Distributed electric vehicle battery management optimization method and system based on deep learning

Abstract

This invention relates to the field of battery management technology, specifically to a deep learning-based method and system for optimizing distributed electric vehicle battery management. The method includes: acquiring current state data of a distributed battery management system during vehicle operation; wherein the current state data includes voltage and current information of individual battery cells, temperature information of the battery module, and the total SOC value of the battery pack; inputting the current state data into a pre-trained deep residual network for feature extraction to obtain a first current health feature of the battery; wherein the first current health feature includes battery consistency status, battery aging degree, and thermal management requirement status. This invention extracts battery health features through a deep residual network, making battery status monitoring and evaluation more accurate, helping to promptly detect battery consistency anomalies and aging problems. Furthermore, by combining temperature information and battery health features, it dynamically adjusts thermal management strategies to reduce the risk of high-temperature cells.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, specifically to a distributed electric vehicle battery management optimization method and system based on deep learning. Background Technology

[0002] Currently, traditional methods typically rely on rules of thumb and simple mathematical models, which leads to inaccurate monitoring and evaluation of battery health status, easily overlooking potential anomalies and aging issues. Furthermore, traditional battery management systems cannot dynamically adjust based on real-time data and often employ fixed control strategies, making it difficult to adapt to different driving conditions and environmental changes, thus affecting battery energy utilization efficiency.

[0003] In addition, traditional methods often use simple temperature control strategies for thermal management, which fail to fully consider the temperature differences and health status of individual battery cells. This results in the failure to manage high temperature risks in a timely and effective manner, increasing battery safety hazards. Furthermore, traditional methods often rely on passive balancing or simple active balancing strategies, which are difficult to respond to performance differences between battery modules in real time, leading to a decline in the overall performance of the battery pack and a shortened lifespan. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: a deep learning-based distributed electric vehicle battery management optimization method, comprising: During vehicle operation, the current status data of the distributed battery management system is acquired; the current status data includes the voltage and current information of individual battery cells, the temperature information of battery modules, and the total SOC value of the battery pack. The current state data is input into a pre-trained deep residual network for feature extraction to obtain the first current health feature of the battery; wherein, the first current health feature includes the battery consistency state, the degree of battery aging, and the thermal management requirement state; In response to a battery being in an abnormal consistency state, the temperature difference data of each battery cell is obtained based on the temperature information of the battery module and the consistency state in the first current health feature to determine the second current health feature of the battery; wherein, the second current health feature includes a high-temperature cell distribution area and a low-temperature cell distribution area. Based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack, the first target control parameter corresponding to the battery is determined. The system acquires the current driving condition information of the vehicle and dynamically adjusts the first target control parameter based on the driving condition information to obtain the second target control parameter. The battery is then managed and controlled based on the second target control parameter to achieve distributed management optimization of each battery module in the electric vehicle.

[0005] Preferably, the current state data is input into a pre-trained deep residual network for feature extraction to obtain the battery's first current health feature, including: The voltage and current information in the current state data is denoised by variational mode decomposition to obtain a denoised state dataset. The denoised state dataset is fused with the temperature information of the battery module to obtain a fused feature matrix; The fused feature matrix is ​​input into a deep residual network for nonlinear mapping, and the output battery consistency status, battery aging degree and thermal management requirement status are used as the first current health feature.

[0006] Preferably, based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack, the first target control parameter corresponding to the battery is determined, including: In response to determining that the battery is in a state of severe aging, the trend of internal resistance change of the battery is identified based on the current state data; Obtain the first power reduction coefficient corresponding to the internal resistance change trend, and then calculate the first target control parameter by weighting the first power reduction coefficient with the total SOC value of the battery pack.

[0007] Preferably, determining the first target control parameter for the battery based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack further includes: When it is determined that the battery is in an emergency state requiring thermal management, the average temperature and temperature change rate of the battery pack are obtained based on temperature information. A second heat dissipation strategy matching the average temperature and temperature change rate is obtained from a pre-built thermal management strategy library, and the second heat dissipation strategy is modified in combination with the total SOC value of the battery pack to obtain the first target control parameter. When it is determined that the battery is in a high-temperature cell distribution area, the heat conduction path impedance from the high-temperature cell to the thermal management system is obtained based on the temperature difference data and the current state data. When the impedance exceeds a preset impedance threshold, the impedance and the current charging current of the battery are input into a pre-built battery thermal runaway prediction model, and the first target control parameter corresponding to the battery is obtained by combining the total SOC value of the battery pack.

