Battery life real-time evaluation method of energy storage system, medium and system

Abstract

The invention provides an energy storage system battery life real-time evaluation method, medium and system, and belongs to the technical field of energy storage system data processing, and the method comprises the steps: collecting the voltage, current, temperature, internal resistance and other data of an energy storage system battery in real time, carrying out the multivariate empirical mode decomposition of the voltage and current data, and obtaining the steady state and change data. And in combination with the physical and chemical reaction equation set of the battery, calculating to obtain a recession characteristic matrix of the battery material. Meanwhile, a battery internal environment characteristic matrix is established, and a comprehensive characteristic matrix of the battery is obtained through fusion operation of the two. And inputting the matrix into a pre-trained battery loss evaluation model, calculating battery loss indexes such as an electrode material structure loss rate, an electrolyte decomposition loss rate, a solid electrolyte interface film growth rate, an active substance loss rate and a capacity fading rate, and evaluating the remaining service life of the battery. The technical problem that it is difficult to evaluate the health state and the remaining service life of the battery online in real time in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy storage system data processing, and in particular, relates to a method, medium and system for real-time evaluation of battery life of an energy storage system. Background Art

[0002] In recent years, with the rapid development of renewable energy and the widespread application of electric vehicles and energy storage systems, the demand for high-performance electrochemical energy storage devices has increased. Among them, lithium-ion batteries, as the most mainstream electrochemical energy storage technology, have great advantages in key performance indicators such as energy density, power density, and cycle life, and are widely used in mobile electronic devices, power systems, rail transportation and other fields. However, in actual use, batteries will be affected by various factors such as charging and discharging frequency, temperature environment, mechanical stress, etc., and various chemical, electrochemical and mechanical degradations will occur, resulting in the gradual attenuation of battery capacity, the continuous increase of internal resistance, and even safety hazards. Therefore, how to accurately evaluate the remaining service life of the battery is of great significance for guiding users to use the battery reasonably, preventing safety accidents, and improving the reliability of the energy storage system. At present, the commonly used methods for battery life assessment are mainly the following:

[0003] 1. Life prediction based on accelerated aging test. The battery is subjected to accelerated aging test under harsh conditions such as high temperature, high voltage, and high current. The battery life under normal use conditions is estimated based on the attenuation curve of battery performance with the number of cycles. This method can quickly obtain battery life data, but due to the difference between accelerated test conditions and actual use environment, the prediction results often have a large deviation from the actual life.

[0004] 2. Life prediction based on empirical models. By statistically analyzing a large amount of measured data, an empirical relationship model between battery life and influencing factors is established, such as Reynolds distribution and Weibull distribution. This method is simple and easy to implement, but it requires a lot of experimental data support, and the model parameters are difficult to directly relate to the actual operating status of the battery.

[0005] 3. Life prediction based on physical and chemical mechanisms. Physical and chemical models such as chemical kinetics, electrochemical reaction kinetics, stress and strain of electrode materials are used to simulate various degradation mechanisms of batteries during operation, and then predict the remaining life of the battery. This method can better reflect the essential characteristics of the battery, but the model is highly complex and requires a lot of parameter calibration work.

[0006] The above life assessment methods all have certain limitations, and it is difficult to evaluate the health status and remaining service life of the battery online in real time. On the one hand, accelerated aging tests and empirical models cannot timely capture the dynamic degradation characteristics of the battery in the actual operating environment; on the other hand, although the method based on physical and chemical mechanisms is closer to the essence of the battery, the model is highly complex and difficult to quickly deploy and apply in actual systems. Therefore, there is an urgent need for a comprehensive battery life assessment method that integrates battery operation data analysis, physical and chemical modeling, and machine learning to meet the urgent needs of battery status monitoring and life prediction in the fields of power storage, electric vehicles, etc. Summary of the invention

[0007] In view of this, the present invention provides a method, medium and system for real-time evaluation of battery life of an energy storage system, which can solve the technical problem in the prior art that it is difficult to evaluate the health status and remaining service life of the battery online in real time.

[0008] The present invention is achieved in that:

[0009] A first aspect of the present invention provides a method for real-time evaluation of battery life of an energy storage system, comprising the following steps:

[0010] S10, real-time collection of voltage data, current data, temperature data at multiple collection points, and internal resistance data at multiple collection points during the operation of the energy storage system battery; specifically: collect real-time voltage data, current data, temperature data at multiple collection points, and internal resistance data during the operation of the energy storage system battery. First, the voltage and current sensors set on the energy storage system battery are used to record the voltage and current changes of the battery in real time. At the same time, multiple temperature sensors are arranged inside or on the surface of the battery to collect temperature data inside and on the surface of the battery in real time. In addition, the resistance detection device is used to measure the internal resistance changes at multiple locations inside the battery in real time. These real-time collected voltage data, current data, temperature data, and internal resistance data constitute a complete battery operation status data set. These data are the basis for subsequent physical and chemical model calculations and battery loss assessments. By collecting these parameter data in real time, the changes in the electrochemical, thermal, and mechanical properties of the battery during operation can be fully characterized.

[0011] S20. Perform multivariate empirical mode decomposition on the voltage data and current data to obtain a voltage steady-state data matrix, a voltage variation data matrix, a current steady-state data matrix, and a current variation data matrix; specifically: perform multivariate empirical mode decomposition (MEMD) on the collected voltage data and current data. First, input the voltage time series signal and the current time series signal into the MEMD algorithm respectively. The MEMD algorithm is a multidimensional signal decomposition method based on empirical mode decomposition (EMD), which can decompose complex time series signals into several intrinsic mode functions (IMFs) and a residual term. Through MEMD decomposition, the voltage steady-state data matrix V can be obtained. s , voltage change data matrix V d , current steady-state data matrix I s and current change data matrix I d Among them, the voltage steady-state data matrix contains the low-frequency steady-state component and residual term of the voltage, and the voltage variation data matrix contains the high-frequency variation component of the voltage. The current steady-state data matrix and the current variation data matrix also reflect the steady-state and variation characteristics of the current. These decomposed data matrices can better describe the dynamic changes in the electrochemical reaction process of the battery.

