating conditions based on the scene labels and the multidimensional operating condition feature vectors, and predicts the battery health status degradation rate based on the microscopic aging features. The cloud-based system generates a globally optimal control strategy adapted to the current operating conditions by iteratively optimizing the solution through a trained deep reinforcement learning model based on the scene label, the multi-dimensional operating condition feature vector, and the battery health state decay rate. The globally optimal control strategy includes battery charge and discharge regulation parameters and battery thermal management parameters. The cloud-based system tailors the globally optimal control strategy based on scene tags generated and reported by the vehicle-side in real-time clustering, generates a vehicle-side control strategy, and sends it to the vehicle-side system. The vehicle-side system then dynamically adjusts the battery charging and discharging control parameters and battery thermal management parameters according to the vehicle-side control strategy.

[0006] Optionally, the step of collecting multi-dimensional operating data of commercial vehicles from the vehicle end, and performing operating condition scenario clustering on the multi-dimensional operating data to obtain scenario labels and multi-dimensional operating condition feature vectors includes: Vehicle kinematics data, vehicle load data, battery status data, and environmental data are collected by the vehicle end according to a preset sampling period to form multi-dimensional operating data. The vehicle kinematics data includes vehicle speed and acceleration; the load data includes vehicle weight; the battery status data includes cell voltage, surface temperature, and state of charge; and the environmental data includes ambient temperature and humidity. The multidimensional operational data is preprocessed, and driver behavior features are extracted based on time-series statistics; the driver behavior features include at least one of start-stop frequency, state of charge fluctuation amplitude, and microcirculation depth frequency distribution. An improved DBSCAN time-series clustering algorithm is used to divide the multidimensional operational data into working conditions and scenarios, generating scenario labels; the scenario labels include at least one of trunk logistics, urban delivery, and heavy-duty mining. The scene labels are one-hot encoded, and the one-hot encoded scene labels are concatenated with the driver behavior features to construct a multi-dimensional working condition feature vector.

[0007] Optionally, when using the improved DBSCAN temporal clustering algorithm to divide the working conditions, a preset neighborhood radius and a preset minimum number of samples are used as density determination parameters, and a time window is introduced as a temporal constraint condition; when performing neighborhood search, only data points with a time difference less than a preset threshold are included in the neighborhood calculation range, so as to identify densely connected data points in similar time intervals and realize working condition clustering under temporal coupling.

[0008] Optionally, the step of extracting microscopic aging features under corresponding operating conditions by the cloud based on the scene label and the multidimensional operating condition feature vector through a cloud-constructed battery digital twin model, and predicting the degradation rate of battery health status based on the microscopic aging features, includes: The cloud-based digital twin model of the battery, which integrates the PINN electrochemical mechanism, is used to perform working condition alignment processing on the multi-dimensional operating data by combining the scene labels and the multi-dimensional working condition feature vectors, and to extract the micro-aging features under the corresponding working conditions. The micro-aging features include at least one of the following: micro-cycle depth frequency distribution, micro-cycle mechanical stress index, high current pulse polarization voltage, lithium plating overpotential at the end of charging, and cell temperature gradient variance. Based on the aforementioned microscopic aging characteristics, a degradation trajectory prediction algorithm is used to calculate the battery health state degradation rate and remaining battery life under the corresponding operating conditions.

[0009] Optionally, before the step of collecting multi-dimensional operating data of commercial vehicles from the vehicle end, performing operating condition scenario clustering on the multi-dimensional operating data, and obtaining scenario labels and multi-dimensional operating condition feature vectors, the method further includes: A Markov decision process framework for battery life extension control is constructed in the cloud, a reinforcement learning model is built, and a state space, action space, and reward function are defined. The state space includes battery health state, state of charge, battery gridded temperature field distribution, operating condition feature vector, and scene labels encoded by unique thermal encoding. The action space includes charging cutoff voltage adjustment, discharge power limit coefficient, and liquid cooling pump operating speed. The reward function includes a lifespan degradation penalty term, an operational efficiency benefit term, and a safety penalty term, used to balance the degree of battery aging and vehicle operating efficiency, and to solve for the Pareto optimal control solution. Based on the aforementioned Markov decision process framework, using historical operating data uploaded from the vehicle and simulation data generated by the battery digital twin model, all state samples and action samples matching the aforementioned state space and action space definitions are extracted, and the reinforcement learning model is trained offline. After the model converges, a deep reinforcement learning model capable of online reasoning is obtained.

[0010] Optionally, the step of tailoring the globally optimal control strategy by the cloud based on the scene labels generated and reported by the vehicle-side in real-time clustering, generating a vehicle-side control strategy, and distributing it to the vehicle-side includes: The scene labels generated and reported by the cloud based on real-time clustering at the vehicle end are used to cut out local policy parameter packages or lightweight inference models that are adapted to the scene labels from the global optimal control strategy, and these are used as the vehicle end control strategy. The vehicle-side control strategy is sent to the vehicle using a differential OTA transmission method.

[0011] Optionally, the step of dynamically adjusting the battery charging / discharging control parameters and battery thermal management parameters by the vehicle-side according to the vehicle-side control strategy includes: The vehicle-side control strategy is loaded by the vehicle-side battery management system, and the charging cut-off voltage adjustment amount, discharge power limit coefficient and liquid cooling pump operating speed are dynamically adjusted through real-time interpolation calculation or lightweight model inference. The vehicle continuously monitors the communication link status with the cloud. When communication with the cloud is interrupted, it automatically switches to a pre-installed local safety degradation strategy to continuously implement battery safety management.

[0012] Secondly, the present invention also provides a commercial vehicle battery life extension control system, the system comprising: On the vehicle side, multi-dimensional operating data of commercial vehicles is collected, and the multi-dimensional operating data is clustered according to working conditions to obtain scene labels and multi-dimensional working condition feature vectors, which are then uploaded to the cloud. In the cloud, the system extracts microscopic aging features under corresponding operating conditions based on the scene labels and the multidimensional operating condition feature vectors through a cloud-constructed battery digital twin model, and predicts the battery health status degradation rate based on the microscopic aging features. The cloud platform is also used to generate a globally optimal control strategy adapted to the current operating conditions by iteratively optimizing and solving the problem through a trained deep reinforcement learning model based on the scene label, the multi-dimensional operating condition feature vector, and the battery health state decay rate; wherein, the globally optimal control strategy includes battery charge and discharge regulation parameters and battery thermal management parameters. The cloud is also used to tailor the global optimal control strategy based on the scene labels generated and reported by the vehicle terminal in real time clustering, generate the vehicle terminal control strategy, and send it to the vehicle terminal. The vehicle-side device is also used to dynamically adjust the battery charging and discharging parameters and battery thermal management parameters according to the vehicle-side control strategy.

[0013] Thirdly, the present invention provides an electronic device comprising: a memory and a processor; the memory for storing a computer program; and the processor for executing the computer program stored in the memory to cause the electronic device to perform the commercial vehicle battery life extension control method as described above.

[0014] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the commercial vehicle battery life extension control method as described above.

[0015] The above-described one or more technical solutions provided by this invention can have the following advantages or at least achieve the following technical effects: This invention collects multi-dimensional operational data from commercial vehicles at the vehicle end, clusters this data into operating scenarios to obtain scenario labels and multi-dimensional operating scenario feature vectors, and uploads them to the cloud. The cloud then uses a pre-built digital twin model of the battery, combined with the aforementioned operating scenario information, to extract microscopic aging features of the battery and deduce the battery's health status degradation rate. The cloud uses the operating scenario features and battery degradation parameters as input, and iteratively optimizes a trained deep reinforcement learning model to generate a globally optimal control strategy that balances vehicle operating efficiency and battery life extension performance. Based on the scenario labels reported by the vehicle in real time, the cloud tailors the global strategy, generates a lightweight vehicle-side control strategy, and distributes it, allowing the vehicle to dynamically adjust battery charging, discharging, and thermal management parameters. This solution relies on a vehicle-cloud collaborative architecture, combining digital twins and reinforcement learning to achieve intelligent management and control of the entire battery lifecycle. It is adaptable to complex commercial vehicle operating conditions, effectively suppresses battery aging and loss, and extends battery life while ensuring operational efficiency. Attached Figure Description

[0016] Figure 1 The diagram shown is a flowchart of a commercial vehicle battery life extension control method according to an embodiment of the present invention.

