Distributed energy management system for prolonging cycle life of lithium battery

Abstract

The invention relates to a distributed energy management system for prolonging the cycle life of a lithium battery. The system comprises an acquisition unit used for acquiring battery state characteristic data of single batteries in a lithium battery pack; the identification unit is used for carrying out charge and discharge abnormal mode identification on the battery state characteristic data to obtain charge and discharge abnormal event data; the decomposition unit is used for performing hierarchical characteristic decomposition and multi-dimensional fusion based on the charging and discharging abnormal event data to obtain a cycle life attenuation characteristic matrix; and the equalization control unit is used for carrying out life equalization control based on the cycle life attenuation characteristic matrix to obtain a target charge and discharge coefficient. According to the invention, the local over-charge and over-discharge phenomena are effectively inhibited, the service life loss of the battery pack is reduced, and the energy balance adjustment of the multi-node control unit is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium batteries, and in particular to a distributed energy management system for improving the cycle life of lithium batteries. Background Art

[0002] Traditional battery management systems mainly adopt a centralized architecture and rely on global battery pack parameters for status monitoring and balancing control. In large-scale battery pack management, they face problems such as heavy computing load, slow response speed, and high risk of single-point failure.

[0003] In recent years, the scale and power density of battery packs have been continuously improved, and the consistency differences and operating conditions of battery cells have become increasingly prominent, which has led to local overcharge, overdischarge and temperature abnormalities in the battery pack, accelerating the battery's cycle life decay. The existing management method based on voltage balancing is difficult to effectively identify and suppress the chain effects of these abnormal events, and cannot achieve active optimization and control of the battery pack life. Summary of the invention

[0004] The main purpose of the present invention is to provide a distributed energy management system for improving the cycle life of lithium batteries. The present invention effectively suppresses local overcharging and over-discharging phenomena, reduces the life loss of battery packs, and realizes energy balance regulation of multi-node control units.

[0005] To achieve the above objectives, the first aspect of the present invention provides a distributed energy management system for improving the cycle life of lithium batteries, the system comprising: A collection unit, used to collect battery status characteristic data of battery cells in a lithium battery pack; an identification unit, configured to identify abnormal charging and discharging patterns on the battery status characteristic data to obtain abnormal charging and discharging event data; A decomposition unit, used for performing hierarchical feature decomposition and multi-dimensional fusion based on the abnormal charge and discharge event data to obtain a cycle life attenuation feature matrix; A balancing control unit is used to perform life balancing control based on the cycle life attenuation characteristic matrix to obtain a target charge and discharge coefficient.

[0006] In combination with the first aspect, in a first implementation of the first aspect of the present invention, the acquisition unit is specifically used for: The voltage data of the battery cells in the lithium battery pack are collected through the voltage collection module, the temperature data are collected through the temperature collection module, and the internal resistance data are collected through the internal resistance measurement module to obtain the original collected data; Eliminating high-frequency noise from the original collected data to obtain filtered processed data, and calculating the voltage change rate, temperature gradient, and internal resistance increment of the filtered processed data to obtain characteristic quantity data; Packing the characteristic data into data frames and transmitting them to the upper control unit via the CAN bus to obtain a transmission data packet; Calculating the change rate of the voltage data and the temperature data in the transmission data packet, dynamically adjusting the sampling frequency according to the change rate value, and obtaining an adaptive sampling parameter; The adaptive sampling parameters are returned to the voltage acquisition module, the temperature acquisition module and the internal resistance measurement module, and the sampling frequency is updated and configured to obtain the battery status characteristic data.

[0007] In combination with the first aspect, in a second implementation of the first aspect of the present invention, the identification unit is specifically configured to: Based on the battery state characteristic data, the battery terminal voltage, surface temperature, and internal resistance change are constructed as observation vectors, and the state of charge, health state, and polarization voltage are constructed as state vectors to obtain a Kalman filter state equation; Calculating the mean and variance of the Kalman filter state equation, updating the process noise covariance matrix and the observation noise covariance matrix through recursive operation, and obtaining noise characteristic parameters; Substituting the noise characteristic parameters into the strong tracking filter algorithm, calculating the optimal attenuation factor through the exponential attenuation function, correcting the prediction error covariance matrix, and obtaining the state prediction result; Perform multi-dimensional comparison on the state prediction result with preset voltage threshold, temperature threshold, and internal resistance threshold to obtain abnormal deviation data; The abnormal deviation data is cumulatively counted, and when the abnormal duration of any state quantity exceeds the corresponding time threshold, an abnormal trigger signal is obtained, and the abnormal trigger signal is graded according to the deviation magnitude to obtain the charging and discharging abnormal event data.

[0008] In combination with the first aspect, in a third implementation of the first aspect of the present invention, the decomposition unit further includes: A division subunit is used to divide the abnormal charge and discharge event data into groups of N adjacent battery cells, and calculate the voltage differential capacity, temperature distribution and internal resistance characteristics of each group of divided battery cells to obtain feature vector data; An extraction subunit, used to calculate the eigenvector mutual correlation coefficients between the calculation subdomains based on the eigenvector data to obtain a spatial correlation matrix, and to extract the frequency band energy distribution of the voltage data and the temperature data to obtain a time-frequency feature matrix; A calculation subunit, used for calculating core tensor coefficients and a three-dimensional factor matrix according to the spatial correlation matrix and the time-frequency feature matrix to obtain a multimodal feature tensor; A construction subunit is used to perform low-rank approximate calculation on the multimodal feature tensor to obtain feature component data, and to construct an attenuation matrix of the spatiotemporal evolution characteristics based on the feature component data to obtain a cycle life attenuation feature matrix.

[0009] In combination with the first aspect, in a fourth implementation manner of the first aspect of the present invention, the computing subunit is specifically configured to: Rearranging the spatial correlation matrix according to the physical position relationship of the battery cells, and reorganizing the time-frequency feature matrix according to the time sequence to obtain initial tensor data; The initial tensor data is normalized, and normalization coefficients of the time dimension, the space dimension, and the feature dimension are calculated to obtain a normalized tensor; Decomposing the normalized tensor into three orthogonal basis matrices, calculating a characteristic basis matrix of a time dimension, a characteristic basis matrix of a space dimension, and a characteristic basis matrix of a characteristic dimension, and obtaining an orthogonal decomposition result; Performing Tucker decomposition operation on the orthogonal decomposition result to obtain a tensor decomposition equation, and substituting the tensor decomposition equation into an alternating least squares method for iterative solution to obtain optimized tensor parameters; A tensor reconstruction operation is performed on the optimized tensor parameters to obtain a multimodal feature tensor.

[0010] In combination with the first aspect, in a fifth implementation manner of the first aspect of the present invention, the balancing control unit is specifically configured to: Substituting the cycle life attenuation characteristic matrix into the Thevenin equivalent circuit model, calculating the polarization resistance, polarization capacitance and state of charge value of each battery cell, and obtaining the underlying state data; Input the bottom state data and the influencing relationship parameters between the battery cells into the particle swarm filter calculation model, calculate the charge state distribution curve of the entire battery pack, and obtain the middle state data; Performing difference analysis on the state of charge and health state of each battery cell in the middle-level state data, calculating the charge and discharge weight factor of each cell, and obtaining the distribution weight coefficient; Substituting the distribution weight coefficient into the current optimization model, calculating the corresponding optimal charge and discharge current limit for the battery cells that deviate from the average value, and obtaining the balancing control parameter; The balance control parameter and the electrochemical characteristic correlation matrix are cross-calculated to obtain risk distribution data, and the charge and discharge rate of the battery pack is planned by region based on the risk distribution data to obtain the target charge and discharge coefficient.

[0011] In combination with the first aspect, in a fifth implementation of the first aspect of the present invention, the distributed energy management system for improving the cycle life of lithium batteries further includes: The real-time constraint unit is used to input the target charge and discharge coefficient and the charge and discharge depth, rate and temperature parameters into the electrochemical stress model, calculate the fatigue stress distribution of the single cell, and obtain electrochemical damage data; perform time-series accumulation operation on the electrochemical damage data according to the linear fatigue damage criterion, construct the life attenuation curve of each battery cell, and obtain a fatigue damage degree matrix; divide the fatigue damage degree matrix into gears according to the degree of damage, set corresponding charge and discharge rate upper limits for different gears, and obtain stratified restriction parameters; calculate the power attenuation ratio of each single cell based on the stratified restriction parameters and the battery surface temperature data, and obtain the temperature correction coefficient; calculate the power allocation strategy inside the battery pack in combination with the temperature correction coefficient and the total power demand of the system, and obtain the power optimization parameter; perform charge and discharge stage-by-stage processing on the power optimization parameter, set a dynamic current limit interval for each battery cell, and obtain the life protection control parameter.