[0012] S30. Establish a set of physical and chemical reaction equations for batteries that take into account voltage, current, temperature and internal resistance, including: electrochemical reaction balance equation, ion diffusion mass transfer equation, electrode material stress-strain equation, solid-liquid interface reaction kinetics equation, electrolyte conductivity equation; by establishing such a set of physical and chemical reaction equations covering electrochemical, thermal and mechanical properties, it is possible to comprehensively describe the various degradation mechanisms of the battery during operation, laying the foundation for subsequent battery loss assessment.

[0013] S40, substituting the voltage steady-state data matrix, voltage variation data matrix, current steady-state data matrix, current variation data matrix, the temperature data of the plurality of acquisition points, and the internal resistance data of the plurality of acquisition points into the battery physical and chemical reaction equations to calculate the battery material degradation characteristic matrix; specifically: substituting the voltage steady-state data matrix V obtained in step S20 into the battery material degradation characteristic matrix s , voltage change data matrix V d , current steady-state data matrix I s , current change data matrix I d The temperature data and internal resistance data collected in step S10 are substituted into the battery physical and chemical reaction equations established in step S30 for calculation. By solving these partial differential equations, various degradation characteristic values ​​of battery materials can be obtained, such as changes in electrode material structure, degree of electrolyte decomposition, growth rate of solid electrolyte interface film, loss rate of active material, etc. These characteristic values ​​can form a battery material degradation characteristic matrix M, where each element m ijIt represents the influence of the i-th degradation mechanism on the j-th performance index, which is obtained by weighted summation.

[0014] S50, based on the temperature data of the plurality of acquisition points and the internal resistance data of the plurality of acquisition points, establish a battery internal environment characteristic matrix; specifically: construct a battery internal environment characteristic matrix H using the temperature data and internal resistance data acquired in step S10. Each row of the matrix contains the temperature T of a measurement point i and internal resistance R i , reflecting the internal environmental conditions of the battery at different locations. Temperature and internal resistance are key factors affecting battery performance and life, and their spatial distribution directly determines the overall degradation characteristics of the battery. Therefore, the battery internal environment characteristic matrix H is an important input parameter for evaluating the comprehensive loss of the battery.

[0015] S60, fusing the battery material degradation characteristic matrix with the battery internal environment characteristic matrix to obtain a battery comprehensive characteristic matrix; specifically: fusing the battery material degradation characteristic matrix M obtained in step S40 and the battery internal environment characteristic matrix H obtained in step S50 to obtain a battery comprehensive characteristic matrix. Specifically, the two matrices can be fused in a weighted average manner:

[0016] G = w1M + w2H;

[0017] Among them, w1 and w2 are the fusion weight coefficients of the two matrices, which can be adjusted according to the actual situation. The battery comprehensive characteristic matrix G reflects the various degradation mechanisms and internal environmental conditions of the battery during operation, providing comprehensive input data for subsequent battery life evaluation.

[0018] S70, inputting the battery comprehensive characteristic matrix into a pre-trained battery loss assessment model, and calculating the battery loss vector, including the electrode material structure loss rate, electrolyte decomposition loss rate, solid electrolyte interface film growth rate, active material loss rate, and capacity decay rate; specifically:

[0019] The battery comprehensive feature matrix G obtained in step S60 is input into a pre-trained battery loss assessment model to calculate the battery loss vector D. This battery loss assessment model adopts a method based on deep learning, and the specific structure is as follows:

[0020] The input layer uses a one-dimensional convolutional layer to extract time series features; the middle layer contains two long short-term memory (LSTM) layers to capture long-term dependencies; the output layer is a fully connected layer that outputs five battery loss factors, namely, electrode material structure loss rate d1, electrolyte decomposition loss rate d2, solid electrolyte interface film growth rate d3, active material loss rate d4 and capacity attenuation rate d5.

[0021] In order to train this model, we first need to establish a training data set. The specific approach is: conduct cycle life tests on multiple groups of different types of batteries, collect battery status data every 100 cycles and calculate physical and chemical parameters, use 80% of the data for the training set and 20% for the test set. The Adam optimizer is used in the training process, with a learning rate of 0.001, a batch size of 32, and a loss function of mean square error plus L2 regularization. The early stopping strategy is used in the training process to monitor the performance of the validation set until the model converges. Finally, the cross-validation method is used to evaluate the generalization performance of the model.

[0022] Through the trained battery loss assessment model, the battery comprehensive feature matrix G obtained in step S60 can be input, and the output is various battery loss factors d1-d5, that is, the battery loss vector D. These loss factors provide a key basis for subsequent battery life prediction.

[0023] S80: Calculate the battery remaining life evaluation value using a preset battery life empirical equation according to the battery loss vector.

[0024] On the basis of the above technical solution, the real-time evaluation method of battery life of an energy storage system of the present invention can also be improved as follows:

[0025] Among them, the steps of establishing the training data set of the battery loss assessment model include: first, conducting cycle life tests on multiple groups of different types of batteries, and the test process continues until the capacity decays to 80% of the rated capacity; for each group of test processes, the battery status data is collected every 100 cycles and the physical and chemical parameters are calculated; the calculated battery material decay characteristic matrix and the battery internal environment characteristic matrix are divided according to the ratio of 80% for the training set and 20% for the test set; the training data is normalized and standardized to eliminate the dimension effect; finally, the processed characteristic data is paired with the corresponding actual capacity decay data to construct a training sample.