[0017] Figure 2 The diagram shows a functional module schematic of a commercial vehicle battery life extension control system according to an embodiment of the present invention.

[0018] Figure 3 The diagram shown is a schematic representation of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0019] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention 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 the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0020] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention 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] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0022] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0023] Unless otherwise stated, the term "multiple" means two or more.

[0024] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0025] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0026] The technical solutions of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] Please see Figure 1 An embodiment of the present invention provides a method for controlling the life extension of commercial vehicle batteries, the method including the following steps S10~S40: Step S10: Collect multi-dimensional operating data of commercial vehicles from the vehicle terminal, perform working condition scenario clustering on the multi-dimensional operating data to obtain scenario labels and multi-dimensional working condition feature vectors, and upload them to the cloud.

[0028] Among them, multidimensional operational data refers to the raw operational data collected by the vehicle end, which includes vehicle motion information, load information, battery status information and environmental information, and is used to characterize the real-time operating conditions of commercial vehicles.

[0029] Scene tags refer to the operation condition type identifiers generated by operation condition clustering, which are used to distinguish between trunk logistics, urban delivery, and heavy-load operation scenarios in mining areas.

[0030] Multidimensional operating condition feature vectors refer to structured feature vectors constructed based on scene labels that have undergone one-hot encoding. These vectors are used to identify the current operating condition of a vehicle in the cloud, in order to support the extraction of micro-aging features and the generation of globally optimal control strategies.

[0031] In practical implementation, the vehicle-side collects multi-dimensional operational data from commercial vehicles, including vehicle operation information, load information, battery status information, and environmental information. This multi-dimensional operational data undergoes operational scenario clustering processing to obtain scenario labels that include at least trunk logistics, urban delivery, and heavy-duty mining operations. These scenario labels are then one-hot encoded, and a multi-dimensional operational scenario feature vector is constructed based on the encoding results. Subsequently, the scenario labels and the multi-dimensional operational scenario feature vector are uploaded to the cloud, allowing the cloud to extract micro-aging features and iteratively generate a globally optimal control strategy.

[0032] Step S20: The cloud extracts the micro-aging features under the corresponding operating conditions through the battery digital twin model constructed in the cloud based on the scene label and the multi-dimensional operating condition feature vector, and predicts the battery health status degradation rate based on the micro-aging features.

[0033] Among them, the battery digital twin model refers to a digital simulation model built in the cloud that accurately maps to the physical characteristics of the battery in the vehicle. This model is a physical information neural network (PINN) that integrates electrochemical mechanisms and data-driven methods. It is used to reproduce the internal operating state of the battery in combination with the real-time operating conditions of the vehicle, analyze the micro-aging behavior of the battery, and deduce the health evolution law of the battery.

[0034] Microscopic aging characteristics refer to multidimensional microscopic feature parameters that are strongly correlated with battery aging, extracted from high-frequency data collected from the vehicle by the cloud-based battery digital twin model. These parameters are used to characterize the internal electrochemical loss and thermal stress evolution of the battery, reflect the degree of battery aging, and predict the trend of battery health degradation.

[0035] Battery health status degradation rate refers to the rate at which the battery health status changes with operating conditions and operating time. It is used to quantify the battery aging rate under different operating conditions and to provide core data basis for generating the global optimal control strategy.

[0036] In its implementation, the cloud receives scene tags and multi-dimensional operating condition feature vectors uploaded from the vehicle, and calls a pre-constructed battery digital twin model that is precisely mapped to the vehicle's physical battery. Based on the scene tags and multi-dimensional operating condition feature vectors, the cloud inputs multi-dimensional operational data collected from the vehicle into this battery digital twin model, thereby analyzing the battery's internal operating state and extracting microscopic aging features under the corresponding operating conditions. These microscopic aging features include multi-dimensional parameters such as micro-circulation depth frequency distribution, high-current pulse polarization voltage, cell temperature gradient variance, and lithium plating overpotential at the charging end, which can characterize the battery's internal minute losses and stress evolution. Based on this, the cloud, using the extracted microscopic aging features and a degradation trajectory prediction algorithm, deduces the battery aging evolution law, calculates the battery health state decay rate under the corresponding operating conditions, and simultaneously estimates the battery's remaining lifespan, providing reliable data support for subsequently generating a globally optimal control strategy.

[0037] Step S30: The cloud uses the scene label, the multi-dimensional operating condition feature vector, and the battery health state decay rate to iteratively optimize and solve the problem through a trained deep reinforcement learning model to generate a globally optimal control strategy adapted to the current operating condition; wherein, the globally optimal control strategy includes battery charge and discharge regulation parameters and battery thermal management parameters.

[0038] Among them, the global optimal control strategy refers to the integrated battery management strategy obtained by combining the battery's micro-aging characteristics and health state decay rate in the cloud and iteratively solving through a trained deep reinforcement learning model. It is suitable for various vehicle operating conditions. This strategy covers battery charging and discharging parameters and thermal management parameters, which can suppress battery aging and extend battery life while ensuring vehicle operating efficiency.

[0039] Battery charge and discharge control parameters are core control parameters used to manage the input and output states of battery power. These include the adjustment of the charging cut-off voltage and the discharge power limit coefficient. They are used to regulate the charging and discharging behavior of the battery, avoid battery loss caused by abnormal operating conditions, and slow down the decline in battery health.

[0040] Battery thermal management parameters are core control parameters used to regulate the battery's operating temperature state and optimize the battery's thermal environment. These include the liquid cooling pump's operating speed, which is used to balance the battery's temperature field distribution, suppress the accumulation of battery thermal stress, and ensure the battery's thermal safety and stable operation.

[0041] In the specific implementation, the deep reinforcement learning model trained by the cloud application takes scene labels, multi-dimensional operating condition feature vectors, battery health state decay rate and battery body state parameters as state inputs. Iteratively optimizes the solution based on the multi-objective reward function built into the model, balances battery life extension performance and vehicle operating efficiency to obtain Pareto optimal solution, and generates a globally optimal control strategy adapted to the current operating conditions in real time. The globally optimal control strategy specifically includes battery charge and discharge regulation parameters and battery thermal management parameters.

[0042] In step S40, the cloud performs a trimming of the global optimal control strategy based on the current scene labels generated and reported by the vehicle terminal in real time clustering, generates a vehicle terminal control strategy, and sends it to the vehicle terminal; the vehicle terminal dynamically adjusts the battery charging and discharging control parameters and battery thermal management parameters according to the vehicle terminal control strategy.

[0043] Among them, the vehicle-side control strategy refers to the localized battery management scheme generated by the cloud based on the current scenario label and tailoring the globally optimal control strategy, which is used by the vehicle to adjust the battery charging, discharging and thermal management parameters in real time.

[0044] In the specific implementation, the cloud obtains the current scene label reported by the vehicle. This current scene label is generated by real-time clustering on the vehicle and is used to characterize the vehicle's real-time operating conditions. Based on this current scene label, the cloud tailors the globally optimal control strategy, extracts local strategy parameter packages or lightweight inference models adapted to the current operating conditions from the global strategy matrix, generates the vehicle-side control strategy, and sends it to the vehicle. The vehicle receives and loads this control strategy, and dynamically adjusts the battery charging and discharging control parameters and battery thermal management parameters through real-time interpolation calculations or lightweight model inference. This adapts to the actual operating conditions of the vehicle, suppresses battery micro-aging and thermal stress accumulation, and achieves optimized control for battery life extension while ensuring normal vehicle operating efficiency.