[0012] In combination with the first aspect, in a sixth implementation of the first aspect of the present invention, the distributed energy management system for improving the cycle life of lithium batteries further includes: The energy balance adjustment unit is used to: input the life protection control parameter into the model prediction controller, calculate the local protection control instruction sequence of the single cell, and obtain the target control data; input the target control data into the Coulomb efficiency compensation model, calculate the charge and discharge compensation current of each single cell, and obtain the balance compensation parameter; input the balance compensation parameter into the data interaction network between adjacent control nodes, calculate the optimal control quantity of each control node, and obtain the coordinated control variable; substitute the coordinated control variable into the Lyapunov function, calculate the convergence speed of the state consistency between the nodes, and obtain the stability evaluation data; based on the stability evaluation data, calculate the reconstruction strategy of the control network, and obtain the fault-tolerant control parameter; substitute the fault-tolerant control parameter into the equivalent circuit network, calculate the energy transmission path and power size between the nodes, and obtain the balance control instruction.

[0013] In summary, the technical solution provided by the present invention improves the recognition accuracy of abnormal charging and discharging events through distributed data acquisition and multiple threshold Kalman filtering algorithms, and realizes the rapid capture and positioning of abnormal states of battery cells; adopts hierarchical feature decomposition and tensor decomposition algorithms to construct an electrochemical feature association matrix, accurately characterizes the spatiotemporal evolution characteristics of abnormal events, and provides a reliable data basis for life balance control; based on distributed state recursion and dynamic weight allocation strategies, adaptive balance control inside the battery pack is realized, which effectively suppresses local overcharge and over-discharge phenomena and reduces the life loss of the battery pack; combined with the electrochemical stress model and hierarchical restriction function, a cumulative damage assessment system is established, and the cycle life attenuation rate of the battery pack is reduced by real-time constraints on charging and discharging conditions; a distributed collaborative learning network is introduced to realize energy balance adjustment of multi-node control units, ensure the reliable operation of the system under single-point failure conditions, and significantly improve the stability and robustness of battery pack management. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of a distributed energy management system for improving the cycle life of lithium batteries in one embodiment of the present invention.

[0015] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] Reference Figure 1 This embodiment provides a distributed energy management system for improving the cycle life of lithium batteries, including the following steps: The acquisition unit 101 is used to acquire the battery status characteristic data of the battery cells in the lithium battery pack; Specifically, the voltage acquisition module is used to monitor the voltage change of each battery cell in the lithium battery pack in real time to ensure the accurate measurement of the battery voltage fluctuation. At the same time, the temperature acquisition module detects the temperature of the battery cell and provides high-precision thermal management data, while the internal resistance measurement module is used to measure the internal impedance of the battery cell to evaluate the health status and aging trend of the battery. The acquisition of these data constitutes the original acquisition data of the system. The original acquisition data is subjected to high-frequency noise elimination. For example, signal filtering technology, such as low-pass filtering, Kalman filtering and other methods, is used to effectively eliminate high-frequency noise to ensure that the acquired voltage, temperature and internal resistance data are more stable and reliable. After filtering, the key characteristic quantities of the battery state are calculated, including the voltage change rate, temperature gradient and internal resistance increment. Among them, the voltage change rate reflects the severity of the voltage fluctuation of the battery during the charging and discharging process, which is used to evaluate the load adaptability of the battery, while the temperature gradient reveals the thermal distribution of the battery during operation. The calculation of the internal resistance increment can identify the trend of battery aging, help predict the battery life and take balancing measures in advance. The characteristic quantity data is packaged into data frames and transmitted to the upper control unit through the CAN bus according to the standardized communication protocol to ensure the real-time and reliability of the data. The data packets transmitted on the CAN bus contain key information such as voltage, temperature, and internal resistance to support the intelligent decision-making of the system. In order to optimize the data acquisition efficiency, an adaptive sampling mechanism is introduced to dynamically adjust the sampling frequency by calculating the change rate of the voltage data and temperature data in the transmission data packet. When the voltage or temperature changes more drastically, the sampling frequency is automatically increased to capture more detailed data, thereby enhancing the abnormality detection capability. When the battery state is relatively stable, the sampling frequency is reduced accordingly to reduce data redundancy and power consumption. In order to optimize the acquisition process, the adaptive sampling parameters are fed back to the voltage acquisition module, the temperature acquisition module, and the internal resistance measurement module, so that each acquisition unit is updated and configured according to the latest dynamic parameters, so that the system is always in the best working state. The battery status characteristic data is obtained.

[0018] The identification unit 102 is used to identify the abnormal charging and discharging patterns of the battery status characteristic data to obtain abnormal charging and discharging event data; In this embodiment, the state estimation model is constructed using the battery state characteristic data, wherein the battery terminal voltage, surface temperature and internal resistance change together constitute the observation vector, and the state of charge (SOC), state of health (SOH) and polarization voltage are used as the state vector of the system. Through the relationship between these variables, a state equation based on Kalman filtering is established to dynamically track the operating state of the battery and effectively suppress the accumulation of measurement errors in complex working environments. On this basis, the Kalman filter state equation is mathematically processed, the mean and variance of the state variables are calculated, and the process noise covariance matrix and the observation noise covariance matrix are updated in real time using recursive operations, so as to accurately characterize the system noise characteristics and ensure the robustness of the state estimation. In this way, the weight of the filter is dynamically adjusted to adapt to the changes in the battery state under different working conditions and improve the sensitivity to abnormal conditions. The noise characteristic parameters are substituted into the strong tracking filter algorithm, and the exponential decay function is introduced to calculate the optimal decay factor, so that the prediction error covariance matrix is ​​corrected in the state estimation process, so that it can quickly adjust the response under sudden changes, reduce the impact of error accumulation, and obtain the state prediction result. The strong tracking filter algorithm combined with the dynamic noise characteristic adjustment enables the system to maintain a high prediction accuracy during long-term operation and identify potential abnormal trends at an early stage. The state prediction results are compared with the set safety thresholds in multiple dimensions, including voltage thresholds, temperature thresholds, and internal resistance thresholds. The deviation data is analyzed to determine whether there is an abnormality in the battery's charge and discharge state. If the predicted value deviates from the normal working range, the deviation data is recorded as abnormal deviation data, and a trend analysis is formed after further calculation to identify the direction and degree of abnormal development. In order to ensure the reliability of anomaly detection, the abnormal deviation data is cumulatively statistically analyzed, and the abnormal duration of the state variables is calculated. Once the abnormal duration of any state variable exceeds the set time threshold, the system triggers an abnormal alarm, and the abnormal trigger signal is graded according to the severity of the deviation, and finally the charge and discharge abnormal event data is generated.

[0019] A decomposition unit 103 is used to perform hierarchical feature decomposition and multi-dimensional fusion based on the abnormal charge and discharge event data to obtain a cycle life attenuation feature matrix; Specifically, in order to improve the efficiency of data processing, a division sub-unit is introduced. By dividing the adjacent N battery cells into a group, a reasonable calculation sub-domain is formed. On this basis, the voltage, temperature and internal resistance data of each group of battery cells are calculated, and the voltage differential capacity, temperature distribution and internal resistance characteristics are extracted to construct the feature vector data. Among them, the voltage differential capacity can effectively reflect the charge and discharge characteristics of the battery under different charge states, the temperature distribution can characterize the battery thermal management status, and the internal resistance characteristics can reveal the degree of battery aging. Entering the calculation stage of the extraction sub-unit, the spatial correlation between different calculation sub-domains is explored. Based on the feature vector data, the eigenvector mutual correlation coefficient between each calculation sub-domain is calculated to construct a spatial correlation matrix, which can effectively reflect the state coupling characteristics between different battery cells and reveal the balance and consistency within the battery pack. The frequency band energy distribution of the voltage data and temperature data is extracted to generate a time-frequency feature matrix, which can capture the energy distribution pattern of the battery in different frequency bands, especially under fast charging and discharging and high load conditions, and can effectively identify potential abnormal fluctuations. Based on the spatial correlation matrix and the time-frequency feature matrix, the system enters the core calculation link of the calculation subunit. At this stage, the core tensor coefficients are calculated and the three-dimensional factor matrix is ​​solved to obtain a multimodal feature tensor, which can simultaneously characterize the spatial, temporal and frequency domain characteristics of the battery state. Through multimodal data fusion, the state evolution process of the battery cell is described. The tensor modeling method can effectively reduce the complexity of the data dimension while retaining the key relationship between different feature variables, making the life decay modeling more accurate and efficient. Entering the calculation stage of the construction subunit, low-rank approximate calculations are performed on the multimodal feature tensor to extract the most representative feature component data. This process can effectively remove redundant information and retain the most physically meaningful state characteristics. On this basis, the feature component data is used to construct the decay matrix of the spatiotemporal evolution characteristics, and finally a cycle life decay feature matrix is ​​formed. This matrix is ​​used to monitor the overall life decay trend of the battery pack and identify the main factors affecting the life.