[0026] The specific structure of the battery loss evaluation model is as follows: the input layer uses a one-dimensional convolutional layer with 64 neurons to extract time series features; the middle layer contains two long short-term memory layers, each with 128 neurons to capture long-term dependencies; the output layer is a fully connected layer with 5 neurons, corresponding to the electrode material structure loss rate, electrolyte decomposition loss rate, solid electrolyte interface film growth rate, active material loss rate, and capacity decay rate; a dropout layer is set between each layer, and the dropout rate is 0.2 to prevent overfitting. The training steps of the battery loss evaluation model include: first using the Adam optimizer, setting the learning rate to 0.001, and the batch size to 32; using the mean square error as the loss function, and introducing the L2 regularization term, with a regularization coefficient of 0.01; using an early stopping strategy during training, and stopping training when the validation set loss does not decrease for 5 consecutive epochs; after each epoch, using the test set to evaluate the model performance, recording indicators such as accuracy and mean square error; when the model converges, using the cross-validation method to evaluate the generalization performance of the model.

[0027] Furthermore, the multivariate empirical mode decomposition of the voltage data and the current data is specifically expressed as follows:

[0028]

[0029]

[0030] Where: V(t) is the time domain voltage signal; I(t) is the time domain current signal; c vi (t) is the voltage i-th intrinsic mode function; c ii (t) is the i-th eigenmode function of current; r vn (t) and r in (t) are the voltage and current residual terms respectively; n v and n i is the decomposition order;

[0031] The obtained voltage steady-state data matrix is ​​specifically expressed as:

[0032]

[0033] In the formula, c v1 represents the first intrinsic mode function after the voltage signal is decomposed, representing the main steady-state component of the voltage; r vn represents the residual term after the voltage signal is decomposed, representing the long-term trend component; t1, t2, ..., t n ,t m Represents the time series sampling point; the voltage change data matrix is ​​specifically expressed as:

[0034]

[0035] In the formula, c v2 ,c v3 ,...,c vn Represents the 2nd to nth intrinsic mode functions of the voltage signal, representing the fluctuation components of different frequencies; each column corresponds to the value of an intrinsic mode function at different time points; each row of the matrix represents different frequency components at the same time point;

[0036] The current steady-state data matrix is ​​specifically expressed as:

[0037]

[0038] In the formula, c i1 represents the first intrinsic mode function after the current signal is decomposed, representing the main steady-state component of the current; r in Represents the residual term after the current signal is decomposed; the current change data matrix is ​​specifically expressed as:

[0039]

[0040] In the formula, c i2 , c i3 , ..., c in It represents the 2nd to nth intrinsic mode functions of the current signal, representing the fluctuation components of different frequencies; each column corresponds to the value of an intrinsic mode function at a different time point; each row of the matrix represents a different frequency component at the same time point.

[0041] In the battery physicochemical reaction equation group, the electrochemical reaction balance equation is specifically expressed as follows:

[0042]

[0043]

[0044] Where: C s is the solid phase lithium ion concentration; C e is the liquid phase lithium ion concentration; D s is the solid phase diffusion coefficient; D e is the liquid phase diffusion coefficient; r is the radial coordinate of the spherical particle; x is the coordinate in the electrode thickness direction; t + is the lithium ion migration number; F is the Faraday constant; j Li is the lithium ion insertion and extraction reaction rate;

[0045] The ion diffusion mass transfer equation is specifically expressed as follows:

[0046]

[0047]

[0048] Where: i0 is the exchange current density; k c , k a is the reaction rate constant; is the maximum lithium ion concentration; is the surface lithium ion concentration; η is the overpotential; α a is the anode transfer coefficient; α c is the cathode transfer coefficient; R is the gas constant, which is 8.314 J / (mol·K); T is the temperature, in K.

[0049] The stress-strain equation of the electrode material is specifically expressed as follows:

[0050]

[0051]

[0052] Where: σ is the stress tensor; E is Young's modulus; ε is the strain tensor; Ω is the partial molar volume; is the initial lithium ion concentration; ε xx , ε yy , ε zz is the strain component; γ xy , γ yz , γ zx is the shear strain component.

[0053] The solid-liquid interface reaction kinetic equations are specifically expressed as follows:

[0054]

[0055]

[0056] Where: SEI is the thickness of the solid electrolyte interface film; k sei is the growth rate constant; E a is the activation energy; α is the reaction order; R sei is the interface impedance; L sei is the film thickness; κ sei is the membrane conductivity;

[0057] The electrolyte conductivity equation is specifically expressed as follows:

[0058]

[0059]

[0060] Where: is the reference diffusion coefficient; is the diffusion activation energy; κ is the conductivity of the electrolyte; z i is the ion charge number; Di is the ion diffusion coefficient; C i is the ion concentration.

[0061] The battery material degradation characteristic matrix calculation is specifically expressed as follows:

[0062]

[0063] In the formula, m ij =w1a ij +w2b ij +w3e ij ;

[0064] Where: a ij , b ij , e ij are electrochemical, mechanical and electrolyte characteristic values ​​respectively; w1, w2, w3 are weight coefficients;

[0065] The battery internal environment characteristic matrix is ​​specifically expressed as follows:

[0066]

[0067] Where: T i is the temperature of the i-th measuring point; R i is the internal resistance of the ith measuring point;

[0068] The battery loss vector is specifically expressed as follows:

[0069] D = [d1, d2, d3, d4, d5] T ;

[0070]

[0071]

[0072]

[0073]

[0074]

[0075] Where: d1 is the electrode material structure loss rate; d2 is the electrolyte decomposition loss rate; d3 is the SEI film growth rate; d4 is the active material loss rate; d5 is the capacity attenuation rate; k1~k6 are weight coefficients; V0, R0 are the initial voltage and internal resistance; is the maximum SEI film thickness; σ yield is the yield stress; is the initial active substance content; Q0 is the initial capacity.