[0045] In this embodiment, multi-dimensional operational data of commercial vehicles is collected from the vehicle-side, and scene labels and multi-dimensional operating condition feature vectors are obtained through operating condition scenario clustering and uploaded to the cloud. The cloud uses a pre-built battery digital twin model to extract the battery's micro-aging characteristics based on the above operating condition information and deduce the battery's health status degradation rate. The cloud uses the operating condition features and battery degradation parameters as inputs and iteratively optimizes the trained deep reinforcement learning model to generate a globally optimal control strategy that balances vehicle operating efficiency and battery life extension performance. The cloud tailors the global strategy based on the scene labels reported by the vehicle-side in real time, generates a lightweight vehicle-side control strategy, and distributes it, allowing the vehicle-side to dynamically adjust battery charging, discharging, and thermal management parameters. This solution relies on a vehicle-cloud collaborative architecture, combining digital twins and reinforcement learning to achieve intelligent management and control of the entire battery chain. It can adapt to complex commercial vehicle operating conditions, effectively suppress battery aging and loss, and extend battery life while ensuring operational efficiency.

[0046] Based on the foregoing embodiments, a second embodiment of the commercial vehicle battery life extension control method of the present invention is proposed. In this embodiment, step S10 may include the following sub-steps S101~S104: In sub-step S101, the vehicle terminal collects vehicle kinematic data, vehicle load data, battery status data, and environmental data according to a preset sampling period to form multi-dimensional operating data. The vehicle kinematic data includes vehicle speed and acceleration; the load data includes vehicle weight; the battery status data includes cell voltage, surface temperature, and state of charge; and the environmental data includes ambient temperature and humidity.

[0047] Among them, vehicle kinematic data, which characterizes the vehicle's motion state, includes vehicle speed and acceleration.

[0048] Load data, which characterizes the vehicle's load-bearing status, includes the vehicle's weight.

[0049] Battery status data, which characterizes the real-time operating status of a battery, includes cell voltage, surface temperature, and state of charge.

[0050] Environmental data, which characterizes the external environmental conditions in which the battery is located, includes ambient temperature and humidity.

[0051] In the specific implementation, the vehicle collects vehicle kinematic data, vehicle load data, battery status data and environmental data according to a preset sampling period, and summarizes them into multi-dimensional operating data. Among them, the vehicle kinematic data includes vehicle speed and acceleration, the load data includes vehicle load, the battery status data includes cell voltage, surface temperature and state of charge, and the environmental data includes ambient temperature and humidity. The collected multi-dimensional operating data is cached locally on the vehicle for subsequent clustering calculations of operating conditions.

[0052] Sub-step S102 involves preprocessing the multidimensional operational data and extracting driver behavior features based on time-series statistics; the driver behavior features include at least one of start-stop frequency, state of charge fluctuation amplitude, and microcirculation depth frequency distribution.

[0053] Among them, driver behavior characteristics refer to the characteristic parameters extracted from multi-dimensional operational data that can reflect the driver's operating habits.

[0054] Start-stop frequency refers to the statistical number of times a vehicle starts and stops within a set time period.

[0055] The state of charge fluctuation range refers to the range of changes in the battery's state of charge during vehicle operation.

[0056] Microcycle depth frequency distribution refers to the statistical distribution pattern of discharge depth corresponding to small charge-discharge cycles of a battery.

[0057] In the specific implementation, the vehicle preprocesses the multi-dimensional operating data, sequentially performing outlier removal, missing data completion, and data normalization. Based on the continuous running sequence window statistical method, the preprocessed multi-dimensional operating data is subjected to time-series traversal and feature quantification calculation to extract driver behavior features. The driver behavior features include at least one of the following: start-stop frequency, state of charge fluctuation amplitude, and microcirculation depth frequency distribution, which are used to support subsequent working condition scenario clustering operations.

[0058] Sub-step S103: The improved DBSCAN time-series clustering algorithm is used to divide the multidimensional operating data into working conditions and scenarios, and generate scenario labels; the scenario labels include at least one of trunk logistics, urban delivery and heavy mining.

[0059] Among them, the improved DBSCAN temporal clustering algorithm refers to the improved density clustering algorithm that introduces temporal constraints on the basis of traditional DBSCAN density clustering and relies on multi-dimensional operating data to carry out working condition classification. It takes into account the sample feature distance and time continuity, avoids mis-clustering of working conditions across time periods, and automatically classifies working condition categories and generates scene labels.

[0060] Trunk logistics refers to cross-regional transportation operations involving commercial vehicles over long distances at a high rate of continuous, constant speed.

[0061] Urban delivery refers to the short-distance, frequent start-stop, and low-speed freight transportation in urban areas using commercial vehicles.

[0062] Heavy-duty mining operations refer to the transportation conditions within mining areas where commercial vehicles are continuously under heavy loads and road conditions fluctuate drastically.

[0063] In the specific implementation, the vehicle-side uses an improved DBSCAN time-series clustering algorithm to perform time-series operating condition clustering on multi-dimensional operating data. Based on data density features and combined with time series continuity, similar operating conditions are selected to complete the division of operating conditions and generate scene labels. These scene labels are used to identify the operating condition category to which the vehicle belongs, including at least one of trunk logistics, urban delivery, and heavy-duty mining. Subsequently, they are uploaded to the cloud along with the multi-dimensional operating condition feature vector to support cloud-based analysis of battery micro-aging characteristics.

[0064] Furthermore, in one embodiment, sub-step S103 may further include the following sub-step S1031: In sub-step S1031, when using the improved DBSCAN temporal clustering algorithm to divide the working conditions into scenarios, a preset neighborhood radius and a preset minimum number of samples are used as density determination parameters, and a time window is introduced as a temporal constraint. When performing neighborhood search, only data points with a time difference less than a preset threshold are included in the neighborhood calculation range, so as to identify densely connected data points in similar time intervals and realize working condition clustering under temporal coupling.

[0065] The preset neighborhood radius refers to the preset distance threshold used in the improved DBSCAN temporal clustering algorithm to determine the spatial proximity of samples. In this embodiment, the preset neighborhood radius ε = 0.35.

[0066] The preset minimum sample size refers to the minimum amount of data required to determine the core cluster samples within the neighborhood. In this embodiment, the preset minimum sample size MinPts = 50.

[0067] The preset threshold is the maximum allowable time difference used to constrain the temporal correlation of samples.

[0068] In its implementation, the vehicle-side uses an improved DBSCAN temporal clustering algorithm to classify multi-dimensional operating data into different operating scenarios. The algorithm uses a preset neighborhood radius and a preset minimum number of samples as the basis for determining data density, and adds a time window as a temporal constraint. During the neighborhood search process, only sample data with a time difference less than a preset threshold are included in the neighborhood calculation range, and only densely connected data samples are selected within similar time intervals. This completes the identification and classification of operating scenario clusters under temporal coupling constraints, ultimately achieving accurate classification of vehicle operating conditions and generating corresponding scenario labels. This effectively avoids interference from irrelevant samples across time periods and improves the accuracy of operating condition identification.

[0069] Sub-step S104: Perform one-hot encoding on the scene label, and concatenate the one-hot encoded scene label with the driver behavior features to construct a multi-dimensional working condition feature vector.