[0020] The spatial correlation matrix is ​​rearranged according to the physical position relationship of the battery cells, so that the organization of the matrix can more intuitively reflect the spatial distribution of the battery cells in the battery pack, thereby enhancing the parsability of the spatial data. At the same time, the time-frequency feature matrix is ​​reorganized in chronological order to ensure the data coherence in the time dimension, forming the initial tensor data, which not only contains the spatial characteristics of the battery state, but also retains the time change trend and frequency domain characteristics. The initial tensor data is standardized, and the normalization coefficients of the time dimension, space dimension and feature dimension are calculated. The tensor is normalized accordingly to ensure that the numerical range of all features is consistent and avoid calculation deviations caused by different scales. The normalized tensor is orthogonally decomposed to extract the core feature information of different dimensions. The normalized tensor is decomposed into three orthogonal basis matrices, and the feature basis matrix of the time dimension, the feature basis matrix of the space dimension and the feature dimension are calculated respectively. In this way, the complex information in the tensor is decoupled, and the independent characteristics of the time, space and feature variables are extracted respectively, and a clearer data representation is obtained. Perform Tucker decomposition on the orthogonal decomposition results to extract the core characteristic patterns of the tensor. Through the Tucker decomposition method, the high-dimensional tensor is mapped to a low-dimensional potential structure, and the tensor decomposition equation is established to model the main components of the data. In order to optimize the solution of the tensor decomposition equation, the alternating least squares method is used for iterative calculation. This method can alternately optimize parameters of different dimensions in each iteration, thereby gradually approaching the optimal solution and improving the convergence speed and accuracy of the calculation. Through this optimization process, the optimized tensor parameters are obtained. The tensor reconstruction operation is performed on the optimized tensor parameters, and the original tensor data is reversely calculated using the optimized parameters to generate the final multimodal feature tensor.

[0021] The balancing control unit 104 is used to perform life balancing control based on the cycle life attenuation characteristic matrix to obtain a target charge and discharge coefficient.

[0022] In this embodiment, the state of the battery cell is finely modeled using a cycle life attenuation characteristic matrix, and the characteristic matrix is ​​substituted into the Thevenin equivalent circuit model to calculate the polarization resistance, polarization capacitance and state of charge value of each battery cell to obtain the underlying state data of the battery, wherein the polarization resistance can reflect the electrochemical resistance encountered by the battery during the charging and discharging process, and the polarization capacitance is used to describe the transient charging and discharging characteristics of the battery. At the same time, the state of charge value (SOC) is an important indicator for measuring the remaining power of the battery. Through this calculation process, the dynamic electrochemical behavior of the single cell is characterized. After obtaining the underlying state data, the particle swarm filter calculation model is used to analyze the state of charge distribution characteristics of the entire battery pack. The parameters of the influence relationship between the underlying state data and the battery cells are input into the particle swarm filter model. The algorithm is used to perform iterative optimization in the multidimensional state space, and the state of charge distribution curve of the battery pack is calculated through the dynamic weight adjustment mechanism. In this process, the particle swarm filter algorithm can effectively integrate multiple uncertain factors, improve the estimation accuracy of the complex system state, and provide a theoretical basis for the formulation of battery balancing strategies by tracking the change trend of the state of charge. The middle-level state data obtained can comprehensively characterize the SOC distribution of the entire battery pack under different working conditions. The middle-level state data is analyzed, focusing on the differences between battery cells, and the state of charge and health status of each battery cell are evaluated differently, and the charge and discharge weight factor of each battery cell is calculated. In this process, the aging degree, internal resistance change and dynamic response capability of the battery cell are comprehensively considered based on SOC, SOH and historical charge and discharge characteristics, and the balancing requirements of the battery cell are calculated in combination with the data-driven method to obtain the distribution weight coefficient, which reflects the charge and discharge compensation required for each battery cell under the current working state. The allocation weight coefficient is substituted into the current optimization model, and the optimal charge and discharge current limit is calculated for the battery cells whose state of charge deviates from the average value. In this process, the constrained optimization method is adopted to minimize the state of charge deviation between battery cells, while considering the charge and discharge efficiency, temperature distribution and safety constraints, so as to calculate the optimal current regulation parameter and obtain the balancing control parameter. This parameter is used to guide the battery management system (BMS) to dynamically adjust the charge and discharge current of each battery cell to ensure the overall balance of the battery pack.The balancing control parameters are cross-calculated with the electrochemical characteristic association matrix to obtain the risk distribution data of the battery pack. The risk distribution data is used to evaluate the failure risk faced by each battery cell under different charging and discharging conditions, and based on this, the charge and discharge rate of the battery pack is planned by region. According to the health status and balancing requirements of the batteries in different regions, the charge and discharge rate is dynamically adjusted to finally obtain the target charge and discharge coefficient. This coefficient can optimize the overall energy management strategy of the battery pack, improve the balance during the charging and discharging process, prevent the premature attenuation of the life of some battery cells due to overcharging or over-discharging, and at the same time improve the cycle life and safety of the entire battery system.

[0023] The target charge and discharge coefficient and the working state parameters of the battery, including charge and discharge depth, rate and temperature parameters, are input into the electrochemical stress model to calculate the fatigue stress distribution of the single cell under different working conditions. The electrochemical stress model simulates the stress changes inside the electrode material according to the dynamic process of charge and discharge of the lithium-ion battery, thereby revealing the changes in the electrode structure and crack propagation caused by the cycle charge change, obtaining electrochemical damage data, and quantifying the aging degree of the battery under different load conditions. After obtaining the electrochemical damage data, it is accumulated in time series based on the linear fatigue damage criterion to construct the life decay curve of the single cell. In this process, by modeling the damage accumulation effect, the nonlinear change trend of the battery life under different charge and discharge conditions is analyzed, and the fatigue damage degree matrix is ​​gradually formed, which describes the damage evolution of the battery cell throughout its life cycle. In order to more effectively manage the health status of the battery and dynamically adjust the working parameters according to the fatigue degree of the battery, the fatigue damage degree matrix is ​​graded, that is, the battery cell is divided into multiple different damage gears according to the degree of damage, and the corresponding charge and discharge rate upper limit is set for each gear, thereby generating the hierarchical limit parameters. This strategy can effectively prevent the accelerated aging of highly damaged battery cells due to excessive charging and discharging, while ensuring the balance and safety of the entire battery pack. After obtaining the stratification limit parameters, in order to optimize the operating state of the battery, the temperature parameters of the battery are combined for dynamic correction. By calculating the power attenuation ratio of each single battery based on the stratification limit parameters and the battery surface temperature data, and thus determining the temperature correction coefficient, this correction coefficient can reflect the performance degradation characteristics of the battery under different temperature conditions, and ensure that the battery can be charged and discharged in a safer working mode under high or low temperature environments, thereby reducing the risk of electrochemical failure caused by temperature changes. Combined with the temperature correction coefficient and the total power demand of the system, the power allocation strategy within the entire battery pack is calculated to optimize the power regulation between battery cells and obtain power optimization parameters. These parameters enable the battery pack to maintain efficient energy output while minimizing the overload of individual battery cells and improving the life stability of the entire system. In order to ensure that the charging and discharging process of the battery pack meets the requirements of life optimization, the power optimization parameters are charged and discharged in stages, that is, according to different operating conditions, the current limit interval is dynamically set for each battery cell to obtain the life protection control parameters. This parameter dynamically adjusts the maximum charge and discharge current according to the actual health status and power demand of the battery to ensure that the stress level of the battery remains low throughout its life cycle and reduce the adverse effects of over-discharge or over-charge on battery life.