[0076] The empirical equation for battery life is as follows:

[0077]

[0078] The parameters are described as follows:

[0079] 1. SOL indicates the remaining battery life percentage, with a value range of [0, 100] and a unit of %;

[0080] 2.d1 represents the structural loss rate of the electrode material, with a value range of [0, 1] and dimensionless;

[0081] 3.d2 represents the electrolyte decomposition loss rate, with a value range of [0, 1] and dimensionless;

[0082] 4.d3 represents the growth rate of the solid electrolyte interface film, with a value range of [0, 1] and dimensionless;

[0083] 5.d4 represents the loss rate of active substances, with a value range of [0, 1] and dimensionless;

[0084] 6.d5 represents the capacity attenuation rate, with a value range of [0, 1] and is dimensionless;

[0085] 7.λ1 represents the electrode structure loss weight coefficient, the value range is [0.8, 1.2], the optimal value is 1.0, dimensionless;

[0086] 8.λ2 represents the electrolyte loss weight coefficient, with a value range of [0.6, 1.0], an optimal value of 0.8, and dimensionless;

[0087] 9.λ3 represents the interface film growth weight coefficient, ranging from [0.5, 0.8], with an optimal value of 0.6, dimensionless;

[0088] 10.λ4 represents the active material loss weight coefficient, with a value range of [1.0, 1.5], an optimal value of 1.2, and dimensionless;

[0089] 11.λ5 represents the capacity attenuation weight coefficient, with a value range of [1.2, 1.8], an optimal value of 1.5, and dimensionless;

[0090] 12.β1 represents the nonlinear coefficient of electrode structure loss, ranging from [1.5, 2.0], with an optimal value of 1.8, dimensionless;

[0091] 13.β2 represents the nonlinear coefficient of electrolyte loss, ranging from [1.2 to 1.8], with an optimal value of 1.5, dimensionless;

[0092] 14.β3 represents the nonlinear coefficient of interfacial film growth, ranging from [1.0, 1.5], with an optimal value of 1.2, dimensionless;

[0093] 15.β4 represents the nonlinear coefficient of active material loss, ranging from [1.8, 2.5], with an optimal value of 2.0, dimensionless;

[0094] 16.β5 represents the capacity attenuation nonlinear coefficient, with a value range of [2.0, 3.0], an optimal value of 2.5, and dimensionless;

[0095] Special instructions:

[0096] 1. The exponential function exp represents the natural exponential function e x ;

[0097] 2. All weight coefficients λ i and the nonlinear coefficient β i It is necessary to obtain the optimal value through experimental data calibration;

[0098] 3. When any loss factor reaches a critical value, the exponential function value of the corresponding item will quickly approach zero, resulting in a significant reduction in the remaining life;

[0099] 4. Each loss factor reflects the coupling effect in the form of a product. Severe degradation of any factor will lead to a significant reduction in lifespan.

[0100] 5. The parameter range and optimal value are summarized based on the experimental data of lithium iron phosphate batteries. Other types of batteries need to be recalibrated.

[0101] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed in a computer, they are used to execute the above-mentioned method for real-time evaluation of battery life of an energy storage system.

[0102] A third aspect of the present invention provides a real-time battery life assessment system for an energy storage system, which includes the above-mentioned computer-readable storage medium.

[0103] Compared with the prior art, the method, medium and system for real-time evaluation of battery life of an energy storage system provided by the present invention have the following beneficial effects:

[0104] 1. Real-time collection of multi-dimensional battery operation data. By deploying voltage, current, temperature and internal resistance detection devices, multi-dimensional state data of the battery during operation can be obtained in real time, providing comprehensive basic data for subsequent physical and chemical analysis and machine learning. This multi-sensor fusion method can better capture various degradation characteristics during the dynamic operation of the battery.

[0105] 2. Establish a battery physicochemical model covering multiple degradation mechanisms. The inventors have established a complete set of reaction equations for complex physicochemical processes such as electrochemical reactions, ion mass transfer, stress and strain that occur during battery operation. These equations can not only describe various degradation mechanisms of batteries, but also perform parameter identification and model update based on real-time monitoring data, greatly improving the accuracy and reliability of the model.

[0106] 3. Use deep learning to evaluate battery loss. The inventors designed a deep learning model based on a one-dimensional convolutional network and LSTM, which can take the comprehensive characteristic matrix of the battery as input and output multiple battery loss factors including the electrode material structure loss rate, electrolyte decomposition loss rate, solid electrolyte interface film growth rate, etc. This data-driven machine learning method can effectively capture the complex nonlinear degradation laws of batteries and realize automatic learning and adaptive updates.

[0107] 4. Establish an empirical equation for battery life that takes into account the coupling effect. The inventors proposed an empirical equation for battery life that takes into account the coupling effect of various degradation mechanisms. Through the product form of exponential functions, it can accurately reflect the characteristic that severe degradation of any loss factor will lead to a significant decrease in battery life. The precise calibration of the empirical equation parameters ensures the accuracy of the remaining battery life assessment.

[0108] In general, the real-time battery life assessment method proposed in the present invention fully integrates battery operation status monitoring, physical and chemical modeling analysis and machine learning technology, realizes the refined characterization and prediction of the dynamic degradation process of the battery, provides strong technical support for the safe operation of the energy storage system, and solves the technical problem that the existing technology is difficult to evaluate the health status and remaining service life of the battery online in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] Figure 1 A flow chart of the method provided by the present invention; DETAILED DESCRIPTION

[0110] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0111] like Figure 1 FIG. 1 is a flow chart of a method for real-time evaluation of battery life of an energy storage system provided by the present invention. The method comprises the following steps:

[0112] S10, real-time collection of voltage data, current data, temperature data at multiple collection points, and internal resistance data at multiple collection points during the operation of the energy storage system battery;

[0113] S20, performing multivariate empirical mode decomposition on the voltage data and the current data to obtain a voltage steady-state data matrix, a voltage variation data matrix, a current steady-state data matrix, and a current variation data matrix;

[0114] S30, establishing a battery physical and chemical reaction equation group taking into account voltage, current, temperature and internal resistance, including: electrochemical reaction balance equation, ion diffusion mass transfer equation, electrode material stress strain equation, solid-liquid interface reaction kinetics equation, electrolyte conductivity equation;

