Battery management system, battery pack and electric vehicle

Abstract

The invention provides a battery management system, a battery pack and an electric vehicle, and relates to the technical field of batteries. The system comprises a plurality of slave control battery management layers, each slave control battery management layer is used for collecting battery working parameters of a corresponding single battery through a battery cell management chip arranged on each single battery, and single battery feature extraction is carried out from the battery working parameters; a plurality of master control battery management layers, each master control battery management layer correspondingly obtaining the characteristics of each single battery extracted by one slave control battery management layer, performing SOC estimation and SOH estimation of the battery pack according to the characteristics of the single batteries, and performing battery pack management on the corresponding battery pack; and the battery cluster management layer is used for acquiring the SOC estimation data, the SOH estimation data and the battery pack management data estimated by each master control battery management layer, and performing battery health state self-inspection according to the battery pack management data. According to the invention, the battery state can be comprehensively monitored, and the estimation precision of the battery state is improved.

Description

Technical Field

[0001] The present application relates to the technical field of batteries, and particularly to a battery management system, a battery pack, and an electric vehicle. Background Art

[0002] With the rapid development of electric vehicles, renewable energy storage systems, and smart grids, the battery, as the core energy carrier, its safety, lifespan, and efficiency have become the key to technological innovation. The battery management system (Battery Management System, abbreviated as "BMS"), as the core technology for monitoring and managing the state of the battery pack, mainly functions to prevent overcharging / overdischarging, extend the battery lifespan, ensure safe operation, etc. However, the existing battery management systems still have significant deficiencies in terms of technical architecture, data acquisition breadth, and management accuracy, making it difficult to meet the growing high-performance requirements.

[0003] Currently, the mainstream battery management systems mainly perform state estimation and management based on three basic parameters: voltage, current, and temperature. Although these parameters can reflect the operating state of the battery to a certain extent, their limitations are becoming increasingly prominent. For example, the non-linear relationship between voltage and state of charge (SOC) is easily disturbed by factors such as temperature and aging, resulting in a large SOC estimation error; the prediction of state of health (SOH) usually relies on the capacity decay model and is also difficult to accurately capture the battery aging characteristics. In addition, the existing systems mostly adopt a centralized architecture, and data acquisition and processing highly rely on the central server, resulting in high data transmission latency, wasted computing resources, and the inability to achieve fine-grained control of individual batteries. These problems exacerbate the inconsistency between individual cells in the battery pack, thereby triggering the risk of thermal runaway, shortening the overall lifespan, and increasing the operation and maintenance costs.

[0004] Therefore, there is an urgent need for an improved technical solution to solve at least one of the above technical problems. Summary of the Invention

[0005] The purpose of the present application is to provide a battery management system, a battery pack, and an electric vehicle to solve the above problems.

[0006] To achieve the above purpose, in a first aspect, the present application proposes a battery management system, which includes:

[0007] Multiple slave battery management layers, each slave battery management layer is used to collect the battery operating parameters of each individual battery in the corresponding battery pack through the cell management chips set on each individual battery in the corresponding battery pack, and extract the individual battery characteristics from the battery operating parameters, where each individual battery is uniquely configured with a cell management chip;

[0008] Multiple master battery management layers, each master battery management layer respectively obtains each single battery characteristic extracted from a slave battery management layer, estimates the state of charge (SOC) and the state of health (SOH) of the battery pack according to the single battery characteristic, and performs battery pack management on the corresponding battery pack;

[0009] A battery cluster management layer is used to obtain the SOC estimation data, SOH estimation data and battery pack management data estimated by each master battery management layer, and perform self-check on the battery health state according to the battery pack management data.

[0010] In some embodiments, the battery operating parameters include voltage data, current data, temperature data and pressure data of the battery. Extracting the single battery characteristics from the battery operating parameters includes:

[0011] Perform correlation analysis between the voltage data and the SOC of the corresponding battery at the corresponding discharge moment to obtain a first voltage correlation coefficient, and screen the first voltage data corresponding to the first voltage correlation coefficient whose first voltage correlation coefficient exceeds the first voltage coefficient threshold;

[0012] Perform differential processing on the current data, perform correlation analysis between the differentially processed current data and the SOC of the corresponding battery at the corresponding moment to obtain a first current correlation coefficient, and screen the first current data corresponding to the first current correlation coefficient whose first current correlation coefficient exceeds the first current coefficient threshold;

[0013] Perform correlation analysis between the temperature data and the SOC of the corresponding battery at the corresponding moment to obtain a first temperature correlation coefficient, and screen the first temperature data corresponding to the first temperature correlation coefficient whose first temperature correlation coefficient exceeds the first temperature coefficient threshold;

[0014] Perform correlation analysis between the pressure data and the SOC of the corresponding battery at the corresponding moment to obtain a first pressure correlation coefficient, and screen the first pressure data corresponding to the first pressure correlation coefficient whose first pressure correlation coefficient exceeds the first pressure coefficient threshold;

[0015] The estimating the SOC and SOH of the battery pack according to the single battery characteristic includes:

[0016] Perform first fusion processing on the selected first voltage data, the selected first current data, and the selected first temperature data to form a first fusion feature, and estimate the SOC of the battery pack based on the first fusion feature and the selected first pressure data.

[0017] In some embodiments, the estimating the SOC of the battery pack based on the first fusion feature and the selected first pressure data includes:

[0018] Take the first fusion feature and the selected first pressure data as the inputs of a preset unscented Kalman model, and call the unscented Kalman model to estimate the SOC of the battery pack.