[0008] Preferably, determining the first target control parameter for the battery based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack further includes: In response to determining that the battery is in a normal cycle state, the maximum allowable charging voltage of the battery pack is obtained from the current state data; The first target control parameter for the battery is obtained by multiplying the maximum allowable charging voltage with the total SOC value of the battery pack.

[0009] Preferably, the current vehicle driving condition information is obtained, and the first target control parameter is dynamically adjusted based on the driving condition information to obtain the second target control parameter, including: The vehicle's dynamic load level is determined based on the current vehicle operating condition information. Historical charge and discharge data of each battery module are obtained, and a remaining life prediction model for each battery module is constructed based on the historical charge and discharge data, so as to obtain a life prediction sequence based on the remaining life prediction model. The first target control parameter is adjusted downwards based on the dynamic load level, and the adjusted parameter is corrected by combining the lifetime prediction sequence to obtain the second target control parameter of the battery.

[0010] Preferably, the value of the first target control parameter is adjusted downward according to the dynamic load level, and the adjusted parameter is corrected by combining the lifetime prediction sequence to obtain the second target control parameter of the battery, including: The dynamic load level is input into a pre-trained neural network model, which outputs an initial reduction coefficient. The first target control parameter is initially adjusted based on the initial reduction coefficient to obtain the intermediate control parameter; The aging rate features are extracted from the lifetime prediction sequence, and the intermediate control parameters are corrected twice based on the aging rate features to obtain the second target control parameters.

[0011] Preferably, the battery is managed and controlled according to the second target control parameters, including: Obtain the current actual output current of the battery and compare it with the second target control parameter; When the actual output current exceeds the second target control parameter, real-time status data, real-time temperature data, real-time voltage data, and real-time current data of each battery module in the electric vehicle after equalization optimization are collected as training samples to train the deep residual network and output the load demand change rate. Based on the load demand change rate, the differential evolution algorithm is used to optimize the balance correction amount of each battery module, resulting in a balance correction optimization sequence.

[0012] Preferably, based on the load demand change rate, a differential evolution algorithm is used to optimize the balancing correction amount of each battery module, resulting in a balanced correction optimization sequence, which includes: The ratio of the load demand change rate of each battery module to its maximum discharge current is used as the compensation coefficient. The balance correction optimization sequence is weighted by compensation coefficients to obtain the balance compensation amount; Each corresponding battery module is controlled to perform three equalization optimizations based on its own equalization compensation amount, and the current state data of the battery is updated after the optimization is completed, so as to realize the distributed management optimization closed loop of each battery module in the electric vehicle.

[0013] A deep learning-based distributed electric vehicle battery management optimization system, applicable to the aforementioned deep learning-based distributed electric vehicle battery management optimization method, including: The data acquisition unit is used to acquire the current status data of the distributed battery management system during vehicle operation; the current status data includes the voltage and current information of individual battery cells, the temperature information of battery modules, and the total SOC value of the battery pack. The feature extraction unit is used to input the current state data into a pre-trained deep residual network for feature extraction to obtain the first current health feature of the battery; wherein, the first current health feature includes the battery consistency state, the degree of battery aging, and the thermal management requirement state; The state detection unit is used to respond to the battery being in an abnormal consistency state by obtaining the temperature difference data of each battery cell based on the temperature information of the battery module and the consistency state in the first current health feature to determine the second current health feature of the battery; wherein, the second current health feature includes a high temperature cell distribution area and a low temperature cell distribution area. The parameter determination unit is used to determine the first target control parameter corresponding to the battery based on the high temperature cell distribution area in the second current health feature, the battery aging degree in the first current health feature, and the total SOC value of the battery pack. The management optimization unit is used to acquire the current driving condition information of the vehicle, and dynamically adjust the first target control parameter according to the driving condition information to obtain the second target control parameter; and manage and control the battery according to the second target control parameter to realize the distributed management optimization of each battery module in the electric vehicle.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention extracts battery health characteristics through deep residual networks, making the monitoring and evaluation of battery status more accurate and helping to detect battery inconsistencies and aging problems in a timely manner. Moreover, by combining temperature information and battery health characteristics, it dynamically adjusts thermal management strategies to reduce the risk of high-temperature cells, thereby improving battery safety and lifespan. Furthermore, by monitoring the status and load demand changes of battery modules in real time, it uses differential evolution algorithms for balance optimization to ensure performance consistency between battery modules and improve the overall battery pack efficiency. This invention constructs a remaining life prediction model using historical charge and discharge data, formulates corresponding control parameters, effectively reduces the aging rate of the battery, thereby extending the battery's service life. Furthermore, by dynamically adjusting the control parameters based on vehicle driving condition information, it can better adapt to different driving conditions, improve the battery's energy utilization efficiency, and form a closed-loop management system through real-time data acquisition and deep learning model training. This system can intelligently identify and respond to changes in battery status, thereby enhancing the overall intelligence level of electric vehicles. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of the overall method in one embodiment of the present invention; Figure 2 This is a schematic diagram of the overall system architecture in one embodiment of the present invention.