[0024] The life protection control parameters are input into the model predictive controller to calculate the local protection control instruction sequence of each battery cell and obtain the target control data. Based on the dynamic optimization method, the model predictive controller adjusts the battery charging and discharging strategy in real time through iterative calculation and optimization constraints within a certain prediction time domain to ensure that the health state of each battery cell is maintained at the optimal level, and minimize the impact of external factors such as load changes and ambient temperature fluctuations, so that the battery can maintain a stable working state throughout its life cycle. The target control data is input into the coulomb efficiency compensation model to calculate the charge and discharge compensation current of each single battery, and thus obtain the balanced compensation parameters. Since lithium batteries have irreversible energy loss during long-term charging and discharging, and there are differences in coulomb efficiency between battery cells, the compensation model can effectively correct the imbalance of the state of charge (SOC) caused by coulomb efficiency deviation, thereby ensuring that the energy of each battery cell is kept within a reasonable range, avoiding premature aging of some batteries due to overcharging or over-discharging, and improving the energy utilization efficiency of the entire battery pack. The balanced compensation parameters are input into the data interaction network between adjacent control nodes to calculate the optimal control quantity of each control node and obtain the coordinated control variable. Under the distributed energy management architecture, control nodes need to continuously exchange data to achieve the overall optimal energy scheduling. Therefore, based on the data interaction network, energy state information is shared between nodes, and the charging and discharging power of each battery cell is adjusted based on the optimal control strategy, so that the charge state distribution of the entire battery pack is more uniform, the overall stability of the system is improved, and the local overload caused by the imbalance of battery state is reduced. After obtaining the cooperative control variables, in order to ensure the state consistency of the system, the convergence speed between the control nodes is calculated, and the cooperative control variables are substituted into the Lyapunov function to evaluate the convergence speed of the state consistency between the nodes, and the stability evaluation data is obtained accordingly. The Lyapunov function is a mathematical method used to analyze the stability of a dynamic system. In this system, it is used to determine whether the state of the battery pack under the action of the balanced control strategy can quickly converge to a steady state, and analyze the control effect of the energy management strategy. If the state deviation of some control nodes is large, the system will adjust the balanced control strategy to accelerate the state convergence and improve the dynamic response capability of the system. After completing the stability assessment, the reconstruction strategy of the control network is calculated to ensure that the normal operation of the system can be maintained under abnormal conditions. Based on the stability evaluation data, the reconstruction strategy of the control network is calculated, and the fault-tolerant control parameters are obtained. In practical applications, due to communication failures, sensor errors or sudden failures of some battery cells, the control network needs to be dynamically adjusted, and the calculation of fault-tolerant control parameters can ensure that in the case of partial control node failure, the system can still be effectively scheduled through the remaining nodes and maintain energy balance on a global scale, thereby improving the reliability and robustness of the system.In order to implement the fault-tolerant control strategy into specific energy management operations, the fault-tolerant control parameters are substituted into the equivalent circuit network, and the energy transmission path and power size between nodes are calculated to finally obtain the balancing control instructions. The equivalent circuit network can simulate the energy flow inside the battery pack, and based on the optimized control strategy, determine the optimal energy transmission path and power adjustment parameters, so that energy can be reasonably distributed among different battery cells, thereby ensuring that the balanced state of the entire system is maintained. The balancing control instructions are used to guide the battery management system (BMS) to accurately control the charging and discharging of the battery pack, ensuring that the health status of each battery cell is optimized throughout the life cycle, thereby improving the cycle life of the lithium battery and improving the safety and reliability of the entire energy management system.

[0025] Optionally, the collection unit 101 is specifically used for: The voltage data of the battery cells in the lithium battery pack are collected through the voltage collection module, the temperature data are collected through the temperature collection module, and the internal resistance data are collected through the internal resistance measurement module to obtain the original collected data; Eliminate high-frequency noise from the original collected data to obtain filtered processed data, and calculate the voltage change rate, temperature gradient, and internal resistance increment of the filtered processed data to obtain characteristic quantity data; Pack the characteristic data into data frames and transmit them to the upper control unit via the CAN bus to obtain a transmission data packet; The change rate of the voltage data and the temperature data in the transmission data packet is calculated, and the sampling frequency is dynamically adjusted according to the change rate value to obtain the adaptive sampling parameter; The adaptive sampling parameters are returned to the voltage acquisition module, the temperature acquisition module and the internal resistance measurement module, the sampling frequency is updated and configured, and the battery status characteristic data is obtained.

[0026] In this embodiment, the voltage acquisition module, temperature acquisition module and internal resistance measurement module respectively collect the voltage, temperature and internal resistance data of the single cell to obtain the original collected data. The voltage acquisition module reads the voltage signal at both ends of the battery through a high-precision analog-to-digital converter (ADC) to ensure that the sampling error is within an acceptable range and sets the voltage sampling value of the single cell to ,in Represents the battery number, is a time variable. At the same time, the temperature acquisition module uses a temperature sensor (such as a thermistor or a thermocouple) to measure the surface temperature of the battery, which is recorded as The internal resistance measurement module measures the battery internal resistance by AC impedance or DC pulse method, and defines its measurement value as Through this step, we can obtain , and The original collected data is filtered to improve the reliability of the data. Suppose the voltage data after filtering is , the temperature data is , the internal resistance data is , a first-order low-pass filter is used to smooth the signal, and its mathematical expression is:

[0027] in, Represents the original collected data, , or , is the filter coefficient, take 0.3, used to control the filtering effect, smaller Reduce the impact of high frequency noise, while the larger After filtering, the characteristic data is extracted for subsequent intelligent analysis and balanced control. The voltage change rate is defined as:

[0028] This quantity describes the change trend of battery voltage over time and can reflect the change of charge and discharge state. Similarly, the temperature gradient is calculated by the following formula:

[0029] This feature is used to evaluate the thermal characteristics and heat dissipation capabilities of the battery, and the internal resistance increment is given by the following formula:

[0030] It can reveal the degree of internal aging of the battery and indicate possible battery failure when it increases abnormally. These calculated feature data are used to evaluate the battery health status and serve as key inputs for balancing control and life prediction. After the feature calculation is completed, these feature data are packaged into data frames and transmitted to the upper control unit through the controller area network (CAN bus). Assuming that the data frame includes battery number, filter voltage, filter temperature, filter internal resistance and corresponding change rate information, the data frame format is expressed as:

[0031] Among them, ID is the unique number of the battery cell. The upper control unit dynamically adjusts the energy management strategy of the battery pack and performs health status assessment by parsing these data frames. While data is being transmitted, the sampling frequency is dynamically adjusted according to the changes in the battery status to improve the timeliness of the data. The dynamic adjustment strategy of the sampling frequency is determined by calculating the change rate of the voltage data and temperature data. If the voltage change rate or temperature change rate exceeds the set threshold and , then the sampling period Need to be shortened to capture more fine-grained changes:

[0032] in, is the minimum sampling period, To dynamically adjust the coefficient, this formula ensures that when the voltage or temperature changes rapidly, the system automatically increases the sampling frequency to avoid the loss of key status information. The calculated adaptive sampling parameters need to be returned to the voltage acquisition module, temperature acquisition module and internal resistance measurement module to update the sampling frequency configuration, thereby optimizing the data acquisition process. After the above data acquisition, processing, analysis and dynamic adjustment, the system can continuously optimize the acquisition process of battery status characteristic data, improve the accuracy and real-time performance of battery status monitoring, ensure that the lithium battery pack can maintain the best working state under various working conditions, and effectively extend its cycle life.

[0033] Optionally, the identification unit 102 is specifically configured to: Based on the battery status characteristic data, the battery terminal voltage, surface temperature, and internal resistance change are constructed as observation vectors, and the state of charge, health state, and polarization voltage are constructed as state vectors to obtain the Kalman filter state equation; The mean and variance of the Kalman filter state equation are calculated, and the process noise covariance matrix and the observation noise covariance matrix are updated through recursive operation to obtain the noise characteristic parameters; Substitute the noise characteristic parameters into the strong tracking filter algorithm, calculate the optimal attenuation factor through the exponential attenuation function, correct the prediction error covariance matrix, and obtain the state prediction result; Perform a multi-dimensional comparison between the state prediction results and the preset voltage threshold, temperature threshold, and internal resistance threshold to obtain abnormal deviation data; The abnormal deviation data is accumulated and counted. When the abnormal duration of any state quantity exceeds the corresponding time threshold, an abnormal trigger signal is obtained. The abnormal trigger signal is graded according to the deviation magnitude to obtain the abnormal charging and discharging event data.

