Lithium battery SOP estimation optimization method

Abstract

The invention relates to the technical field of energy storage battery management, and discloses a lithium battery SOP estimation optimization method, and the method comprises the steps: building a dual-polarization second-order RC equivalent circuit model of a lithium battery, and carrying out the online identification of resistance and capacitance parameters in the model through employing a least square method with a forgetting factor; based on the identified model parameters, estimating the SOC in real time by using an H-infinity filtering algorithm; then, under the SOC constraint, the OCV constraint and the battery allowable current constraint, continuous peak current is calculated respectively, the minimum value of the three is used as effective continuous peak current, and then an SOP preliminary estimation value under multiple constraints is obtained; and finally, inputting the parameters influencing the SOP estimation precision and a model method estimation result into a neural network model, and outputting a final SOP estimation value through training optimization. According to the method, the precision and robustness of lithium battery SOP estimation can be remarkably improved under the multi-parameter constraint condition, and the method is suitable for lithium ion battery power state real-time estimation in the fields of new energy automobiles, energy storage systems and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage batteries, and in particular to a lithium battery SOP estimation and optimization method. Background Art

[0002] With the rapid development of new energy vehicles, the demand for lithium-ion battery performance management systems (BMS) is growing. These include state of charge (SOC), state of health (SOH), and state of power (SOP). SOP is defined as the peak power that a lithium-ion battery can provide to or absorb from the powertrain over a period of time, ensuring safe operation. It is a key parameter characterizing battery power performance, assessing its ability to operate properly under extreme operating conditions, similar to the engine displacement in traditional vehicles. Like SOC, SOP is an internal state parameter that cannot be directly measured and is affected by various factors, including ambient temperature, operating conditions, and SOC. Currently, commonly used methods for estimating SOP include experimental, model-based, and data-driven approaches. Model-based approaches have attracted widespread attention due to their solid theoretical foundation and high computational accuracy. However, the estimation accuracy of traditional model-based approaches remains to be improved when considering multiple constraints. Summary of the Invention

[0003] To overcome the current problems and defects, the present invention proposes a lithium battery SOP (state of power) estimation and optimization method, comprising the following steps:

[0004] S1. Establish an equivalent circuit model of the lithium battery and perform online identification of the model parameters;

[0005] S2. Estimate the state of charge (SOC) of the lithium battery based on the model parameters after online identification;

[0006] S3. Calculate a preliminary estimate of the lithium battery's SOP based on an equivalent circuit model with identified parameters, subject to constraints on SOC, open-circuit voltage (OCV), and battery current.

[0007] S4. Input the parameters related to SOP estimation and the estimation results of the model method into the neural network model, and output the SOP optimization estimation results.

[0008] Furthermore, the equivalent circuit model in step S1 is a dual-polarization second-order RC network model, and the dual-polarization second-order RC network model consists of a voltage source, an ohmic internal resistance, a first resistance-capacitance branch, and a second resistance-capacitance branch.

[0009] Furthermore, the online identification of the model parameters in step S2 adopts the least square method with a forgetting factor, which specifically includes the following steps:

[0010] S11. Setting the initial parameter vector estimate and the initial covariance matrix;

[0011] S12. At each sampling moment, collect battery terminal current and terminal voltage data and calculate the estimated error at the current moment;

[0012] S13. Calculate the algorithm gain, where the algorithm gain is determined by the current covariance matrix, the system input matrix, and the forgetting factor;

[0013] S14. Update the current covariance matrix, wherein the covariance matrix is ​​weighted recursively calculated according to the forgetting factor;

[0014] S15. Update the estimated value of the parameter vector, and correct the parameter vector using the algorithm gain and the estimated error;

[0015] S16. According to the estimated value of the parameter vector, the resistance and capacitance parameters in the dual-polarization model are identified in real time.

[0016] Furthermore, the forgetting factor has a value range of 0-1.

[0017] Furthermore, in step S2, the SOC estimation adopts the H∞ filtering algorithm. Based on the model structure and the parameters obtained by online identification, the battery current and voltage data are input into the state space equation to achieve robust estimation of the SOC. The specific steps include:

[0018] S21. Construct a discrete linear system equation based on a dual-polarization equivalent circuit model;

[0019] S22. Set the filter initial state and initial covariance matrix;

[0020] S23. Execute the time update step to predict the system state and covariance at the current moment;

[0021] S24. Execute the measurement update step, use the measurement value to correct the system state and covariance, and obtain the SOC estimation value.

[0022] Furthermore, in step S3, the preliminary SOP estimate is obtained as follows:

[0023] S31. Calculate the corresponding continuous peak current based on the SOC constraint, OCV constraint, and the maximum / minimum allowable operating current constraint of the battery;

[0024] S32. Take the minimum value of the continuous peak current under the above three constraints and calculate the preliminary SOP estimate based on the battery terminal voltage.

[0025] Furthermore, the neural network in step S4 is a BP neural network or an RBF neural network, the input parameters include charge and discharge current, terminal voltage, SOC estimation value, continuous peak current under various constraints and model method SOP estimation value, and the output is the optimized SOP estimation value.

