Novel chaos firefly-particle optimization battery health state evaluation method

Abstract

The invention relates to the technical field of new energy measurement and control, in particular to a novel chaos firefly-particle optimization battery health state evaluation method, which comprises the following steps of: S1, acquiring updated particles by adopting a particle filtering algorithm; s2, optimizing the updated particles by using a chaos firefly optimization method to obtain a global optimal solution; s3, performing weight updating, normalization and resampling on particles in the globally optimal solution to obtain an SOC estimated value of the current time step; and S4, inputting the SOC estimation value of the current time step into a Kalman filter, obtaining an updated battery estimation capacity value and an SOH estimation value, inputting the updated battery estimation capacity value into a control matrix of a particle filtering algorithm, obtaining an SOC estimation value of a next time step, and returning to S1 for iteration until an optimal SOH estimation value is obtained. The method provides powerful support for stable operation and battery maintenance of a large-scale energy storage system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy measurement and control technology, in particular to a new chaotic firefly-particle optimization battery state of health evaluation method. BACKGROUND

[0002] With the rapid development of renewable energy, large-scale energy storage technology has become one of the key technologies to realize efficient use of energy and stable operation of power grid. However, the health state of the battery in the energy storage system directly affects its performance and life, therefore, how to accurately evaluate the battery health state has become a hot research topic. The traditional battery health state evaluation method often has the problems of low precision, large calculation amount, etc., and it is difficult to meet the actual needs of large-scale energy storage system. Specifically, SOH (State of Health) can be quantitatively represented by the ratio of the remaining capacity to the initial capacity. When the SOH value is too low, it will significantly increase the power consumption, reduce the performance of the battery and increase the temperature, thereby increasing the risk of safety accidents. Therefore, accurate estimation of SOH is of great significance to ensure the stable operation of the entire power system and the battery.

[0003] The research methods for SOH estimation are mainly divided into data-driven method and model-based method. The data-driven method estimates the battery health state model according to the recorded data under the working state of the battery, and then uses the model to estimate the SOH of the battery. The model can be established and optimized using machine learning, artificial neural network, etc. to improve the estimation accuracy and stability. Neural network method is a commonly used data-driven method, its basic idea is to use historical data to train network parameters, during the training of the network, a large amount of battery working state data and corresponding SOH data are needed to improve the accuracy and accuracy of the network. The model-based method establishes a battery-related model, and then relies on filtering algorithms, intelligent optimization algorithms, etc. to complete the calculation of the battery SOH. Kalman filtering algorithm and its improved algorithm, particle filtering algorithm and its improved algorithm are commonly used methods for estimating SOH. This kind of method has targetedly improved the inherent defects of traditional filtering algorithm, and can improve the SOH estimation accuracy to a certain extent. Moving average filtering is a commonly used signal processing technology, which can smooth the signal curve, remove noise and outliers. Some researchers have improved the traditional moving average filtering algorithm and proposed a new filtering algorithm, which can more effectively remove noise and outliers, thereby improving the accuracy of SOH estimation.

[0004] In summary, the SOH estimation research of large-scale energy storage battery has made certain progress, but the current commonly used model-based method and data-driven method for evaluating battery health state have the problems of complex data calculation process, slow speed and poor precision. SUMMARY

[0005] The purpose of this invention is to provide a novel chaotic firefly-particle optimized method for battery state of health assessment. This method transforms linear variables into chaotic variables and leverages the ergodicity and randomness of chaos to optimize the search process, thereby achieving high-precision estimation of the state of health (SOH) of lithium batteries. This results in high accuracy and stability, providing strong support for the stable operation of large-scale energy storage systems and battery maintenance.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A novel chaotic firefly-particle optimized method for battery health assessment includes:

[0008] S1. Initialize the migration model and parameters, obtain input parameters, input the input parameters into the particle filtering algorithm, obtain several initial SOH particles according to the initial probability distribution of battery SOH, predict and update the initial SOH particles, and obtain the updated particles, wherein the input parameters are current, battery capacity value at the current time step, and battery voltage;

[0009] S2. The chaotic firefly optimization method is used to optimize several of the updated particles and update their weights to obtain the global optimal solution. The chaotic firefly optimization method is to introduce chaotic mapping into the firefly optimization algorithm.

[0010] S3. Normalize the particles in the global optimal solution, and determine whether the resampling condition is met based on the normalized particles. If the resampling condition is met, obtain the resampled particle set, and obtain the SOC estimate of the current time step based on the resampled particle set. If the resampling condition is not met, obtain the SOC estimate of the current time step based on the normalized particles.

