A charging regulation method and system for repairing SEI film of a lithium iron phosphate battery

Abstract

This application relates to a charging regulation method and system for repairing the SEI film in lithium iron phosphate batteries, belonging to the field of battery regulation technology. The method includes: constructing an SEI state identification model and an SEI loss prediction plugin on a cloud server based on historical SEI film loss data of similar lithium iron phosphate batteries; deploying the identification model to the battery management unit; collecting a first cell feature data sequence via sensors at a preset frequency, inputting it into the identification model of the battery management unit, and outputting the SEI state identification result; if the state is one of loss, collecting a second cell feature data sequence and uploading it to the cloud server, calling the prediction plugin, and outputting the predicted loss type and degradation coefficient; based on the optimized battery charging repair strategy, outputting an adaptation scheme and controlling battery charging to complete the active repair of the SEI film. This invention solves the problem that traditional methods struggle to monitor the loss in the dynamic evolution of the SEI film in real time, leading to poor repair results and affecting battery life, safety, and efficiency.

Description

Technical Field

[0001] This application relates to the field of battery regulation, and in particular to a charging regulation method and system for repairing the SEI film of lithium iron phosphate batteries. Background Technology

[0002] With the rapid development of new energy technologies, the performance stability and lifespan of lithium iron phosphate batteries have an increasingly significant impact on energy storage efficiency. Optimizing battery management to cope with the dynamic changes in the SEI film has become a key technical issue for improving battery reliability. Currently, traditional battery management methods rely solely on fixed charging strategies, which are difficult to adapt to the complex dynamic evolution of the SEI film.

[0003] Existing battery management methods lack accurate identification and analysis of SEI film loss status, resulting in insufficient targeted adjustment of repair strategies. This not only makes it difficult to effectively repair the SEI film, but also exacerbates battery performance degradation, affecting battery lifespan, safety performance and energy conversion efficiency, and increasing the complexity and cost of battery maintenance. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a charging regulation method and system for repairing the SEI film of lithium iron phosphate batteries. This method improves upon the shortcomings of traditional methods, which lack accurate identification of the loss state and targeted repair strategies when dealing with the dynamic evolution of the SEI film, resulting in poor SEI film repair performance, shortened battery life, and compromised safety and efficiency.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] In a first aspect, embodiments of this application provide a charging regulation method for repairing the SEI film of a lithium iron phosphate battery, the method comprising:

[0007] On the cloud server, an SEI state identification model and an SEI loss prediction plugin are constructed based on historical experimental data of SEI film loss of the same type of lithium iron phosphate battery, and the SEI state identification model is deployed to the battery management unit.

[0008] According to the preset data acquisition frequency, the first cell feature data sequence is collected by the micro sensor configured on the cell, input into the SEI state recognition model of the battery management unit for analysis, and output the SEI state recognition result.

[0009] If the SEI state identification result is a loss state, the second cell feature data sequence is collected by the micro sensor according to the preset data collection frequency and uploaded to the cloud server. The SEI loss prediction plugin is called to perform SEI film loss inference based on the second cell feature data sequence and output the predicted loss type and predicted degradation coefficient.

[0010] Based on the predicted loss type and predicted degradation coefficient, the battery charging repair strategy is optimized, an adapted charging scheme is output, and battery charging control is performed according to the adapted charging scheme to complete the active repair of the SEI film.

[0011] Secondly, embodiments of this application provide a charging regulation system for repairing the SEI film of a lithium iron phosphate battery, the system comprising:

[0012] The SEI model building and deployment module is used to build an SEI state identification model and an SEI loss prediction plugin on a cloud server based on historical experimental data of SEI film loss of the same type of lithium iron phosphate battery, and to deploy the SEI state identification model to the battery management unit.

[0013] The SEI state monitoring module is used to collect the first cell feature data sequence through the micro sensor configured on the cell according to the preset data acquisition frequency, input it into the SEI state recognition model of the battery management unit for analysis, and output the SEI state recognition result.

[0014] The SEI loss inference module is used to collect the second cell feature data sequence through the micro sensor according to the preset data acquisition frequency if the SEI state identification result is a loss state, and upload it to the cloud server. It then calls the SEI loss prediction plugin to perform SEI film loss inference based on the second cell feature data sequence and outputs the predicted loss type and predicted degradation coefficient.

[0015] The charging repair control module is used to optimize the battery charging repair strategy based on the predicted loss type and predicted degradation coefficient, output an adaptive charging scheme, and control the battery charging according to the adaptive charging scheme to complete the active repair of the SEI film.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] This application proposes a charging regulation method and system for repairing the SEI film in lithium iron phosphate batteries. By introducing a cloud-edge collaborative architecture and combining collaborative operations of SEI film state identification, loss inference, and charging repair strategy optimization, adaptive repair management of the SEI film in lithium iron phosphate batteries is achieved. First, on a cloud server, an SEI state identification model and an SEI loss prediction plugin are constructed based on historical experimental data of SEI film loss in similar batteries, and the SEI state identification model is deployed to the battery management unit. Then, a cell characteristic data sequence is collected through a micro-sensor, and the SEI state identification model is used to determine the SEI film state. If it is in a loss state, more comprehensive data is collected and uploaded to the cloud, and the SEI loss prediction plugin is called to infer the loss type and degradation coefficient. Finally, based on the inference results, a repair cycle analysis is performed, and charging parameter adjustment thresholds are obtained. The optimal charging scheme is sought with the goal of minimizing the SEI film degradation coefficient and side reaction rate, and an adaptive charging scheme is obtained to achieve active repair of the SEI film.

[0018] This application's technical solution addresses the issues of insufficient targeting and low efficiency in SEI film repair in traditional lithium iron phosphate battery management by integrating a layered computing architecture that combines cloud and edge collaboration, collecting and analyzing multi-dimensional feature data, identifying and predicting state based on deep learning models, and optimizing charging and repair strategies with multi-objective optimization. It achieves adaptive repair of the SEI film, providing technical support for improving battery performance and lifespan. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic flowchart of a charging regulation method for repairing the SEI film of a lithium iron phosphate battery, provided in an embodiment of this application;

[0021] Figure 2 This is a schematic diagram of a charging control system for repairing the SEI film of a lithium iron phosphate battery, provided in an embodiment of this application.

[0022] The components represented by each number in the attached diagram are explained below:

[0023] SEI Model Construction and Deployment Module 01, SEI State Monitoring Module 02, SEI Loss Inference Module 03, and Charging Repair and Control Module 04. Detailed Implementation

[0024] This application provides a charging regulation method and system for repairing the SEI film of lithium iron phosphate batteries, which solves the technical problems in the prior art that lack accurate monitoring and analysis of the dynamic evolution process of the SEI film, and cannot formulate appropriate repair strategies according to its loss type and degree of degradation, resulting in poor SEI film repair effect and affecting battery life, safety and efficiency.

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0027] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0028] Example 1, as shown in the appendix Figure 1 As shown, this application provides a charging regulation method for repairing the SEI film of a lithium iron phosphate battery, the method comprising the following steps:

[0029] S110: On the cloud server, an SEI state identification model and an SEI loss prediction plugin are constructed based on historical experimental data of SEI film loss of the same type of lithium iron phosphate battery, and the SEI state identification model is deployed to the battery management unit.