[0019] In some embodiments, the battery operating parameters include battery voltage data, current data, temperature data, and pressure data. Extracting single-cell characteristics from the battery operating parameters includes:

[0020] Perform a correlation analysis between the voltage data and the SOH of the corresponding battery at the corresponding discharge time to obtain a second voltage correlation coefficient, and select the second voltage data corresponding to the second voltage correlation coefficient whose second voltage correlation coefficient exceeds the second voltage coefficient threshold;

[0021] Differentiate the current data, perform a correlation analysis between the differentiated current data and the SOH of the corresponding battery at the corresponding time to obtain a second current correlation coefficient, and select the second current data corresponding to the second current correlation coefficient whose second current correlation coefficient exceeds the second current coefficient threshold;

[0022] Perform a correlation analysis between the temperature data and the SOH of the corresponding battery at the corresponding time to obtain a second temperature correlation coefficient, and select the second temperature data corresponding to the second temperature correlation coefficient whose second temperature correlation coefficient exceeds the second temperature coefficient threshold;

[0023] Perform a correlation analysis between the pressure data and the SOH of the corresponding battery at the corresponding time to obtain a second pressure correlation coefficient, and select the second pressure data corresponding to the second pressure correlation coefficient whose second pressure correlation coefficient exceeds the second pressure coefficient threshold;

[0024] The SOC estimation and SOH estimation of the battery pack based on the single-cell characteristics include:

[0025] Perform a second fusion process on the selected second voltage data, the selected second current data, and the selected second temperature data to form a second fusion feature, and perform SOH estimation of the battery pack based on the second fusion feature and the selected second pressure data.

[0026] In some embodiments, the SOH estimation of the battery pack based on the second fusion feature and the second pressure correlation coefficient includes:

[0027] Based on the second fusion feature, the selected second pressure data, and the SOC estimation data, perform SOH estimation of the battery pack through a dynamic graph constraint and a physical memory gating network.

[0028] In some embodiments, the battery cluster management layer is further configured to, when detecting an abnormality in one of the single cells, send a control signal to the cell management chip uniquely corresponding to the single cell with the abnormality, so that the single cell with the abnormality is adjusted according to the control signal.

[0029] In some embodiments, when the abnormality is an excessive battery balancing difference, if the capacity of the single cell with the abnormality is too high, the cell management chip uniquely corresponding to the single cell with the abnormality performs battery capacity consumption in a passive balancing manner with full-load operation based on the control signal;

[0030] if the capacity of the single cell with the abnormality is too low, the cell management chip uniquely corresponding to the single cell with the abnormality performs battery capacity consumption in a low-load operation mode of maintaining dormancy based on the control signal.

[0031] In some embodiments, the battery pack management includes thermal management, charge and discharge management, and balancing management;

[0032] The slave battery management layer communicates with the master battery management layer via Bluetooth;

[0033] The master battery management layer communicates with the battery cluster management layer via a CAN bus;

[0034] The model of the cell management chip is BAT100X, and the cell management chips communicate with each other and with the master battery management layer via Bluetooth;

[0035] The SOC estimation and SOH estimation of the battery pack are performed according to the characteristics of the single cells, and battery pack management is performed on the corresponding battery pack, including: performing thermal management on the battery pack according to the temperature calculation result obtained from the characteristics of the single cells, and performing balancing management and charge and discharge management of the battery pack according to the SOC estimation result and the SOH estimation result.

[0036] In addition, to achieve the above object, in a second aspect, the present application further provides a battery pack of the battery management system as described in any one of the above.

[0037] In a third aspect, the present application further provides an electric vehicle with the battery pack as described above.

[0038] Compared with the prior art, the beneficial effects of the present application include:

[0039] By setting up multiple slave battery management layers, multiple master battery management layers, and one battery cluster management layer, a hierarchical distributed architecture is formed. Among them, the slave battery management layer can collect key data such as the voltage, temperature, pressure, and current of individual batteries in real time. These data are transmitted to the master battery management layer through Bluetooth MESH and then to the battery cluster management layer via the CAN bus, ensuring the efficiency and stability of data transmission. In terms of data processing and management, the slave battery management layer is responsible for preliminary data collection and feature extraction, the master battery management layer conducts SOC estimation, SOH estimation, and battery pack management, and the battery cluster management layer realizes data storage, transmission, processing, and battery health status self-check. This hierarchical processing mechanism significantly reduces the computing power and storage pressure of the data center. At the same time, the system configures a unique cell management chip for each individual battery, achieving "one chip, one management", which can accurately monitor and control the status of each individual battery, effectively solve the problem of individual consistency within the battery pack, greatly improve the safety, consistency, and service life of the battery pack, and ensure the efficient operation of the battery pack. Description of the Drawings

[0040] To more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application and should not be regarded as limiting the scope of the present application.

[0041] Figure 1 Schematic diagram of the overall architecture of the battery management system in one embodiment;

[0042] Figure 2 Schematic diagram of the internal interaction of the battery management system in one embodiment.

[0043] Explanation of the reference numerals in the drawings: 10, battery management system; 100, slave battery management layer; 110, individual battery; 120, cell management chip; 200, master battery management layer; 300, battery cluster management layer. Detailed Embodiments

[0044] To make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings of the embodiments of the present application. However, it should be understood that the described embodiments are only some exemplary embodiments of the present application, rather than all embodiments. Therefore, the following detailed description of the embodiments of the present application is not intended to limit the scope of the present application to be protected. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0045] It should be noted that the terms "first", "second", etc. in the description and claims of this application are only used to distinguish and describe similar objects, rather than to describe a specific order or sequence, nor can they be understood as indicating or implying relative importance.

[0046] As mentioned above, the current mainstream battery management systems mainly perform state estimation and management based on three basic parameters: voltage, current, and temperature. Although these parameters can reflect the operating state of the battery to a certain extent, their limitations are becoming increasingly prominent. For example, the non-linear relationship between voltage and state of charge (SOC) is easily affected by factors such as temperature and aging, resulting in a large SOC estimation error; the prediction of state of health (SOH) usually relies on a capacity decay model and it is also difficult to accurately capture the battery aging characteristics. In addition, existing systems mostly adopt a centralized architecture, and data collection and processing highly rely on a central server, resulting in high data transmission latency, waste of computing resources, and inability to achieve refined control of individual batteries. These problems exacerbate the inconsistency between individual cells in the battery pack, thereby triggering the risk of thermal runaway, shortening the overall lifespan, and increasing the operation and maintenance costs. Therefore, there is an urgent need for an improved technical solution to solve at least one of the above technical problems. For this purpose, this application proposes a battery management system, a battery pack, and an electric vehicle, which can comprehensively monitor the battery state and improve the estimation accuracy of the battery state.