[0011] S4. Input the SOC estimate of the current time step into the Kalman filter to obtain the updated battery capacity estimate and SOH estimate. Input the updated battery capacity estimate into the control matrix in the particle filter algorithm to obtain the SOC estimate of the next time step. Return to S1 for iteration until the preset number of iterations is reached or convergence is achieved, and obtain the optimal SOH estimate.

[0012] Optionally, in step S2, the chaotic firefly optimization method is used to optimize several of the updated particles and update their weights to obtain the global optimal solution, including:

[0013] S21. Based on the updated particles, construct an initial firefly population, wherein the individual firefly is represented by its SOH value, and the brightness of the individual firefly is represented by the weight of the updated particles;

[0014] S22. Use a chaotic mapping algorithm to optimize and update the positions of fireflies, and calculate the fluorescence intensity and attractiveness of fireflies in the current group;

[0015] S23. Based on the fluorescence brightness and attractiveness, return to S22 until the number of iterations is satisfied and the global optimal solution is obtained.

[0016] Optionally, updating the firefly's position in step S22 by combining a chaotic mapping algorithm for optimization includes:

[0017] The individuals in the current firefly population are sorted according to their fluorescence intensity, and the best individuals within a preset range are used as the chaotic search space.

[0018] Select a firefly from the preset range and use the SOH value of the selected firefly as the basic solution;

[0019] The basic solution is mapped to a preset interval according to the Tent chaotic mapping, and a chaotic variable sequence is iteratively generated.

[0020] The chaotic variable sequence is inversely solved to the original solution space to obtain a new solution sequence in the original solution space;

[0021] The fluorescence intensity of each firefly in the new solution sequence is calculated to obtain the current optimal solution;

[0022] The weights of the current optimal solution and the basic solution are compared. If the weight of the current optimal solution is greater than the weight of the basic solution, the current optimal solution is updated to the global optimal solution. If the weight of the current optimal solution is not greater than the weight of the basic solution, the current optimal solution is not updated.

[0023] Optionally, iteratively generating the chaotic variable sequence includes: obtaining initial values ​​for the chaotic variable sequence; and obtaining the chaotic variable sequence Tent(m), where m = 1, 2, ..., ITERmax, based on the initial values ​​and the Tent mapping model. The Tent mapping model is ITERmax is the maximum number of iterations in the chaotic search, X A As the fundamental solution, X min and X max x represents the minimum and maximum values ​​among individuals within a preset range. n Let x represent the output value of the Tent chaotic map at the nth iteration. n+1 This represents the output value of the Tent chaotic map at the (n+1)th iteration.

[0024] Optionally, the control matrix of the particle filter algorithm is the control matrix of the state transition equation in the particle filter algorithm.

[0025] Optionally, after obtaining the SOC estimate at the current time step, the process includes: transferring the parameters in the state transition equation and observation equation of the particle filter algorithm.

[0026] Optionally, the parameters in the state transition equation and observation equation of the particle filter algorithm are transferred to:

[0027] ;

[0028] in, The SOC value of the battery model at time step k+1 after transfer is... Let k be the SOC value at time step k. The step size between adjacent time steps. The actual capacity value at time step k. Let k be the current at time step k. For current, For process noise in SOC estimation, This refers to process noise in polarization voltage estimation. For time step k+1, R P C P The terminal voltage of the circuit. For time step R, k P C P The terminal voltage of the circuit. C represents the polarization resistance. P Indicates polarization capacitor, It is a time constant. The open-circuit voltage of the battery model at time step k+1 after migration is given. The internal ohmic resistance of the battery after migration of the battery model. This represents the observation noise at time step k. The voltage at time step k+1 is the closed-circuit voltage formed after the battery is connected to the external circuit in the transferred battery model. The superscript M_battery indicates the parameters of the model after the transfer, and the time constant τ after the transfer is... M =R P M C P M R P M C represents the polarization resistance after migration. P M This indicates the polarization capacitance after migration.

[0029] The beneficial effects of this invention are as follows: The novel chaotic firefly-particle optimized battery health status assessment method proposed in this invention has strong applicability. For the goal of accurate SOH estimation of large-scale energy storage batteries, this invention establishes a dual-filter mutual correction mechanism. By constructing a dual-layer filter structure, the SOC estimate obtained from the first layer through improved optimization is used as the input to the second-layer SOH filter. Through iterative updates, the SOH assessment is continuously corrected. A population-intelligent optimized firefly algorithm is introduced to improve the optimization accuracy and speed of the traditional PF algorithm and improve the resampling process. Furthermore, the Tent chaotic mapping algorithm is introduced to seek better optimal solutions, helping the firefly algorithm escape local optima and achieving accurate SOH estimation. This invention's method is applicable to large-scale energy storage systems, enabling rapid assessment of the health status of a large number of batteries, providing strong support for battery maintenance and stable system operation. It can provide a methodological reference for battery SOH estimation in different application scenarios, and has the advantages of simple calculation, strong adaptability, and high accuracy. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart of a novel chaotic firefly-particle optimized battery health status assessment method according to an embodiment of the present invention;