[0030] In this embodiment of the application, in the scenario of SEI film state management of lithium iron phosphate batteries, in order to achieve accurate identification and loss analysis of SEI film state, it is necessary to construct an SEI state identification model and an SEI loss prediction plugin to support the formulation of subsequent repair solutions.

[0031] Specifically, the first step is to configure simplified cell characteristic monitoring indicators, including three core monitoring indicators: temperature, voltage, and impedance, in order to reduce the data acquisition and calculation pressure at the edge and achieve a rapid preliminary judgment of the SEI film status.

[0032] Furthermore, based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, samples were collected according to the above three types of simplified cell characteristic monitoring indicators, and these were summarized to obtain a sample simplified cell characteristic data sequence set. At the same time, the historical SEI state under each sequence was obtained to form a sample SEI state set.

[0033] The historical SEI state is either in a normal state or a damaged state, reflecting whether there is damage to the SEI film at present, and providing a basis for judging whether to conduct subsequent loss analysis.

[0034] Based on this, a simplified set of sample cell feature data sequences is used as input, and a sample SEI state set is used as output labels. These are then fed into a backpropagation (BP) neural network for training. By iteratively optimizing the network parameters until the model converges, an SEI state recognition model capable of rapidly identifying the SEI film state is generated.

[0035] The SEI state recognition model can accurately determine whether the SEI film is in a normal state or a loss state based on the input simplified cell feature data.

[0036] Furthermore, a cell characteristic monitoring index including temperature, voltage, current, impedance, and carbon dioxide concentration is configured. Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, data is collected according to the above five cell characteristic monitoring indexes to obtain a sample cell characteristic data sequence set.

[0037] Simultaneously, the historical loss type and historical degradation coefficient of the SEI film corresponding to the characteristic data sequence of each sample cell are recorded to form a sample loss type set and a sample degradation coefficient set, so as to comprehensively characterize the loss details of the SEI film.

[0038] Based on this, using the sample cell feature data sequence set as input and the sample loss type set and sample degradation coefficient set as output, an SEI loss prediction plugin is constructed and trained based on a graph neural network, enabling it to infer the loss type and degradation coefficient of the SEI film based on the input feature data.

[0039] Finally, the trained SEI state recognition model is deployed to the battery management unit, while the SEI loss prediction plugin is stored on the cloud server, thus building a cloud-edge collaborative SEI film state analysis architecture, providing reliable model support for the subsequent development of repair solutions for SEI film loss.

[0040] Step S110 in the method provided in this application embodiment includes:

[0041] Configure simplified cell characteristic monitoring indicators, wherein the simplified cell characteristic monitoring indicators include temperature, voltage and impedance;

[0042] Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a set of simplified cell feature data sequences was collected according to the simplified cell feature monitoring index, and the historical SEI state under different sample simplified cell feature data sequences was obtained to obtain a sample SEI state set, where the historical SEI state is a normal state or a loss state.

[0043] Using the sample simplified cell feature data sequence set and sample SEI state set, a BP neural network is trained until convergence to generate an SEI state recognition model.

[0044] Configure cell characteristic monitoring indicators, wherein the cell characteristic monitoring indicators include at least temperature, voltage, current, impedance and carbon dioxide concentration;

[0045] Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a set of sample cell characteristic data sequences was collected according to the cell characteristic monitoring index, and the historical loss type and historical degradation coefficient of SEI film under different sample cell characteristic data sequences were obtained to obtain a set of sample loss type and a set of sample degradation coefficient, wherein the degradation coefficient is positively correlated with the degree of loss.

[0046] Using the sample cell feature data sequence set, sample loss type set, and sample degradation coefficient set as training data, an SEI loss prediction plugin is constructed based on a graph neural network.

[0047] In this embodiment of the application, in order to achieve efficient monitoring and accurate analysis of the SEI film state of lithium iron phosphate batteries, it is necessary to construct a basic analysis framework by configuring core monitoring indicators.

[0048] Specifically, a simplified set of cell characteristic monitoring indicators was first configured, with temperature, voltage, and impedance identified as the core monitoring items. These three types of monitoring indicators can intuitively reflect the energy state and internal electrochemical characteristics of the cell.

[0049] For example, an abnormally high temperature may indicate a localized rupture of the SEI film triggering side reactions, voltage fluctuations may be associated with changes in SEI film stability, and impedance changes directly reflect the susceptibility of ion conduction pathways. Selecting these three types of cell characteristic indicators can reduce data acquisition and edge computing load while ensuring monitoring effectiveness, thus adapting to the lightweight operation requirements of the battery management unit.

[0050] Furthermore, based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a set of simplified cell characteristic data sequences was collected according to the above three categories of simplified cell characteristic monitoring indicators.

[0051] For example, for a batch of lithium iron phosphate batteries that have been cycled 1000 times, time-series data of temperature (e.g., 25℃, 32℃), voltage (e.g., 3.2V, 3.0V), and impedance (e.g., 5mΩ, 8mΩ) are recorded once per hour to form a continuous feature sequence. At the same time, the SEI film state at the corresponding time is detected by equipment such as an electrochemical workstation and marked as normal state or loss state. Finally, the data are summarized to form a sample SEI state set.

[0052] In this context, a normal SEI film condition refers to a film with an intact structure, uniform thickness, and stable performance. This effectively isolates the electrolyte from direct reaction with the electrode materials, ensuring stable lithium-ion transport between the electrodes. Under these conditions, characteristic data such as temperature, voltage, and impedance are within normal fluctuation ranges. For example, the temperature is maintained at 25-30℃, the voltage is stable at around 3.2V, and the impedance remains at 5-6mΩ.

[0053] Conversely, SEI film loss status encompasses situations where the SEI film undergoes cracking, peeling, abnormal thickening or thinning, leading to a weakened protective effect and potentially triggering side reactions such as electrolyte decomposition and active material loss. Corresponding characteristic data will show significant anomalies. For example, a sudden temperature rise above 35°C, voltage fluctuations exceeding 0.2V, and impedance increases to 8mΩ or higher.

[0054] Based on this, a simplified cell feature data sequence set is used as input and a sample SEI state set is used as output label. A BP neural network is used as the basic framework to train and generate an SEI state recognition model.

[0055] Specifically, a backpropagation (BP) neural network framework consisting of an input layer, hidden layers, and an output layer is constructed. The input layer has three neurons, corresponding to three simplified cell feature data: temperature, voltage, and impedance. The hidden layer has two layers, each containing 16 neurons, and uses the ReLU activation function to enhance the model's nonlinear fitting ability and avoid the gradient vanishing problem. The output layer has two neurons, corresponding to the normal state and the loss state of the SEI membrane, and uses the Softmax activation function to output the state probability distribution.

[0056] First, the simplified cell feature data sequence set of the samples is standardized (e.g., normalized to the 0-1 range) and divided into training and validation sets in an 8:2 ratio. For example, 4000 sets are selected from 5000 sets as the training set for iterative optimization of model parameters, and 1000 sets are selected as the validation set for real-time monitoring of the model's generalization ability.

[0057] Secondly, during training, the deviation between the predicted state and the actual state (sample SEI state set) is calculated using the cross-entropy loss function, and the network weights are updated through the Adam optimizer. The initial learning rate is set to 0.001, and it decays to half of the current learning rate every 100 iterations to balance the model's convergence speed and accuracy.