[0016] In the diagram: 1. Data acquisition unit; 2. Feature extraction unit; 3. Status detection unit; 4. Parameter determination unit; 5. Management optimization unit. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 This invention provides a technical solution: a deep learning-based distributed electric vehicle battery management optimization method, comprising: S1. During vehicle operation, acquire the current status data of the distributed battery management system; the current status data includes the voltage and current information of individual battery cells, the temperature information of battery modules, and the total SOC value of the battery pack. S2. Input the current state data into a pre-trained deep residual network for feature extraction to obtain the first current health feature of the battery; wherein, the first current health feature includes the battery consistency state, battery aging degree and thermal management requirement state; S3. In response to the battery being in an abnormal consistency state, based on the temperature information of the battery module and the consistency state in the first current health feature, the temperature difference data of each battery cell is obtained to determine the second current health feature of the battery; wherein, the second current health feature includes the high temperature cell distribution area and the low temperature cell distribution area. S4. Determine the first target control parameter corresponding to the battery based on the high temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack. S5. Obtain the current driving condition information of the vehicle, and dynamically adjust the first target control parameter according to the driving condition information to obtain the second target control parameter; manage and control the battery according to the second target control parameter to achieve distributed management optimization of each battery module in the electric vehicle.

[0019] It should be noted that during the operation of the electric vehicle, the system collects the status data of the distributed battery management system in real time; this data includes the voltage and current information of each battery cell, the temperature of different battery modules, and the remaining charge (SOC) of the entire battery pack. The collected state data is input into a pre-trained deep residual network, a deep learning model specifically designed to extract data features. After processing, the system generates first current health features that reflect the battery's health status, including the battery's consistency (i.e., the performance consistency between individual cells), the degree of battery aging (showing the degradation of the battery after use), and thermal management requirements (whether the battery needs cooling or heating). When an inconsistency is detected in the battery (e.g., the performance of some battery cells is significantly lower than that of others), it obtains the temperature difference data of each battery cell based on the temperature information of the battery module and the consistency status in the first current health feature. This process helps to determine the second current health feature, which includes the distribution area of ​​high-temperature cells and low-temperature cells, revealing which cells may have overheating or overcooling problems. Based on the high-temperature cell distribution area identified in the second current health feature, the battery aging degree in the first current health feature, and the total SOC value of the battery pack, the system can determine the first target control parameters; these parameters guide how to manage the battery to optimize its performance and safety. The system acquires information about the vehicle's current driving conditions, such as speed and acceleration, and dynamically adjusts the first target control parameters based on this information to derive the second target control parameters. This step ensures that the battery management strategy can change flexibly according to the actual operating environment. By utilizing the second target control parameters, the individual battery modules within the electric vehicle are managed and optimized; this process enables distributed battery management, aiming to improve battery efficiency and extend its lifespan, while ensuring the overall performance and safety of the electric vehicle.

[0020] In an optional embodiment, the current state data is input into a pre-trained deep residual network for feature extraction to obtain a first current health feature of the battery, including: The voltage and current information in the current state data is denoised by variational mode decomposition to obtain a denoised state dataset. The denoised state dataset is fused with the temperature information of the battery module to obtain a fused feature matrix; The fused feature matrix is ​​input into a deep residual network for nonlinear mapping, and the output battery consistency status, battery aging degree and thermal management requirement status are used as the first current health feature.