[0032] Figure 2 This is a schematic diagram of the mutual correction mechanism of the dual filters SOC and SOH of a large-scale energy storage battery according to an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] A novel chaotic firefly-particle optimized method for battery health assessment includes:

[0036] S1. Initialize the transfer model and parameters, obtain the input parameters, input the input parameters into the particle filter algorithm, obtain several initial SOH particles according to the initial probability distribution of the battery SOH, predict and update the initial SOH particles, and obtain the updated particles. The input parameters are current, battery capacity value at the current time step, and battery voltage.

[0037] S2. Use the chaotic firefly optimization method to optimize several updated particles and update their weights to obtain the global optimal solution. The chaotic firefly optimization method is to introduce chaotic mapping into the firefly optimization algorithm.

[0038] S3. Normalize the particles in the global optimal solution. Determine whether the resampling condition is met based on the normalized particles. If the resampling condition is met, obtain the set of resampled particles. Obtain the SOC estimate for the current time step based on the set of resampled particles. If the resampling condition is not met, obtain the SOC estimate for the current time step based on the normalized particles.

[0039] S4. Input the SOC estimate of the current time step into the Kalman filter to obtain the battery capacity estimate and SOH estimate updated at the current time step. Input the battery capacity estimate updated at the current time step into the control matrix of the particle filter algorithm to obtain the SOC estimate of the next time step. Return to S1 to iterate until the preset number of iterations is reached or convergence is achieved, and obtain the optimal SOH estimate.

[0040] Specifically, this embodiment introduces a novel chaotic firefly-particle optimization method for assessing the health status of large-scale energy storage batteries. By constructing a two-layer filter structure, the SOC estimate obtained by the first layer through improved optimization is used as the input to the second layer SOH filter, i.e., the Kalman filter. Through iterative updates, the SOH assessment is continuously corrected. This embodiment uses the firefly algorithm to estimate the battery's SOH, and through a cyclic optimization process, the battery's SOH is assessed.

[0041] Furthermore, in S2, the chaotic firefly optimization method is used to optimize several updated particles and update their weights to obtain the global optimal solution, including:

[0042] S21. Based on the updated particles, construct an initial firefly population, where the individual firefly is represented by its SOH value and the brightness of the individual firefly is represented by the weight of the updated particles.

[0043] S22. Use a chaotic mapping algorithm to optimize and update the positions of fireflies, and calculate the fluorescence intensity and attractiveness of fireflies in the current group;

[0044] S23. Based on the relative fluorescence intensity and attractiveness, return to S22 until the number of iterations is satisfied and the global optimal solution is obtained.

[0045] Specifically, to address the shortcomings of traditional particle filtering algorithms, such as particle degeneration and reduced particle diversity, a swarm-based intelligent optimization firefly algorithm is introduced. To effectively improve the firefly algorithm's dependence on the initial solution, slow later convergence speed, and susceptibility to local optimization, and to further enhance the accuracy of SOH estimation, a chaotic mapping algorithm is introduced, forming a chaotic firefly-particle optimization algorithm.

[0046] Furthermore, in S22, the optimization of the firefly's position using a chaotic mapping algorithm includes:

[0047] The individuals in the current firefly population are sorted according to their fluorescence intensity, and the best individuals within a preset range are used as the chaotic search space.

[0048] Select a firefly from the preset range and use the SOH value of the selected firefly as the basic solution;

[0049] The basic solutions are mapped to a preset interval based on the Tent chaotic mapping, and a sequence of chaotic variables is iteratively generated.

[0050] The chaotic variable sequence is inversely solved to the original solution space to obtain a new solution sequence in the original solution space;

[0051] The fluorescence intensity of each firefly in the new solution sequence is calculated to obtain the current optimal solution;

[0052] The weights of the current optimal solution and the basic solution are compared. If the weight of the current optimal solution is greater than the weight of the basic solution, the current optimal solution is updated to the global optimal solution. If the weight of the current optimal solution is not greater than the weight of the basic solution, the current optimal solution is not updated.

[0053] Furthermore, the iterative generation of the chaotic variable sequence includes: obtaining the initial values ​​of the chaotic variable sequence; and obtaining the chaotic variable sequence Tent(m), where m = 1, 2, ..., ITERmax, based on the initial values ​​and the Tent mapping model. The Tent mapping model is ITERmax is the maximum number of iterations in the chaotic search, X A As the fundamental solution, X min and X max x represents the minimum and maximum values ​​among individuals within a preset range. n Let x represent the output value of the Tent chaotic map at the nth iteration. n+1 This represents the output value of the Tent chaotic map at the (n+1)th iteration.