[0058] Furthermore, when the state recognition accuracy of the validation set remains stable above 95% for 20 consecutive rounds (e.g., in a certain round of validation, 952 out of 1000 samples have recognition results that are consistent with the actual state), and the loss function value tends to stabilize (fluctuation amplitude is less than 0.001), the model is determined to have converged, training is stopped, and the final SEI state recognition model is generated.

[0059] Ultimately, the generated SEI state identification model can quickly output the determination result of whether the SEI film is in a normal state or a loss state based on the input temperature, voltage, and impedance time series data.

[0060] For example, when a set of time-series data for temperature, voltage, and impedance are “26℃, 3.2V, 5.5mΩ”, after inputting them into the SEI state recognition model, the model, through the computational processing of neurons in each layer, outputs a probability of 98% for the normal state and a probability of 2% for the damaged state, and the judgment result is a normal state, which is consistent with the actual detection result of the SEI membrane structure being intact at that moment.

[0061] In addition, after inputting another set of time series data "34℃, 2.9V, 9.2mΩ", the SEI state recognition model output a loss state probability of 96%, which corresponds to the actual measured situation of local rupture of the SEI film, thus verifying the recognition accuracy of the model.

[0062] At the same time, it is equipped with cell characteristic monitoring indicators including temperature, voltage, current, impedance and carbon dioxide concentration.

[0063] Among them, current can reflect charge transfer efficiency, and carbon dioxide concentration can indirectly characterize the gaseous products generated by the decomposition of SEI membrane. Combining these two with other indicators can more comprehensively depict the dynamic evolution of SEI membrane.

[0064] Furthermore, based on the historical experimental data, a sample cell feature data sequence set is collected, and the corresponding historical loss type and historical degradation coefficient of the SEI film are obtained simultaneously to form a sample loss type set and a sample degradation coefficient set. Using this as training data, an SEI loss prediction plugin is constructed based on a graph neural network.

[0065] Specifically, firstly, according to the configured cell characteristic monitoring indicators (temperature, voltage, current, impedance, carbon dioxide concentration), time-series data are extracted from the historical experimental data of SEI film loss of similar lithium iron phosphate batteries to form a sample cell characteristic data sequence set. For example, the changes of the above indicators of a certain battery during the cycling process are recorded every 5 minutes and summarized into a continuous characteristic sequence.

[0066] Secondly, through physical characterization (such as electron microscopy observation) and chemical analysis (such as gas detection), the historical loss type (such as physical pyrolysis, chemical corrosion, and overgrowth) and historical degradation coefficient (values ​​from 0 to 1, where 0 indicates no loss and 1 indicates severe loss) of the SEI film corresponding to each sample sequence are determined, and the sample loss type set and sample degradation coefficient set are compiled.

[0067] The degradation coefficient is positively correlated with the degree of loss; that is, the more severe the loss, the larger the degradation coefficient. For example, when the SEI film has large-area cracking, the degradation coefficient may reach 0.8, while it may be 0.3 when there is slight corrosion.

[0068] Finally, the sample cell feature data sequence set is used as input, and the sample loss type set and sample degradation coefficient set are used as output to construct a graph neural network model for training. The graph convolutional layer captures the correlation between features (such as the synergistic effect of current change and temperature rise on the SEI film) until the model converges, generating an SEI loss prediction plugin.

[0069] For example, in a certain sample cell characteristic data sequence, the temperature continuously rises to 38°C, the current fluctuation range reaches 15%, and the carbon dioxide concentration rises to 100ppm. After the SEI loss prediction plugin inputs this sequence, the output loss type is "chemical corrosion" with a degradation coefficient of 0.7, which is consistent with the actual detected corrosion state of the SEI film caused by electrolyte decomposition.

[0070] This step, through the layered construction of an SEI condition identification model and an SEI loss prediction plugin, forms a complete technical system from initial condition assessment to in-depth loss analysis. The lightweight SEI condition identification model is deployed on the battery management unit, enabling real-time and rapid monitoring of the SEI film condition; while the complex SEI loss prediction plugin resides on the cloud server, accurately inferring the loss type and degree of degradation. The two work together in a cloud-edge collaborative architecture, ensuring both real-time monitoring and improved analysis accuracy, providing solid model support for subsequent adaptive repair of the SEI film.

[0071] S120: According to the preset data acquisition frequency, the first cell feature data sequence is acquired through the micro sensor configured on the cell, input into the SEI state recognition model of the battery management unit for analysis, and output the SEI state recognition result;

[0072] In this embodiment of the application, in the scenario of real-time monitoring of the SEI film state of lithium iron phosphate batteries, in order to grasp the state changes of the SEI film in a timely manner, it is necessary to collect feature data through sensors and perform rapid analysis in combination with the deployed SEI state recognition model to achieve real-time judgment of the SEI film state.

[0073] Specifically, the process begins by activating the micro-sensors installed on the battery cell according to the preset data acquisition frequency and the configured simplified cell characteristic monitoring indicators (temperature, voltage, impedance).

[0074] These micro-sensors capture the temperature changes, voltage fluctuations, and impedance values ​​of the target battery cell in real time during operation, and organize this data in chronological order to form the first battery cell characteristic data sequence.

[0075] Furthermore, the acquired first cell feature data sequence is directly input into the SEI state recognition model deployed in the battery management unit. This model uses a built-in BP neural network algorithm to quickly analyze the input feature data and determine the current state of the SEI film.

[0076] Finally, the SEI state identification model outputs the SEI state identification result, that is, whether the SEI film is in a normal state or a damaged state, providing a direct basis for whether to initiate the deep damage analysis and repair process.

[0077] Step S120 in the method provided in this application embodiment includes:

[0078] According to the preset data acquisition frequency and the simplified cell feature monitoring indicators, the micro sensors configured on the cell are activated to acquire the first cell feature data sequence of the target cell;

[0079] The first cell feature data sequence is input into the SEI state recognition model of the battery management unit for SEI film state analysis, and the SEI state recognition result is output.

[0080] In this embodiment of the application, in the scenario of real-time monitoring of the SEI film status of lithium iron phosphate batteries, in order to capture the state changes of the SEI film in a timely manner and quickly determine whether there is any loss, it is necessary to collect core feature data by sensors at a predetermined frequency, and to perform efficient analysis with the help of models deployed at the edge, forming a complete closed loop from data acquisition to status output.

[0081] Specifically, the first step is to determine the preset data collection frequency. This frequency can be flexibly set according to the battery's usage scenario and characteristics. For example, it can be set to once every 10 minutes during the high-frequency charging and discharging phase of the battery, and once every hour during the resting phase, so as to reduce unnecessary energy consumption and data redundancy while ensuring the timeliness of monitoring.

[0082] At the same time, the pre-deployed micro sensors on the battery cell are triggered to enter the working state strictly according to the configured simplified cell characteristic monitoring indicators (temperature, voltage, impedance).

[0083] These micro-sensors are characterized by high precision and low power consumption. They can sense and record the temperature (e.g., 23℃, 28℃), voltage changes (e.g., 3.15V, 3.22V), and impedance (e.g., 5.2mΩ, 5.8mΩ) of the target battery cell in real time during operation. They integrate these discrete data into a continuous sequence of first battery cell characteristic data in chronological order to fully reflect the state evolution of the battery cell during the monitoring period.