[0070] In the specific implementation, the vehicle-side performs one-hot encoding on the clustered operating condition scene labels to achieve the digital and vectorized conversion of discrete operating condition labels. Then, the scene label vectors that have completed one-hot encoding are concatenated with the previously extracted driver behavior features to merge the operating condition category information and driver operation behavior information, thereby constructing a multi-dimensional operating condition feature vector with complete dimensions and full information coverage. This effectively enriches the dimensions of operating condition feature representation and ensures the accuracy of subsequent cloud-based battery aging analysis and strategy optimization.

[0071] In this embodiment, multi-dimensional operation data is constructed by collecting various types of vehicle operation-related data at preset intervals on the vehicle side; the multi-dimensional operation data is preprocessed and driver behavior features are extracted through time-series statistics; the improved DBSCAN time-series clustering algorithm is used to complete the division of operating conditions and generate corresponding scene labels; the scene labels are one-hot encoded and then spliced ​​and fused with driver behavior features to construct a multi-dimensional operating condition feature vector, providing accurate operating condition data support for subsequent cloud-based battery aging analysis and control strategy generation.

[0072] Based on the foregoing embodiments, a third embodiment of the commercial vehicle battery life extension control method of the present invention is proposed. In this embodiment, step S20 may include the following sub-steps S201~S202: In sub-step S201, the cloud utilizes a battery digital twin model that integrates the PINN electrochemical mechanism, and combines the scene labels and the multi-dimensional operating condition feature vector to perform operating condition alignment processing on the multi-dimensional operating data, extracting the micro-aging features under the corresponding operating conditions; the micro-aging features include at least one of the following: micro-cycle depth frequency distribution, micro-cycle mechanical stress index, high current pulse polarization voltage, lithium plating overpotential at the end of charging, and cell temperature gradient variance.

[0073] The semi-empirical kinetic equation for electrochemical aging embedded in the Physical Information Neural Network (PINN) can be expressed as:

[0074] In the formula, Indicates the rate of change of battery health status over time; k 1、 k2 represents the aging coefficient to be identified, which is obtained from aging test calibration: k1 = 0.0023, k2 = 0.015; I represents the battery current amplitude; E a R represents the activation energy of the aging reaction; T represents the universal gas constant; T represents the thermodynamic temperature of the battery cell in K; and SOC represents the state of charge.

[0075] Microcirculation depth frequency distribution refers to the statistical distribution pattern of discharge depth corresponding to small charge and discharge cycles of the battery within a time interval.

[0076] The microcycle mechanical stress index is a quantitative indicator of the degree of stress accumulation on the electrodes caused by repeated small-scale charging and discharging of a battery. The formula for calculating the microcycle mechanical stress index is as follows:

[0077] In the formula, MSI is the microcirculation mechanical stress index; n is the total number of microcirculation events identified within the statistical period; Let be the SOC fluctuation amplitude of the i-th microcycle; This index corresponds to the frequency of small SOC fluctuations. It characterizes the cumulative mechanical stress caused by repeated expansion and contraction of the electrodes due to frequent small SOC fluctuations. In this embodiment, the MSI is 28.7 for urban delivery and 5.2 for trunk logistics; the micro-circulation stress in urban delivery is approximately 450% higher than that in trunk logistics. The reason for this is that urban delivery trucks frequently start and stop, and operate at low speeds, continuously generating small charging and discharging behaviors, significantly accelerating electrode mechanical fatigue aging.

[0078] High-current pulse polarization voltage refers to the voltage polarization difference generated when a battery is subjected to a high-current pulse.

[0079] The lithium plating overpotential at the end of charging refers to the additional potential threshold required for lithium ion deposition at the end of charging.

[0080] Cell temperature gradient variance refers to the statistical quantification of the degree of temperature unevenness among individual battery cells. In this embodiment, under heavy-load conditions in the mining area, the temperature difference variance between the cell surface and the core area of ​​the cell can reach 4.2℃². This large temperature unevenness will cause inconsistent SEI film growth rates in different areas of the battery, with the local SEI film growth rate increasing by up to 2.1 times, accelerating local aging of the cell.

[0081] During the extraction of micro-aging features, target features can be selected according to the differences in vehicle operation scenarios: For urban delivery scenarios of commercial vehicles (frequent start-stop), the digital twin model focuses on extracting the "micro-circulation depth (Micro-DOD)" feature to identify high-frequency micro fluctuations in SOC between 40% and 45% and calculate the fatigue damage caused by the cumulative mechanical stress to the electrode coating; For trunk logistics scenarios (long-term fast charging), the focus is on extracting "charging end polarization voltage" and "temperature gradient variance" to assess the risk of lithium plating under high current.

[0082] In its implementation, a cloud-based digital twin model of the battery, integrating the electrochemical mechanism of PINN, is used to perform condition alignment and correction processing on multi-dimensional operating data by combining scene labels and multi-dimensional operating condition feature vectors. This achieves accurate matching between the operating data and the actual operating conditions. Based on the effective data after condition alignment, the battery's micro-aging parameters are calculated, and the battery's micro-aging characteristics are extracted. These micro-aging characteristics include at least one of the following: micro-cycle depth frequency distribution, micro-cycle mechanical stress index, high-current pulse polarization voltage, lithium plating overpotential at the end of charging, and cell temperature gradient variance. This accurately characterizes the internal aging mechanism of the battery under different operating conditions, providing reliable micro-feature support for subsequent prediction of the battery's healthy degradation rate.

[0083] Sub-step S202: Based on the microscopic aging characteristics, a degradation trajectory prediction algorithm is used to calculate the battery health state decay rate and the remaining battery life under the corresponding operating conditions.

[0084] Among them, the degradation trajectory prediction algorithm refers to a time-series prediction algorithm that takes the micro-aging characteristics of the battery as input, fits the aging evolution curve of the battery health status with time / cumulative operating conditions, quantitatively infers the SOH decay rate, and predicts the remaining service life of the battery; at least one of Gaussian process regression, long short-term memory network (LSTM), and Wiener process degradation model can be selected.

[0085] Battery remaining lifespan refers to the usable time / number of cycles that a battery can continue to work normally under its current operating conditions and aging state until its health deteriorates to a preset failure threshold.

[0086] In the specific implementation, the cloud relies on the extracted microscopic aging characteristics of various batteries to input the degradation trajectory prediction algorithm to construct a battery aging evolution model; based on the aging model, the battery health state decay rate under the corresponding operating conditions is deduced and solved, and the remaining service life of the battery is continuously predicted iteratively, so as to realize the quantitative assessment of the long-term aging trend of the battery under multiple operating conditions, and provide accurate battery life constraints for subsequent deep reinforcement learning strategy optimization.

[0087] In this embodiment, the cloud relies on a battery digital twin model that integrates the PINN electrochemical mechanism to complete the alignment of multi-dimensional operating data and extract the battery's micro-aging characteristics. Based on the micro-aging characteristics, a degradation trajectory prediction algorithm is used to calculate the battery's health state decay rate and remaining service life, providing quantitative aging boundary conditions for optimizing the cloud control strategy.

[0088] Based on the foregoing embodiments, a fourth embodiment of the commercial vehicle battery life extension control method of the present invention is proposed. In this embodiment, before step S10, the method may further include the following steps S50~S60: Step S50: A Markov decision process framework for battery life extension control is constructed in the cloud, a reinforcement learning model is built, and a state space, action space, and reward function are defined. The state space includes battery health state, state of charge, battery gridded temperature field distribution, operating condition feature vector, and scene labels encoded by unique thermal coding. The action space includes charging cutoff voltage adjustment, discharge power limit coefficient, and liquid cooling pump operating speed. The reward function includes a lifespan degradation penalty term, an operational efficiency benefit term, and a safety penalty term, used to balance the degree of battery aging and vehicle operating efficiency, and to solve for the Pareto optimal control solution.

[0089] The Markov decision process framework refers to a time-series decision model framework that uses state space, action space, and reward function as core elements to construct the decision logic for battery life extension optimization control.