[0026] Beneficial effects of the present invention:

[0027] The present invention adopts the least square method with a forgetting factor to perform online identification on the parameters of the dual-polarization equivalent circuit model, which can reflect the dynamic changes of the battery parameters in real time and improve the accuracy and adaptability of the model.

[0028] The present invention uses the H∞ filtering algorithm to perform SOC estimation without making strict assumptions about noise characteristics, and can maintain high estimation accuracy under complex working conditions and strong noise environments.

[0029] The present invention integrates the triple constraints of SOC, OCV and battery allowable current, adopts the minimum continuous peak current method, and ensures the safety and practicality of SOP estimation.

[0030] The present invention introduces a BP or RBF neural network to perform nonlinear optimization compensation on the SOP estimation result of the model method, which significantly reduces the SOP estimation error and improves the accuracy and generalization ability of the estimation.

[0031] This method is suitable for real-time estimation of the SOP of lithium-ion batteries in new energy vehicle power batteries and energy storage systems, providing more reliable power status data support for battery management systems and improving the safety and economy of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0033] Figure 1 Schematic diagram of the SOP optimization method for energy storage batteries in the present invention;

[0034] Figure 2 This is a diagram of a dual-polarization equivalent circuit model of the energy storage battery of the present invention;

[0035] Figure 3 This is the SOP estimation graph obtained using the traditional model method in the present invention;

[0036] Figure 4 The error diagram of the SOP estimated value and the true value obtained using the traditional model method in the present invention;

[0037] Figure 5 This is the SOP estimation optimization diagram obtained by combining the traditional model method with the BP neural network in the present invention;

[0038] Figure 6 The error diagram of the SOP estimated optimization value and the true value obtained by combining the traditional model method with the BP neural network in the present invention;

[0039] Figure 7 This is the SOP estimation optimization diagram obtained by combining the traditional model method with the RBF neural network in the present invention;

[0040] Figure 8 This is the error diagram between the SOP estimation optimization value and the true value obtained by combining the traditional model method with the RBF neural network in the present invention. DETAILED DESCRIPTION

[0041] The present application is described below with reference to specific embodiments:

[0042] Example 1:

[0043] The present invention discloses a method for estimating and optimizing the SOP of an energy storage battery. Figure 1 As shown, including:

[0044] S1. Establish a dual-polarization equivalent circuit model for the energy storage battery, and use the least squares method with forgetting factor to identify the resistance and capacitance parameters of the equivalent circuit model.

[0045] The dual polarization equivalent circuit model (second-order RC network, representing the polarization characteristics of lithium batteries) is as follows Figure 2 As shown, it consists of a voltage source, a battery ohmic internal resistance R0 and a second-order RC network (including a first equivalent resistor R1, a first equivalent capacitor C1, a second equivalent resistor R2 and a second equivalent capacitor C1). The open circuit voltage is , terminal voltage is . Open circuit voltage There is a mapping relationship with the charge state SOC of the energy storage battery, and the available function Indicates that U1 and U2 represent the voltages across the first equivalent resistor R1 and the second equivalent resistor R2 respectively.

[0046] According to the dual-polarization equivalent circuit model, the state equation and observation equation of the energy storage battery are obtained:

[0047]

[0048]

[0049] in, express Voltage at all times The derivative of express The voltage across the first equivalent resistor R1 at the moment, express The terminal current at the moment, Indicates Voltage at all times The derivative of express The voltage across the second equivalent resistor R2 at the moment, express The state of charge at the moment, express The state of charge at the moment, Indicates the rated capacity of the energy storage battery; Represents the Coulomb efficiency, usually taken as 1; Indicates the terminal voltage, represents the open circuit voltage, Indicates the battery's internal resistance in ohms.

[0050] The above state equation and observation equation are transformed into the discrete equation of the second-order RC equivalent circuit model through bilinear transformation:

[0051]

[0052] in, express System output at each moment; and express System input at time , and , for The terminal current at the moment, express The terminal voltage at the moment, express The open circuit voltage at the moment; express Parameters that need to be identified at all times include 、 、 、 and .

[0053] On this basis, the least squares method with forgetting factor is used to identify the resistance and capacitance parameters in the equivalent circuit model. The identification process includes:

[0054] Depend on Estimated value of the parameter vector at time ,calculate Time estimation error :

[0055]

[0056] Calculating algorithm gain :

[0057]

[0058] in, represents the forgetting factor, express The covariance matrix at time t, Represents the system input matrix The transposed matrix of

[0059] renew Moment covariance matrix :

[0060]

[0061] Update parameter vector estimates :

[0062]

[0063] Get vector estimates After that, the resistance and capacitance in the dual polarization model can be identified in real time, specifically, the impedance parameters about The expression is as follows:

[0064]

[0065] Where T represents the sampling period.

[0066] The resistance and capacitance parameters are obtained from the impedance parameters:

[0067]

[0068] Among them, the parameters ,parameter .