[0047] As Figure 1 shown, the battery management system 10 of the embodiment of this application includes:

[0048] Multiple slave battery management layers 100, each slave battery management layer 100 is used to collect the battery operating parameters of each individual battery 110 in the corresponding battery pack through the cell management chip 120 set on each individual battery 110, and extract individual battery characteristics from the battery operating parameters. Among them, each individual battery 110 is uniquely configured with a cell management chip 120.

[0049] Multiple master battery management layers 200, each master battery management layer 200 correspondingly obtains each individual battery characteristic extracted by a slave battery management layer 100, estimates the SOC and SOH of the battery pack according to the individual battery characteristics, and performs battery pack management on the corresponding battery pack.

[0050] A battery cluster management layer 300, which is used to obtain the SOC estimation data, SOH estimation data, and battery pack management data estimated by each master battery management layer 200, and perform a self-check on the battery health state according to the battery pack management data.

[0051] In this embodiment, the slave battery management layer 100 refers to the module responsible for managing each single battery 110 in a single battery pack. The battery cell management chip 120 refers to the chip installed on each single battery 110 for real-time monitoring and management of the state of the single battery 110, and its model can be BAT100X. The battery operating parameters refer to various physical quantities used to describe and monitor the operating state of the battery, and specifically may include one or more of the original features such as the voltage data, current data, temperature data, and pressure data of the battery collected. The single battery feature is a feature quantity used to measure the correlation between the battery operating parameters and the SOC.

[0052] In one embodiment, the single battery feature is a correlation feature obtained after processing the battery operating parameters, which can be the battery operating parameters themselves. For example, it can be the operating parameters retained after screening from the battery operating parameters, and the retained operating parameters are suitable for SOC estimation and SOH estimation.

[0053] Specifically, the battery operating parameters include the voltage data, current data, temperature data, and pressure data of the battery. Correlation analysis can be performed between these battery operating parameters and the SOC and SOH to obtain the corresponding correlation coefficients, and the battery operating parameters corresponding to the higher correlation coefficients are screened as the single battery features. For example, the single battery features may include one or more of the following selected first voltage data, first current data, first temperature data, first pressure data, second voltage data, second current data, second temperature data, and second pressure data, etc.

[0054] Correlation analysis can be performed between the voltage data and the SOC of the corresponding battery at the corresponding discharge moment to obtain the first voltage correlation coefficient, and the first voltage data corresponding to the first voltage correlation coefficient exceeding the first voltage coefficient threshold is screened. Among them, SOC refers to the ratio of the remaining battery capacity to the total battery capacity, usually expressed as a percentage, and is used to reflect the charge and discharge state of the battery. This is because there is usually a significant non-linear relationship between the voltage of the battery and the SOC. During the charging or discharging process, the voltage change rate of the battery is closely related to the remaining battery capacity. Especially during the discharging process, the voltage of the battery gradually decreases with the SOC. Person coefficient and spearman coefficient correlation analysis can be performed between the discharge voltage of the battery selected from the same type of battery dataset and the corresponding SOC obtained by the ampere-hour integration method to obtain the first voltage correlation coefficient.

[0055] Differentiate the current data, perform a correlation analysis between the differentiated current data and the SOC of the corresponding battery at the corresponding moment to obtain a first current correlation coefficient, and screen the first current data corresponding to the first current correlation coefficient whose first current correlation coefficient exceeds the first current coefficient threshold. Considering that the change in current will also affect the SOC, by differentiating the current data, the current change rate within a certain moment is obtained, and then a correlation analysis is performed between the current change rate and the corresponding SOC value to obtain the first current correlation coefficient.

[0056] Perform a correlation analysis between the temperature data and the SOC of the corresponding battery at the corresponding moment to obtain a first temperature correlation coefficient, and screen the first temperature data corresponding to the first temperature correlation coefficient whose first temperature correlation coefficient exceeds the first temperature coefficient threshold. This is because high temperature may cause an increase in the internal resistance of the battery, which in turn affects the voltage output of the battery, thereby affecting the estimation of the SOC. Perform a correlation analysis between temperature data such as the temperature change rate, maximum value, minimum value, and average value and the SOC of the battery to obtain the first temperature correlation coefficient.

[0057] Perform a correlation analysis between the pressure data and the SOC of the corresponding battery at the corresponding moment to obtain a first pressure correlation coefficient, and screen the first pressure data corresponding to the first pressure correlation coefficient whose first pressure correlation coefficient exceeds the first pressure coefficient threshold. This is because during the charge and discharge process of the battery, the internal pressure will change with the insertion and extraction of lithium ions. By recording different pressure values and performing a correlation analysis with the corresponding SOC values, the first pressure correlation coefficient can be obtained.

[0058] It should be noted that the above-mentioned various correlation analyses can be performed by using the person coefficient and / or the spearman coefficient to obtain the corresponding correlation coefficients. The magnitudes of the first voltage coefficient threshold, the first current coefficient threshold, the first temperature coefficient threshold, and the first pressure coefficient threshold can be the same or different. For example, they can be set according to the absolute values of the person coefficient (Pearson correlation coefficient) and the spearman coefficient (Spearman rank correlation coefficient). The person coefficient is an index to measure the linear correlation degree between two variables, and its value range is between -1 and 1. A value of 1 indicates a perfect positive correlation, a value of -1 indicates a perfect negative correlation, and a value of 0 indicates no linear correlation. The spearman coefficient is an index to measure the monotonic correlation degree between two variables, and its value range is also between -1 and 1. Different from the person coefficient, the spearman coefficient does not require a linear relationship between variables, but calculates the correlation based on the ranks of the variables (i.e., the order of the data). For example, the above-mentioned coefficient thresholds can all be set at 0.8 to screen out features with strong correlations, thereby improving the prediction accuracy of the unscented Kalman model.