[0084] Furthermore, the integrated first cell feature data sequence is directly input into the SEI state recognition model deployed in the battery management unit. This model, as a lightweight edge computing core, has been trained to convergence using a large number of historical samples and can quickly process the input feature data, ultimately outputting analysis results indicating whether the SEI film is in a normal or damaged state.

[0085] For example, during the normal charging and discharging process of a lithium iron phosphate battery, a micro-sensor collects data every 30 minutes. The first cell characteristic data sequence obtained is "temperature 26℃, voltage 3.2V, impedance 5.3mΩ". After inputting this sequence into the SEI state recognition model, the model analyzes it through its internal algorithm and determines that the SEI film is in a normal state, and outputs "normal state", which is consistent with the actual detection of the battery's intact SEI film structure.

[0086] Furthermore, when the first cell characteristic data sequence of another lithium iron phosphate battery of the same type is "temperature 36℃, voltage 2.8V, impedance 9.1mΩ", the SEI state identification model outputs "loss state", which corresponds to the detected local rupture phenomenon of the SEI film, verifying the accuracy of the model's analysis.

[0087] This step, by setting flexible acquisition frequencies and precise feature indicators, combined with the real-time data acquisition capabilities of micro-sensors and the rapid analysis capabilities of the edge SEI state recognition model, enables real-time monitoring and efficient judgment of the SEI film state. This lays the foundation for taking corresponding measures based on the SEI film state, ensuring the timeliness and effectiveness of battery management.

[0088] S130: If the SEI state identification result is a loss state, the second cell feature data sequence is collected by the micro sensor according to the preset data acquisition frequency and uploaded to the cloud server. The SEI loss prediction plugin is called to perform SEI film loss inference based on the second cell feature data sequence and output the predicted loss type and predicted degradation coefficient.

[0089] In this embodiment of the application, when the SEI status identification result is a loss status, it is necessary to collect more comprehensive feature data and use cloud models for refined analysis in order to clarify the specific loss situation and provide accurate basis for the formulation of subsequent repair strategies.

[0090] Specifically, the process begins by activating the micro-sensors on the battery cell according to the preset data acquisition frequency and the configured cell characteristic monitoring indicators (temperature, voltage, current, impedance, and carbon dioxide concentration).

[0091] These micro-sensors capture data such as temperature, voltage, current, impedance, and carbon dioxide concentration of the target cell in real time under loss conditions, and organize this data in chronological order to form a second cell characteristic data sequence.

[0092] Furthermore, the acquired second cell feature data sequence is uploaded to a cloud server, and the SEI loss prediction plugin is invoked. This plugin, based on a graph neural network algorithm, performs deep computational analysis on the input feature data to infer the loss type and degradation coefficient of the SEI film.

[0093] Finally, the SEI loss prediction plugin outputs the predicted loss type and predicted degradation coefficient, providing detailed information for the subsequent development of targeted charging repair solutions.

[0094] Step S130 in the method provided in this application embodiment includes:

[0095] According to the preset data acquisition frequency and the cell characteristic monitoring indicators, the second cell characteristic data sequence of the target cell is acquired by the micro sensor.

[0096] In this embodiment of the application, in the scenario of deep analysis of SEI film loss in lithium iron phosphate batteries, when the SEI state identification result is a loss state, it is necessary to collect more comprehensive feature data to support subsequent refined loss reasoning, and to provide a data basis for clarifying the specific type and degree of SEI film loss.

[0097] Specifically, the preset data acquisition frequency is first determined. Considering that it is necessary to capture cell state changes more intensively at this time, the frequency is usually higher than the data acquisition frequency in the first stage. For example, it is set to once every 5 minutes to ensure that enough detailed data can be obtained to reflect the dynamic characteristics of the SEI film loss process.

[0098] Similarly, strictly adhering to the configured cell characteristic monitoring indicators, including at least temperature, voltage, current, impedance, and carbon dioxide concentration, the micro-sensors configured on the cell are activated into a high-intensity operating mode. These sensors possess multi-parameter synchronous acquisition capabilities, enabling them to accurately sense and record various characteristic parameters of the target cell under loss conditions.

[0099] For example, in terms of temperature, the microsensor can detect local temperature rises or fluctuations (e.g., 32°C) caused by SEI film loss; in terms of voltage, the microsensor can record voltage drops or unstable fluctuations related to loss (e.g., 3.0V); in terms of current, the microsensor can monitor current changes during charging and discharging (e.g., 1.2A); in terms of impedance, the microsensor can track impedance changes caused by changes in SEI film structure in real time (e.g., 7.8mΩ); and in terms of carbon dioxide concentration, the microsensor can detect changes in gas concentration that may be generated by SEI film decomposition (e.g., 55ppm).

[0100] Furthermore, the micro-sensors integrate these discrete data collected in chronological order to form a continuous second cell feature data sequence. This sequence fully records the multi-dimensional feature evolution process of the cell under loss conditions. Compared with the first cell feature data sequence, it contains more comprehensive information and richer details, providing sufficient data support for subsequent SEI film loss inference on the cloud server.

[0101] For example, after a lithium iron phosphate battery is identified as being in a state of SEI film loss, a micro-sensor collects data every 5 minutes, forming a second cell characteristic data sequence of "temperature 34℃→35℃→33℃, voltage 2.9V→2.85V→2.9V, current 1.4A→1.6A→1.5A, impedance 8.1mΩ→8.3mΩ→8.2mΩ, carbon dioxide concentration 62ppm→65ppm→64ppm". This sequence clearly shows the changing trends of various characteristics of the cell in a state of loss.

[0102] Based on this, the obtained second cell feature data sequence is uploaded to the cloud server, and the SEI loss prediction plugin is called. This plugin is based on graph neural network algorithm, which can deeply explore the correlation between various feature parameters and accurately infer the loss type of SEI film (such as physical cracking, chemical corrosion, overgrowth, etc.) and degradation coefficient (value range 0-1, the larger the value, the more severe the loss).

[0103] For example, the characteristic data sequence of the second cell of a lithium iron phosphate battery is "temperature 36℃→37℃→35℃, voltage 2.8V→2.75V→2.8V, current 1.8A→2.0A→1.9A, impedance 9.0mΩ→9.2mΩ→9.1mΩ, carbon dioxide concentration 90ppm→95ppm→92ppm". After analyzing this sequence, the SEI loss prediction plugin outputs "loss type: physical cracking, degradation coefficient: 0.75", which is consistent with the state of large-area cracks in the SEI film that was actually detected.

[0104] Ultimately, the predicted loss type and predicted degradation coefficient output by the SEI loss prediction plugin provide a precise basis for the subsequent development of targeted charging repair strategies, ensuring that the repair scheme can effectively adapt to the damage status of the SEI film and effectively improve the repair effect.

[0105] S140: Optimize the battery charging repair strategy based on the predicted loss type and predicted degradation coefficient, output an adapted charging scheme, and control battery charging according to the adapted charging scheme to complete the active repair of the SEI film.

[0106] In this embodiment of the application, in the active repair scenario of the SEI film of lithium iron phosphate battery, in order to repair the SEI film accurately and efficiently, it is necessary to optimize and control the charging repair scheme based on the predicted loss type and predicted degradation coefficient obtained in the early stage.

