A new energy battery intelligent maintenance method based on a charging pile and a medium

Abstract

The application relates to the field of new energy battery management, in particular to a new energy battery intelligent maintenance method based on a charging pile and a medium. The method comprises the following steps: obtaining data of a battery through the charging pile; uploading the data to a cloud database, matching an optimal charging parameter interval of the battery of the same model under historical working conditions; based on the matching result, extracting voltage-current curve shape characteristics, analyzing the correlation between long sequence parameters, and generating an optimal target current value considering charging efficiency and battery health; when the voltage deviation of the battery exceeds a dynamic balance threshold, dynamic balance charging control is performed through a charging pile communication interface according to the optimal target current value. The core contradiction that the prior art cannot cooperatively optimize the charging speed and the battery life is effectively solved, so that each charging can adapt to the current battery state and respond to the voltage deviation in time, and can preventively inhibit the aging trend based on historical health data, so that the battery service life is significantly prolonged while the charging efficiency is ensured.

Description

Technical Field

[0001] This application relates to the field of new energy battery management, and in particular to a method and medium for intelligent maintenance of new energy batteries based on charging piles. Background Technology

[0002] With the popularization of new energy vehicles, the performance degradation of power batteries, as core components, is becoming increasingly prominent. During the cycle of charging and discharging, batteries inevitably experience capacity decay and increased internal resistance. This process is significantly affected by charging strategies—especially the high current surge and temperature rise effect in fast charging scenarios, which accelerates the aging of electrode materials. Currently, the industry generally relies on on-board battery management systems (BMS) for charge and discharge control. However, BMS has limited local computing power and can only execute simple balancing strategies based on basic parameters such as battery cell voltage and temperature.

[0003] In existing technologies, a few solutions attempt to optimize charging strategies through cloud platforms, but these solutions have fundamental limitations: First, they rely on battery test data from laboratory environments and do not integrate multi-dimensional operating condition data from real vehicle operation (such as dynamic current fluctuations and transient ambient temperature changes); second, they lack the ability to deeply mine real-time data collected by charging piles, and can only generate static charging curves, unable to dynamically correct control commands during charging. The most critical flaw is that existing technologies cannot link the real-time heterogeneous data such as current, voltage, and temperature available from charging piles with the historical degradation characteristics of the battery for analysis, thereby achieving a dynamic balance between immediate charging efficiency and long-term health maintenance. This results in the charging process either sacrificing speed for battery life or accelerating battery aging in pursuit of speed, making it difficult to balance the contradiction between the two. Summary of the Invention

[0004] To address the problem that existing technologies cannot link and analyze the real-time heterogeneous data (current, voltage, temperature, etc.) obtained from charging piles with the historical degradation characteristics of batteries, this application provides a smart maintenance method and medium for new energy batteries based on charging piles. The method provided in this application includes the following steps:

[0005] When charging new energy batteries, real-time charging current, power, voltage, external ambient temperature and internal core temperature data of the battery are obtained through charging piles;

[0006] The data is uploaded to a cloud database to match the optimal charging parameter range of batteries of the same model under historical operating conditions. The matching process uses a multi-dimensional parameter matching matrix constructed by combining historical battery operating data and aging test results.

[0007] Based on the matching results, the morphological features of the voltage-current curves are extracted, and the correlation between long-sequence parameters is analyzed to generate the optimal target current value that balances charging efficiency and battery health.

[0008] When the voltage deviation of the battery exceeds the dynamic balancing threshold, dynamic balancing charging control is performed through the charging pile communication interface according to the optimal target current value.

[0009] Specifically, the multi-dimensional parameter matching matrix constructed by combining historical battery operating data and aging test results includes:

[0010] Based on the battery's charge-discharge cycle count, capacity decay rate, and internal resistance change data, a two-dimensional parametric plane is constructed with current as the horizontal axis and voltage as the vertical axis.

[0011] In the two-dimensional parameter plane, the temperature dimension is superimposed to form a three-dimensional parameter space;

[0012] Based on the battery aging test results, the safety boundary surfaces under different SOH health states are marked in the three-dimensional parameter space;

[0013] The optimal charging parameter range is located in the overlapping region between the safety boundary surface and the maximum charging efficiency isosurface.

[0014] Specifically, the method for generating the optimal target current value that balances charging efficiency and battery health includes:

[0015] The local impulse response slope and the variance of voltage plateau fluctuation during the constant current phase of the voltage-current curve are extracted using a convolutional neural network.