[0059] In another embodiment, correlation analysis can be performed between the voltage data and the SOH of the corresponding battery at the corresponding discharge moment to obtain a second voltage correlation coefficient, and the second voltage data corresponding to the second voltage correlation coefficient exceeding the second voltage coefficient threshold is screened; the current data is differentiated, and correlation analysis is performed between the differentiated current data and the SOH of the corresponding battery at the corresponding moment to obtain a second current correlation coefficient, and the second current data corresponding to the second current correlation coefficient exceeding the second current coefficient threshold is screened; correlation analysis is performed between the temperature data and the SOH of the corresponding battery at the corresponding moment to obtain a second temperature correlation coefficient, and the second temperature data corresponding to the second temperature correlation coefficient exceeding the second temperature coefficient threshold is screened; correlation analysis is performed between the pressure data and the SOH of the corresponding battery at the corresponding moment to obtain a second pressure correlation coefficient, and the second pressure data corresponding to the second pressure correlation coefficient exceeding the second pressure coefficient threshold is screened. It should be understood that the aging of the battery will cause changes in the internal mechanical stress distribution, and the pressure data can reflect the internal state of the battery, especially the deformation of the battery. By recording different pressure values and performing correlation analysis with the corresponding SOH values, the second pressure correlation coefficient is obtained.

[0060] In this embodiment, the master battery management layer refers to the module responsible for managing the entire battery pack, which is used to receive the data transmitted by the slave battery management layer 100, estimate the SOC (state of charge) and SOH (state of health) of the battery pack, perform battery pack management on the battery pack, and transmit the data to the master battery cluster management layer 300. Among them, SOH refers to the ratio of the current performance of the battery to the performance of a new battery, usually expressed as a percentage, and is used to reflect the aging degree and remaining life of the battery. SOH refers to the ratio of the current performance of the battery to the performance of a new battery, usually expressed as a percentage, and is used to reflect the aging degree and remaining life. Battery pack management can include thermal management, balancing management, and charge and discharge management.

[0061] In some embodiments, the SOC estimation and SOH estimation of the battery pack are performed according to the characteristics of the single battery, and battery pack management is performed on the corresponding battery pack, including: performing thermal management on the battery pack according to the temperature calculation result obtained from the characteristics of the single battery, and performing balancing management and charge and discharge management on the battery pack according to the SOC estimation result and the SOH estimation result.

[0062] As a feasible implementation for estimating the state of charge (SOC) of a battery pack based on the characteristics of individual cells, the first fused feature can be formed by performing first fusion processing on the selected first voltage data, the selected first current data, and the selected first temperature data, and the SOC of the battery pack can be estimated based on the first fused feature and the selected first pressure data.

[0063] In some embodiments, the first fused feature and the selected first pressure data can be used as the input of a preset unscented Kalman model, and the unscented Kalman model can be called to estimate the SOC of the battery pack.

[0064] In some embodiments, based on the first fused feature and the selected first pressure data, the state vector of the unscented Kalman model is extended; a set of Sigma points is generated based on the extended state vector and a preset covariance; each Sigma point is propagated through a state transition function to generate the predicted state mean (SOC estimation data) and state covariance of the state vector at the next time step; each Sigma point is propagated through an observation function to generate the predicted observation mean and observation covariance of the predicted observations. Based on the predicted state mean, the state covariance, the predicted observation mean, and the observation covariance, the Kalman gain is calculated; the state vector of the unscented Kalman model is updated based on the Kalman gain and the actual observation value (the first fused feature) at the next time step.

[0065] Specifically, based on the first fused feature F k , the state vector of the unscented Kalman model is extended, and the extended state vector is as shown in (1.1):

[0066] x k =[SOC k ,V 1,k ,V 2,k ,R 0,k ,C n.k ,S k ,F k T (1.1)

[0067] where the subscript k represents the time step k, indicating the k-th moment. SOC k is the state of charge of the battery at time step k (the k-th moment), V 1,k , V 2,k are used to represent the polarization voltages of the two RC networks in the second-order equivalent circuit model at time step k, R 0,k represents the internal resistance of the battery at time step k, C n.k represents the capacity of the battery at time step k, S k ​Denote the input of the first pressure data at time step k as an additional state variable. It should be understood that the second-order equivalent circuit model is a model used to describe the electrical behavior of a battery, which predicts the voltage response of the battery by simulating parameters such as the internal resistance and polarization voltage of the battery. This model usually includes two RC (resistance-capacitance) networks to simulate the polarization effect of the battery.

[0068] The dynamic equation of SOC is shown in (1.2):

[0069]

[0070] where the subscript k + 1 also represents time step k + 1, specifically representing the next moment at the k-th moment (i.e., the (k + 1)-th moment), η is the Coulomb efficiency representing the charge and discharge efficiency, Δt is the time interval between adjacent time steps, and I k is the current at time step k.

[0071] The dynamic equation of the polarization voltage of the first RC network is shown in (1.3):

[0072]

[0073] where R1 is the resistance of the first-order network and C1 is the capacitance of the first-order network.

[0074] The dynamic equation of the polarization voltage of the second RC network is shown in (1.4):

[0075]

[0076] where R2 is the resistance of the second-order network and C2 is the capacitance of the second-order network.

[0077] The internal resistance and capacity attenuation formulas are shown in (1.5) and (1.6):

[0078] R 0,k+1 = R 0,k + f(SOC k , I k , T k ) (1.5)

[0079] C n,k+1 = C n,k + f(SOC k , I k , T k )(1.6)

[0080] The dynamic equation of the first pressure data is shown in (1.7):

[0081] S k+1 = S k + h(SOC k , Ik ,T k )(1.7)

[0082] The dynamic equation of the fusion feature is shown in (1.8):

[0083] F k+1 = ω1V k + ω2T k + ω3I k (1.8)

[0084] where ω1, ω2, and ω3 are the weights of the voltage, temperature, and current features respectively.

[0085] The algorithm steps for generating Sigma points are shown in (1.9):

[0086]

[0087] where P is the preset covariance.

[0088] The weights of the Sigma points are shown in (1.10):

[0089]

[0090] where the subscript m represents the mean value, and c represents the variance. The parameter l is used to adjust the scaling ratio parameter of the overall estimation error. The selection of the parameter α determines the distribution state of the Sigma points; β is a weight coefficient greater than or equal to 0.

[0091] The weighted sum of the prediction results of each Sigma point is calculated to obtain the state covariance and the predicted state mean at the k+1 moment as shown in (1.11) and (1.12):

[0092]

[0093]

[0094] The predicted observation mean and the observation covariance are shown in (1.13) and (1.14):

[0095]

[0096]

[0097] where Y k+1 = h[X k+1 (i) = [V t (i) ,S (i) T (1.15)

[0098] ​

[0099]

[0100] wherein, R is the observation noise depending on the actual situation, OCV is the terminal voltage of the battery, and ω (i) m is the weight of each Sigma point.