[0107] Specifically, the repair cycle is first analyzed based on the predicted loss type and predicted degradation coefficient output by the SEI loss prediction plugin, and the repair cycle is output using the preset degradation coefficient and repair cycle mapping rules.

[0108] The adaptation and repair cycle is a periodic charging and repair schedule designed to restore battery performance to an adapted state under the predicted loss repair needs. It includes several charging and repair nodes, and the number of charging and repair nodes is positively correlated with the predicted degradation coefficient.

[0109] Furthermore, charging parameter adjustment thresholds are obtained based on cell characteristics, SEI film repair safety boundaries, and historical repair experience data. These thresholds include charging voltage adjustment thresholds and current rate adjustment thresholds, in order to limit the adjustment range of charging parameters and ensure the repair process is safe and effective.

[0110] Based on this, the threshold and charging repair node are adjusted according to the obtained charging parameters. The battery charging repair strategy is optimized with the goal of minimizing the SEI film degradation coefficient and minimizing the side reaction rate.

[0111] Finally, through an iterative optimization process that continuously generates new charging schemes, evaluates their fitness, and dynamically updates the current optimal scheme within a preset number of optimization iterations, the current optimal charging scheme is determined, and an adapted charging scheme is output to clarify the charging voltage, current rate, and other parameters of each charging repair node, which can be used to guide the actual battery charging control.

[0112] This step involves accurately planning the repair cycle based on the prediction results, constraining the range of charging parameter adjustments, and then determining the optimal charging strategy through multi-objective optimization to achieve proactive and efficient repair of the SEI film, ensuring the recovery of cell performance and safe operation.

[0113] Step S140 in the method provided in this application embodiment includes:

[0114] Repair cycle analysis is performed based on the predicted loss type and predicted degradation coefficient, and an adaptive repair cycle is output. The adaptive repair cycle includes several charging repair nodes, and the number of charging repair nodes is positively correlated with the predicted degradation coefficient.

[0115] Obtain the charging parameter adjustment threshold, where the charging parameters include charging voltage and current rate;

[0116] Based on the charging parameter adjustment threshold and several charging repair nodes, the battery charging repair strategy is optimized with the goal of minimizing the SEI film degradation coefficient and minimizing the side reaction rate, and an adapted charging scheme is output.

[0117] In this embodiment of the application, in the scenario of active repair of SEI film in lithium iron phosphate battery, in order to achieve accurate and efficient SEI film repair, it is necessary to construct a complete process from repair cycle planning to charging strategy optimization based on the previous SEI loss prediction results, forming a data-driven active repair closed loop.

[0118] First, the repair cycle is analyzed based on the predicted loss type and predicted degradation coefficient output by the SEI loss prediction plugin to output a suitable repair cycle. Specifically, the repair cycle is accurately calculated and planned by combining the built-in degradation coefficient-repair cycle mapping rules with historical battery repair data, repair experience with similar batteries, and multi-dimensional information such as battery usage conditions.

[0119] Specifically, the basic repair cycle coefficient corresponding to the predicted loss type is first extracted. Different loss types (such as physical pyrolysis, chemical corrosion, etc.) correspond to different basic repair cycle coefficients due to differences in repair difficulty and characteristics. For example, the basic repair cycle coefficient for physical pyrolysis is 1.2, and for chemical corrosion it is 1.1.

[0120] Furthermore, the repair cycle is dynamically adjusted based on the predicted degradation coefficient. For every 0.1 increase in the degradation coefficient, the repair cycle is extended by a certain percentage from the base repair cycle, such as by 10%. Through this calculation method, the final adaptive repair cycle duration is determined, and charging repair nodes are defined.

[0121] The adaptation and repair cycle is a periodic charging and repair schedule designed to restore battery performance to an adapted state under the predicted loss repair needs (divided into time dimensions such as days and weeks). It includes several charging and repair nodes, and the number of charging and repair nodes is positively correlated with the predicted degradation coefficient. That is, the larger the degradation coefficient, the more charging and repair nodes there are. For example, a degradation coefficient of 0.6 corresponds to 3 charging and repair nodes, and 0.3 corresponds to 1 charging and repair node.

[0122] For example, a lithium iron phosphate battery, analyzed by the SEI loss prediction plugin, is predicted to have a loss type of "chemical corrosion" and a predicted degradation coefficient of 0.5. First, the base repair cycle coefficient for "chemical corrosion" is 1.1. Based on a degradation coefficient of 0.5, and following the rule of extending the repair cycle by 10% for every 0.1 degradation coefficient, the repair cycle needs to be extended by 50% (0.5 × 10%). If the base repair cycle (without degradation) is calculated as 1 week, then the current suitable repair cycle is 1.5 weeks (1 + 1 × 0.5).

[0123] Meanwhile, since the degradation coefficient is 0.5, there are two corresponding charging repair nodes. That is, within the 1.5-week repair cycle, charging repair nodes need to be set on the 3rd and 7th days. Subsequently, charging strategy optimization and repair control will be carried out based on these nodes, which is consistent with the actual scenario where the SEI film of the battery is damaged due to chemical corrosion and needs to be repaired in stages to restore performance.

[0124] Furthermore, the charging parameter adjustment threshold is obtained based on the cell's electrochemical characteristics, the SEI film repair safety boundary, and historical repair experience data.

[0125] Specifically, by analyzing the charging tolerance limits of the battery cell under different health conditions and different loss types, and combining the safe charging parameter range of laboratory tests, the charging parameter adjustment thresholds are determined, including the charging voltage adjustment threshold and the current rate adjustment threshold.

[0126] For example, the charging voltage adjustment threshold is set to ±0.2V and the current rate adjustment threshold is set to ±0.3A to limit the adjustment range of the charging parameters, so as to ensure the safety and effectiveness of the repair process and provide charging parameter constraints for subsequent charging strategy optimization.

[0127] Based on this, the battery charging repair strategy is optimized by adjusting the threshold based on the obtained charging parameters and several charging repair nodes, with the comprehensive optimization objectives of minimizing the SEI film degradation coefficient and minimizing the side reaction rate.

[0128] The method provided in this application embodiment includes the step of "optimizing the battery charging repair strategy based on the charging parameter adjustment threshold and several charging repair nodes, with minimizing the SEI film degradation coefficient and minimizing the side reaction rate as comprehensive optimization objectives":

[0129] A first charging scheme is randomly generated based on the charging parameter adjustment threshold and several charging repair nodes;

[0130] An SEI film repair predictor is constructed, which performs SEI film repair prediction based on the predicted loss type, predicted degradation coefficient and first charging scheme, and outputs the first predicted SEI film degradation coefficient and the first predicted side reaction rate. The SEI film repair predictor is constructed based on a deep learning model and trained to convergence using sample data.

[0131] A scheme evaluation function is constructed with minimizing the SEI membrane degradation coefficient and minimizing the side reaction rate as the comprehensive optimization objectives. The first fitness is determined based on the first predicted SEI membrane degradation coefficient and the first predicted side reaction rate. The fitness is negatively correlated with the predicted SEI membrane degradation coefficient and the predicted side reaction rate.

[0132] Continue to adjust the threshold based on the charging parameters and randomly generate a second charging scheme using several charging repair nodes, and evaluate and obtain the second fitness.