[0090] The state space is the set of all observed states of the reinforcement learning model. It is used to represent the battery health status, electrical status, temperature field distribution, and vehicle operating conditions in real time, providing complete environmental input information for control decisions.

[0091] Battery health status refers to a quantitative parameter used to characterize the current overall aging and degradation level and available capacity of a battery.

[0092] State of charge (SOC) refers to the ratio of a battery's current remaining usable capacity to its rated capacity, reflecting the battery's real-time energy reserve level.

[0093] Battery gridded temperature field distribution refers to the real-time temperature distribution data of each region obtained after gridding the battery pack area, which is used to characterize the temperature uniformity and thermal state characteristics of the battery cell.

[0094] Action space refers to the set of all controllable execution variables that a reinforcement learning model can output and that can be used to regulate the battery system.

[0095] The charging cut-off voltage adjustment refers to the dynamic offset adjustment relative to the standard charging cut-off voltage. It is used to flexibly adjust the upper limit of battery charging to adapt to different aging conditions and suppress the risks of overcharging, aging, and lithium plating.

[0096] The discharge power limiting factor is a proportional factor used to dynamically limit the maximum discharge power of a battery. It is used to avoid damage from high-current pulse polarization and to balance power output with battery life extension requirements.

[0097] The operating speed of the liquid cooling pump refers to the real-time operating speed of the liquid cooling pump in the battery thermal management system. It is used to adjust the coolant circulation rate, optimize the uniformity of the cell temperature field, and suppress local aging induced by temperature differences.

[0098] The reward function refers to a multi-objective evaluation function used to quantify the quality of control actions and construct a balance between battery life extension, operational safety, and vehicle operating efficiency.

[0099] The lifespan benefit term refers to the evaluation component constructed based on the real-time aging and degradation rate of the battery, which is used to guide intelligent strategies to suppress accelerated battery aging under extreme operating conditions and achieve battery life extension optimization.

[0100] The operational efficiency benefit component refers to the positive benefit component constructed based on the vehicle's power output and the degree of adaptation to operating conditions, which is used to take into account the operational efficiency of commercial vehicles under battery-constrained control conditions.

[0101] Safety penalty items refer to risk penalty components constructed based on battery voltage, current, temperature safety thresholds, and lithium plating boundary conditions. These components are used to avoid battery abuse conditions and ensure battery operation safety.

[0102] The reward function can be expressed as:

[0103] In the formula, For lifetime gains, it is inversely proportional to the rate of SOH decay caused by the current action; For efficiency gains, it is related to the time and energy consumption of the vehicle to complete the predetermined transportation task, ensuring that commercial vehicles do not sacrifice operational efficiency due to excessive protection; For safety penalties, a severe negative penalty is imposed when an action causes the cell temperature to exceed the safety threshold or triggers the lithium plating boundary; α, β, and γ are weight coefficients that are dynamically adjusted according to the operating condition scenario label (for example, the weight of β is increased in heavy-load mining scenarios, and the weight of α is increased in trunk logistics scenarios).

[0104] In one set of embodiments, the reward function takes a fixed-weight form, expressed as follows:

[0105] In the formula, The amount of battery health state degradation per unit time under the baseline control strategy; This represents the predicted battery health degradation amount corresponding to the current control action to be evaluated. The target time for the vehicle to complete the transportation task; The actual runtime corresponding to the current control action; For safety penalties.

[0106] Reinforcement learning models refer to intelligent decision-making models that rely on the Markov decision process framework, continuously optimize control decision strategies through offline iterative training of samples, and achieve autonomous reasoning for optimal battery regulation actions with battery safety, operational efficiency, and life extension performance as optimization objectives.

[0107] This model weights and fuses the lifespan degradation penalty, operational efficiency benefit, and safety penalty to construct the overall reward function of the reinforcement learning model. By quantifying battery aging losses, operational benefits, and operational safety risks, it comprehensively evaluates the merits of battery control actions, weighs battery life extension, vehicle operational efficiency, and battery operational safety, and supports the model in iteratively solving for the optimal control strategy that is safe, efficient, and adapted to battery life extension.

[0108] In the specific implementation, a Markov decision process framework for battery life extension control is constructed in the cloud, and a deep reinforcement learning model is built, clarifying the model constraint dimensions and optimization objectives. Specifically, the state space includes battery health state, state of charge, battery gridded temperature field distribution, operating condition feature vector, and unique thermal encoding scene label; the action space includes charging cutoff voltage adjustment, discharge power limit coefficient, and liquid cooling pump operating speed; and a multi-dimensional reward function includes life decay penalty term, operational efficiency benefit term, and safety penalty term. Through a multi-objective reward and punishment mechanism, the battery aging loss and vehicle operating efficiency are weighed, and the Pareto optimal control solution that takes into account safety, energy efficiency, and life extension is obtained through iterative solution, providing a global optimal control benchmark for the subsequent vehicle-side lightweighting strategy.

[0109] Step S60: Based on the Markov decision process framework, using the historical operation data uploaded by the vehicle and the simulation data generated by the battery digital twin model, extract all state samples and action samples that match the above-mentioned state space and action space definitions, and train the reinforcement learning model offline. After the model converges, a deep reinforcement learning model that can be used for online reasoning is obtained.

[0110] Among them, state samples refer to model input sample data that conforms to the preset state space dimension, is extracted from measured data and simulation data, and is used to characterize the real-time characteristics of the battery and operating conditions.

[0111] Action samples refer to model output sample data that fits the preset action space dimension and corresponds to battery regulation behavior, and are used to guide the iterative optimization of the reinforcement learning model's control strategy.

[0112] In its implementation, the system relies on the established Markov decision process framework, integrates historical battery operation data uploaded from the vehicle with high-precision simulation data generated from the battery digital twin model, and batches extracts all state and action samples that match the preset state and action space definitions. It then uses a massive dataset of mixed measured and simulated samples to conduct offline iterative training of the reinforcement learning model, continuously optimizing model parameters until the model converges. This ultimately generates a deep reinforcement learning model with high-precision online inference capabilities, providing reliable model support for subsequent real-time battery optimization control strategy output.

[0113] In this embodiment, a Markov decision process framework for battery life extension control is constructed in the cloud, and a reinforcement learning model is built. The state space, action space, and reward function adapted to battery optimization control are defined. Based on the Markov decision process framework, state samples and action samples are extracted by integrating measured data and simulation data. The offline training and convergence of the reinforcement learning model are completed, and a deep reinforcement learning model that can be reasoned online is obtained, providing model support for solving the optimal battery life extension control strategy.

[0114] Based on the foregoing embodiments, a fifth embodiment of the commercial vehicle battery life extension control method of the present invention is proposed. In this embodiment, in order to adaptively tailor the optimal control strategy according to the real-time scenario on the vehicle and complete the remote deployment, step S40 may include the following sub-steps S401~S402: In sub-step S401, the cloud generates and reports scene labels based on real-time clustering on the vehicle side, and then cuts out a local policy parameter package or lightweight inference model that is adapted to the scene labels from the global optimal control strategy, and uses it as the vehicle control strategy.

[0115] Among them, the local strategy parameter package refers to the set of reinforcement learning control parameters extracted for a specific operating scenario, which can be directly invoked by the vehicle to execute the control logic.

[0116] Lightweight inference models refer to small inference models that are trimmed and compressed from complete deep reinforcement learning models in the cloud and adapted to run on the edge computing power of vehicles.

[0117] In the specific implementation, the cloud receives the scene label information uploaded by the vehicle in real time through clustering calculation, retrieves the previously solved global optimal control strategy, completes strategy pruning based on the operating condition characteristics corresponding to the scene label, and generates a local strategy parameter package or lightweight inference model adapted to the current scene. The pruned local strategy parameter package or lightweight inference model is then sent to the vehicle and used as the real-time battery control strategy, reducing the computing load on the vehicle edge devices and enabling the edge execution of battery life extension optimization strategies under different operating conditions.