[0021] It should be noted that the voltage and current information contained in the current state data may be affected by various noises, thereby reducing the accuracy of the data. To solve this problem, a method called variational mode decomposition is used to denoise these voltage and current data. After denoising, a cleaner and more reliable state dataset is obtained, which can better reflect the actual operating state of the battery. The denoised state dataset is fused with the temperature information of the battery module. This process combines two different types of data to form a more comprehensive feature matrix. This fusion can provide richer information to help subsequent analysis more accurately assess the health status of the battery. The obtained fused feature matrix is ​​input into a pre-trained deep residual network. Deep residual networks are a type of deep learning model that excels at handling complex data and performing nonlinear mappings. In this process, the network performs in-depth analysis and learning of the input data, extracting important features hidden behind the data. After processing by a deep residual network, the system outputs three key first current health characteristics: battery consistency status (indicating whether the performance of each battery cell is consistent), battery aging level (indicating the degradation of the battery during use), and thermal management demand status (indicating whether the battery needs cooling or heating under the current environment).

[0022] In an optional embodiment, a first target control parameter corresponding to the battery is determined based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack, including: In response to determining that the battery is in a state of severe aging, the trend of internal resistance change of the battery is identified based on the current state data; Obtain the first power reduction coefficient corresponding to the internal resistance change trend, and then calculate the first target control parameter by weighting the first power reduction coefficient with the total SOC value of the battery pack.

[0023] It should be noted that if the battery is detected to be in a severely aged state, this means that the battery's performance may be significantly degraded. After confirming that the battery is in a severely aged state, the system will further analyze the current status data to identify the trend of changes in the battery's internal resistance. Internal resistance is an important indicator of battery performance, which can reflect the battery's health status. By monitoring changes in internal resistance, the system can determine potential problems that may occur in the battery during the aging process, such as reduced charging efficiency or weakened discharge capacity. Based on the trend of internal resistance change, the system will calculate a parameter called the first power reduction factor; this factor indicates the degree to which the battery's usable power will be affected due to battery aging and internal resistance changes; in other words, it reflects the extent to which the battery's maximum output power is reduced in the current state. The obtained first power reduction factor is weighted and calculated with the total SOC value of the battery pack; this process takes into account the remaining capacity of the battery in order to adjust the power output of the battery more accurately; for example, when the battery SOC is low, a more conservative power control strategy may be needed to prevent over-discharge or damage to the battery.

[0024] In an optional embodiment, determining the first target control parameter corresponding to the battery based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack further includes: When it is determined that the battery is in an emergency state requiring thermal management, the average temperature and temperature change rate of the battery pack are obtained based on temperature information. A second heat dissipation strategy matching the average temperature and temperature change rate is obtained from a pre-built thermal management strategy library, and the second heat dissipation strategy is modified in combination with the total SOC value of the battery pack to obtain the first target control parameter. When it is determined that the battery is in a high-temperature cell distribution area, the heat conduction path impedance from the high-temperature cell to the thermal management system is obtained based on the temperature difference data and the current state data. When the impedance exceeds a preset impedance threshold, the impedance and the current charging current of the battery are input into a pre-built battery thermal runaway prediction model, and the first target control parameter corresponding to the battery is obtained by combining the total SOC value of the battery pack.

[0025] It should be noted that, based on the previously mentioned second current health characteristic (high temperature cell distribution area) and first current health characteristic (battery aging degree and total SOC value), the system can assess the overall health status of the battery; if the battery is found to be in a state of urgent thermal management needs, it means that the battery temperature may be too high and corresponding measures need to be taken to prevent overheating. In this case, the system calculates the average temperature and temperature change rate of the battery pack based on real-time temperature information; this information is the basis for formulating appropriate heat dissipation strategies, as it directly reflects the thermal state and cooling requirements of the battery. Find a second heat dissipation strategy that matches the current average temperature and its rate of change from a pre-built thermal management strategy library; this strategy will be used to guide how to effectively dissipate heat to reduce the battery temperature; at the same time, the heat dissipation strategy needs to be modified in combination with the total SOC value of the battery pack to ensure that the best cooling effect can be achieved under different charging states. When the battery is in a high-temperature cell distribution area, the system will further analyze the temperature difference data and current status data to obtain the impedance of the heat conduction path between the high-temperature cell and the thermal management system. This impedance value can indicate the efficiency of heat conduction. The greater the impedance, the more difficult it is to transfer heat, which may lead to local overheating. If the measured impedance value exceeds the preset threshold, it indicates insufficient heat dissipation capacity, which may lead to safety hazards. In this case, the system will input this impedance value and the current charging current into a pre-built battery thermal runaway prediction model. The model will combine the total SOC value of the battery pack to predict potential thermal runaway risks and help generate the corresponding first target control parameters.