[0034] In this embodiment, abnormal pattern recognition is performed based on the battery state characteristic data, and a state observation model is constructed to accurately describe the operating characteristics of the lithium battery. The battery terminal voltage, surface temperature and internal resistance change are defined as observation vectors, which are set as:

[0035] in, Indicates time The collected battery terminal voltage, Indicates the battery surface temperature, In order to describe the internal state of the battery, the state of charge (SOC), state of health (SOH) and polarization voltage ( ) is constructed as a state vector defined as:

[0036] in, Reflects the remaining battery power. Represents the health of the battery. is the polarization voltage, which represents the combined effect of ohmic polarization and concentration polarization inside the battery. In order to establish a mathematical model for the evolution of the battery state, the Kalman filter method is used to estimate the system state, and the state transfer equation is assumed to be:

[0037] in, is the state transfer matrix, which describes the dynamic evolution characteristics of the system state. B is the input influence matrix, which represents the influence of external input on the system state. is the input vector, such as charging and discharging current, is the process noise, assuming it obeys a zero-mean Gaussian distribution, that is: ; in, is the process noise covariance matrix, and the observation equation is expressed as: ; in, is the observation matrix, For the observation noise, it is assumed to obey a zero-mean Gaussian distribution, that is: ; in, To observe the noise covariance matrix, in order to improve the estimation accuracy, the noise characteristic parameters are dynamically updated. Therefore, the mean and variance of the Kalman filter state equation are calculated, and the process noise covariance matrix is ​​updated by recursive operation. and the observation noise covariance matrix , the specific calculation is as follows:

[0038] in, is the state prediction error covariance matrix, and the Kalman gain Calculated by the following formula: ; And update the state estimate using the following formula: ; In order to enhance the sensitivity to abnormal patterns, the updated noise characteristic parameters are substituted into the strong tracking filter algorithm, and an exponential decay function is introduced to calculate the optimal decay factor. , to adaptively adjust the filter gain so that the system can quickly track state changes. The calculation formula is as follows: ; in, The attenuation factor is used to adjust the coefficient. It is used to correct the prediction error covariance matrix to improve the response speed of the filter. , Temperature threshold and internal resistance threshold Perform multi-dimensional comparison and calculate abnormal deviation data: ; When any abnormal deviation exceeds the time threshold When , the condition is met:

[0039] The system triggers an abnormal signal and classifies it according to the deviation magnitude to generate abnormal charging and discharging event data.

[0040] The noise characteristic parameters are substituted into the strong tracking filter algorithm, the optimal attenuation factor is calculated by the exponential decay function, and the prediction error covariance matrix is ​​corrected to obtain the state prediction result, including: based on the noise characteristic parameters, the measurement prediction value is obtained by multiplying the observation matrix with the state prediction value, and the measurement prediction value is subtracted from the actual observation value, and the residual sequence is constructed in time sequence, and the autocorrelation matrix of the residual sequence is calculated to obtain the residual eigenvector; the residual eigenvector is subjected to Schmidt orthogonal transformation, an orthogonal basis vector group is constructed, and the residual sequence is projected and decomposed on the orthogonal basis, the variance of each orthogonal component is calculated, and the comprehensive tracking error index is obtained by weighted summation; a nonlinear optimization equation is established according to the comprehensive tracking error index, an exponential decay function is set, and an iterative solution is performed by the variable scale least squares method. When the difference between the attenuation factors of two adjacent iterations is less than 10 -6 The iteration is stopped when , and the optimal attenuation factor is obtained; the optimal attenuation factor is multiplied by the diagonal elements of the prediction error covariance matrix, and the non-diagonal elements are kept unchanged to construct a diagonal weighted matrix, and the corrected prediction error covariance matrix is ​​obtained by matrix multiplication; based on the corrected prediction error covariance matrix, the sliding time window length is set, and the state variables are recursively predicted. Each prediction uses the latest observation data for incremental update to obtain an L-step prediction sequence; the L-step prediction sequence is weighted averaged using exponentially decreasing weights, and the state prediction result is obtained by weighted summation.

[0041] Optionally, the decomposition unit 103 further includes: A division subunit is used to divide the abnormal charge and discharge event data into groups of N adjacent battery cells, and calculate the voltage differential capacity, temperature distribution and internal resistance characteristics of each group of divided battery cells to obtain feature vector data; The extraction subunit is used to calculate the eigenvector mutual correlation coefficients between the calculation subdomains based on the eigenvector data to obtain the spatial correlation matrix, and to extract the frequency band energy distribution of the voltage data and the temperature data to obtain the time-frequency feature matrix; A computing subunit is used to calculate the core tensor coefficients and the three-dimensional factor matrix according to the spatial correlation matrix and the time-frequency feature matrix to obtain the multimodal feature tensor; A subunit is constructed to perform low-rank approximate calculations on the multimodal feature tensor to obtain feature component data, and an attenuation matrix of the spatiotemporal evolution characteristics is constructed based on the feature component data to obtain a cycle life attenuation feature matrix.

[0042] In this embodiment, the data is structured and divided into sub-units, and the abnormal charge and discharge event data is divided into adjacent sub-units. The battery cells are grouped into partitions to improve the local consistency of data processing. The data of each group of battery cells includes voltage changes , Temperature distribution And the internal resistance characteristics , for each group of data after division, calculate the voltage differential capacity To reflect the dynamic characteristics of battery charging and discharging, the calculation formula is as follows:

[0043] in, is the charge and discharge capacity of the battery, Represents the amount of charge change corresponding to a unit voltage change. This indicator is used to characterize the charge state change characteristics of the battery, while the temperature distribution data The local temperature difference is estimated by sampling data from multiple thermal sensors and used to analyze the thermal management of the battery under high load conditions. At the same time, the internal resistance change characteristics of each battery cell are calculated. To monitor the health status of the battery, it is calculated as follows:

[0044] in, is the initial reference internal resistance value, is the current measured internal resistance value, increment The larger it is, the more serious the aging of the battery is. All these feature quantities constitute feature vector data:

[0045] After acquiring the feature vector data, the extraction sub-unit calculation stage is entered. Based on these data, the feature vector correlation coefficients between the calculation sub-domains are calculated to form a spatial correlation matrix , this matrix is ​​used to describe the mutual influence relationship between different battery cells, and its calculation method is as follows:

[0046] in, Represents battery With battery Correlation on the feature vector data, and are their means respectively, when When it approaches 1, it means that the two states are highly correlated, and when it approaches 0, it means that the two states are basically unrelated. In order to capture the timing characteristics of the battery state, the frequency band energy distribution of the voltage data and temperature data is extracted to form a time-frequency feature matrix , the calculation formula is as follows:

[0047] in, Indicates that the characteristic signal is at a frequency The energy distribution on the battery is analyzed by Fourier transform, and the dynamic change characteristics of the battery state on different time scales are extracted, and the spatial correlation matrix is ​​combined After entering the calculation subunit stage, the spatial correlation matrix is ​​used And the time-frequency feature matrix Calculate the core tensor coefficients and construct a three-dimensional factor matrix to form a multimodal feature tensor , defined as follows:

[0048] in, represents the rank of tensor decomposition, is the weight coefficient of the feature tensor, , The tensor model is used to effectively extract the potential relationship between multi-dimensional features in order to model the overall life state of the battery pack. In order to reduce the dimension of the data and improve the computational efficiency, the multi-modal feature tensor is constructed. Perform low-rank approximation calculations to extract the most critical feature component data and use the Tucker decomposition method to decompose the tensor To reduce the dimension, the calculation formula is as follows:

[0049] in, is the low-rank core tensor, They are the dimension reduction matrices of time, space and characteristic patterns respectively. This calculation process can effectively reduce redundant data and extract the most representative state information. Based on the characteristic component data, the attenuation matrix of the spatiotemporal evolution characteristics is constructed to describe the long-term change trend of the battery state. Its mathematical expression is as follows:

[0050] in, Represents battery and battery The life decay correlation between and is a weight factor used to balance the influence of spatial correlation and timing characteristics. The attenuation matrix is ​​used to construct the cycle life attenuation characteristic matrix to predict the battery life trend.

[0051] Optionally, the computing subunit is specifically used for: The spatial correlation matrix is ​​rearranged according to the physical position relationship of the battery cells, and the time-frequency feature matrix is ​​reorganized in time sequence to obtain the initial tensor data; The initial tensor data is standardized, and the normalization coefficients of the time dimension, space dimension, and feature dimension are calculated to obtain a normalized tensor; Decompose the normalized tensor into three orthogonal basis matrices, calculate the characteristic basis matrix of the time dimension, the characteristic basis matrix of the space dimension, and the characteristic basis matrix of the characteristic dimension, and obtain the orthogonal decomposition result; Perform Tucker decomposition operation on the orthogonal decomposition result to obtain the tensor decomposition equation, and substitute the tensor decomposition equation into the alternating least squares method for iterative solution to obtain the optimized tensor parameters; A tensor reconstruction operation is performed on the optimized tensor parameters to obtain a multimodal feature tensor.