[0118] In sub-step S402, the vehicle-side control strategy is sent to the vehicle-side using differential OTA transmission.

[0119] Among them, differential OTA transmission mode refers to the incremental upgrade transmission mechanism that only transmits the data fragments that differ between the old and new version strategies. Compared with the distribution of complete data packets, it can effectively reduce transmission traffic and transmission latency.

[0120] In the specific implementation, the cloud adopts differential OTA transmission to incrementally encapsulate and transmit the target vehicle-side control strategy, only sending data segments that have changed compared to the existing strategy on the vehicle side; the differential data packets are remotely sent to the vehicle terminal to complete the strategy update deployment, reducing the amount of data transmitted, shortening the time for strategy delivery, and ensuring that the vehicle side can efficiently and stably complete the online update of the battery optimization control strategy.

[0121] Furthermore, in one embodiment, after completing sub-steps S401 to S402, which adaptively tailor the optimal control strategy based on the real-time scenario on the vehicle and complete the remote deployment, in order to enable the vehicle to dynamically adjust the battery operating parameters based on the deployed strategy and ensure the safe and stable operation of the battery system under abnormal communication conditions, step S40 may further include the following sub-steps S403 to S404: In sub-step S403, the vehicle-side battery management system loads the vehicle-side control strategy and dynamically adjusts the charging cut-off voltage adjustment amount, discharge power limit coefficient, and liquid cooling pump operating speed through real-time interpolation calculation or lightweight model inference. In sub-step S404, the vehicle continuously monitors the communication link status with the cloud. When communication with the cloud is interrupted, it automatically switches to the pre-fixed local safety degradation strategy to continuously implement battery safety management.

[0122] Among them, the safety degradation strategy refers to the fallback control strategy that is pre-fixed locally on the vehicle. It is automatically triggered when communication is lost and ensures the safe and stable operation of the battery system through conservative battery charging and discharging and thermal management control logic.

[0123] The specific manifestations of the dynamic execution of the vehicle-side battery management system (BMS) are as follows: When the vehicle-side BMS receives the strategy sent from the cloud, during the charging phase, if it is detected that the current environment is high temperature and the driver has a habit of parking for a long time after the battery is fully charged, the BMS dynamically reduces the charging cut-off voltage (e.g., from 100% to 95%) and starts a low-power liquid cooling cycle to slow down the battery calendar aging; during the discharging phase, if it is detected that there is a long downhill or heavy load start-up condition, the BMS dynamically relaxes the instantaneous discharge power limit, and at the same time fine-tunes the operating speed of the liquid cooling pump to intervene in heat dissipation in advance to suppress the temperature rise peak caused by high current.

[0124] In its implementation, the vehicle-side battery management system loads the vehicle-side control strategy issued from the cloud. Relying on real-time interpolation calculation or lightweight model online inference, it dynamically and adaptively adjusts the battery charging cutoff voltage adjustment, discharge power limit coefficient, and liquid cooling pump operating speed to achieve precise optimization and control of battery charging and discharging parameters and thermal management parameters. At the same time, the vehicle-side continuously monitors the communication link status with the cloud in real time. In case of abnormal operating conditions such as communication interruption, it automatically switches to the locally pre-stored and solidified safety degradation strategy to continuously carry out battery safety management and control, effectively avoiding the risk of control failure caused by cloud disconnection, and ensuring uninterrupted, safe and stable operation of the battery system under all operating conditions.

[0125] In this embodiment, the vehicle control strategy is constructed by using the cloud to generate local strategy parameter packages or lightweight inference models adapted to the current operating conditions from the global optimal control strategy based on the scene tags reported by the vehicle in real time clustering. The vehicle control strategy is then sent to the vehicle terminal using a differential OTA transmission mechanism, realizing incremental remote deployment of the scene-matching control strategy and providing strategy support for real-time optimization and control of the vehicle battery.

[0126] The following describes a specific embodiment of the verification of the present invention: active life extension control verification in an urban delivery scenario.

[0127] This embodiment describes the specific implementation process and verification results of applying the technical solution of this invention under typical urban delivery micro-circulation conditions. This embodiment selects a 6×4 electric light truck equipped with a lithium iron phosphate battery pack as the test vehicle. The battery pack has a total energy of 142kWh, a rated voltage of 618.24V, and consists of four modules connected in series, with each module containing 48 cells. The test uses the CLTC-B (China Heavy-duty Commercial Vehicle Test Cycle-Bus) urban delivery cycle condition to simulate urban delivery road conditions, simulating an average daily mileage of 200km. The vehicle-side BMS collects multi-dimensional operating data from GPS and CAN bus, obtaining the following operating condition characteristics: average vehicle speed 28km / h, start-stop frequency reaching 22 times / km; affected by the frequent start-stop of the vehicle, the battery SOC fluctuates frequently within the range of 40% to 80%. Based on the aforementioned working condition clustering identification method, the operating condition was determined to be a high-frequency micro-circulation scenario, and the micro-feature parameters were extracted: micro-circulation depth frequency f = 152 times / 100 km, and micro-circulation mechanical stress index MSI = 28.7.

[0128] The commercial vehicle battery life extension control method of the present invention, which operates collaboratively between the vehicle and the cloud, includes the following specific implementation steps: Step A1: Dynamic Adjustment of Charging Strategy Driven by Microscopic Aging Characteristics: A cloud-based digital twin model of the battery, integrating the electrochemical mechanism of PINN, continuously monitors the battery's operating status. When the cumulative micro-cycle mechanical stress index (MSI) exceeds a preset threshold of 20, it determines that the risk of mechanical fatigue of the battery electrodes has increased and sends an adaptive charging control command to the vehicle. The vehicle's BMS dynamically lowers the standard charging cutoff voltage of the single cell from 3.65V to 3.55V, reducing the charging upper limit by approximately 2.7%. Simultaneously, current limiting is implemented in the SOC range above 90%, reducing the charging current from 1C to 0.5C to mitigate electrode grain boundary stress damage caused by lithium-ion intercalation.

[0129] Step A2, Active Thermal Management Control Based on Cell Temperature Gradient: The vehicle-side BMS collects cell temperature data in real time. When the cell temperature gradient variance σT² > 3℃² is detected, it is determined that the battery has a risk of local overheating. The operating speed of the liquid cooling pump is increased from the conventional 2500rpm to 3800rpm, and the condenser fan is turned on simultaneously to assist in heat dissipation, keeping the temperature difference on the cell surface within 2℃ and suppressing the excessive growth of the SEI film (Solid Electrolyte Interphase) induced by uneven thermal stress.

[0130] The aforementioned adaptive life-extending control strategy was maintained, and 300 complete charge-discharge cycle tests were conducted to simulate approximately three months of actual vehicle operation. After the cycle tests, battery performance and microstructure were evaluated. Specific test results were as follows: After the cycle tests, the battery pack's state of equilibrium (SOH) decreased to 94.2%; the battery pack cells were disassembled, and the microstructure of the negative electrode was observed using SEM (Scanning Electron Microscope), revealing a crack density of 12.3 cracks / μm² in the electrode coating; side reaction analysis was performed using ICA (Incremental Capacity Analysis), and no significant lithium plating characteristic peaks were observed in the curves, resulting in a battery lithium plating risk probability assessment of 5.2%.