[0115] S40, substituting the voltage steady-state data matrix, the voltage variation data matrix, the current steady-state data matrix, the current variation data matrix, the temperature data of multiple collection points, and the internal resistance data of multiple collection points into the battery physical and chemical reaction equation group to calculate the battery material degradation characteristic matrix;

[0116] S50, establishing a battery internal environment characteristic matrix based on the temperature data of multiple collection points and the internal resistance data of multiple collection points;

[0117] S60, fusing the battery material degradation characteristic matrix with the battery internal environment characteristic matrix to obtain a battery comprehensive characteristic matrix;

[0118] S70, inputting the battery comprehensive characteristic matrix into a pre-trained battery loss assessment model, and calculating the battery loss vector, including the electrode material structure loss rate, the electrolyte decomposition loss rate, the solid electrolyte interface film growth rate, the active material loss rate, and the capacity decay rate;

[0119] S80: Calculate the battery remaining life evaluation value based on the battery loss vector using a preset battery life empirical equation.

[0120] The specific implementation methods of the above steps are described in detail below:

[0121] The specific implementation method of step S10 is to collect real-time voltage data, current data, temperature data and internal resistance data of multiple collection points during the operation of the energy storage system battery. First, the voltage and current sensors set on the energy storage system battery are used to record the voltage and current changes of the battery in real time. At the same time, multiple temperature sensors are arranged inside or on the surface of the battery to collect temperature data inside and on the surface of the battery in real time. In addition, a resistance detection device is used to measure the internal resistance changes at multiple locations inside the battery in real time. These real-time collected voltage data, current data, temperature data and internal resistance data constitute a complete battery operation status data set. These data are the basis for subsequent physical and chemical model calculations and battery loss assessments. By collecting these parameter data in real time, the changes in the electrochemical, thermal and mechanical properties of the battery during operation can be fully characterized.

[0122] The specific implementation of step S20 is to perform multivariate empirical mode decomposition (MEMD) on the collected voltage data and current data. First, the voltage time series signal and the current time series signal are respectively input into the MEMD algorithm. The MEMD algorithm is a multidimensional signal decomposition method based on empirical mode decomposition (EMD), which can decompose complex time series signals into several intrinsic mode functions (IMFs) and a residual term. Through MEMD decomposition, the voltage steady-state data matrix V can be obtained. s , voltage change data matrix V d , current steady-state data matrix I s and current change data matrix I d Among them, the voltage steady-state data matrix contains the low-frequency steady-state component and residual term of the voltage, and the voltage variation data matrix contains the high-frequency variation component of the voltage. The current steady-state data matrix and the current variation data matrix also reflect the steady-state and variation characteristics of the current. These decomposed data matrices can better describe the dynamic changes in the electrochemical reaction process of the battery.

[0123] The specific implementation method of step S30 is to establish a battery physical and chemical reaction equation group taking into account voltage, current, temperature and internal resistance. This equation group includes the following parts:

[0124] 1. The electrochemical reaction balance equation. This equation describes the solid phase lithium ion concentration C inside the battery. s and the liquid phase lithium ion concentration C e The relationship between the change of the electrolyte and the electrolyte over time includes the diffusion process of lithium ions in the electrode material and the electrolyte.

[0125]

[0126]

[0127] 2. Ion diffusion mass transfer equation. This equation describes the kinetics of lithium ion insertion and extraction reactions on the electrode surface, including key parameters such as exchange current density i0 and overpotential η.

[0128]

[0129]

[0130] 3. Electrode material stress-strain equation. This equation describes the stress-strain changes of electrode materials during lithium ion insertion and extraction, including stress tensor σ and strain tensor ε.

[0131]

[0132]

[0133] 4. Solid-liquid interface reaction kinetic equation. This equation describes the growth kinetics of the solid electrolyte interface film (SEI), including the film thickness SEI and the interface impedance R sei .

[0134]

[0135]

[0136] 5. Electrolyte conductivity equation. This equation describes the ion conductivity of the electrolyte, including conductivity κ and ion diffusion coefficient D i .

[0137]

[0138]

[0139] By establishing such a set of physicochemical reaction equations covering electrochemical, thermal and mechanical properties, the various degradation mechanisms of the battery during operation can be fully described, laying the foundation for subsequent battery loss assessment.

[0140] The specific implementation of step S40 is to convert the voltage steady-state data matrix V obtained in step S20 into s , voltage change data matrix V d , current steady-state data matrix I s , current change data matrix I d The temperature data and internal resistance data collected in step S10 are substituted into the battery physical and chemical reaction equations established in step S30 for calculation. By solving these partial differential equations, various degradation characteristic values ​​of battery materials can be obtained, such as changes in electrode material structure, degree of electrolyte decomposition, growth rate of solid electrolyte interface film, loss rate of active material, etc. These characteristic values ​​can form a battery material degradation characteristic matrix M, where each element m ij It represents the influence of the i-th degradation mechanism on the j-th performance index, which is obtained by weighted summation.

[0141] The specific implementation of step S50 is to construct the battery internal environment characteristic matrix H using the temperature data and internal resistance data collected in step S10. Each row of the matrix contains the temperature T of a measuring point. i and internal resistance R i , reflecting the internal environmental conditions of the battery at different locations. Temperature and internal resistance are key factors affecting battery performance and life, and their spatial distribution directly determines the overall degradation characteristics of the battery. Therefore, the battery internal environment characteristic matrix H is an important input parameter for evaluating the comprehensive loss of the battery.

[0142] The specific implementation of step S60 is to fuse the battery material degradation characteristic matrix M obtained in step S40 and the battery internal environment characteristic matrix H obtained in step S50 to obtain a battery comprehensive characteristic matrix. Specifically, the two matrices can be fused in a weighted average manner:

[0143] G = w1M + w2H;

[0144] Among them, w1 and w2 are the fusion weight coefficients of the two matrices, which can be adjusted according to the actual situation. The battery comprehensive characteristic matrix G reflects the various degradation mechanisms and internal environmental conditions of the battery during operation, providing comprehensive input data for subsequent battery life evaluation.