[0054] Furthermore, the control matrix of the particle filter algorithm is the control matrix of the state transition equation in the particle filter algorithm.

[0055] Furthermore, after obtaining the SOC estimate for the current time step, the following steps are taken: the parameters in the state transition equation and observation equation of the particle filter algorithm are transferred.

[0056] Furthermore, the parameters in the state transition equation and observation equation of the particle filter algorithm are transformed by migration as follows:

[0057] ;

[0058] in, The SOC value of the battery model at time step k+1 after transfer is... Let k be the SOC value at time step k. The step size between adjacent time steps. The actual capacity value at time step k. Let k be the current at time step k. For current, For process noise in SOC estimation, This refers to process noise in polarization voltage estimation. For time step k+1, R P C P The terminal voltage of the circuit. For time step R, k P C P The terminal voltage of the circuit. C represents the polarization resistance. P Indicates polarization capacitor, It is a time constant. The open-circuit voltage of the battery model at time step k+1 after migration is given. The internal ohmic resistance of the battery after migration of the battery model. This represents the observation noise at time step k. The voltage at time step k+1 is the closed-circuit voltage formed after the battery is connected to the external circuit in the transferred battery model. The superscript M_battery indicates the parameters of the model after the transfer, and the time constant τ after the transfer is... M =R P M C P M R P M C represents the polarization resistance after migration. P M This indicates the polarization capacitance after migration.

[0059] The method of this embodiment will be further described below with reference to the accompanying drawings:

[0060] This embodiment provides a novel chaotic firefly-particle optimization method for battery health state assessment. Through a dual-filter mutual correction mechanism, it achieves effective characterization of battery state of health (SOH) estimation. The first-layer particle filter estimates the state of charge (SOC), and this estimated value is used as input to the second-layer filter to achieve progressive SOH estimation. The SOH estimate obtained from the second-layer Kalman filter is then used as input to the next iteration, further correcting the SOC estimate for the next time step, forming a closed loop until the iteration ends. The firefly algorithm is then applied to the SOH particle optimization process, enabling the particles to approach the optimal value, thus achieving the goal of the state particles further approximating the true value. Based on this, a chaotic mapping algorithm is added, linearly mapping variables to chaotic variables. Then, the search process is optimized based on the ergodicity and randomness of chaos, thereby achieving high-precision estimation of battery SOH. Specifically, the method includes the following steps:

[0061] The overall SOC / SOH co-estimation framework is divided into two main parts: offline initial model construction and online model transfer. Offline initial model construction involves determining the initial model through theoretical analysis and standard tests such as HPPC (Hybrid Pulse Power Characterization) operating condition data, without actual battery operating data. The main task at this stage is to construct the Thevenin equivalent circuit model and perform offline parameter identification. Through this process, the functional relationships between various battery parameters and SOC (State of Charge), as well as the SOC-OCV (Open Circuit Voltage) curve, can be obtained. This offline initial model provides the initial variables and parameters required for the system state in the online model transfer, including the functional relationships between battery parameters and SOC. The next step is to initialize the transfer model and parameters, obtaining the input parameters. These input parameters include the current flowing through the battery directly measured by the current sensor in the Battery Management System (BMS), the remaining capacity value, and the voltage across the battery terminals directly measured by the voltage sensor. In the second part, the online migration model is built upon the offline initial model. The online migration model considers the state changes of the battery during actual operation, including factors such as aging. The form of the state equation of the migration model is derived from general linear dynamic system theory and is used to describe the changes in battery state over time.