[0101] Finally, calculate the Kalman gain as shown in (1.18):

[0102] K k+1 = P XY P -1 YY (1.18)

[0103] Update the state and covariance as shown in (1.19) and (1.20):

[0104]

[0105]

[0106] wherein, Y k+1 is the actual observation value (the first fusion feature at the (k + 1)th moment), and X k+1 is the updated value.

[0107] Based on the above formulas (1.1) to (1.20), in this application, first, the first fusion feature and the first pressure data are used as the inputs of the unscented Kalman model, and the state vector is extended to include elements such as SOC, the polarization voltages of the two RC networks in the second-order equivalent circuit model, the battery internal resistance, the battery capacity, and the first pressure data. Then, a set of Sigma points is generated based on the extended state vector and the preset covariance, and these Sigma points are propagated through the state transition function and the observation function to obtain the predicted state mean (i.e., the SOC estimation data) of the state vector at the next time step and the state covariance, as well as the predicted observation mean and the observation covariance of the predicted observation values. Next, the Kalman gain is calculated using these predicted values, and the state vector of the unscented Kalman model is updated in combination with the actual observation value (the first fusion feature) at the next time step. This series of steps is repeated at each time step. By continuously updating the state vector and the covariance, the true SOC value is gradually approximated, and finally, the SOC estimation data at each moment is obtained, forming an SOC vector, where each element represents the SOC estimation value at the corresponding moment.

[0108] In this embodiment, a feature vector is extracted based on the first fusion feature and the selected first pressure data. A second-order equivalent circuit model is used to model the dynamic characteristics of the battery. This model takes into account key parameters such as the ohmic internal resistance, polarization resistance, and polarization capacitance of the battery, and can accurately describe the working state of the battery. Based on the unscented Kalman filter (UKF) algorithm, the state of charge (SOC) of the battery is estimated in real time. UKF avoids the linearization error of the traditional Kalman filter by performing an unscented transform on the nonlinear characteristics of the system state, improving the estimation accuracy. At the same time, the particle swarm optimization algorithm (PSO) is introduced to optimize the unknown parameters in the second-order equivalent circuit model, such as the ohmic internal resistance, polarization resistance, and capacitance, to ensure a high degree of matching between the model parameters and the actual characteristics of the battery, further improving the accuracy of SOC estimation. Through this solution, real-time high-precision estimation of the SOC of the battery pack is achieved, with the error controlled within 2%. The output SOC estimation data provides an important input for the subsequent estimation of the state of health (SOH) of the battery pack, laying a solid foundation for the assessment of the health state of the battery and providing reliable data support for the management of the battery system.

[0109] As a feasible implementation method for estimating the SOH of a battery pack based on the characteristics of single cells, second fusion processing can be performed on the selected second voltage data, selected second current data, and selected second temperature data to form a second fusion feature, and the SOH of the battery pack is estimated based on the second fusion feature and the selected second pressure data.

[0110] In some embodiments, the SOH of the battery pack can be estimated through a dynamic graph constraint and a physical memory gating network based on the second fusion feature, the selected second pressure data, and the SOC estimation data.

[0111] Specifically, all signals are processed according to channel normalization, and the calculation formula is as shown in (1.21):

[0112]

[0113] where, X norm i is the dimensionless data after normalization, i represents the signal type, σ i is the standard deviation, indicating the degree of data dispersion, and X raw i is the unprocessed data.

[0114] Definition of graph nodes, the node feature vector at each moment within the sliding window τ, the calculation formula is as shown in (1.22):

[0115]

[0116] where, R represents the dimension number of the vector.

[0117] Adaptive adjacency matrix calculation, dynamically calculate the edge weights through GATv2, and the calculation formula is shown in (1.23):

[0118] e ij = a T ·LeakyReLU(W q h i (0) + W k h j (0) )(1.23)

[0119] Among them, e ij represents the unnormalized attention score of node i to node j, quantifies the intensity of interaction between two nodes, and the larger the value, the stronger the correlation. α represents the trainable attention parameter vector, controls the non-linear expression ability of the attention mechanism, and extracts the high-order combination relationship between features. W w is the query change matrix, used to actively retrieve the information of other nodes. W k is the key transformation matrix, used to match the correlation when being queried by other nodes

[0120] , is the initial feature vector of node i, is the initial feature vector of node j.

[0121] The normalized attention weight, and the calculation formula is shown in (1.24):

[0122]

[0123] Among them, W q , W k are learnable parameter matrices, and α is the attention coefficient vector. α ij is the normalized edge weight, N(i) is the dynamic neighbor set of node i. In the time-space graph structure, it dynamically captures the relevant nodes or adjacent battery cells in the current time step and its front and back windows. exp is the exponential function, which maps the original score to the positive space, preserves the monotonicity of the size relationship, and at the same time amplifies the difference of high score values to prevent negative values from interfering.

[0124] Based on the time window for graph convolution update, splice all node features within the window to generate the supernode matrix, and the calculation formula is shown in (1.25):

[0125] H [t-τ:t] = Concat(h t-τ (0) ,…,h t (0) ) ∈ R τ×5 (1.25)

[0126] Among them, H [t-τ:t] is the super-node matrix, which is formed by horizontally concatenating the node feature vectors at all times within the time window. Concat(·) is the operation symbol, and vertically stacks multiple feature vectors into a matrix in the time dimension.

[0127] Extract the temporal correlation features through 1D convolution, and the calculation formula is as shown in (1.26):

[0128] A t = Conv1D(H [t-τ:t] ; kernel = 3) ∈ R 5×5 (1.26)

[0129] Among them, Conv1D represents the one-dimensional convolution operation, slides the convolution kernel along the time axis to extract local patterns, captures the temporal correlation between sensor variables, and converts the time context information into the dynamic weights between nodes; kernel = 3 represents the convolution kernel size, which controls the local time range perceived by the model. A longer window can capture slow changes, and a shorter window is more sensitive to instantaneous fluctuations.

[0130] Temporal feature extraction, that is, fusing the second current data, the second voltage data, and the second temperature data to extract the feature vector, and the calculation formula is as shown in (1.27):

[0131] f1 = Bi-LSTM(Conv1D([I, V, T])) (1.27)

[0132] It should be noted that the parameters of the 1D-CNN and Bi-LSTM layers can be set according to specific situations.