[0052] In this embodiment, the spatial correlation matrix of the battery cell is optimized to improve the accuracy of subsequent calculations. Rearrange the data according to the physical position of the battery cells, that is, readjust the rows and columns of the matrix through the topological structure of the battery pack to make it more consistent with the layout of the battery in the physical space, ensuring that the spatial distribution of the data conforms to the actual battery pack structure. Suppose the original spatial correlation matrix is ,in is the number of battery cells, then the rearrangement matrix Obtained by the following transformation: ; in, is a rearrangement matrix, which redistributes the relationship between battery cells according to the physical topological structure. At the same time, in order to make the distribution of time series data more orderly, the time-frequency feature matrix Reorganized in chronological order, Represents the time step, the reorganized matrix Ensure that the time information remains consistent for subsequent modeling. And the reorganized time-frequency feature matrix Combine and construct the initial tensor data : ; in, Represents the characteristic dimension, such as voltage, temperature, internal resistance and other state variables. The initial tensor data is standardized and the normalization coefficients of the time dimension, space dimension and characteristic dimension are calculated. , and are the means in time, space and feature dimensions respectively, , and is the corresponding standard deviation, then the normalization formula is as follows:

[0053] Normalized tensor data It can avoid the influence of the characteristic dimension and make the data of different dimensions be calculated on the same scale. The normalized tensor is orthogonally decomposed into three orthogonal basis matrices, corresponding to the time dimension, space dimension and characteristic dimension respectively. The characteristic basis matrices after decomposition are , and , then the orthogonal decomposition results are as follows:

[0054] in, , , are the ranks in time, space and feature dimensions respectively, is the nuclear tensor coefficient, and Represents the outer product operation. This decomposition process can decompose multidimensional data into low-dimensional feature vectors, making data analysis more efficient. After obtaining the orthogonal decomposition result, Tucker decomposition is performed to optimize the quality of tensor modeling. The mathematical expression of Tucker decomposition is as follows:

[0055] in, is the core tensor, which represents the low-rank tensor after dimensionality reduction, and , , Respectively represent the modular product operations along the time, space and feature dimensions. In order to optimize the decomposition process, the alternating least squares method is used for iterative solution. That is, in each iteration, two basis matrices are fixed and the third basis matrix is ​​updated to minimize the error. The iterative update formula is as follows:

[0056] in, Represents the number of iteration steps. This method can effectively approximate the optimal tensor representation and improve computational efficiency. Based on the optimized tensor parameters, tensor reconstruction operation is performed to obtain the final multimodal feature tensor , and its calculation formula is as follows:

[0057] in, As the optimized feature basis matrix, the multimodal feature tensor can characterize the spatiotemporal characteristics of the lithium battery pack.

[0058] Optionally, the balancing control unit 104 is specifically configured to: Substitute the cycle life attenuation characteristic matrix into the Thevenin equivalent circuit model, calculate the polarization resistance, polarization capacitance and state of charge value of each battery cell, and obtain the underlying state data; The bottom-level state data and the influencing relationship parameters between battery cells are input into the particle swarm filter calculation model, the state of charge distribution curve of the entire battery pack is calculated, and the middle-level state data is obtained; Perform difference analysis on the state of charge and health status of each battery cell in the middle-level status data, calculate the charge and discharge weight factor of each cell, and obtain the distribution weight coefficient; Substitute the allocation weight coefficient into the current optimization model, calculate the corresponding optimal charge and discharge current limit for the battery cells that deviate from the average value, and obtain the balancing control parameters; The balance control parameters and the electrochemical characteristic correlation matrix are cross-calculated to obtain the risk distribution data, and the charge and discharge rate of the battery pack is planned by region based on the risk distribution data to obtain the target charge and discharge coefficient.

[0059] In this embodiment, an equivalent circuit model of the battery is constructed based on the cycle life attenuation characteristic matrix, and the characteristic matrix is ​​substituted into the Thevenin equivalent circuit model to calculate the polarization resistance, polarization capacitance and state of charge (SOC) of each battery cell. These parameters together constitute the underlying state data of the battery. The Thevenin equivalent circuit model uses an open circuit voltage source , an ohmic resistor in series and a parallel RC component, where represents the polarization resistance, Represents polarized capacitance, and the specific circuit equation is expressed as: ; in, is the battery terminal voltage, is the charge and discharge current of the battery cell, and the polarization voltage It is described by the following differential equation:

[0060] By solving the differential equation and combining the actual charging and discharging effects of the battery, the polarization resistance of each battery cell is calculated. , polarized capacitor and state of charge , these parameters constitute the underlying state data of the battery, which is used to optimize the balancing control strategy. The underlying state data and the influencing relationship parameters between battery cells are input into the particle swarm filter calculation model to calculate the state of charge distribution curve of the entire battery pack and obtain the middle-level state data. Particle swarm filtering (PSO-PF) combines the advantages of particle filtering (PF) and particle swarm optimization (PSO) algorithms for nonlinear estimation of battery state of charge (SOC). Assume that the SOC distribution curve of the battery pack is:

[0061] in, is the number of battery cells, and the particle swarm filter algorithm uses the state prediction equation:

[0062] in, is the Coulomb efficiency, is the battery capacity, is the charge and discharge current of the battery cell, combined with the measurement update step, by calculating the measurement error To correct the estimated value and finally converge to the optimal SOC estimate:

[0063] in, is the number of particles, To normalize the weight, this method is used to obtain the state of charge distribution curve of the battery pack and form the middle-level state data. The SOC and SOH (health state) of each battery cell in the middle-level state data are analyzed for differences, and the charge and discharge weight factor of each battery cell is calculated. , to determine a reasonable energy allocation strategy. The calculation method of the charge and discharge weight factor is as follows:

[0064] in, is the adjustment coefficient, is the average SOC of the battery pack, Reflects the relative weight of the battery in the balancing control. If the SOC of a battery cell is far from the average value, the weight factor is large, indicating that the battery needs a large energy compensation, thus obtaining the final allocation weight coefficient. Substitute the allocation weight coefficient into the current optimization model, calculate the optimal charge and discharge current limit for the battery cell that deviates from the average value, and obtain the balancing control parameter. Assume that the total current of the battery pack is , then the optimal charge and discharge current of each battery is calculated according to the allocation weight coefficient: ; At the same time, make sure the current does not exceed the battery's charge and discharge capacity: ; in, and The optimization process ensures that the charging and discharging behavior of the battery cell meets the health status requirements and avoids overcharging or over-discharging. In order to optimize the charging and discharging strategy of the battery pack, the balance control parameters and the electrochemical characteristic correlation matrix are cross-calculated in combination with the electrochemical characteristics of the battery to analyze the overall risk distribution of the battery pack. The electrochemical characteristic correlation matrix is ​​assumed to be , where each element Represents battery and battery The electrochemical correlation between the two, the risk distribution data is calculated as follows: ; in, is the risk distribution matrix, which indicates the aging risk that each battery cell may face under the current charge and discharge strategy. This matrix is ​​used to guide the charge and discharge rate planning of the battery pack. Based on the risk distribution data, the charge and discharge rate of the battery pack is planned in different regions to calculate the final target charge and discharge coefficient. Assume that the battery pack is divided into The rate planning formula for each area is as follows:

[0065] in, Representative area The target charge and discharge coefficient is finally adjusted to maintain the SOC of the entire battery pack within a reasonable range by adjusting the rate of each area.

[0066] Optional, distributed energy management systems for improved lithium battery cycle life also include: The real-time constraint unit is used to input the target charge and discharge coefficient and the charge and discharge depth, rate and temperature parameters into the electrochemical stress model, calculate the fatigue stress distribution of the single cell, and obtain the electrochemical damage data; perform time-series accumulation operation on the electrochemical damage data according to the linear fatigue damage criterion, construct the life attenuation curve of each battery cell, and obtain the fatigue damage degree matrix; divide the fatigue damage degree matrix into gears according to the degree of damage, set the corresponding charge and discharge rate upper limit for different gears, and obtain the stratified restriction parameters; calculate the power attenuation ratio of each single cell based on the stratified restriction parameters and the battery surface temperature data, and obtain the temperature correction coefficient; calculate the power allocation strategy inside the battery pack in combination with the temperature correction coefficient and the total power demand of the system, and obtain the power optimization parameters; perform charge and discharge stage-by-stage processing on the power optimization parameters, set a dynamic current limit interval for each battery cell, and obtain the life protection control parameters.