[0062] A population-based intelligent optimization firefly algorithm is introduced to improve the optimization accuracy and speed of the traditional particle filter algorithm (PF) and to improve the resampling process. Furthermore, a Tent chaotic mapping algorithm is introduced to seek better optimal solutions, helping the firefly algorithm escape local optima. The proposed Chaotic Firefly-Particle Filtering Algorithm (CF-PF) implements iterative updates of the migration factor in the battery migration model. A particle filter-Kalman filter dual-filter is constructed to achieve collaborative estimation and mutual correction of SOC and SOH in lithium-ion batteries. The overall algorithm flow is as follows: Figure 1 As shown, the lithium battery SOH estimation process based on the PF algorithm first generates N initial SOH particles using a prior probability P(x0). According to the system update equation, the prior probability sample for the next time step is obtained, and the particle weights are updated simultaneously. After the system obtains new observations, a new particle set is generated through the state equation, and the predicted value of the observation is obtained through the observation equation. Then, the error between the observed value and the predicted value of each particle is calculated, and the particle weight is obtained through the error. The weights of the N SOH particles after the state update are normalized, the effective particle number is calculated, and particle diversity is judged. If the effective particle number is lower than a certain threshold, it indicates that the particle distribution is too concentrated, and particle diversity is reduced. In this case, resampling is required to restore particle diversity. Specifically, resampling is performed when the effective particle number is lower than a preset threshold. This threshold is usually set based on experience to ensure that the algorithm can effectively explore the state space. In this embodiment, the total number of particles is set to N, and the effective particle number threshold may be set to 10% of N. If the effective particle number is lower than this threshold, the particle distribution is considered too concentrated, and resampling is required. Next, resampling is performed using the new random sample distribution generated in the previous steps. The number of effective particles is calculated, and it is determined whether the particles meet the resampling conditions. Finally, the estimated value is output based on the particle set and particle weights obtained after resampling. To solve the problem of reduced particle diversity caused by the resampling process in the PF algorithm, the firefly optimization algorithm is introduced. The initial firefly population is constructed based on N SOH particles with updated weights. The relative fluorescence brightness of firefly i and firefly j is shown in equation (1).

[0063] (1);

[0064] In equation (1), It represents the relative fluorescence intensity of firefly i and firefly j. γ represents the maximum fluorescence brightness of firefly i. The better the objective function value, the higher the brightness of the firefly itself. This is applied in the PF algorithm to implement the second-layer filtering estimation of the SOH of a lithium-ion battery; that is, the higher the particle weight, the higher the particle brightness. γ is the light intensity absorption coefficient, set to 0.98 in this embodiment. is the spatial distance between fireflies i and j. When this parameter is applied in the PF algorithm, it is considered to be the difference between the estimated values ​​of SOC particles i and j. In the firefly algorithm, the relative attraction of firefly i to j is shown in equation (2).

[0065] (2);

[0066] In equation (2), It represents the relative attraction of firefly i to firefly j. It is the attraction at the light source (r=0), that is, the maximum attraction of the firefly at the light source. The position updates of the attracted firefly j and the globally optimal firefly i are shown in Equation (3).

[0067] (3);

[0068] In equation (3), and The spatial positions of fireflies i and j are represented. When applied to the PF algorithm to estimate the SOH of the battery, they represent the SOH values ​​of particles i and j in the PF algorithm. The step size factor α is a constant on [0,1], and its value is set to 0.05 in this embodiment. rand is a random factor on [0,1] and follows a uniform distribution. Since other fireflies cannot attract the brightest individual firefly, it is assumed that the individual moves randomly within a small area. The position update of the brightest firefly is performed according to the expression in the second part of equation (3). After the firefly algorithm completes particle optimization, PF starts particle resampling. In this process, the firefly algorithm has the inherent defects of slow optimization speed and easy to get trapped in local optima. The introduction of the chaotic mapping algorithm, which updates the firefly position with the optimal solution generated in the chaotic sequence, can make the algorithm jump out of local optima. The excellent two-dimensional chaotic characteristics and uniform distribution function exhibited by the Tent mapping are used to achieve fast optimization speed and high optimization efficiency. This embodiment innovatively uses the Tent mapping to generate chaotic sequences in order to optimize the performance of the Firefly algorithm in the process of searching for the optimal solution. The mathematical expression of the Tent mapping is shown in equation (4).

[0069] (4);

[0070] in, This represents the output value of the Tent chaos map at the nth iteration. This represents the output value of the Tent chaotic map at the (n+1)th iteration.

[0071] In one iteration, in firefly population X iIn the process, all fireflies are sorted from largest to smallest based on their individual fluorescence intensity (particle weight), and the top 5% of the better individuals are selected. The minimum value X among these 5% of individuals is then obtained. min and maximum value X max As a chaotic search space, a firefly is randomly selected from the top 5% of individuals, and its spatial location (SOH value) is X. A As a fundamental solution, equation (5) is used to convert X... A Map to the interval (0,1). Let the initial value of the chaotic sequence Tent (m) be Tent (1) as shown in equation (5).

[0072] (5);

[0073] Substituting equation (5) into the Tent chaotic mapping in equation (4), a chaotic variable sequence Ten(m) (m=1, 2, ..., ITERmax) is generated iteratively. ITERmax is the maximum number of iterations for the chaotic search, which is set to ITERmax = 200 in this embodiment. The generated sequence Ten(m) is then deduced back into the original solution space, and a new solution sequence x(m) in the original solution space is generated using equation (6).