[0133] If the second fitness is greater than or equal to the first fitness, then the second charging scheme is set as the current optimal charging scheme; if the second fitness is less than the first fitness, then the second charging scheme is set as the current optimal charging scheme according to probability, wherein the probability is the ratio of the difference between the preset number of optimization iterations and the current number of optimization iterations to the preset number of optimization iterations.

[0134] Continue iterative optimization until the preset number of optimization iterations is reached, and output the current optimal charging scheme as the adapted charging scheme.

[0135] In this embodiment of the application, in the scenario of optimizing the active repair strategy of the SEI film of lithium iron phosphate battery, in order to accurately find a charging scheme that can effectively repair the SEI film and has few side reactions, it is necessary to use intelligent algorithms and prediction models to screen out the optimal solution from a large number of potential charging schemes, so as to provide accurate strategy support for SEI film repair.

[0136] Specifically, the first step is to generate a first charging scheme. This involves adjusting the threshold based on the determined charging parameters and several charging repair nodes. By generating random numbers, each charging repair node is assigned a charging voltage and current rate value that meets the threshold requirements. The parameters of these nodes are then combined to form the first charging scheme.

[0137] For example, if the charging voltage threshold is [3.2V, 3.6V] and the current rate threshold is [0.5C, 1.2C], there are 3 charging repair nodes, each corresponding to a different time point within the repair cycle. The charging scheme for these 3 repair nodes might be: "Node 1: Charging voltage 3.3V, current rate 0.6C; Node 2: Charging voltage 3.4V, current rate 0.8C; Node 3: Charging voltage 3.2V, current rate 1.0C," thus constructing the initial prototype of the charging strategy.

[0138] Furthermore, an SEI membrane repair predictor was constructed. The SEI membrane repair predictor was built using a fusion architecture of LSTM (Long Short-Term Memory) and CNN (Convolutional Neural Network), and was trained using a large amount of sample data.

[0139] The sample data covers different SEI film loss types (physical pyrolysis, chemical corrosion, etc.), different degradation coefficients (0.1-1.0 range), and corresponding charging schemes, as well as the actual changes in SEI film degradation coefficients and side reaction rates after the implementation of the charging schemes.

[0140] In addition, the LSTM layer is used to capture the temporal dependencies of the cell feature data sequence, such as the dynamic impact of temperature and voltage changes at different time points on SEI film repair; the CNN layer extracts local key features in the data through convolution operations, such as the correlation pattern between current mutations and side reaction rates.

[0141] Specifically, the model input layer has 5 neurons, corresponding to the predicted loss type, predicted degradation coefficient, and charging voltage, current rate, and repair node timing in the charging scheme, respectively; the hidden layer has 3 layers, each containing 32 neurons, and uses the LeakyReLU activation function to solve the neuron death problem and enhance the model's feature expression ability; the output layer has 2 neurons, corresponding to the predicted SEI membrane degradation coefficient and predicted side reaction rate, respectively, and uses a linear activation function to output specific values.

[0142] First, the training sample data is preprocessed. This involves normalizing the sample data (e.g., mapping the degradation coefficient to the range of 0-1 and scaling the side reaction rate to the maximum value) and dividing it into a training set and a validation set in a 7:3 ratio. For example, 7,000 sets are selected from 10,000 historical repair cases as the training set for model parameter learning, and 3,000 sets are selected as the validation set to evaluate the model's predictive performance.

[0143] The sample data includes historical SEI film loss types, degradation coefficients, different charging schemes and their corresponding post-repair degradation coefficients and side reaction rates.

[0144] Secondly, during the training process, mean squared error (MSE) is used as the loss function to calculate the deviation between the predicted SEI membrane degradation coefficient and the side reaction rate and the actual values. The network weights are updated through the RMSprop optimizer. The initial learning rate is set to 0.0005, and it decays to 1 / 3 of the current learning rate every 200 iterations to balance the model convergence speed and prediction accuracy.

[0145] Furthermore, when the MSE of the validation set remains stable below 0.001 for 30 consecutive rounds (e.g., in a certain round of validation, the average deviation between the predicted and actual values ​​of 3000 samples is 0.0008), and the difference in loss between the model on the training set and the validation set is less than 0.0005, the model is considered to have converged, training is stopped, and the final SEI membrane repair predictor is generated.

[0146] Ultimately, this SEI film repair predictor can accurately output the predicted SEI film degradation coefficient and side reaction rate based on the input predicted loss type, predicted degradation coefficient, and charging scheme.

[0147] For example, when the input predicted loss type is "chemical corrosion", the predicted degradation coefficient is 0.6, and the charging scheme is "Node 1: charging voltage 3.3V, current rate 0.7C; Node 2: charging voltage 3.2V, current rate 0.5C", the SEI film repair predictor outputs a first predicted SEI film degradation coefficient of 0.35 and a first predicted side reaction rate of 15ppm / min. The deviation from the actual measured degradation coefficient of 0.37 and side reaction rate of 14ppm / min after repair is small, which verifies the accuracy of the predictor.

[0148] Furthermore, an evaluation function for the scheme is constructed with the combined optimization objectives of minimizing the SEI membrane degradation coefficient and minimizing the side reaction rate.

[0149] Specifically, the first predicted SEI membrane degradation coefficient and the first predicted side reaction rate are first processed to be dimensionless, that is, the degradation coefficient is mapped to the [0,1] interval through maximum-minimum normalization, and the side reaction rate is processed in the same way to eliminate the influence of the difference in the dimensions of different indicators.

[0150] Furthermore, weights are assigned to the two indicators based on actual repair needs. The weight assignment can be determined based on the battery's application scenario, usage stage, and priority requirements for SEI film repair effect and side reaction control.

[0151] For example, if more emphasis is placed on the SEI membrane repair effect, the degradation coefficient weight can be set to 0.6 and the side reaction rate weight can be set to 0.4 to construct the scheme evaluation function, and the comprehensive evaluation value can be obtained by weighted summation. The specific expression formula of the scheme evaluation function is "Comprehensive evaluation value = (first predicted SEI membrane degradation coefficient × 0.6) + (first predicted side reaction rate × 0.4)".

[0152] Fitness is negatively correlated with the comprehensive evaluation value; that is, the smaller the comprehensive evaluation value, the higher the fitness. The first fitness is determined based on this value, and the specific calculation formula can be expressed as "first fitness = 1 - comprehensive evaluation value".

[0153] For example, if the first predicted SEI film degradation coefficient is 0.35 and the first predicted side reaction rate is 0.3 (both are normalized values), then the comprehensive evaluation value = 0.35 × 0.6 + 0.3 × 0.4 = 0.33, and the corresponding first fitness can be set to 1 - 0.33 = 0.67 (fitness range is 0-1), indicating that the charging scheme has a medium to high adaptability under the current evaluation standard.

[0154] Similarly, a second charging scheme is generated by adjusting the threshold based on the charging parameters and randomly generating several charging repair nodes. The generation logic is the same as that of the first charging scheme to ensure that the parameter combination meets the threshold constraints.

[0155] Furthermore, the second charging scheme is input into the constructed SEI film repair predictor to obtain the corresponding second predicted degradation coefficient and second predicted side reaction rate, and then the second fitness is calculated through the above scheme evaluation function.

[0156] Furthermore, by comparing the first fitness and the second fitness, if the second fitness is greater than or equal to the first fitness, it indicates that the second charging scheme is better in terms of overall repair effect, and it is directly set as the current optimal charging scheme; if the second fitness is less than the first fitness, then the current optimal charging scheme is replaced according to probability.