[0131] Comparative Example: Validation of Traditional Curing BMS Control Strategy To verify the technical effect of this invention, a test vehicle platform, lithium iron phosphate battery pack specifications, CLTC-B urban delivery operation conditions, and an environmental test condition of 25±2℃ were set up, completely identical to those in the above embodiments. The traditional, fixed BMS control strategy was adopted, without access to a cloud platform or the ability to perceive microscopic aging characteristics and perform dynamic optimization through reinforcement learning. The specific control process and test procedure are as follows: Step B1, Fixed Parameter Charging Control Process: This comparative example uses conventional constant current and constant voltage fixed charging control logic, without adaptive parameter adjustment based on operating conditions or battery aging status. Specifically, the charging cutoff voltage of each battery cell is fixed at 3.65V, and charging stops immediately once the battery voltage reaches the fixed threshold. The entire charging process uses a conventional constant current and constant voltage charging mode, without adding dynamic current limiting logic for the high-frequency micro-circulation and frequent start-stop characteristics of urban delivery operations; the charging parameters remain constant throughout the process.

[0132] Step B2, Passive Thermal Management Control Process: This comparative example adopts the industry's traditional passive heat dissipation control logic, lacking cell temperature gradient sensing, early prediction, and active adjustment capabilities. It only sets a single temperature protection threshold, activating the liquid cooling device only when the highest cell temperature exceeds 45°C, with the liquid cooling pump maintaining a constant speed of 3000 rpm. When the battery temperature is within the limit, the thermal management system remains dormant, unable to proactively intervene in battery temperature rise and uneven temperature differences within the cell.

[0133] Step B3, Cyclic Testing and Data Acquisition: Using the same testing process as the above embodiment, complete 300 full charge-discharge cycles. During the test, collect battery SOH decay data and cell operating status data throughout the process. After the test, use SEM scanning micro-morphology detection and ICA analysis to complete a comprehensive evaluation of the battery's micro-aging characteristics and lithium plating risk.

[0134] Step B4, Aging Result Comparison Analysis: After the cycle test, the core performance parameters of the comparative battery were collected. The specific test results are as follows: the battery pack's SOH eventually decayed to 88.5%, and the overall life decay rate was nearly doubled compared to the embodiment of the commercial vehicle battery life extension control method; SEM micro-morphology detection results showed that the battery negative electrode sheet suffered large-area fatigue damage, and the electrode coating crack density reached 37.6 cracks / μm², indicating that high-frequency micro-cycle conditions caused significant damage to the electrode structure; ICA incremental capacity analysis results showed that the battery had a significant lithium plating response peak in the 3.65V voltage range, and the overall lithium plating risk probability of the battery could reach 18.7%.

[0135] The test indicators of the examples and comparative examples are summarized in the following table: Table 1. Comparison of the overall effects of the comparative examples and the embodiments, and conclusions. SOH decay rate (% / cycle) 0.0383 0.0193 50% decrease The results demonstrate that dynamically adjusting the charging limit effectively suppresses irreversible capacity loss caused by high-frequency shallow charging and discharging. Cell temperature difference (°C) 5.2 1.8 65% decrease The results demonstrate that temperature gradient prediction and active thermal management based on digital twins significantly improve cell consistency. Lithium plating risk probability 18.70% 5.20% 72% decrease This demonstrates that lowering the charging cutoff voltage effectively prevents the graphite negative electrode potential from falling into the lithium plating range (<0.15Vvs Li / Li+). Electrode crack density (cracks / μm2) 37.6 12.3 67% decrease Microscopic evidence directly supports the improvement in macroscopic lifetime data, verifying the feasibility of "microscopic features driving lifetime extension". Total lifecycle TCO (ten thousand yuan) 142.5 118.3 Down 17% The increased residual value of used cars and reduced replacement costs resulting from extended battery life demonstrate significant commercial value. A comprehensive comparison of the comparative and implementation examples shows that traditional fixed BMS management strategies are ill-suited to the differentiated aging characteristics of urban delivery micro-circulation conditions, and cannot effectively suppress electrode mechanical fatigue and battery lithium plating. This invention, through a closed-loop management system of operating condition clustering identification, microscopic aging feature extraction, cloud-based intelligent strategy optimization, and precise vehicle-side execution, achieves customized life-extending control adapted to operating conditions, significantly improving battery operational consistency and service stability, effectively verifying the effectiveness and engineering practical value of this solution.

[0136] Based on the same inventive concept, the seventh embodiment of this invention also provides a commercial vehicle battery life extension control system corresponding to the commercial vehicle battery life extension control method of the foregoing embodiments. Since the principle of the system in the seventh embodiment of this invention is similar to the aforementioned commercial vehicle battery life extension control method of the foregoing embodiments, the implementation of the system can be referred to the implementation of the method, and repeated details will not be elaborated further. Please refer to... Figure 2 The present invention relates to a commercial vehicle battery life extension control system, wherein the system may include: Vehicle terminal 10 is used to collect multi-dimensional operating data of commercial vehicles, perform working condition scenario clustering on the multi-dimensional operating data, obtain scenario labels and multi-dimensional working condition feature vectors, and upload them to the cloud. Cloud 20 is used to extract micro-aging features under corresponding working conditions through a battery digital twin model constructed in the cloud based on the scene label and the multi-dimensional working condition feature vector, and to predict the battery health status degradation rate based on the micro-aging features. The cloud 20 is also used to generate a globally optimal control strategy adapted to the current operating condition by iteratively optimizing and solving the problem through a trained deep reinforcement learning model based on the scene label, the multi-dimensional operating condition feature vector, and the battery health state decay rate; wherein, the globally optimal control strategy includes battery charge and discharge regulation parameters and battery thermal management parameters. The cloud 20 is also used to trim the global optimal control strategy based on the scene labels generated and reported by the vehicle terminal in real time clustering, generate the vehicle terminal control strategy and send it to the vehicle terminal. The vehicle terminal 10 is also used to dynamically adjust the battery charging and discharging parameters and battery thermal management parameters according to the vehicle terminal control strategy.

[0137] In addition, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described commercial vehicle battery life extension control method.

[0138] Figure 3 This is a schematic block diagram of the electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device includes at least one processor 401, a memory 402, at least one network interface 403, and a user interface 405. The various components in the electronic device are coupled together via a bus system 404. It is understood that the bus system 404 is used to implement communication between these components. In addition to a data bus, the bus system 404 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in… Figure 3 The general will label all buses as bus systems.

[0139] The user interface 405 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0140] It is understood that memory 402 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this invention are intended to include, but are not limited to, these and any other suitable categories of memory.

[0141] In this embodiment of the invention, the memory 402 is used to store various types of data to support the operation of the electronic device 400. Examples of this data include: any executable program for operation on the electronic device 400, such as the operating system 4021 and application program 4022; the operating system 4021 contains various system programs, such as the framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. The application program 4022 may contain various applications, such as a media player, browser, etc., for implementing various application services. The commercial vehicle battery life extension control method provided in this embodiment of the invention can be included in the application program 4022.

[0142] The methods disclosed in the above embodiments of the present invention can be applied to processor 401, or implemented by processor 401. Processor 401 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 401 or by instructions in the form of software. The processor 401 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 401 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. General-purpose processor 401 may be a microprocessor or any conventional processor, etc. The steps of the commercial vehicle battery life extension control method provided in the embodiments of the present invention can be directly reflected as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in a memory. The processor reads the information in the memory and combines it with its hardware to complete the steps of the aforementioned method.

[0143] In an exemplary embodiment, the electronic device 400 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to perform the aforementioned method.