[0016] The long sequence parameter association is established through the self-attention mechanism of the Transformer model. The long sequence parameter association includes: the cross-influence weight of the ambient temperature sequence and the SOC change rate and the constraint coefficient of the historical capacity decay trend on the current maximum allowable current.

[0017] By fusing the outputs of the convolutional neural network and the Transformer model, the optimal current value is solved using an objective function, which is:

[0018]

[0019] in, or For charging efficiency, ΔT For the temperature rise, SOH For a healthy state, α , β , c These are weighting coefficients, and α + β + c =1;

[0020] The objective function is optimized under constraints using the gradient descent method. These constraints include voltage constraints, temperature constraints, and state of health (SOH) constraints.

[0021] Specifically, the method further includes, during the dynamic equalization charging control, adjusting the dynamic equalization threshold based on the SOC (State of Charge), wherein the dynamic equalization threshold adjustment includes:

[0022] When the SOC (State of Charge) is <20%, the equalization threshold is relaxed to ±5%.

[0023] When 20% ≤ the SOC state of charge ≤ 95%, the balancing threshold is limited to ±2%;

[0024] When the SOC state of charge is >95%, the equalization threshold is limited to <1%.

[0025] Specifically, the method further includes correcting the optimal target current value during the dynamic equalization charging control, the correction including:

[0026] For batteries with a state of health (SOH) below the health threshold, reduce their balancing current while satisfying the formula:

[0027] ,

[0028] in, I bal This represents the adjusted target current value, where k is the proportional coefficient used to prevent overcharging. I opt −I real | represents the absolute deviation between the target current and the actual current, where the target current is the optimal target current value. SOH i This is the value of the battery's State of Health (SOH).

[0029] Specifically, the method further includes:

[0030] After performing dynamic equalization charging control, battery response data is collected in real time;

[0031] Calculate actual charging efficiency or actual With prediction efficiency or pred deviation Δ or ;

[0032] When Δ or >Δ or max Then, the voltage-current curve features are re-extracted, the historical data in the parameter matching matrix is ​​updated, and the optimal target current value is regenerated to establish the feedback coefficient. It is used to correct the target current value in the next control cycle, where the preset upper limit of the charging efficiency deviation allowed by the system is.

[0033] In addition to the above, this application also includes a computer-readable storage medium storing a computer program that can be executed by at least one processor to implement the method described above.

[0034] This application has the following technical advantages:

[0035] This invention collects battery operating data in real time through charging piles and uploads it to the cloud. It intelligently integrates historical operating conditions and aging characteristics to generate the optimal charging strategy and dynamically adjusts the balancing current during the charging process. This effectively solves the core contradiction that existing technologies cannot coordinate the optimization of charging speed and battery life. It enables each charge to adapt to the current battery state and respond to voltage deviations in a timely manner, while also preventing aging trends based on historical health data. Ultimately, it significantly extends battery life while ensuring charging efficiency. Attached Figure Description

[0036] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. Several embodiments of this application are illustrated in the drawings by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts.

[0037] Figure 1 This is a flowchart of a smart maintenance method for new energy batteries based on charging piles, as described in an embodiment of this application. Detailed Implementation

[0038] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0039] To address battery degradation, real-time monitoring is necessary during the battery charging process, particularly for State of Charge (SOC) and State of Health (SOH). Conventional SOC and SOH testing typically requires laboratory testing using the vehicle's Battery Management System (BMS). This embodiment, however, utilizes a data acquisition layer integrated into the charging station. This layer directly acquires high-precision raw data streams during charging via sensors and their communication modules, achieving a sampling frequency of up to 1000 times per second. This data forms the foundation for a "digital twin" of the battery.

[0040] In this embodiment, the acquired data includes: real-time charging current, power, voltage, external ambient temperature, and internal core temperature, such as... Figure 1 As shown, the steps of the intelligent maintenance method for new energy batteries based on this data include:

[0041] The data is uploaded to a cloud database to match the optimal charging parameter range of batteries of the same model under historical operating conditions. The matching process uses a multi-dimensional parameter matching matrix constructed by combining historical battery operating data and aging test results.

[0042] Based on the matching results, the morphological features of the voltage-current curves are extracted, and the correlation between long-sequence parameters is analyzed to generate the optimal target current value that balances charging efficiency and battery health.

[0043] When the voltage deviation of the battery exceeds the dynamic balancing threshold, dynamic balancing charging control is performed through the charging pile communication interface according to the optimal target current value.