[0157] The probability calculation formula is (preset number of optimization iterations - current number of optimization iterations) / preset number of optimization iterations. For example, if the preset number of iterations is 100 and the current iteration is the 20th iteration, the probability is (100-20) / 100=0.8, meaning there is an 80% probability of replacement. As the number of iterations increases, the probability gradually decreases to avoid missing the optimal solution due to random fluctuations in later iterations.

[0158] Furthermore, the process of randomly generating new charging schemes, predicting and evaluating, and comparing and updating the current optimal scheme is continuously repeated until the preset number of optimization iterations is reached. At this point, the current optimal charging scheme retained during the iteration process is output as the adaptive charging scheme. This scheme can achieve the best balance between reducing the SEI film degradation coefficient and controlling the side reaction rate, providing precise guidance for actual charging and repair operations.

[0159] Finally, the battery is charged according to the output adaptive charging scheme, and the corresponding charging voltage and current rate parameters are precisely executed at each charging repair node. Through this series of targeted charging operations, the repair and regeneration of the SEI film is gradually promoted, while the occurrence of side reactions is suppressed to the maximum extent. This achieves active and efficient repair of the SEI film, restores the battery's performance and safety, and ensures that the battery maintains a stable working state during subsequent use.

[0160] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects:

[0161] This application proposes a charging regulation method for SEI film repair in lithium iron phosphate batteries. First, based on historical experimental data of SEI film loss from similar lithium iron phosphate batteries, an SEI state identification model and an SEI loss prediction plugin are constructed on a cloud server. The SEI state identification model is generated through a BP neural network training process, while the SEI loss prediction plugin is built based on a graph neural network. The former is deployed to the battery management unit. Next, at a preset acquisition frequency, a first cell characteristic data sequence is collected using miniature sensors based on simplified cell characteristic monitoring indicators such as temperature, voltage, and impedance. This data is input into the model to obtain the SEI state identification result. If the cell is in a loss state, a second cell characteristic data sequence containing current, carbon dioxide concentration, etc., is collected at the same frequency and uploaded to the cloud server. The SEI loss prediction plugin is then invoked to output the predicted loss type and degradation coefficient. Finally, based on this, the repair cycle is analyzed, and charging parameter adjustment thresholds are obtained. The optimal charging scheme is then sought to minimize the SEI film degradation coefficient and side reaction rate, thus completing the active repair of the SEI film.

[0162] The method provided in this application, through the technical solution of "model construction - data acquisition and state recognition - loss inference - repair strategy optimization", solves the problems of lack of specificity and low efficiency in SEI film repair in traditional lithium iron phosphate battery management. It realizes intelligent control of the entire process from SEI film state monitoring to active repair, improves battery life and safety, and provides reliable technical support for the efficient management of lithium iron phosphate batteries.

[0163] Example 2, as shown in the appendix Figure 2As shown, based on the inventive concept of a charging regulation method for repairing the SEI film of a lithium iron phosphate battery provided in Embodiment 1, this application also provides a charging regulation system for repairing the SEI film of a lithium iron phosphate battery, specifically including:

[0164] SEI model building and deployment module 01 is used to build an SEI state identification model and an SEI loss prediction plugin on a cloud server based on historical experimental data of SEI film loss of the same type of lithium iron phosphate battery, and to deploy the SEI state identification model to the battery management unit.

[0165] SEI state monitoring module 02 is used to collect the first cell feature data sequence through the micro sensor configured on the cell according to the preset data acquisition frequency, input it into the SEI state recognition model of the battery management unit for analysis, and output the SEI state recognition result.

[0166] SEI loss inference module 03 is used to collect the second cell feature data sequence through the micro sensor according to the preset data acquisition frequency if the SEI state identification result is a loss state, and upload it to the cloud server, call the SEI loss prediction plugin, perform SEI film loss inference based on the second cell feature data sequence, and output the predicted loss type and predicted degradation coefficient.

[0167] The charging repair control module 04 is used to optimize the battery charging repair strategy based on the predicted loss type and predicted degradation coefficient, output an adaptive charging scheme, and control the battery charging according to the adaptive charging scheme to complete the active repair of the SEI film.

[0168] In one embodiment, the SEI model building and deployment module 01 is also used for:

[0169] Configure simplified cell characteristic monitoring indicators, wherein the simplified cell characteristic monitoring indicators include temperature, voltage and impedance;

[0170] Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a set of simplified cell feature data sequences was collected according to the simplified cell feature monitoring index, and the historical SEI state under different sample simplified cell feature data sequences was obtained to obtain a sample SEI state set, where the historical SEI state is a normal state or a loss state.

[0171] Using the sample simplified cell feature data sequence set and sample SEI state set, a BP neural network is trained until convergence to generate an SEI state recognition model.

[0172] Configure cell characteristic monitoring indicators, wherein the cell characteristic monitoring indicators include at least temperature, voltage, current, impedance and carbon dioxide concentration;

[0173] Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a set of sample cell characteristic data sequences was collected according to the cell characteristic monitoring index, and the historical loss type and historical degradation coefficient of SEI film under different sample cell characteristic data sequences were obtained to obtain a set of sample loss type and a set of sample degradation coefficient, wherein the degradation coefficient is positively correlated with the degree of loss.

[0174] Using the sample cell feature data sequence set, sample loss type set, and sample degradation coefficient set as training data, an SEI loss prediction plugin is constructed based on a graph neural network.

[0175] In one embodiment, the SEI status monitoring module 02 is further configured to:

[0176] According to the preset data acquisition frequency and the simplified cell feature monitoring indicators, the micro sensors configured on the cell are activated to acquire the first cell feature data sequence of the target cell;

[0177] The first cell feature data sequence is input into the SEI state recognition model of the battery management unit for SEI film state analysis, and the SEI state recognition result is output.

[0178] In one embodiment, the SEI loss inference module 03 is further configured to:

[0179] According to the preset data acquisition frequency and the cell characteristic monitoring indicators, the second cell characteristic data sequence of the target cell is acquired by the micro sensor.

[0180] In one embodiment, the charging repair control module 04 is further configured to:

[0181] Repair cycle analysis is performed based on the predicted loss type and predicted degradation coefficient, and an adaptive repair cycle is output. The adaptive repair cycle includes several charging repair nodes, and the number of charging repair nodes is positively correlated with the predicted degradation coefficient.

[0182] Obtain the charging parameter adjustment threshold, where the charging parameters include charging voltage and current rate;

[0183] Based on the charging parameter adjustment threshold and several charging repair nodes, the battery charging repair strategy is optimized with the goal of minimizing the SEI film degradation coefficient and minimizing the side reaction rate, and an adapted charging scheme is output.