[0074] (6);

[0075] The fluorescence intensity (particle weight) of each firefly in the new solution sequence x(m) is calculated sequentially to generate a new optimal solution X. B and with X A The particle weights are compared, and the optimal solution is retained to allow the algorithm to escape the defect of local optima. Specific steps include:

[0076] Step 1. Generation of chaotic sequences:

[0077] First, a sequence of chaotic variables is generated using the Tent chaotic map. This sequence is used to explore the solution space and help the algorithm escape local optima.

[0078] Step 2. Solve the space mapping:

[0079] The generated chaotic sequence is then inverted back to the original solution space to generate a new solution sequence x(m). This step is achieved through mathematical transformation, with the aim of finding possible solutions in the original problem space.

[0080] Step 3. Calculate fluorescence intensity (particle weight):

[0081] For each firefly (solution) in the new solution sequence x(m), its fluorescence intensity is calculated. In this embodiment, the fluorescence intensity corresponds to the particle weight, where fluorescence intensity (or particle weight) is an indicator for evaluating the quality of the solution.

[0082] Step 4. Compare and select the optimal solution:

[0083] Compare the solutions in the new solution sequence x(m) with the current optimal solution X. A The particle weights are compared. X A It is the basic solution among the top 5% of the better individuals, representing a relatively good position in the current search space.

[0084] Step 5. Update the optimal solution:

[0085] If the fluorescence intensity (particle weight) of the new solution is better than X A The fluorescence brightness is then updated for X. A This is a new solution, meaning the algorithm has found a better solution and thus escaped the local optimum. If the fluorescence intensity of the new solution is not better than X... A Then keep X A The search process continues unchanged. Based on the fluorescence intensity and attractiveness of the fireflies in the current swarm, the positions of the fireflies are updated using the attraction rules and random factors defined in the firefly algorithm. This typically involves calculating the positional change of each firefly relative to other fireflies, as well as the attractiveness between them. After updating the positions, return to step 1 to continue the next round of iterations. This includes generating a new sequence of solutions, calculating the fluorescence intensity (particle weights), and comparing it with the current best solution.

[0086] Step 6. Iterative Optimization: Repeat steps 1-6 until the maximum number of iterations is reached or the solution falls below a preset convergence threshold in multiple consecutive iterations. This ensures the algorithm fully explores the solution space and finds the global optimum as much as possible. In this way, the algorithm combines the randomness and ergodicity of chaotic mappings to continuously update and optimize solutions, effectively avoiding getting trapped in local optima and improving the algorithm's global search capability and solution quality. This process is the core mechanism by which the Chaotic Firefly Algorithm escapes local optima.

[0087] Incorporating the migration factor update into the PF algorithm to estimate the state of charge (SOC) involves considering the internal ohmic resistance R0 and polarization resistance R within the battery in the system state transition equation and observation equation. P Polarization capacitor C P Closed-circuit voltage U L and open circuit voltage U OC The initial SOC estimate obtained from PF is considered inaccurate and is therefore also transferred to the value after migration, as shown in formula (7).

[0088] (7);

[0089] in, The SOC value of the battery model at time step k+1 after transfer is... Let k be the SOC value at time step k. The step size between adjacent time steps. The actual capacity value at time step k. Let k be the current at time step k. For current, For process noise in SOC estimation, This refers to process noise in polarization voltage estimation. For time step k+1, R P C P The terminal voltage of the circuit. For time step R, k P C P The terminal voltage of the circuit. C represents the polarization resistance. P Indicates polarization capacitor, It is a time constant. The open-circuit voltage of the battery model at time step k+1 after migration is given. The internal ohmic resistance of the battery after migration of the battery model. This represents the observation noise at time step k. The voltage at time step k+1 is the closed-circuit voltage formed after the battery is connected to the external circuit in the transferred battery model. The superscript M_battery indicates the parameters of the model after the transfer, and the time constant τ after the transfer is... M =R P M C P M R P M C represents the polarization resistance after migration. P M This indicates the polarization capacitance after migration.

[0090] The parameters after migration and the SOC value are used together as update variables for the PF algorithm. The weights of the particles are determined based on the quality of the SOC particle estimates, and then the final estimate for the next time step is obtained based on the weights.

[0091] In the SOC estimation process based on the migration model, the original system state variables are expanded into a migration matrix X=[x1, x2, x3, ... x10], where x1, x2, x3, ... x10 represent different internal parameters or state variables in the battery model. Each battery internal parameter is adaptively adjusted at each sampling point to adapt to different battery aging conditions, resulting in a more accurate SOC value that better reflects the actual situation.

[0092] After the migration matrix is ​​transferred, a conditional judgment is made. Specifically, this judgment condition involves the calculation of the effective particle count. The following is a detailed description of this judgment condition:

[0093] 1. Calculation of effective particle count: In the particle filtering (PF) algorithm, the effective particle count is calculated after each update of particle weights.