[0133] The graph structure features of the dynamic graph attention, stacking 3 layers of GAT to generate graph features, and the calculation formula is as shown in (1.28):

[0134] f2 = GAT (3) (H (0) , A t ) (1.28)

[0135] Among them, GAT is the graph attention network, A t is the dynamic adjacency matrix, which describes the dynamic association strength between sensor variables, and f2 is the graph attention output.

[0136] SOC pre The time Transformer encodes the historical SOC sequence, and the calculation formula is as shown in (1.29):

[0137]

[0138] Among them, Softmax is a row-wise normalization function that converts the similarity scores into a probability distribution with a sum of 1, and d k is the dimension of the key vector, which is used to scale the dot product to prevent the gradient from vanishing due to an overly large magnitude. Q soc K T is the query-key association matrix that calculates the similarity scores for each pair of queries and keys, reflecting the basis of the attention weights. V soc is the value matrix that carries the actual information to be transmitted, and f3 is the final attention output.

[0139] Based on gated multi-modal fusion, a dynamic weighted gating vector is calculated using the formula shown in (1.30):

[0140] g = σ(W g ·[f1; f2; f3] + b g ) ∈ R 3 (1.30)

[0141] Among them, g is the gating vector, σ is the sigmoid function that controls the weights of each branch in the interval [0, 1]. b g is the gating bias term that adjusts the output baseline of the activation function to prevent the weights of some modalities from being completely suppressed in the initial stage; W g is the gating weight matrix that learns the collaborative weights between modalities and determines the importance of each modality.

[0142] The above features are fused using the formula shown in (1.31):

[0143] (1.31)

[0144] Among them, is the fused feature vector that synthesizes multi-modal information for further estimating the SOH, where ⊙ represents element-wise multiplication.

[0145] The embedding of physical constraints, the correlation constraint between pressure and capacity decay, and the consistency term based on the lithium-ion concentration gradient are shown in the calculation formula (1.32):

[0146]

[0147] Among them, T(t) is the temperature sequence that affects the lithium-ion diffusion rate and the kinetics of electrochemical reactions. λ1, λ2, and λ3 are physical coefficients calibrated through experiments, and ΔSOH pred is the difference between the t-th moment and the (t - 1)-th moment.

[0148] The final feature is mapped to a specified interval to obtain the SOH estimation data, as shown in the calculation formula (1.33):

[0149] SOH pred=0.7+0.3·σ(W out f fusion +b out ) (1.33)

[0150] The default battery SOH is in the range of [0.7, 1], which can be changed according to the actual situation. out is the weight matrix of the output layer, b out is the bias term of the output layer, which is used to adjust the baseline prediction of the model.

[0151] In this embodiment, a battery SOH estimation method based on a multimodal dynamic graph network is proposed. A multimodal graph attention network (PM-GAT) algorithm based on physical constraints is proposed. High-precision SOH estimation is achieved through multi-source signal collaborative perception and dynamic interaction modeling. Current, voltage, temperature, pressure and historical SOC estimation data are used as nodes, and the correlation weights between parameters are dynamically calculated through an adaptive graph attention mechanism (GATv2); the graph structure is updated based on a sliding window to capture the coupling relationship between pressure fluctuations and electrochemical parameters in real time (such as the correlation between pressure gradient and capacity attenuation in the full charge stage). Based on multi-level feature fusion, time series features, and graph features, a gated cross-modal attention mechanism is used to hierarchically fuse multi-source features to obtain features for estimating SOH, and a physical constraint embedding mechanism is introduced, and a modified Sigmoid function is used to constrain the prediction results to a reasonable range.

[0152] In this embodiment, the battery cluster management layer 300 refers to a module responsible for managing the entire battery cluster, and is used to aggregate and analyze the SOC (state of charge), SOH (state of health) estimation data and battery pack management data from each main control battery management layer 200, to implement self-checking of the battery management system 10, and to ensure the safe operation of each battery pack. The battery pack management data includes management information such as thermal management, balancing management, and charge and discharge management of the battery pack.

[0153] In some embodiments, the self-check of the battery management system 10 includes: monitoring the voltage, temperature, pressure, current and other data of the single battery 110 in each battery pack, and checking the total voltage, balancing current, insulation condition, etc. of the battery pack. Through these checks, the battery cluster management layer can promptly detect abnormal conditions in the battery pack, such as overcharge, over discharge, abnormal temperature, abnormal pressure, etc.

[0154] In some embodiments, the battery cluster management layer 300 is also used to collect all alarm information and determine whether the multi-level alarm information is not empty. If there is alarm information, a fault self-check reset is performed, and each level of alarm information is checked from high to low level by level, and the system is reset and the alarm information is cleared after the fault is eliminated.

[0155] In some embodiments, the battery cluster management layer 300 is further configured to send a control signal to the battery cell management chip 120 uniquely corresponding to the single battery cell 110 with an abnormality when detecting an abnormality in one of the single battery cells 110, so that the single battery cell 110 with the abnormality is adjusted according to the control signal. Among them, the abnormality may include temperature abnormality, pressure abnormality, overcharge, over-discharge, excessive battery balancing difference, etc.

[0156] In some embodiments, when the abnormality is an excessive battery balancing difference, if the capacity of the single battery cell 110 with the abnormality is too high, the battery cell management chip 120 uniquely corresponding to the single battery cell 110 with the abnormality consumes the battery capacity in a passive balancing manner with full-load operation based on the control signal; if the capacity of the single battery cell 110 with the abnormality is too low, the battery cell management chip 120 uniquely corresponding to the single battery cell 110 with the abnormality consumes the battery capacity in a low-load operation manner of maintaining dormancy based on the control signal.