[0184] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0185] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0186] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A charging regulation method for repairing the SEI film of a lithium iron phosphate battery, characterized in that, The methods include: On the cloud server, an SEI state identification model and an SEI loss prediction plugin are constructed based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries. The SEI state identification model is then deployed to the battery management unit, including: configuring simplified cell characteristic monitoring indicators, wherein the simplified cell characteristic monitoring indicators include temperature, voltage and impedance. Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a set of simplified cell feature data sequences was collected according to the simplified cell feature monitoring index, and the historical SEI state under different sample simplified cell feature data sequences was obtained to obtain a sample SEI state set, where the historical SEI state is a normal state or a loss state. Using the sample simplified cell feature data sequence set and sample SEI state set, a BP neural network is trained until convergence to generate an SEI state recognition model; Configure cell characteristic monitoring indicators, wherein the cell characteristic monitoring indicators include at least temperature, voltage, current, impedance and carbon dioxide concentration; Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a sample cell characteristic data sequence set was collected according to the cell characteristic monitoring index, and the historical loss type and historical degradation coefficient of SEI film under different sample cell characteristic data sequences were obtained to obtain a sample loss type set and a sample degradation coefficient set, wherein the degradation coefficient is positively correlated with the degree of loss. Using the sample cell feature data sequence set, sample loss type set, and sample degradation coefficient set as training data, an SEI loss prediction plugin is constructed based on a graph neural network. According to the preset data acquisition frequency, the first cell feature data sequence is collected by the micro sensor configured on the cell, input into the SEI state recognition model of the battery management unit for analysis, and output the SEI state recognition result. If the SEI state identification result is a loss state, the second cell feature data sequence is collected by the micro sensor according to the preset data collection frequency and uploaded to the cloud server. The SEI loss prediction plugin is called to perform SEI film loss inference based on the second cell feature data sequence and output the predicted loss type and predicted degradation coefficient. Based on the predicted loss type and predicted degradation coefficient, the battery charging repair strategy is optimized, an adapted charging scheme is output, and battery charging control is performed according to the adapted charging scheme to complete the active repair of the SEI film.

2. The charging regulation method for repairing the SEI film of a lithium iron phosphate battery according to claim 1, characterized in that, According to a preset data acquisition frequency, the first cell characteristic data sequence is collected through a micro-sensor configured on the cell, and input into the SEI state recognition model of the battery management unit for analysis, including: According to the preset data acquisition frequency and the simplified cell feature monitoring indicators, the micro sensors configured on the cell are activated to acquire the first cell feature data sequence of the target cell; The first cell feature data sequence is input into the SEI state recognition model of the battery management unit for SEI film state analysis, and the SEI state recognition result is output.

3. The charging regulation method for repairing the SEI film of a lithium iron phosphate battery according to claim 2, characterized in that, According to the preset data acquisition frequency and the cell characteristic monitoring indicators, the second cell characteristic data sequence of the target cell is acquired by the micro sensor.

4. The charging regulation method for repairing the SEI film of a lithium iron phosphate battery according to claim 1, characterized in that, Based on the predicted loss type and predicted degradation coefficient, the battery charging repair strategy is optimized, and an adapted charging scheme is output, including: Repair cycle analysis is performed based on the predicted loss type and predicted degradation coefficient, and an adaptive repair cycle is output. The adaptive repair cycle includes several charging repair nodes, and the number of charging repair nodes is positively correlated with the predicted degradation coefficient. Obtain the charging parameter adjustment threshold, where the charging parameters include charging voltage and current rate; Based on the charging parameter adjustment threshold and several charging repair nodes, the battery charging repair strategy is optimized with the goal of minimizing the SEI film degradation coefficient and minimizing the side reaction rate, and an adapted charging scheme is output.

5. The charging regulation method for repairing the SEI film of a lithium iron phosphate battery according to claim 4, characterized in that, Based on the aforementioned charging parameter adjustment thresholds and several charging repair nodes, and with the comprehensive optimization objectives of minimizing the SEI film degradation coefficient and minimizing the side reaction rate, a battery charging repair strategy is optimized, including: A first charging scheme is randomly generated based on the charging parameter adjustment threshold and several charging repair nodes; An SEI film repair predictor is constructed, which performs SEI film repair prediction based on the predicted loss type, predicted degradation coefficient and first charging scheme, and outputs the first predicted SEI film degradation coefficient and the first predicted side reaction rate. The SEI film repair predictor is constructed based on a deep learning model and trained to convergence using sample data. A scheme evaluation function is constructed with minimizing the SEI membrane degradation coefficient and minimizing the side reaction rate as the comprehensive optimization objectives. The first fitness is determined based on the first predicted SEI membrane degradation coefficient and the first predicted side reaction rate. The fitness is negatively correlated with the predicted SEI membrane degradation coefficient and the predicted side reaction rate. Continue to adjust the threshold based on the charging parameters and randomly generate a second charging scheme using several charging repair nodes, and evaluate and obtain the second fitness. If the second fitness is greater than or equal to the first fitness, then the second charging scheme is set as the current optimal charging scheme; if the second fitness is less than the first fitness, then the second charging scheme is set as the current optimal charging scheme according to probability, wherein the probability is the ratio of the difference between the preset number of optimization iterations and the current number of optimization iterations to the preset number of optimization iterations. Continue iterative optimization until the preset number of optimization iterations is reached, and output the current optimal charging scheme as the adapted charging scheme.

6. A charging regulation system for repairing the SEI film of a lithium iron phosphate battery, characterized in that, The system is used to execute the charging regulation method for repairing the SEI film of a lithium iron phosphate battery according to any one of claims 1-5, the system comprising: The SEI model building and deployment module is used to build an SEI state identification model and an SEI loss prediction plugin on a cloud server based on historical experimental data of SEI film loss of the same type of lithium iron phosphate battery, and to deploy the SEI state identification model to the battery management unit. The module includes configuring simplified cell characteristic monitoring indicators, wherein the simplified cell characteristic monitoring indicators include temperature, voltage and impedance. Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a set of simplified cell feature data sequences was collected according to the simplified cell feature monitoring index, and the historical SEI state under different sample simplified cell feature data sequences was obtained to obtain a sample SEI state set, where the historical SEI state is a normal state or a loss state. Using the sample simplified cell feature data sequence set and sample SEI state set, a BP neural network is trained until convergence to generate an SEI state recognition model; Configure cell characteristic monitoring indicators, wherein the cell characteristic monitoring indicators include at least temperature, voltage, current, impedance and carbon dioxide concentration; Based on historical experimental data of SEI film loss of similar lithium iron phosphate batteries, a sample cell characteristic data sequence set was collected according to the cell characteristic monitoring index, and the historical loss type and historical degradation coefficient of SEI film under different sample cell characteristic data sequences were obtained to obtain a sample loss type set and a sample degradation coefficient set, wherein the degradation coefficient is positively correlated with the degree of loss. Using the sample cell feature data sequence set, sample loss type set, and sample degradation coefficient set as training data, an SEI loss prediction plugin is constructed based on a graph neural network. The SEI state monitoring module is used to collect the first cell feature data sequence through the micro sensor configured on the cell according to the preset data acquisition frequency, input it into the SEI state recognition model of the battery management unit for analysis, and output the SEI state recognition result. The SEI loss inference module is used to collect the second cell feature data sequence through the micro sensor according to the preset data acquisition frequency if the SEI state identification result is a loss state, and upload it to the cloud server. It then calls the SEI loss prediction plugin to perform SEI film loss inference based on the second cell feature data sequence and outputs the predicted loss type and predicted degradation coefficient. The charging repair control module is used to optimize the battery charging repair strategy based on the predicted loss type and predicted degradation coefficient, output an adaptive charging scheme, and control the battery charging according to the adaptive charging scheme to complete the active repair of the SEI film.

Images