[0094] 2. Resampling judgment condition: The condition for determining whether resampling is needed is: if the number of effective particles is lower than a certain preset threshold, it is considered that the particle distribution is too concentrated and resampling is needed to restore particle diversity.

[0095] 3. Resampling operation: If the resampling conditions are met, then the resampling operation is performed.

[0096] 4. Update the transition matrix: After resampling, update the transition matrix and use the new set of particles obtained from the resampling for subsequent state estimation and transition model updates.

[0097] In this embodiment, after obtaining the SOC estimate for the current time step using the chaotic firefly algorithm, this value is used as the input to the SOH estimation filter, forming the second layer of filtering. The purpose of this is to correct the previously estimated capacity value Q derived from the state transition equation. Once the capacity value Q for the current time step is obtained, it is fed back into the chaotic firefly algorithm. By updating the control matrix B in the state transition equation, the SOC estimate for the next time step is calculated, thereby updating the input to the SOH estimation filter at the next time step. This allows Q to be updated and corrected in reverse, achieving accurate estimation of the battery's SOH. The control matrix B is a matrix in the state-space model used to describe the system's state transitions. Specifically, the control matrix B defines how the battery state (such as SOC) changes over time and how it is affected by control inputs (such as battery charging and discharging current).

[0098] like Figure 2 The left-hand flow shown primarily estimates the battery's State of Charge (SOC), predicting the battery's state through the state transition equation and control matrix B. The right-hand flow focuses on estimating and correcting the State of Charge (SOH), adjusting the SOH based on changes in SOC, which helps to more accurately reflect the battery's actual health condition. Reverse correction is an iterative process, adjusting the SOC estimate through SOH updates, forming a closed-loop control system, thereby improving the accuracy and robustness of the entire system. Equation (8) is the state transition equation, describing the process of the system transitioning from one state to another. In battery health assessment, this involves how the battery's State of Charge (SOC) and other internal parameters change over time.

[0099] Equation (9) is the observation equation, which describes how to obtain observation data from the system state. In battery health status assessment, this involves how to measure voltage, current, etc. from the internal state of the battery.

[0100] (8);

[0101] (9);

[0102] In the above formula, k represents time step k, and k+1 represents time step k+1. This represents the actual capacity value of the battery at time step k+1. This represents the actual capacity value of the battery at time step k. U is the SOC value at time step k+1. L,k+1 The closed-circuit voltage formed after the battery is connected to the external circuit at time step k+1 is given. This represents the observation noise at time step k. R represents the relative distance between firefly particles at time step k. P Q represents the polarization resistance. k_update This indicates the estimated capacity value updated at time step k in the EKF, instead of a constant value. It is the product of the control matrix B in the battery state transition equation at the k-th time step and the battery's state of charge.

[0103] In summary, this invention addresses the estimation of State of Health (SOH) of large-scale energy storage batteries. Considering estimation accuracy, computational complexity, and algorithm stability, it proposes a novel chaotic firefly-particle optimization method for assessing the SOH of large-scale energy storage batteries. Based on a thorough consideration of the battery's complex operating environment, a dual-filter mutual correction mechanism is constructed. The estimated SOC value of the first-layer filter is obtained through chaotic firefly-particle optimization and used as the input to the second-layer SOH estimation filter to correct the estimated capacity value Q. The SOC estimate for the next time step is calculated by updating the coefficient matrix B, which in turn updates the input to the SOH estimation filter at the next time step, allowing Q to be updated and corrected in reverse. This achieves accurate SOH estimation for large-scale energy storage batteries, providing a foundation for SOH assessment and real-time monitoring of their operating status, and has broad application prospects in large-scale energy storage systems.

[0104] This invention addresses the issue of reduced particle diversity caused by the resampling process in the PF algorithm by introducing a population-based intelligent optimization firefly algorithm. Furthermore, by incorporating a chaotic mapping algorithm, it improves the firefly algorithm's tendency to get trapped in local optima, thereby enhancing the estimation accuracy of battery SOH.