[0026] In an optional embodiment, determining the first target control parameter corresponding to the battery based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack further includes: In response to determining that the battery is in a normal cycle state, the maximum allowable charging voltage of the battery pack is obtained from the current state data; The first target control parameter for the battery is obtained by multiplying the maximum allowable charging voltage with the total SOC value of the battery pack.

[0027] It should be noted that the overall health status of the battery is assessed based on the second current health characteristic (such as the distribution area of ​​high-temperature cells) and the first current health characteristic (battery aging degree and total SOC value). If the battery is identified as being in a normal cycle state, it indicates that the battery has not overheated or other performance problems and can be safely charged. After confirming that the battery is in a normal cycle state, the system will extract the maximum allowable charging voltage of the battery pack from the current status data. This voltage value is the highest voltage that the battery can withstand during charging. Exceeding this range may cause battery damage or safety hazards. The maximum permissible charging voltage is multiplied by the total SOC value of the battery pack; the SOC value represents the current state of charging of the battery, and the result of the product reflects the maximum charging capacity that the battery can safely accept under the current charging state.

[0028] In an optional embodiment, current vehicle driving condition information is obtained, and the first target control parameter is dynamically adjusted based on the driving condition information to obtain the second target control parameter, including: The vehicle's dynamic load level is determined based on the current vehicle operating condition information. Historical charge and discharge data of each battery module are obtained, and a remaining life prediction model for each battery module is constructed based on the historical charge and discharge data, so as to obtain a life prediction sequence based on the remaining life prediction model. The first target control parameter is adjusted downwards based on the dynamic load level, and the adjusted parameter is corrected by combining the lifetime prediction sequence to obtain the second target control parameter of the battery.

[0029] It should be noted that the system acquires real-time information on the vehicle's current driving conditions, including vehicle speed, acceleration, road conditions, etc. These factors directly affect the vehicle's dynamic load level. Based on the collected driving condition information, the system evaluates the vehicle's dynamic load level. The dynamic load level reflects the load demand of the vehicle on the battery system under specific driving conditions. For example, the dynamic load level may be higher when driving at high speed or climbing hills. Acquire historical charge and discharge data for each battery module; this data includes the battery's usage under different operating conditions, which can help identify changes in battery performance and health status; Based on historical charge and discharge data, the system builds a remaining life prediction model for each battery module. These models can analyze the battery's usage patterns and predict its future lifespan, resulting in a life prediction sequence that shows the remaining lifespan that the battery can maintain under current usage conditions. After obtaining the dynamic load level, the system will adjust the value of the first target control parameter downward according to this level; this adjustment is to prevent the battery from being over-consumed under high load conditions and to protect the battery's health. By combining the lifespan prediction sequence, the system will further refine the adjusted parameters; this step aims to ensure that the battery lifespan is maximized while meeting current driving needs, and to avoid damage caused by over-discharge or over-charging.

[0030] In an optional embodiment, the value of the first target control parameter is adjusted downward according to the dynamic load level, and the adjusted parameter is corrected by combining the lifetime prediction sequence to obtain the second target control parameter of the battery, including: The dynamic load level is input into a pre-trained neural network model, which outputs an initial reduction coefficient. The first target control parameter is initially adjusted based on the initial reduction coefficient to obtain the intermediate control parameter; The aging rate features are extracted from the lifetime prediction sequence, and the intermediate control parameters are corrected twice based on the aging rate features to obtain the second target control parameters.

[0031] It should be noted that the determined dynamic load level is input into a pre-trained neural network model; this model has been trained based on a large amount of historical data and can effectively analyze battery performance under different load conditions. The neural network outputs an initial reduction factor based on the input dynamic load level; this factor indicates how much the first target control parameter should be reduced under the current load conditions; the calculation of the reduction factor takes into account the impact of the load on battery performance, aiming to protect the battery from overuse. Based on the obtained initial reduction coefficient, the system will make a preliminary adjustment to the first target control parameter; this adjustment generates an intermediate control parameter designed to reduce the stress on the battery under high load conditions, thereby extending its service life; Features related to battery aging rate are extracted from the lifespan prediction sequence; these features can reflect the degradation of the battery over time and under different usage conditions, and further help to assess the battery's health status. Based on the extracted aging rate characteristics, the intermediate control parameters are corrected a second time. This step is to ensure that the final control parameters not only adapt to the current dynamic load, but also fully consider the aging of the battery, so as to achieve better performance and safety.