[0157] In the battery management system 10 proposed in the embodiments of the present application, by setting a plurality of slave battery management layers 100, a plurality of master battery management layers 200, and a battery cluster management layer 300, a hierarchical distributed architecture is formed. Among them, the slave battery management layer 100 can collect key data such as the voltage, temperature, pressure, and current of the single battery cell 110 in real time. These data are transmitted to the master battery management layer 200 through Bluetooth MESH and then transmitted to the battery cluster management layer 300 via the CAN bus, ensuring the efficiency and stability of data transmission. In terms of data processing and management, the slave battery management layer 100 is responsible for preliminary data collection and feature extraction, the master battery management layer 200 performs SOC estimation, SOH estimation, and battery pack management, and the battery cluster management layer 300 realizes data storage, transmission, processing, and battery health status self-check. This hierarchical processing mechanism significantly reduces the computing power and storage pressure of the data center. At the same time, the system configures a unique battery cell management chip 120 for each single battery cell 110, realizing "one chip for one management", which can accurately monitor and control the state of each single battery cell 110, effectively solve the problem of monomer consistency in the battery pack, greatly improve the safety, consistency, and service life of the battery pack, and ensure the efficient operation of the battery pack.

[0158] In one embodiment, as Figure 2As shown, communication between the slave battery management layer 100 and the master battery management layer 200 is carried out via Bluetooth; communication between the master battery management layer 200 and the battery cluster management layer 300 is carried out via a CAN (Controller Area Network) bus; communication between the cell management chips 120 and between the cell management chips 120 and the master battery management layer 200 is carried out via Bluetooth.

[0159] As a networking address configuration strategy for Bluetooth communication, the master battery management layer 200 is responsible for allocating network identifiers to ensure that each single battery 110 node in each slave battery management layer 100 can be uniquely identified in the network. The cell management chip 120 on each single battery 110 generates its own node identifier to ensure the uniqueness of each single battery 110 node in the network. After receiving the network identifier broadcast by the master battery management layer 200, the cell management chip 120 on each single battery 110 generates a unique ID and sends it to the master battery management layer 200 for registration. The master battery management layer 200 maintains an address mapping table to record the addresses and physical locations of each single battery 110 node.

[0160] Exemplarily, when an abnormal battery with abnormal temperature is detected, the master battery management layer 200 precisely directs the control signal to the cell management chip 120 matching the abnormal battery via Bluetooth MESH transmission. When the chip coding IDs match, the control signal is output to make the corresponding temperature control system of the abnormal battery work to keep the battery operating within a suitable temperature range.

[0161] As a networking key configuration for Bluetooth communication, asymmetric encryption is used for key distribution and symmetric encryption is used for data transmission. Specifically, the master battery management layer 200 generates a pair of public and private keys. The public key is used for encryption and the private key is used for decryption. The cell management chip 120 of each single battery 110 node generates a random symmetric key, encrypts it using the public key of the master battery management layer 200, and sends it to the master battery management layer 200. The master battery management layer 200 decrypts it using the private key to obtain the symmetric key of the cell management chip 120 of each single battery 110 node. The master battery management layer 200 stores and maintains the key table and regularly updates the keys to prevent being cracked.

[0162] As a data communication strategy, a mesh topology is adopted. The cell management chips 120 of each single battery 110 node can communicate directly with each other, use lightweight routing protocols such as AODV to implement dynamic routing, and at the same time, dynamically adjust the data transmission priority according to the importance and real-time nature of the data. Fault alarm data is given priority in transmission, and the hash function SHA-256 is used to ensure data integrity.

[0163] In the battery management system 10 proposed in the embodiments of this application, through technical means such as optimizing communication methods, hierarchical address allocation strategies, key management, and dynamic routing, efficient, reliable, and intelligent management of the battery pack is achieved. The system adopts a communication method combining Bluetooth and CAN bus to ensure the flexibility and reliability of data transmission. The master battery management layer 200 manages each node in the network through a hierarchical address allocation strategy to avoid address conflicts and improve network stability. At the same time, through a key configuration strategy combining asymmetric encryption and symmetric encryption, the security and integrity of data transmission are ensured. In addition, the system adopts a mesh topology and a lightweight routing protocol to dynamically adjust the data transmission priority to ensure the timely transmission of important data. In particular, each single battery 110 is equipped with a cell management chip 120 to achieve one-chip-one-management, ensuring the consistency of the single batteries 110 in the battery pack and improving the overall performance and lifespan of the battery pack.

[0164] In addition, the embodiments of this application also provide a battery pack based on the aforementioned battery management system 10. Benefiting from the aforementioned battery management system 10, this battery pack realizes precise monitoring and management of the single batteries 110 through the battery management system 10, ensuring the safe operation and performance optimization of the battery pack. Specifically, each single battery 110 in the battery pack is equipped with a cell management chip 120, and these chips communicate with the slave battery management layer 100 through Bluetooth, while the slave battery management layer 100 transmits data to the master battery management layer 200 through Bluetooth. The master battery management layer 200 is responsible for data aggregation and analysis, and transmits the data to the battery cluster management layer 300 through the CAN bus to achieve monitoring and management of the entire battery pack. This hierarchical distributed architecture not only improves the reliability of the system but also reduces the computing power and storage pressure of the data center, realizing refined control and management of the single batteries 110.

[0165] In addition, an embodiment of the present application further provides an electric vehicle, which includes the above battery pack. By integrating an advanced battery management system 10, the electric vehicle achieves efficient management and safety control of the battery pack. Specifically, each single battery 110 in the battery pack of the electric vehicle is equipped with a cell management chip 120, and these chips communicate with the slave battery management layer 100 via Bluetooth, while the slave battery management layer 100 transmits data to the master battery management layer 200 via Bluetooth. The master battery management layer 200 is responsible for summarizing and analyzing the data, and transmits the data to the battery cluster management layer 300 via the CAN bus to achieve monitoring and management of the entire battery pack. This hierarchical distributed architecture not only improves the reliability of the system, but also reduces the computing power and storage pressure of the data center, and realizes fine control and management of the single battery 110. In this way, the electric vehicle can monitor and adjust the state of the battery pack in real time, ensure the safe operation of the vehicle and the long life of the battery, thereby improving the overall performance and user experience of the electric vehicle.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

[0167] In addition, those skilled in the art can understand that although some embodiments herein include certain features included in other embodiments but not other features, the combination of the features of different embodiments means that it is within the scope of the present application and forms different embodiments. For example, in the above claims, any of the claimed embodiments can be used in any combination. The information disclosed in this background art section is only intended to deepen the understanding of the overall background art of the present application, and should not be regarded as an admission or any form of implication that this information constitutes prior art known to those skilled in the art.