[0067] In this embodiment, the target charge and discharge coefficient Combined with key operating parameters of the battery, including depth of charge and discharge , magnification And temperature parameters , and input the electrochemical stress model to calculate the fatigue stress distribution of the single cell and obtain the electrochemical damage data. The electrochemical stress model is used to describe the mechanical stress distribution and electrochemical degradation effect of the battery under different working conditions. Its core expression is as follows:

[0068] in, Indicates battery cells The mechanical stress is the Young's modulus of the active material, is Poisson's ratio, is the volume change caused by electrode expansion, is the electrode thickness. At the same time, the electrochemical stress model also needs to consider the temperature effect and use the Arrhenius equation to describe the battery aging rate:

[0069] in, For battery The aging rate, is an empirical constant, is the activation energy, is the gas constant, is the battery surface temperature. Based on these calculations, the electrochemical damage data of the battery cell is obtained to evaluate its cycle life decay trend. After obtaining the electrochemical damage data, a time series accumulation operation is performed according to the linear fatigue damage criterion to construct the life decay curve of each battery cell and obtain the fatigue damage degree matrix. Based on Miner's linear cumulative damage model, the battery damage accumulation is calculated as follows:

[0070] in, Represents battery The cumulative damage degree under the current charge and discharge cycle, For the The number of charge and discharge cycles experienced in the working mode, is the theoretical maximum cycle life of the battery under this condition. When the battery is in a failed state, it indicates that the battery has entered a failed state. , real-time tracking of battery aging trends, and building a fatigue damage matrix for the entire battery pack. The fatigue damage matrix is ​​divided into gears according to the degree of damage, so as to set the upper limit of the charge and discharge rate for battery cells in different health states, thereby generating hierarchical limit parameters. Assuming that the battery health state is divided into 3 gears, the setting of the upper limit of the charge and discharge rate is expressed as: ; in, Indicates battery cells The maximum allowable charge and discharge rate under the current damage state, is the rated rate of the battery. Through the stratification limitation strategy, the decay rate of high-damage batteries is effectively slowed down and the overall health level of the battery pack is optimized. After determining the stratification limitation parameters, the power decay ratio is calculated in combination with the surface temperature data of the battery to obtain the temperature correction factor. According to the influence of battery temperature on the rate, the temperature correction factor is used. Make adjustments: ; in, is the temperature sensitivity coefficient, For the optimal working temperature, is the current battery cell temperature. When the battery temperature is too high, will decrease, thereby reducing the allowable charge and discharge rate to reduce the risk of thermal runaway. After calculating the temperature correction coefficient, combined with the total power demand, the power allocation strategy inside the battery pack is calculated to obtain the power optimization parameters. Assume that the total power demand is , then the power distribution scheme of each battery cell is as follows:

[0071] in, To allocate to the battery This calculation ensures that the power allocation takes into account both the health status of the battery and the temperature effect, thereby improving the safety and life of the system. The power optimization parameters are processed in stages for charging and discharging, and a dynamic current limit interval is set for each battery cell to obtain the life protection control parameters. , assuming that the charge and discharge current limit is:

[0072]

[0073] in, is the dynamic adjustment coefficient (usually between 0.5 and 0.8), and Together they form the dynamic current limit range of the battery to ensure it operates within a safe range.

[0074] Among them, before the life protection control parameters are input into the distributed collaborative learning network, the following steps are also included: grouping the life protection control parameters according to the differences in node computing capabilities within the battery pack, setting corresponding update cycles for control nodes with different computing capabilities, and obtaining heterogeneous node group data; dividing the heterogeneous node group data into parallel computing tasks, building a task allocation matrix based on the communication bandwidth and computing delay of each node, and obtaining a computing resource allocation plan; inputting the computing resource allocation plan into the ADMM iterative solver, calculating the objective function value and dual variable of each node, and obtaining a local optimization result; performing asynchronous update processing on the local optimization result, dynamically adjusting the update order based on the node response time, and obtaining an asynchronous iteration sequence; inputting the asynchronous iteration sequence into the prediction compensation model, forward predicting the state of the node with a large communication delay, and obtaining state prediction data; constructing anomaly detection rules based on the state prediction data, identifying abnormal nodes that deviate from the predicted trajectory, and obtaining abnormal node tags; substituting the abnormal node tags into the information compensator, correcting and filling the data of the abnormal nodes, and obtaining compensation control data; verifying the convergence of the asynchronous iteration process according to the compensation control data, calculating the convergence error of the global objective function, and obtaining the optimized control parameters.

[0075] Optional, distributed energy management systems for improved lithium battery cycle life also include: The energy balance adjustment unit is used to: input the life protection control parameter into the model prediction controller, calculate the local protection control instruction sequence of the single cell, and obtain the target control data; input the target control data into the Coulomb efficiency compensation model, calculate the charge and discharge compensation current of each single cell, and obtain the balance compensation parameter; input the balance compensation parameter into the data interaction network between adjacent control nodes, calculate the optimal control quantity of each control node, and obtain the coordinated control variable; substitute the coordinated control variable into the Lyapunov function, calculate the convergence speed of the state consistency between the nodes, and obtain the stability evaluation data; based on the stability evaluation data, calculate the reconstruction strategy of the control network, and obtain the fault-tolerant control parameter; substitute the fault-tolerant control parameter into the equivalent circuit network, calculate the energy transmission path and power size between the nodes, and obtain the balanced control instruction.

[0076] In this embodiment, the life protection control parameter Input the model predictive controller to calculate the local protection control instruction sequence of the single battery and obtain the target control data. The model predictive controller is based on the system's dynamic equations and uses future state predictions to optimize the current control decision. Inside, minimize the objective function:

[0077] in, For the The battery is in the prediction step The state of charge, is the reference SOC, is the current at the current prediction moment, is the target current, is the weight coefficient. The optimization goal is to ensure that the SOC of the battery pack is maintained in the optimal range, while balancing the charge and discharge current of each battery to avoid over-discharge or over-charge of some battery cells, thereby outputting the target control data. After obtaining the target control data, the target control data is input into the coulomb efficiency compensation model to calculate the charge and discharge compensation current of each battery cell, taking into account the coulomb efficiency difference between battery cells. , and thus the equilibrium compensation parameters are obtained. Coulomb efficiency Describes the charge loss of the battery during the charging and discharging process, and is defined as follows:

[0078] in, and are the charge amounts during charging and discharging, respectively. For inefficient batteries, the charging current is increased to compensate for the coulomb efficiency loss, so the compensation current calculation formula is:

[0079] The battery with lower coulombic efficiency obtains additional charging current to ensure that its SOC can return to the equilibrium state as soon as possible and obtain the equilibrium compensation parameters. After calculating the equilibrium compensation parameters, the information between the control nodes can be effectively transmitted, so the equilibrium compensation parameters are input into the data interaction network between adjacent control nodes to calculate the optimal control amount of each control node and obtain the coordinated control variable. In a distributed battery management system (BMS), each control node can only exchange data with adjacent batteries, so its control optimization problem is modeled as an adjacency matrix optimization problem, which is defined as follows:

[0080] in, Representatives and batteries The set of adjacent control nodes, is the weight coefficient in the adjacency matrix. This formula ensures that the control quantity is coordinated and adjusted between adjacent nodes to achieve dynamic balance of SOC. After obtaining the cooperative control variables, the stability of the system is evaluated. The cooperative control variables are substituted into the Lyapunov function to calculate the convergence speed of the state consistency between nodes and obtain the stability evaluation data. The Lyapunov function is used to measure the energy attenuation of the system, which is defined as:

[0081] in, is the average SOC of all batteries, and the state consistency of the system is determined by the derivative of the Lyapunov equation:

[0082] in, is the convergence rate coefficient of the system, if , it indicates that the SOC state is gradually becoming consistent, otherwise the balancing control strategy needs to be adjusted to obtain stability evaluation data. Based on the stability evaluation data, the reconstruction strategy of the control network is calculated to adjust the control scheme in the event of a fault or communication anomaly to obtain fault-tolerant control parameters. Assuming that the control instructions of some battery cells are abnormal, the system needs to dynamically adjust the adjacency matrix to bypass the faulty node, and its optimization goal is: ; in, represents the Frobenius norm, is a set of faulty nodes. This optimization problem ensures that the network topology is reconstructed under the premise of minimizing network disturbances, and a new adjacency matrix is ​​obtained. , and then calculate the fault-tolerant control parameters. Substitute the fault-tolerant control parameters into the equivalent circuit network to calculate the energy transmission path and power size between nodes, and finally generate the balancing control instructions. The equivalent circuit model represents the power flow of the battery pack. Assume that the battery cell The power transmission path is , then the calculation formula is as follows:

[0083] in, For battery and The equivalent resistance between the two batteries is used to calculate the power flow between the batteries to ensure that the energy can be effectively balanced. The balancing control command is determined by the following formula:

[0084] The balancing control instructions take into account individual optimization, charge compensation and coordinated control to ensure that the battery pack can achieve optimal energy distribution.