[0032] In an optional embodiment, battery management control is performed based on a second target control parameter, including: Obtain the current actual output current of the battery and compare it with the second target control parameter; When the actual output current exceeds the second target control parameter, real-time status data, real-time temperature data, real-time voltage data, and real-time current data of each battery module in the electric vehicle after equalization optimization are collected as training samples to train the deep residual network and output the load demand change rate. Based on the load demand change rate, the differential evolution algorithm is used to optimize the balance correction amount of each battery module, resulting in a balance correction optimization sequence.

[0033] It should be noted that this refers to obtaining the battery's current actual output current; this is the amount of current provided by the battery under its current operating conditions, reflecting the battery's workload. The current actual output current is compared with the previously calculated second target control parameter; if the current actual output current is greater than the second target control parameter, it means that the battery is under a load exceeding the expectation, which may cause the battery to overheat or be damaged. When the actual output current exceeds the second target control parameter, the system will collect real-time status data of each battery module in the electric vehicle after equalization optimization. This data includes important indicators such as real-time temperature, real-time voltage, and real-time current of each battery module. This information will be used as training samples for further analysis and model training. Using the collected real-time data, the system will train a deep residual network; this network can learn the performance characteristics of different battery modules and analyze changes in load demand; through training, the network can identify patterns of load demand changes and thus output a load demand change rate. Based on the obtained load demand change rate, the system will then use a differential evolution algorithm to optimize the balance correction amount of each battery module; the balance correction amount is used to adjust the power balance between battery modules to ensure that all modules perform consistently during the charging and discharging process. Through the optimization process, the system will obtain a balanced correction optimization sequence; this sequence indicates how to make corresponding adjustments to each battery module to maintain their power balance, which helps to improve the overall performance of the battery pack and extend its service life.

[0034] In an optional embodiment, based on the load demand change rate, a differential evolution algorithm is used to optimize the balancing correction amount of each battery module, resulting in a balanced correction optimization sequence, which includes: The ratio of the load demand change rate of each battery module to its maximum discharge current is used as the compensation coefficient. The balance correction optimization sequence is weighted by compensation coefficients to obtain the balance compensation amount; Each corresponding battery module is controlled to perform three equalization optimizations based on its own equalization compensation amount, and the current state data of the battery is updated after the optimization is completed, so as to realize the distributed management optimization closed loop of each battery module in the electric vehicle.

[0035] It should be noted that the ratio of the load demand change rate of each battery module to its own maximum discharge current is calculated; this ratio is called the compensation coefficient; the compensation coefficient reflects the relative capability of the battery module under the current load conditions and can help determine how much equalization correction the module needs. The calculated compensation coefficients are used to weight the previously obtained balance correction optimization sequence. In this way, the distribution of the balance correction amount can be adjusted according to the relationship between the actual load demand of each battery module and its discharge capacity, so that the power balance of each module is more reasonable. After weighted processing, the system will obtain a balance compensation amount; this amount indicates the specific adjustments that each battery module needs to make during the balancing process to ensure that all modules can maintain a relatively consistent state of charge during charging and discharging. Based on the calculated equalization compensation amount, the system will perform three rounds of equalization optimization on each battery module. Through these optimization steps, the system will gradually adjust the power of each module to achieve a better balance. This multi-optimization method can effectively eliminate performance differences caused by uneven load. After the equalization optimization is completed, the system will update the current status data of each battery module; this data includes information such as battery charge, temperature, and voltage to reflect the actual situation after optimization. Through the above steps, the system achieves a distributed management optimization closed loop for each battery module in an electric vehicle; this means that the system can continuously monitor and adjust the status of each module, thereby maintaining the efficient operation and safety of the battery pack under different working conditions.