[0044] In this embodiment, the multidimensional parameter matching matrix uses the charge-discharge cycle count, capacity decay rate, and internal resistance change data from 100,000 sets of historical charge-discharge cycle data to construct a two-dimensional parameter plane with current as the horizontal axis and voltage as the vertical axis, and then superimposes the temperature dimension to form a three-dimensional parameter space. For example, the safety boundary surface of a certain ternary lithium battery after 500 cycles in a 25℃ environment shows that if the charging current exceeds 120A, it will trigger the risk of lithium plating, while the lower limit of the current corresponding to the highest charging efficiency isosurface (η>95%) is 80A. The overlapping area of ​​the two constitutes the optimal current range [80A, 120A] under this operating condition.

[0045] After matching is completed, the system extracts the voltage-current curve morphology features of the current charging process. In this embodiment, instead of using the simple slope calculation of the traditional BMS, the system uses a convolutional neural network to extract the local impulse response slope of the voltage-current curve and the voltage plateau fluctuation variance during the constant current stage.

[0046] The long sequence parameter association is established through the self-attention mechanism of the Transformer model. The long sequence parameter association includes: the cross-influence weight of the ambient temperature sequence and the SOC change rate and the constraint coefficient of the historical capacity decay trend on the current maximum allowable current.

[0047] By fusing the outputs of the convolutional neural network and the Transformer model, the optimal current value is solved using an objective function, which is:

[0048]

[0049] in, or For charging efficiency, ΔT For the temperature rise, SOH For a healthy state, α , β , c These are weighting coefficients, and α + β + c =1;

[0050] The objective function is optimized under constraints using the gradient descent method. These constraints include voltage constraints, temperature constraints, and state of health (SOH) constraints.

[0051] When the system detects that the voltage deviation of a single cell in the battery pack exceeds the dynamic balancing threshold, it immediately triggers the core control action. During the control process, the threshold is dynamically adjusted based on the state of charge (SOC). When the SOC is <20%, it is in the deep discharge stage, and the balancing threshold is relaxed to ±5% to avoid over-balancing at low charge levels that could lead to charging interruption. When 20% ≤ SOC ≤ 95%, the balancing threshold is limited to ±2%. When the SOC > 95%, the balancing threshold is limited to <1%.

[0052] When performing the dynamic equalization charging control, it is also necessary to correct the optimal target current value, and the correction includes:

[0053] For batteries with a state of health (SOH) below the health threshold, reduce their balancing current while satisfying the formula:

[0054] ,

[0055] in, I bal This represents the adjusted target current value, where k is the proportional coefficient used to prevent overcharging. I opt −I real | represents the absolute deviation between the target current and the actual current, where the target current is the optimal target current value. SOH i This is the value of the battery's State of Health (SOH).

[0056] After executing dynamic balancing control, the system initiates a multi-source data fusion verification mechanism to collect battery response data in real time;

[0057] Calculate actual charging efficiency or actual With prediction efficiency or pred deviation Δ or ;

[0058] When Δ or >Δ or max Then, the voltage-current curve features are re-extracted, the historical data in the parameter matching matrix is ​​updated, and the optimal target current value is regenerated to establish the feedback coefficient. It is used to correct the target current value in the next control cycle, where the preset upper limit of the charging efficiency deviation allowed by the system is.

[0059] By continuously collecting battery data from different regions, vehicle models, and service years, a closed-loop control system was constructed, encompassing data acquisition, feature extraction, decision optimization, and execution feedback. After 200 consecutive fast-charging cycles, the battery capacity retention rate using this solution was significantly improved compared to traditional BMS management solutions, and charging time was greatly reduced, effectively resolving the conflict between battery health and charging efficiency in fast-charging scenarios.

[0060] In addition to the above, the system in this embodiment also includes an application layer. This layer converts the State of Charge (SOC) and State of Health (SOH) into reports and dynamic curves for output. The SOH health status can also serve as a guideline for determining whether a battery needs repair. The SOH threshold for repair varies depending on the type of battery. For example, the threshold for lithium-ion batteries is SOH ≤ 80% of rated capacity; for lead-acid batteries, it is SOH ≤ 50% of rated capacity, at which point it can be considered severely sulfated or corroded; and for nickel-metal hydride batteries, it is SOH ≤ 70% of rated capacity. Furthermore, the need for battery repair can also be determined by phenomena such as increased internal resistance, abnormal charging voltage, excessively high self-discharge rate, abnormal temperature characteristics, and physical damage. It should be noted that due to differences in material properties, aging mechanisms, and failure modes, the repair strategies for different types of batteries are not entirely the same. This embodiment provides a comparative analysis of the core differences between different types of batteries in the table below.