[0105] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A novel chaotic glowworm-particle swarm optimization battery state of health estimation method, characterized in that, The application relates to a method for estimating state of charge (SOC) and state of health (SOH) of a battery, and belongs to the technical field of battery management. S1. initializing a migration model and parameters, obtaining input parameters, inputting the input parameters into a particle filter algorithm, obtaining a plurality of SOH initial particles according to an initial probability distribution of a battery SOH, performing prediction and updating on the SOH initial particles, and obtaining updated particles, wherein the input parameters are current, a battery capacity value at a current time step and battery voltage; S2. performing optimization and updating weight on the plurality of updated particles by using a chaotic firefly optimization method, and obtaining a global optimal solution, wherein the chaotic firefly optimization method is a chaotic mapping introduced in a firefly optimization algorithm; S3. performing normalization on particles in the global optimal solution, judging whether a resampling condition is met according to the normalized particles, obtaining a particle set after resampling if the resampling condition is met, obtaining an SOC estimation value at a current time step according to the particle set after resampling, and obtaining the SOC estimation value at the current time step according to the normalized particles if the resampling condition is not met; S4. inputting the SOC estimation value at the current time step into a Kalman filter, obtaining an updated battery estimated capacity value and an SOH estimation value, inputting the updated battery estimated capacity value into a control matrix of the particle filter algorithm, and obtaining an SOC estimation value at a next time step and returning to S1 for iteration until a preset iteration number is reached or convergence is achieved, and obtaining an optimal SOH estimation value.

2. The method of novel chaotic glowworm-swarm optimization battery state of health estimation as claimed in claim 1, wherein, In the S2, the global optimal solution is obtained by performing optimization and updating weight on the plurality of updated particles by using the chaotic firefly optimization method, and the method comprises the following steps: S21. constructing an initial firefly population according to the updated particles, wherein a firefly individual is an SOH value, and brightness of the firefly individual is weight of the updated particle; S22. updating positions of the fireflies by combining a chaotic mapping algorithm for optimization, and calculating fluorescence brightness and attraction degree of the fireflies in a current group; S23. returning to S22 according to the fluorescence brightness and the attraction degree until an iteration number is met, and obtaining the global optimal solution.

3. The method of novel chaotic glowworm-swarm optimization battery state of health estimation of claim 2, wherein, In the S22, the positions of the fireflies are updated by combining the chaotic mapping algorithm for optimization, and the method comprises the following steps: sorting individuals in a current firefly population according to individual fluorescence brightness, and obtaining excellent individuals in a preset range as a chaotic search space; selecting a firefly from the preset range, and taking an SOH value of the selected firefly as a basic solution; mapping the basic solution to a preset interval according to a Tent chaotic mapping, and iteratively generating a chaotic variable sequence; back-solving the chaotic variable sequence to an original solution space, and obtaining a new solution sequence in the original solution space; calculating fluorescence brightness of each firefly in the new solution sequence, and obtaining a current optimal solution; comparing weight of the current optimal solution with weight of the basic solution, updating the current optimal solution as the global optimal solution if the weight of the current optimal solution is greater than the weight of the basic solution, and not updating the current optimal solution if the weight of the current optimal solution is not greater than the weight of the basic solution.

4. The method of novel chaotic glowworm-swarm optimization battery state of health estimation of claim 3, wherein, Iteratively generating the chaotic variable sequence includes: obtaining initial values ​​for the chaotic variable sequence; obtaining the chaotic variable sequence Tent(m) based on the initial values ​​and the Tent mapping model, where m = 1, 2, ..., ITERmax, wherein the initial values... The Tent mapping model is ITERmax is the maximum number of iterations in the chaotic search, X A As the fundamental solution, X min and X max x represents the minimum and maximum values ​​among individuals within a preset range. n Let x represent the output value of the Tent chaotic map at the nth iteration. n+1 This represents the output value of the Tent chaotic map at the (n+1)th iteration.

5. The method of novel chaotic glowworm-swarm optimization battery state of health estimation as claimed in claim 1, wherein, The control matrix of the particle filter algorithm is a control matrix of a state transition equation in the particle filter algorithm.

6. The method of novel chaotic glowworm-swarm optimization battery state of health estimation as claimed in claim 1, wherein, After obtaining the SOC estimation value of the current time step, the method further includes migrating parameters in a state transition equation and an observation equation of the particle filter algorithm.

7. The method of novel chaotic glowworm-swarm optimization battery state of health estimation of claim 6 wherein, The migrating parameters in the state transition equation and the observation equation of the particle filter algorithm are: ； where, is the migrated SOC value for the k+1 time step battery model, is the k time step SOC value, is the step size for the adjacent time step, is the k time step actual capacity value, is the k time step current, is the current, is the process noise in the SOC estimation, is the process noise in the polarization voltage estimation, is the k+1 time step R P C P terminal voltage of the circuit, is the k time step R P C P terminal voltage of the circuit, denotes the polarization resistance, C P denotes the polarization capacitance, is the time constant, is the k+1 time step battery model migrated open circuit voltage, is the ohmic internal resistance inside the battery after the battery model is migrated, denotes the observation noise at the k time step, is the k+1 time step closed circuit voltage after the battery is connected with the external circuit after the battery model is migrated, the superscript M battery denotes the parameters after the model is migrated, the time constant τ M = R P M C P M , R P M denotes the migrated polarization resistance, C P M denotes the migrated polarization capacitance.

Images