[0036] Example 2, please refer to Figure 2 This invention provides a technical solution: a deep learning-based distributed electric vehicle battery management optimization system, applicable to the aforementioned deep learning-based distributed electric vehicle battery management optimization method, comprising: Data acquisition unit 1 is used to acquire the current status data of the distributed battery management system during vehicle operation; wherein, the current status data includes the voltage and current information of individual battery cells, the temperature information of battery modules, and the total SOC value of the battery pack; Feature extraction unit 2 is used to input the current state data into a pre-trained deep residual network for feature extraction to obtain the first current health feature of the battery; wherein, the first current health feature includes the battery consistency state, the degree of battery aging and the thermal management requirement state; The state detection unit 3 is used to respond to the battery being in an abnormal consistency state by obtaining the temperature difference data of each battery cell based on the temperature information of the battery module and the consistency state in the first current health feature to determine the second current health feature of the battery; wherein, the second current health feature includes a high temperature cell distribution area and a low temperature cell distribution area. The parameter determination unit 4 is used to determine the first target control parameter corresponding to the battery based on the high temperature cell distribution area in the second current health feature, the battery aging degree in the first current health feature, and the total SOC value of the battery pack. The management optimization unit 5 is used to acquire the current driving condition information of the vehicle, and dynamically adjust the first target control parameter according to the driving condition information to obtain the second target control parameter; and manage and control the battery according to the second target control parameter to realize the distributed management optimization of each battery module in the electric vehicle.

[0037] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A distributed electric vehicle battery management optimization method based on deep learning, characterized in that, include: During vehicle operation, the current status data of the distributed battery management system is acquired; the current status data includes the voltage and current information of individual battery cells, the temperature information of battery modules, and the total SOC value of the battery pack. The current state data is input into a pre-trained deep residual network for feature extraction to obtain the first current health feature of the battery; wherein, the first current health feature includes the battery consistency state, the degree of battery aging, and the thermal management requirement state; In response to a battery being in an abnormal consistency state, the temperature difference data of each battery cell is obtained based on the temperature information of the battery module and the consistency state in the first current health feature to determine the second current health feature of the battery; wherein, the second current health feature includes a high-temperature cell distribution area and a low-temperature cell distribution area. Based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack, the first target control parameter corresponding to the battery is determined. The system acquires the current driving condition information of the vehicle and dynamically adjusts the first target control parameter based on the driving condition information to obtain the second target control parameter. The battery is then managed and controlled based on the second target control parameter to achieve distributed management optimization of each battery module in the electric vehicle.

2. The deep learning-based distributed electric vehicle battery management optimization method according to claim 1, characterized in that, The current state data is input into a pre-trained deep residual network for feature extraction to obtain the battery's first current health features, including: The voltage and current information in the current state data is denoised by variational mode decomposition to obtain a denoised state dataset. The denoised state dataset is fused with the temperature information of the battery module to obtain a fused feature matrix; The fused feature matrix is ​​input into a deep residual network for nonlinear mapping, and the output battery consistency status, battery aging degree and thermal management requirement status are used as the first current health feature.

3. The deep learning-based distributed electric vehicle battery management optimization method according to claim 2, characterized in that, Based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack, the first target control parameters corresponding to the battery are determined, including: In response to determining that the battery is in a state of severe aging, the trend of internal resistance change of the battery is identified based on the current state data; Obtain the first power reduction coefficient corresponding to the internal resistance change trend, and then calculate the first target control parameter by weighting the first power reduction coefficient with the total SOC value of the battery pack.

4. The deep learning-based distributed electric vehicle battery management optimization method according to claim 3, characterized in that, Based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack, the first target control parameter corresponding to the battery is determined, which also includes: When it is determined that the battery is in an emergency state requiring thermal management, the average temperature and temperature change rate of the battery pack are obtained based on temperature information. A second heat dissipation strategy matching the average temperature and temperature change rate is obtained from a pre-built thermal management strategy library, and the second heat dissipation strategy is modified in combination with the total SOC value of the battery pack to obtain the first target control parameter. When it is determined that the battery is in a high-temperature cell distribution area, the heat conduction path impedance from the high-temperature cell to the thermal management system is obtained based on the temperature difference data and the current state data. When the impedance exceeds a preset impedance threshold, the impedance and the current charging current of the battery are input into a pre-built battery thermal runaway prediction model, and the first target control parameter corresponding to the battery is obtained by combining the total SOC value of the battery pack.