[0145] The specific implementation of step S70 is to input the battery comprehensive feature matrix G obtained in step S60 into a pre-trained battery loss assessment model to calculate the battery loss vector D. This battery loss assessment model adopts a method based on deep learning, and the specific structure is as follows:

[0146] The input layer uses a one-dimensional convolutional layer to extract time series features; the middle layer contains two long short-term memory (LSTM) layers to capture long-term dependencies; the output layer is a fully connected layer that outputs five battery loss factors, namely, electrode material structure loss rate d1, electrolyte decomposition loss rate d2, solid electrolyte interface film growth rate d3, active material loss rate d4 and capacity attenuation rate d5.

[0147] In order to train this model, we first need to establish a training data set. The specific approach is: conduct cycle life tests on multiple groups of different types of batteries, collect battery status data every 100 cycles and calculate physical and chemical parameters, use 80% of the data for the training set and 20% for the test set. The Adam optimizer is used in the training process, with a learning rate of 0.001, a batch size of 32, and a loss function of mean square error plus L2 regularization. The early stopping strategy is used in the training process to monitor the performance of the validation set until the model converges. Finally, the cross-validation method is used to evaluate the generalization performance of the model.

[0148] Through the trained battery loss assessment model, the battery comprehensive feature matrix G obtained in step S60 can be input, and the output is various battery loss factors d1-d5, that is, the battery loss vector D. These loss factors provide a key basis for subsequent battery life prediction.

[0149] The specific implementation of step S80 is to calculate the battery loss vector D obtained in step S70 using a preset battery life empirical equation to obtain a battery remaining life evaluation value.

[0150] The battery life empirical equation is specifically expressed as follows:

[0151]

[0152] Among them, SOL represents the remaining battery life percentage, with a value range of [0,100], and d1~d5 correspond to the electrode material structure loss rate, electrolyte decomposition loss rate, solid electrolyte interface film growth rate, active material loss rate and capacity decay rate respectively. λ1~λ5 are the weight coefficients of each loss factor, and β1~β5 are the nonlinear coefficients of each loss factor, which need to be calibrated through experimental data to obtain the optimal value.

[0153] This empirical equation reflects the coupling effect between various degradation mechanisms of the battery. Through the product form of exponential functions, it can reflect the characteristic that severe degradation of any loss factor will lead to a significant reduction in battery life. Reasonable setting of the empirical formula parameters can ensure the accuracy and reliability of the remaining battery life assessment.

[0154] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed in a computer, they are used to execute the above-mentioned method for real-time evaluation of battery life of an energy storage system.

[0155] A third aspect of the present invention provides a real-time battery life assessment system for an energy storage system, which includes the above-mentioned computer-readable storage medium.

[0156] Specifically, the principle of the present invention is:

[0157] 1. Real-time collection of battery operation data. By installing voltage, current, temperature and internal resistance detection devices on the battery, multi-dimensional state data of the battery during operation can be recorded in real time, providing a basis for subsequent physical and chemical analysis and machine learning. These data can fully reflect the dynamic changes in the electrochemical, thermal and mechanical characteristics of the battery.

[0158] 2. Use multimodal signal decomposition. Input the collected voltage and current time series signals into the multivariate empirical mode decomposition (MEMD) algorithm to obtain the steady-state characteristic matrix and change characteristic matrix of voltage and current. This signal decomposition method can effectively capture the dynamic change characteristics of the battery electrochemical process.

[0159] 3. Establish a battery physical and chemical reaction model. The inventors have established a complete set of reaction equations for complex physical and chemical processes such as electrochemical reaction balance, ion diffusion and mass transfer, electrode material stress and strain, solid-liquid interface reaction kinetics, and electrolyte ion conduction that occur during battery operation. These equations can fully describe various degradation mechanisms of batteries and provide a theoretical basis for subsequent calculations of material degradation characteristics.

[0160] 4. Extract comprehensive battery characteristics. Substituting real-time monitoring data into the physical and chemical model, various degradation characteristics of battery materials can be calculated, such as changes in electrode structure, degree of electrolyte decomposition, growth rate of solid electrolyte membrane, etc. At the same time, the temperature and internal resistance data are used to construct the internal environment characteristic matrix of the battery. Finally, the comprehensive characteristic data of the battery is obtained by weighted fusion of these two characteristic matrices.

[0161] 5. Use deep learning for loss assessment. Inputting comprehensive battery feature data into a pre-trained deep learning model can output multiple battery loss factors including electrode material structure loss rate, electrolyte decomposition loss rate, solid electrolyte membrane growth rate, etc. This data-driven machine learning method can effectively capture the complex nonlinear degradation laws of batteries.

[0162] 6. Calculate the remaining life based on the empirical equation. The remaining service life of the battery can be calculated by using the battery loss factor output by the deep learning model and combining it with the pre-calibrated battery life empirical equation. This empirical equation reflects the coupling effect between various degradation mechanisms and can accurately reflect the overall attenuation law of battery performance.

[0163] In general, the working principle of the present invention is: by collecting multi-dimensional battery status data in real time, using physical and chemical modeling and machine learning methods, comprehensively analyzing the degradation characteristics of the battery during actual operation, and then calculating the remaining service life of the battery. This comprehensive method that combines data analysis, mechanism modeling and intelligent prediction can overcome the limitations of the existing technology and achieve accurate evaluation and reliable prediction of the dynamic health status of the battery.

[0164] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below: A new energy enterprise is developing a large lithium-ion battery energy storage system for power peak regulation. Since the system needs to operate stably for a long time, battery life assessment and health management are one of the key technologies. To this end, the enterprise decided to adopt the real-time battery life assessment method proposed in the present invention to achieve accurate monitoring and prediction of the dynamic degradation process of the battery.