[0085] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment system can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned systems. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0086] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or system. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, device, article or system including the element.

[0087] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A distributed energy management system for improving the cycle life of lithium batteries, characterized in that: The system includes: A collection unit, used to collect battery status characteristic data of battery cells in a lithium battery pack; an identification unit, configured to identify abnormal charging and discharging patterns on the battery status characteristic data to obtain abnormal charging and discharging event data; A decomposition unit, used for performing hierarchical feature decomposition and multi-dimensional fusion based on the abnormal charge and discharge event data to obtain a cycle life attenuation feature matrix; A balancing control unit is used to perform life balancing control based on the cycle life attenuation characteristic matrix to obtain a target charge and discharge coefficient.

2. The distributed energy management system for improving the cycle life of lithium batteries according to claim 1, characterized in that: The acquisition unit is specifically used for: The voltage data of the battery cells in the lithium battery pack are collected through the voltage collection module, the temperature data are collected through the temperature collection module, and the internal resistance data are collected through the internal resistance measurement module to obtain the original collected data; Eliminating high-frequency noise from the original collected data to obtain filtered processed data, and calculating the voltage change rate, temperature gradient, and internal resistance increment of the filtered processed data to obtain characteristic quantity data; Packing the characteristic data into data frames and transmitting them to the upper control unit via the CAN bus to obtain a transmission data packet; Calculating the change rate of the voltage data and the temperature data in the transmission data packet, dynamically adjusting the sampling frequency according to the change rate value, and obtaining an adaptive sampling parameter; The adaptive sampling parameters are returned to the voltage acquisition module, the temperature acquisition module and the internal resistance measurement module, and the sampling frequency is updated and configured to obtain the battery status characteristic data.

3. The distributed energy management system for improving the cycle life of lithium batteries according to claim 1, characterized in that: The identification unit is specifically used for: Based on the battery state characteristic data, the battery terminal voltage, surface temperature, and internal resistance change are constructed as observation vectors, and the state of charge, health state, and polarization voltage are constructed as state vectors to obtain a Kalman filter state equation; Calculating the mean and variance of the Kalman filter state equation, updating the process noise covariance matrix and the observation noise covariance matrix through recursive operation, and obtaining noise characteristic parameters; Substituting the noise characteristic parameters into the strong tracking filter algorithm, calculating the optimal attenuation factor through the exponential attenuation function, correcting the prediction error covariance matrix, and obtaining the state prediction result; Perform multi-dimensional comparison on the state prediction result with preset voltage threshold, temperature threshold, and internal resistance threshold to obtain abnormal deviation data; The abnormal deviation data is cumulatively counted, and when the abnormal duration of any state quantity exceeds the corresponding time threshold, an abnormal trigger signal is obtained, and the abnormal trigger signal is graded according to the deviation magnitude to obtain the charging and discharging abnormal event data.

4. The distributed energy management system for improving the cycle life of lithium batteries according to claim 1, characterized in that: The decomposition unit also includes: A division subunit is used to divide the abnormal charge and discharge event data into groups of N adjacent battery cells, and calculate the voltage differential capacity, temperature distribution and internal resistance characteristics of each group of divided battery cells to obtain feature vector data; An extraction subunit, used to calculate the eigenvector mutual correlation coefficients between the calculation subdomains based on the eigenvector data to obtain a spatial correlation matrix, and to extract the frequency band energy distribution of the voltage data and the temperature data to obtain a time-frequency feature matrix; A calculation subunit, used for calculating core tensor coefficients and a three-dimensional factor matrix according to the spatial correlation matrix and the time-frequency feature matrix to obtain a multimodal feature tensor; A construction subunit is used to perform low-rank approximate calculation on the multimodal feature tensor to obtain feature component data, and to construct an attenuation matrix of the spatiotemporal evolution characteristics based on the feature component data to obtain a cycle life attenuation feature matrix.

5. The distributed energy management system for improving the cycle life of lithium batteries according to claim 4, characterized in that: The computing subunit is specifically used for: Rearranging the spatial correlation matrix according to the physical position relationship of the battery cells, and reorganizing the time-frequency feature matrix according to the time sequence to obtain initial tensor data; The initial tensor data is normalized, and normalization coefficients of the time dimension, the space dimension, and the feature dimension are calculated to obtain a normalized tensor; Decomposing the normalized tensor into three orthogonal basis matrices, calculating a characteristic basis matrix of a time dimension, a characteristic basis matrix of a space dimension, and a characteristic basis matrix of a characteristic dimension, and obtaining an orthogonal decomposition result; Performing Tucker decomposition operation on the orthogonal decomposition result to obtain a tensor decomposition equation, and substituting the tensor decomposition equation into an alternating least squares method for iterative solution to obtain optimized tensor parameters; A tensor reconstruction operation is performed on the optimized tensor parameters to obtain a multimodal feature tensor.

6. The distributed energy management system for improving the cycle life of lithium batteries according to claim 1, characterized in that: The balancing control unit is specifically used for: Substituting the cycle life attenuation characteristic matrix into the Thevenin equivalent circuit model, calculating the polarization resistance, polarization capacitance and state of charge value of each battery cell, and obtaining the underlying state data; Input the bottom state data and the influencing relationship parameters between the battery cells into the particle swarm filter calculation model, calculate the charge state distribution curve of the entire battery pack, and obtain the middle state data; Performing difference analysis on the state of charge and health state of each battery cell in the middle-level state data, calculating the charge and discharge weight factor of each cell, and obtaining the distribution weight coefficient; Substituting the distribution weight coefficient into the current optimization model, calculating the corresponding optimal charge and discharge current limit for the battery cells that deviate from the average value, and obtaining the balancing control parameter; The balance control parameter and the electrochemical characteristic correlation matrix are cross-calculated to obtain risk distribution data, and the charge and discharge rate of the battery pack is planned by region based on the risk distribution data to obtain the target charge and discharge coefficient.

7. The distributed energy management system for improving the cycle life of lithium batteries according to claim 1, characterized in that: The distributed energy management system for improving the cycle life of lithium batteries also includes: The real-time constraint unit is used to input the target charge and discharge coefficient and the charge and discharge depth, rate and temperature parameters into the electrochemical stress model, calculate the fatigue stress distribution of the single cell, and obtain electrochemical damage data; perform time-series accumulation operation on the electrochemical damage data according to the linear fatigue damage criterion, construct the life attenuation curve of each battery cell, and obtain a fatigue damage degree matrix; divide the fatigue damage degree matrix into gears according to the degree of damage, set corresponding charge and discharge rate upper limits for different gears, and obtain stratified restriction parameters; calculate the power attenuation ratio of each single cell based on the stratified restriction parameters and the battery surface temperature data, and obtain the temperature correction coefficient; calculate the power allocation strategy inside the battery pack in combination with the temperature correction coefficient and the total power demand of the system, and obtain the power optimization parameter; perform charge and discharge stage-by-stage processing on the power optimization parameter, set a dynamic current limit interval for each battery cell, and obtain the life protection control parameter.

8. The distributed energy management system for improving the cycle life of lithium batteries according to claim 1, characterized in that: The distributed energy management system for improving the cycle life of lithium batteries also includes: The energy balance adjustment unit is used to: input the life protection control parameter into the model prediction controller, calculate the local protection control instruction sequence of the single cell, and obtain the target control data; input the target control data into the Coulomb efficiency compensation model, calculate the charge and discharge compensation current of each single cell, and obtain the balance compensation parameter; input the balance compensation parameter into the data interaction network between adjacent control nodes, calculate the optimal control quantity of each control node, and obtain the coordinated control variable; substitute the coordinated control variable into the Lyapunov function, calculate the convergence speed of the state consistency between the nodes, and obtain the stability evaluation data; based on the stability evaluation data, calculate the reconstruction strategy of the control network, and obtain the fault-tolerant control parameter; substitute the fault-tolerant control parameter into the equivalent circuit network, calculate the energy transmission path and power size between the nodes, and obtain the balance control instruction.

Images