5. The deep learning-based distributed electric vehicle battery management optimization method according to claim 4, characterized in that, Based on the high-temperature cell distribution area in the second current health characteristic, the battery aging degree in the first current health characteristic, and the total SOC value of the battery pack, the first target control parameter corresponding to the battery is determined, which also includes: In response to determining that the battery is in a normal cycle state, the maximum allowable charging voltage of the battery pack is obtained from the current state data; The first target control parameter for the battery is obtained by multiplying the maximum allowable charging voltage with the total SOC value of the battery pack.

6. The deep learning-based distributed electric vehicle battery management optimization method according to claim 5, characterized in that, Obtain the current vehicle driving condition information, and dynamically adjust the first target control parameters based on the driving condition information to obtain the second target control parameters, including: The vehicle's dynamic load level is determined based on the current vehicle operating condition information. Historical charge and discharge data of each battery module are obtained, and a remaining life prediction model for each battery module is constructed based on the historical charge and discharge data, so as to obtain a life prediction sequence based on the remaining life prediction model. The first target control parameter is adjusted downwards based on the dynamic load level, and the adjusted parameter is corrected by combining the lifetime prediction sequence to obtain the second target control parameter of the battery.

7. The deep learning-based distributed electric vehicle battery management optimization method according to claim 6, characterized in that, The first target control parameter is adjusted downwards based on the dynamic load level, and then corrected using a lifetime prediction sequence to obtain the second target control parameter for the battery, including: The dynamic load level is input into a pre-trained neural network model, which outputs an initial reduction coefficient. The first target control parameter is initially adjusted based on the initial reduction coefficient to obtain the intermediate control parameter; The aging rate features are extracted from the lifetime prediction sequence, and the intermediate control parameters are corrected twice based on the aging rate features to obtain the second target control parameters.

8. The deep learning-based distributed electric vehicle battery management optimization method according to claim 7, characterized in that, Battery management and control are performed based on the second target control parameters, including: Obtain the current actual output current of the battery and compare it with the second target control parameter; When the actual output current exceeds the second target control parameter, real-time status data, real-time temperature data, real-time voltage data, and real-time current data of each battery module in the electric vehicle after equalization optimization are collected as training samples to train the deep residual network and output the load demand change rate. Based on the load demand change rate, the differential evolution algorithm is used to optimize the balance correction amount of each battery module, resulting in a balance correction optimization sequence.

9. The deep learning-based distributed electric vehicle battery management optimization method according to claim 8, characterized in that, Based on the load demand change rate, a differential evolution algorithm is used to optimize the balancing correction amount of each battery module, resulting in a balanced correction optimization sequence, including: The ratio of the load demand change rate of each battery module to its maximum discharge current is used as the compensation coefficient. The balance correction optimization sequence is weighted by compensation coefficients to obtain the balance compensation amount; Each corresponding battery module is controlled to perform three equalization optimizations based on its own equalization compensation amount, and the current state data of the battery is updated after the optimization is completed, so as to realize the distributed management optimization closed loop of each battery module in the electric vehicle.

10. A deep learning-based distributed electric vehicle battery management optimization system, applicable to the deep learning-based distributed electric vehicle battery management optimization method according to any one of claims 1-9, characterized in that, include: The data acquisition unit is used to acquire the current status data of the distributed battery management system during vehicle operation; the current status data includes the voltage and current information of individual battery cells, the temperature information of battery modules, and the total SOC value of the battery pack. The feature extraction unit is used to input the current state data into a pre-trained deep residual network for feature extraction to obtain the first current health feature of the battery; wherein, the first current health feature includes the battery consistency state, the degree of battery aging, and the thermal management requirement state; The state detection unit is used to respond to the battery being in an abnormal consistency state by obtaining the temperature difference data of each battery cell based on the temperature information of the battery module and the consistency state in the first current health feature to determine the second current health feature of the battery; wherein, the second current health feature includes a high temperature cell distribution area and a low temperature cell distribution area. The parameter determination unit is used to determine the first target control parameter corresponding to the battery based on the high temperature cell distribution area in the second current health feature, the battery aging degree in the first current health feature, and the total SOC value of the battery pack. The management optimization unit is used to acquire the current driving condition information of the vehicle, and dynamically adjust the first target control parameter according to the driving condition information to obtain the second target control parameter; and manage and control the battery according to the second target control parameter to realize the distributed management optimization of each battery module in the electric vehicle.

Images