[0165] The energy storage system consists of n = 100 parallel-connected lithium iron phosphate batteries, with each battery pack having a rated voltage of V0 = 3.2V and a rated capacity of Q0 = 20Ah. Before the system was put into use, the company conducted an accelerated life test of 2,000 charge and discharge cycles on 10 representative battery samples in order to obtain performance data of the battery under different degradation states, in preparation for the subsequent physical and chemical model establishment and machine learning training.

[0166] During the accelerated test, the researchers conducted detailed tests on the battery samples every 100 cycles, collecting multi-dimensional state data including voltage, current, temperature and internal resistance. At the same time, the researchers also measured key parameters such as lithium ion concentration, interface film thickness, and active material content inside the battery through chemical analysis. These experimental data provide basic support for subsequent physical and chemical modeling.

[0167] After data analysis, the researchers found that during 2,000 cycles, the battery capacity gradually decreased from the initial 20Ah to 16Ah, which is about 80% of the initial capacity. At the same time, the internal resistance of the battery also increased from 50mΩ to 100mΩ, which is about twice the initial internal resistance. From the X-ray diffraction pattern and scanning electron microscopy morphology of the electrode material, it can be seen that the electrode structure has obvious changes, such as the loss of active substances and the growth of SEI film. These experimental data provide important basis for subsequent physical and chemical analysis.

[0168] Based on the above experimental results, researchers began to establish a physical and chemical model to describe the battery degradation mechanism. First, researchers used Fick's second law to establish a diffusion equation that describes the dynamic changes in the concentration of lithium ions in the solid and liquid phases:

[0169]

[0170]

[0171] Among them, C s and C e are the solid and liquid lithium ion concentrations, respectively, s and D e is the corresponding diffusion coefficient, r and x are the radial and thickness coordinates, t + is the lithium ion migration number, F is the Faraday constant, j Li is the lithium ion insertion and extraction reaction rate.

[0172] Secondly, the researchers established a mass transfer equation to describe the lithium ion insertion and extraction reaction on the electrode surface based on the Butler-Volmer kinetic equation:

[0173]

[0174]

[0175] Where i0 is the exchange current density, k c and k a is the reaction rate constant, is the maximum lithium ion concentration, is the surface concentration, η is the overpotential, α a and α cis the charge transfer coefficient.

[0176] In addition, the researchers also established an equation to describe the stress-strain relationship of electrode materials:

[0177]

[0178] Here, σ is the stress tensor, E is Young's modulus, ε is the strain tensor, Ω is the partial molar volume, is the initial lithium ion concentration.

[0179] For the growth dynamics of the solid electrolyte interface (SEI), researchers established the following equation:

[0180]

[0181]

[0182] Where, SEI is the film thickness, k sei is the growth rate constant, E a is the activation energy, α is the reaction order, R sei is the interface impedance, L sei is the film thickness, κ sei is the membrane conductivity.

[0183] Finally, based on the ionic conductivity of the electrolyte, the researchers used the Nernst-Einstein formula to establish the conductivity equation:

[0184]

[0185]

[0186] Here, κ is the conductivity of the electrolyte, z i is the ion valence, D i is the diffusion coefficient, is the reference diffusion coefficient, is the diffusion activation energy.

[0187] By fitting the accelerated test data, the researchers determined the values ​​of the parameters in the above physicochemical model, as shown in Table 1:

[0188] Table 1 Parameter value table

[0189]

[0190] Based on the above physical and chemical model, the researchers calculated the various degradation characteristics of the battery at different cycle times, as shown in Table 2.

[0191] Table 2: Degradation characteristics table

[0192]

[0193] It can be seen from Table 2 that with the increase in the number of cycles, the lithium ion concentration distribution, SEI film thickness and active material content inside the battery have undergone significant changes, which ultimately lead to the attenuation of battery capacity and the increase of internal resistance.

[0194] After completing the establishment of the physical and chemical model and parameter calibration, the R&D personnel began to build a deep learning model for battery loss assessment. First, the R&D personnel paired the degradation characteristics calculated by the above physical and chemical model with the measured battery performance indicators (capacity, internal resistance, etc.) to form a training sample set. Then, the R&D personnel designed a deep learning network structure consisting of a 1D convolutional layer, 2 LSTM layers and a fully connected layer. The input is the battery comprehensive feature matrix, and the output is five loss factors, including electrode structure loss rate, electrolyte decomposition loss rate, SEI film growth rate, active material loss rate and capacity decay rate.

[0195] During the training process, the researchers used the Adam optimizer, set the learning rate to 0.001, and the batch size to 32. At the same time, the researchers also introduced the mean square error loss function and the L2 regularization term, and adopted an early stopping strategy. After 50 training cycles, the validation set accuracy of the deep learning model reached more than 95%, meeting the requirements of practical applications. Finally, the researchers used the cross-validation method to further verify the generalization performance of the model.

[0196] Based on the above trained deep learning model, R&D personnel can collect the voltage, current, temperature and internal resistance data of the battery in real time when the energy storage system is put into use, and input them into the model to output various battery loss factors. Then, R&D personnel substitute these loss factors into the pre-set battery life empirical equation:

[0197]

[0198] Among them, SOL represents the remaining battery life percentage, d1~d5 correspond to the electrode material structure loss rate, electrolyte decomposition loss rate, SEI film growth rate, active material loss rate and capacity decay rate, λ1~λ5 are the weight coefficients of each loss factor, and β1~β5 are nonlinear coefficients. Through calculation, the remaining service life of the battery can be obtained.