[0061] Key differences and comparative analysis of different battery types

[0062]

[0063] Obviously, the embodiments described above are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0064] It should be understood that when the terms "first," "second," etc., are used in the claims, description, and drawings of this application, they are only used to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" used in the description and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

Claims

1. A new energy battery intelligent maintenance method based on charging piles, characterized in that, The method comprises: When charging the new energy battery, the real-time charging current, power, voltage, external environment temperature and internal core temperature data of the battery are obtained through the charging pile; The data is uploaded to the cloud database, and the optimal charging parameter interval of the battery under the same historical working condition is matched, and the matching process uses a multi-dimensional parameter matching matrix constructed by combining the battery historical operation data and the aging test results; The multi-dimensional parameter matching matrix constructed by combining the battery historical operation data and the aging test results is based on the charge-discharge cycle number, capacity attenuation rate and internal resistance change data of the battery, and a two-dimensional parameter plane with current as the horizontal axis and voltage as the vertical axis is constructed; In the two-dimensional parameter plane, a temperature dimension is superimposed to form a three-dimensional parameter space; According to the battery aging test results, the safety boundary surface under different SOH health states is labeled in the three-dimensional parameter space; The optimal charging parameter interval is located in the overlapping area between the safety boundary surface and the highest charging efficiency equivalent surface; Based on the matching result, the voltage-current curve shape feature is extracted, and the correlation between long sequence parameters is analyzed to generate the optimal target current value considering charging efficiency and battery health; When the voltage deviation of the battery exceeds the dynamic balancing threshold, dynamic balancing charging control is performed according to the optimal target current value through the charging pile communication interface.

2. The method of claim 1, wherein, The method for generating the optimal target current value considering charging efficiency and battery health comprises: Local impulse response slope and constant current stage voltage platform fluctuation variance of the voltage-current curve are extracted by using a convolutional neural network; The long sequence parameter correlation is established by the self-attention mechanism of the Transformer model, and the long sequence parameter correlation includes: the cross-influence weight of the environmental temperature sequence and the SOC change rate and the constraint coefficient of the historical capacity attenuation trend on the current maximum allowable current; The outputs of the convolutional neural network and the Transformer model are fused, and the optimal current value is solved by a target function, and the target function is: wherein, η is the charging efficiency, ΔT is the temperature rise, SOH is the state of health, α , β , γ is a weighting factor, and α + β + γ = 1. The target function is optimized under the constraint condition by the gradient descent method, and the constraint condition includes voltage constraint, temperature constraint and SOH health state constraint.

3. The method of claim 1, wherein, The method further comprises, when performing the dynamic balancing charging control, adjusting the dynamic balancing threshold according to the SOC state of charge, and the dynamic balancing threshold adjustment comprises: When the SOC state of charge is less than 20%, the balancing threshold is relaxed to ±5%; When 20%≤the SOC state of charge≤95%, the balancing threshold is limited to ±2%; When the SOC state of charge is greater than 95%, the balancing threshold is limited to less than 1%.

4. The method of claim 1, wherein, The method further comprises, when performing the dynamic balancing charging control, correcting the optimal target current value, and the correction comprises: For the battery with SOH health state lower than the health threshold, the balancing current is reduced, and the formula is satisfied: ， wherein, I bal represents an adjusted target current value, k is a proportional coefficient for preventing overcharging, and I opt −I real represents an absolute deviation of the target current from the actual current, the target current being the optimal target current value, SOH i is a value of the SOH state of health of the battery.

5. The method of claim 1, wherein, The method further comprises: After performing the dynamic balancing charging control, the battery response data is collected in real time; Calculate actual charging efficiency η actual With prediction efficiency η pred deviation Δ η ; When Δ η > Δ η max , re-extract the voltage-current curve characteristics, update the historical data in the parameter matching matrix, re-generate the optimal target current value, and establish a feedback coefficient for correcting the target current value in the next control period, wherein the preset system allows the upper limit of the charging efficiency deviation.

6. A computer readable storage medium characterized by The computer readable storage medium stores a computer program, which can be executed by the at least one processor to implement the method according to any one of claims 1 to 5.

Images