Claims

1. A battery management system, characterized in that, The system includes: Multiple slave battery management layers, each of which is used to collect the battery operating parameters of each single cell in the corresponding battery pack through the cell management chips set on each single cell in the corresponding battery pack, and extract the single cell characteristics from the battery operating parameters. Among them, each single cell is uniquely configured with a cell management chip; Multiple master battery management layers, each of which correspondingly obtains each single cell characteristic extracted by a slave battery management layer, estimates the SOC and SOH of the battery pack according to the single cell characteristics, and manages the corresponding battery pack; A battery cluster management layer, which is used to obtain the SOC estimation data, SOH estimation data and battery pack management data estimated by each master battery management layer, and perform self-check on the battery health status according to the battery pack management data.

2. The battery management system according to claim 1, wherein The battery operating parameters include the voltage data, current data, temperature data and pressure data of the battery. Extracting the single cell characteristics from the battery operating parameters includes: Performing correlation analysis between the voltage data and the SOC of the corresponding battery at the corresponding discharge moment to obtain a first voltage correlation coefficient, and screening the first voltage data corresponding to the first voltage correlation coefficient whose first voltage correlation coefficient exceeds the first voltage coefficient threshold; Performing differential processing on the current data, performing correlation analysis between the differentially processed current data and the SOC of the corresponding battery at the corresponding moment to obtain a first current correlation coefficient, and screening the first current data corresponding to the first current correlation coefficient whose first current correlation coefficient exceeds the first current coefficient threshold; Performing correlation analysis between the temperature data and the SOC of the corresponding battery at the corresponding moment to obtain a first temperature correlation coefficient, and screening the first temperature data corresponding to the first temperature correlation coefficient whose first temperature correlation coefficient exceeds the first temperature coefficient threshold; Performing correlation analysis between the pressure data and the SOC of the corresponding battery at the corresponding moment to obtain a first pressure correlation coefficient, and screening the first pressure data corresponding to the first pressure correlation coefficient whose first pressure correlation coefficient exceeds the first pressure coefficient threshold; The estimating the SOC and SOH of the battery pack according to the single cell characteristics includes: Performing a first fusion process on the selected first voltage data, the selected first current data, and the selected first temperature data to form a first fusion feature, and estimating the SOC of the battery pack based on the first fusion feature and the selected first pressure data.

3. The battery management system according to claim 2, characterized in that, The estimating the SOC of the battery pack based on the first fusion feature and the selected first pressure data includes: Taking the first fusion feature and the selected first pressure data as the input of a preset unscented Kalman model, and calling the unscented Kalman model to estimate the SOC of the battery pack.

4. The battery management system according to claim 1, characterized in that, The battery operating parameters include the voltage data, current data, temperature data and pressure data of the battery. Extracting the single cell characteristics from the battery operating parameters includes: Perform a correlation analysis between the voltage data and the SOH of the corresponding battery at the corresponding discharge moment to obtain a second voltage correlation coefficient, and screen the second voltage data corresponding to the second voltage correlation coefficient whose second voltage correlation coefficient exceeds the second voltage coefficient threshold; Perform a differential processing on the current data, perform a correlation analysis between the differentially processed current data and the SOH of the corresponding battery at the corresponding moment to obtain a second current correlation coefficient, and screen the second current data corresponding to the second current correlation coefficient whose second current correlation coefficient exceeds the second current coefficient threshold; Perform a correlation analysis between the temperature data and the SOH of the corresponding battery at the corresponding moment to obtain a second temperature correlation coefficient, and screen the second temperature data corresponding to the second temperature correlation coefficient whose second temperature correlation coefficient exceeds the second temperature coefficient threshold; Perform a correlation analysis between the pressure data and the SOH of the corresponding battery at the corresponding moment to obtain a second pressure correlation coefficient, and screen the second pressure data corresponding to the second pressure correlation coefficient whose second pressure correlation coefficient exceeds the second pressure coefficient threshold; The SOC estimation and SOH estimation of the battery pack according to the characteristics of the single battery include: Perform a second fusion processing on the screened second voltage data, the screened second current data, and the screened second temperature data to form a second fusion feature, and perform an SOH estimation of the battery pack based on the second fusion feature and the screened second pressure data.

5. The battery management system according to claim 4, wherein, The SOH estimation of the battery pack based on the second fusion feature and the second pressure correlation coefficient includes: Based on the second fusion feature, the screened second pressure data, and the SOC estimation data, perform an SOH estimation of the battery pack through a dynamic graph constraint and a physical memory gating network.

6. The battery management system according to claim 1, wherein The battery cluster management layer is further configured to send a control signal to the cell management chip uniquely corresponding to the single battery with an anomaly when detecting an anomaly in one of the single batteries, so that the single battery with the anomaly is adjusted according to the control signal.

7. The battery management system according to claim 6, characterized in that, When the anomaly is that the battery balance difference is too large, If the capacity of the single battery with the anomaly is too high, the cell management chip uniquely corresponding to the single battery with the anomaly consumes the battery capacity in a passive balancing manner with full-load operation based on the control signal; If the capacity of the single battery with the anomaly is too low, the cell management chip uniquely corresponding to the single battery with the anomaly consumes the battery capacity in a low-load operation mode of maintaining dormancy based on the control signal.

8. The battery management system according to claim 1, characterized in that, The battery pack management includes thermal management, charge and discharge management, and balance management; The slave battery management layer communicates with the master battery management layer via Bluetooth; The master battery management layer communicates with the battery cluster management layer via a CAN bus; The model of the cell management chip is BAT100X, and the cell management chips communicate with each other and with the master battery management layer via Bluetooth; Performing SOC estimation and SOH estimation on the battery pack according to the characteristics of the single battery cells, and performing battery pack management on the corresponding battery pack, including: performing thermal management on the battery pack according to the temperature calculation result obtained from the characteristics of the single battery cells, and performing balancing management and charge-discharge management on the battery pack according to the SOC estimation result and the SOH estimation result.

9. A battery pack, characterized in that, The battery pack includes the battery management system according to any one of claims 1 to 8.

10. An electric vehicle, characterized in that, The electric vehicle includes the battery pack according to claim 9.

Images