[0199] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for real-time evaluation of battery life of an energy storage system, characterized in that: The following steps are involved: S10, real-time collection of voltage data, current data, temperature data at multiple collection points, and internal resistance data at multiple collection points during the operation of the energy storage system battery; S20, performing multivariate empirical mode decomposition on the voltage data and the current data to obtain a voltage steady-state data matrix, a voltage variation data matrix, a current steady-state data matrix, and a current variation data matrix; S30, establishing a battery physical and chemical reaction equation group taking into account voltage, current, temperature and internal resistance, including: electrochemical reaction balance equation, ion diffusion mass transfer equation, electrode material stress strain equation, solid-liquid interface reaction kinetics equation, electrolyte conductivity equation; S40, substituting the voltage steady-state data matrix, the voltage variation data matrix, the current steady-state data matrix, the current variation data matrix, the temperature data of the plurality of acquisition points, and the internal resistance data of the plurality of acquisition points into the battery physical and chemical reaction equation group to calculate and obtain the battery material degradation characteristic matrix; S50, establishing a battery internal environment characteristic matrix based on the temperature data of the multiple collection points and the internal resistance data of the multiple collection points; S60, fusing the battery material degradation characteristic matrix with the battery internal environment characteristic matrix to obtain a battery comprehensive characteristic matrix; S70, inputting the battery comprehensive characteristic matrix into a pre-trained battery loss assessment model, and calculating the battery loss vector, including the electrode material structure loss rate, the electrolyte decomposition loss rate, the solid electrolyte interface film growth rate, the active material loss rate, and the capacity decay rate; S80: Calculate the battery remaining life evaluation value using a preset battery life empirical equation according to the battery loss vector.

2. A method for real-time evaluation of battery life of an energy storage system according to claim 1, characterized in that: The steps of establishing the training data set of the battery loss assessment model include: first, performing cycle life tests on multiple groups of different types of batteries, and the test process continues until the capacity decays to 80% of the rated capacity; for each group of test processes, collecting battery status data every 100 cycles and calculating physical and chemical parameters; dividing the calculated battery material decay feature matrix and battery internal environment feature matrix according to the ratio of 80% for training set and 20% for test set; normalizing and standardizing the training data to eliminate the dimension effect; finally, pairing the processed feature data with the corresponding actual capacity decay data to construct a training sample.

3. The method for real-time evaluation of battery life of an energy storage system according to claim 2, characterized in that: The formula for performing multivariate empirical mode decomposition on the voltage data and current data is expressed as follows: Where: V(t) is the time domain voltage signal; I(t) is the time domain current signal; c vi (t) is the voltage i-th eigenmode function; c ii (t) is the i-th eigenmode function of current; r vn (t) and r in (t) are the voltage and current residual terms respectively; n v and n i is the decomposition order; The obtained voltage steady-state data matrix is ​​specifically expressed as: In the formula, c v1 represents the first intrinsic mode function after the voltage signal is decomposed, representing the main steady-state component of the voltage; r vn represents the residual term after the voltage signal is decomposed, representing the long-term trend component; t1, t2, ..., t n ,t m Represents the time series sampling point; the voltage change data matrix is ​​specifically expressed as: In the formula, c v2 ,c v3 ,...,c vn Represents the 2nd to nth intrinsic mode functions of the voltage signal, representing the fluctuation components of different frequencies; each column corresponds to the value of an intrinsic mode function at different time points; each row of the matrix represents different frequency components at the same time point; The current steady-state data matrix is ​​specifically expressed as: In the formula, c i1 represents the first intrinsic mode function after the current signal is decomposed, representing the main steady-state component of the current; r in Represents the residual term after the current signal is decomposed; the current change data matrix is ​​specifically expressed as: In the formula, c i2 ,c i3 ,...,c in It represents the 2nd to nth intrinsic mode functions of the current signal, representing the fluctuation components of different frequencies; each column corresponds to the value of an intrinsic mode function at different time points; each row of the matrix represents different frequency components at the same time point.

4. The method for real-time evaluation of battery life of an energy storage system according to claim 3, characterized in that: The electrochemical reaction equilibrium equation is specifically expressed as: Where: C s is the solid phase lithium ion concentration; C e is the liquid phase lithium ion concentration; D s is the solid phase diffusion coefficient; D e is the liquid phase diffusion coefficient; r is the radial coordinate of the spherical particle; x is the coordinate in the electrode thickness direction; t + is the lithium ion migration number; F is the Faraday constant; j Li is the lithium ion insertion and extraction reaction rate.

5. A method for real-time evaluation of battery life of an energy storage system according to claim 4, characterized in that: The ion diffusion mass transfer equation is specifically expressed as: In the formula, Where: i0 is the exchange current density; k c ,k a is the reaction rate constant; is the maximum lithium ion concentration; is the surface lithium ion concentration; η is the overpotential; α a is the anode transfer coefficient; α c is the cathode transfer coefficient; R is the gas constant, which is 8.314 J / (mol·K); T is the temperature, in K.

6. A method for real-time evaluation of battery life of an energy storage system according to claim 5, characterized in that: The stress-strain equation of the electrode material is specifically expressed as: In the formula, Where: σ is the stress tensor; E is Young's modulus; ε is the strain tensor; Ω is the partial molar volume; is the initial lithium ion concentration; ε xx ,ε yy ,ε zz is the strain component; γ xy ,γ yz ,γ zx is the shear strain component.

7. A method for real-time evaluation of battery life of an energy storage system according to claim 6, characterized in that: The solid-liquid interface reaction kinetic equation is specifically expressed as: Where: SEI is the thickness of the solid electrolyte interface film; k sei is the growth rate constant; E a is the activation energy; α is the reaction order; R sei is the interface impedance; L sei is the film thickness; κ sei is the membrane conductivity.

8. A method for real-time evaluation of battery life of an energy storage system according to claim 7, characterized in that: The electrolyte conductivity equation is specifically expressed as: In the formula, Where: is the reference diffusion coefficient; is the diffusion activation energy; κ is the conductivity of the electrolyte; z i is the ion charge number; D i is the ion diffusion coefficient; C i is the ion concentration.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed in a computer, the method for real-time evaluation of battery life of an energy storage system according to any one of claims 1 to 8 is used to execute.

10. A real-time evaluation system for battery life of an energy storage system, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images