[0144] In summary, this invention collects multi-dimensional operational data from commercial vehicles at the vehicle end, clusters this data under various operating conditions to obtain scene labels and multi-dimensional operating condition feature vectors, and uploads them to the cloud. The cloud then uses a pre-built battery digital twin model, combined with the aforementioned operating condition information, to extract microscopic aging features of the battery and deduce the battery's health degradation rate. The cloud uses the operating condition features and battery degradation parameters as input, and iteratively optimizes a trained deep reinforcement learning model to generate a globally optimal control strategy that balances vehicle operating efficiency and battery life extension. Based on the scene labels reported in real-time by the vehicle, the cloud tailors the global strategy, generates a lightweight vehicle-side control strategy, and distributes it, allowing the vehicle to dynamically adjust battery charging / discharging and thermal management parameters. This solution relies on a vehicle-cloud collaborative architecture, combining digital twins and reinforcement learning to achieve intelligent end-to-end battery management, adaptable to complex commercial vehicle operating conditions, effectively suppressing battery aging and loss, and extending battery life while ensuring operational efficiency.

[0145] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for controlling the life extension of commercial vehicle batteries, characterized in that, The method includes: Multi-dimensional operating data of commercial vehicles are collected from the vehicle end, and the multi-dimensional operating data is clustered according to working conditions to obtain scene labels and multi-dimensional working condition feature vectors, which are then uploaded to the cloud. The cloud-based system extracts microscopic aging features under corresponding operating conditions based on the scene labels and the multidimensional operating condition feature vectors, and predicts the battery health status degradation rate based on the microscopic aging features. The cloud-based system generates a globally optimal control strategy adapted to the current operating conditions by iteratively optimizing the solution through a trained deep reinforcement learning model based on the scene label, the multi-dimensional operating condition feature vector, and the battery health state decay rate. The globally optimal control strategy includes battery charge and discharge regulation parameters and battery thermal management parameters. The cloud-based system tailors the globally optimal control strategy based on scene tags generated and reported by the vehicle-side in real-time clustering, generates a vehicle-side control strategy, and sends it to the vehicle-side system. The vehicle-side system then dynamically adjusts the battery charging and discharging control parameters and battery thermal management parameters according to the vehicle-side control strategy.

2. The method according to claim 1, characterized in that, The multi-dimensional operating data of commercial vehicles collected from the vehicle end is clustered into operating conditions and scenarios to obtain scenario labels and multi-dimensional operating condition feature vectors, including: Vehicle kinematics data, vehicle load data, battery status data, and environmental data are collected by the vehicle end according to a preset sampling period to form multi-dimensional operating data. The vehicle kinematics data includes vehicle speed and acceleration; the load data includes vehicle weight; the battery status data includes cell voltage, surface temperature, and state of charge; and the environmental data includes ambient temperature and humidity. The multidimensional operational data is preprocessed, and driver behavior features are extracted based on time-series statistics; the driver behavior features include at least one of start-stop frequency, state of charge fluctuation amplitude, and microcirculation depth frequency distribution. An improved DBSCAN time-series clustering algorithm is used to divide the multidimensional operational data into working conditions and scenarios, generating scenario labels; the scenario labels include at least one of trunk logistics, urban delivery, and heavy-duty mining. The scene labels are one-hot encoded, and the one-hot encoded scene labels are concatenated with the driver behavior features to construct a multi-dimensional working condition feature vector.

3. The method according to claim 2, characterized in that, When using the improved DBSCAN temporal clustering algorithm to divide work scenarios, a preset neighborhood radius and a preset minimum number of samples are used as density determination parameters, and a time window is introduced as a temporal constraint. When performing neighborhood search, only data points with a time difference of less than a preset threshold are included in the neighborhood calculation range, so as to identify densely connected data points in similar time intervals and realize work scenario clustering under temporal coupling.

4. The method according to claim 1, characterized in that, The step of extracting microscopic aging features under corresponding operating conditions by using a cloud-based digital twin model of the battery based on the scene labels and the multidimensional operating condition feature vectors, and predicting the degradation rate of the battery health status based on the microscopic aging features, includes: The cloud-based digital twin model of the battery, which integrates the PINN electrochemical mechanism, is used to perform working condition alignment processing on the multi-dimensional operating data by combining the scene labels and the multi-dimensional working condition feature vectors, and to extract the micro-aging features under the corresponding working conditions. The micro-aging features include at least one of the following: micro-cycle depth frequency distribution, micro-cycle mechanical stress index, high current pulse polarization voltage, lithium plating overpotential at the end of charging, and cell temperature gradient variance. Based on the aforementioned microscopic aging characteristics, a degradation trajectory prediction algorithm is used to calculate the battery health state degradation rate and remaining battery life under the corresponding operating conditions.

5. The method according to claim 1, characterized in that, Before the step of collecting multi-dimensional operating data of commercial vehicles from the vehicle end, performing operating condition scenario clustering on the multi-dimensional operating data, and obtaining scenario labels and multi-dimensional operating condition feature vectors, the method further includes: A Markov decision process framework for battery life extension control is constructed in the cloud, a reinforcement learning model is built, and a state space, action space, and reward function are defined. The state space includes battery health state, state of charge, battery gridded temperature field distribution, operating condition feature vector, and scene labels encoded by unique thermal encoding. The action space includes charging cutoff voltage adjustment, discharge power limit coefficient, and liquid cooling pump operating speed. The reward function includes a lifespan degradation penalty term, an operational efficiency benefit term, and a safety penalty term, used to balance the degree of battery aging and vehicle operating efficiency, and to solve for the Pareto optimal control solution. Based on the aforementioned Markov decision process framework, using historical operating data uploaded from the vehicle and simulation data generated by the battery digital twin model, all state samples and action samples matching the aforementioned state space and action space definitions are extracted, and the reinforcement learning model is trained offline. After the model converges, a deep reinforcement learning model capable of online reasoning is obtained.

6. The method according to claim 1, characterized in that, The process of tailoring the globally optimal control strategy based on scene labels generated and reported by the cloud from real-time clustering at the vehicle end, generating a vehicle-end control strategy, and then distributing it to the vehicle end includes: The scene labels generated and reported by the cloud based on real-time clustering at the vehicle end are used to cut out local policy parameter packages or lightweight inference models that are adapted to the scene labels from the global optimal control strategy, and these are used as the vehicle end control strategy. The vehicle-side control strategy is sent to the vehicle using a differential OTA transmission method.

7. The method according to claim 1, characterized in that, The dynamic adjustment of battery charging / discharging control parameters and battery thermal management parameters by the vehicle-side according to the vehicle-side control strategy includes: The vehicle-side control strategy is loaded by the vehicle-side battery management system, and the charging cut-off voltage adjustment amount, discharge power limit coefficient and liquid cooling pump operating speed are dynamically adjusted through real-time interpolation calculation or lightweight model inference. The vehicle continuously monitors the communication link status with the cloud. When communication with the cloud is interrupted, it automatically switches to a pre-installed local safety degradation strategy to continuously implement battery safety management.

8. A commercial vehicle battery life extension control system, characterized in that, The system includes: On the vehicle side, multi-dimensional operating data of commercial vehicles is collected, and the multi-dimensional operating data is clustered according to working conditions to obtain scene labels and multi-dimensional working condition feature vectors, which are then uploaded to the cloud. In the cloud, the system extracts microscopic aging features under corresponding operating conditions based on the scene labels and the multidimensional operating condition feature vectors through a cloud-constructed battery digital twin model, and predicts the battery health status degradation rate based on the microscopic aging features. The cloud platform is also used to generate a globally optimal control strategy adapted to the current operating conditions by iteratively optimizing and solving the problem through a trained deep reinforcement learning model based on the scene label, the multi-dimensional operating condition feature vector, and the battery health state decay rate; wherein, the globally optimal control strategy includes battery charge and discharge regulation parameters and battery thermal management parameters. The cloud is also used to tailor the global optimal control strategy based on the scene labels generated and reported by the vehicle terminal in real time clustering, generate the vehicle terminal control strategy, and send it to the vehicle terminal. The vehicle-side device is also used to dynamically adjust the battery charging and discharging parameters and battery thermal management parameters according to the vehicle-side control strategy.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to cause the processor to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method of any one of claims 1 to 7.

Images