[0069] S2. Based on the equivalent circuit model, a linear system equation is constructed, and an H∞ filter is used to estimate the state of charge (SOC) of the collected battery terminal voltage and current data to obtain a real-time estimated value of the SOC.

[0070] H ∞ The filtering algorithm is a classic robust algorithm that only requires that the noise signal energy of the system is limited. Even if it is not Gaussian white noise, it can maintain a relatively small estimation error when the environmental interference noise is large. ∞The theory is based on the game between the estimation algorithm and noise, calculating the innovation matrix and obtaining the state value that minimizes it. The specific steps are as follows:

[0071] Establish a discrete linear system for the dual-polarization equivalent circuit model:

[0072]

[0073] in, = represents the state variable, represents the state of charge at time k, represents the voltage across the first equivalent resistor R1 at time k, represents the voltage across the second equivalent resistor R2 at time k; = , represents the terminal current at time k-1, which is the input variable; is the output variable at time k; = is the state matrix at time k, where is the sampling period, = is the input matrix at time k, where is the total charge capacity of the energy storage battery, = is the observation matrix at time k, represents the estimated state of charge at time k; and are the system noise and observation noise at time k, is the system noise at time k-1.

[0074] In order to avoid the problem caused by the need to pre-assume noise information in the Kalman series algorithm, H ∞ The filtering algorithm defines the following cost function :

[0075]

[0076] Where: and Indicates the initial value of the state quantity and its initial setting value; and Represents the state value and its estimated value at time k, k=0,1,…,N-1; represents the initial state error covariance matrix, and Represents the noise covariance matrix of the state equation and the noise covariance matrix of the observation equation at time k; Indicates the importance of each state quantity at time k, which is pre-set according to the actual situation. 、 、 and are all symmetric positive definite matrices and are usually constant; and are the system noise and observation noise at time k respectively.

[0077] In nature, by choosing the appropriate initial value of the state quantity , system noise , observation noise Implementing the cost function Maximization of H- ∞ For filtering algorithms, the goal is to find a suitable estimate Minimize the cost function . And directly minimize the cost function More difficult, in the application, you can choose an appropriate performance boundary To achieve, that is, to satisfy:

[0078] With this cost function Rewrite the cost function :

[0079]

[0080] Based on this, H ∞ The filtering can be simply described as: when the initial value of the state quantity , system noise , observation noise Make the cost function When the maximum value is Make the cost function Minimum. From the derivation process, H ∞ The filtering does not make any prior assumptions about the statistical properties of the noise, which is the ∞ The biggest advantage of filtering compared to Kalman series filtering.

[0081] By solving the above problems, we can finally get the cost function The recurrence relation:

[0082]

[0083] in, Indicates H ∞ Gain matrix, represents the identity matrix, represents the error covariance matrix at time k.

[0084] H ∞ The detailed filtering process is as follows:

[0085] Setting H ∞ Initial value of the filter (state observer): , ;

[0086] Complete the prior estimation (time update) operation, and update the state and covariance estimates from the previous moment Calculated to the current time , H ∞ The time update equation of the filter is expressed as follows:

[0087] System state estimation:

[0088] H ∞ Feature matrix estimation:

[0089] After completing the a priori estimation (measurement update) operation, use the measurement value Corrected state estimate and covariance estimate, H ∞ The measurement update equation of the filter is expressed as follows:

[0090] Innovation matrix:

[0091] H ∞ Gain matrix:

[0092] System status correction:

[0093] H ∞ Characteristic matrix correction:

[0094] The moment The state and covariance matrix of are taken as the final output, and the The state estimation at the moment. Based on this, the SOC estimation value is obtained.

[0095] S3. Under multiple constraints, based on the above model and SOC estimation, calculate the continuous peak current under multiple constraints respectively, and then calculate the preliminary estimated value of the continuous peak power SOP under multiple constraints.

[0096] (1) Continuous peak current estimation based on real-time SOC constraints

[0097] To ensure the safety of lithium batteries during use and avoid overcharging and overdischarging of lithium batteries.

[0098] It is necessary to ensure that the SOC is within the appropriate range, that is, 𝑆𝑂𝐶 𝑚i𝑛<𝑆𝑂𝐶 𝑘<𝑆𝑂𝐶 𝑚𝑎𝑥. The peak charging current of the lithium battery after discretization at time k is:

[0099]

[0100] Where, Represents the peak discharge current of the lithium battery in the next sampling interval under the SOC constraint at time k;

[0101] Since it is impossible for a lithium battery to be in use only at a certain moment in real life, the peak charge and discharge current within L sampling time periods is:

[0102]

[0103] Where, Represents the continuous peak discharge current of the lithium battery in the multi-sampling interval subject to SOC constraints at time k; in this patent, L is taken as 300.

[0104] (2) Continuous peak current estimation based on real-time OCV constraints

[0105] Since the terminal voltage of a lithium battery cannot exceed its rated value during use, the terminal voltage of the lithium battery at time k+1 is as follows:

[0106]

[0107] right In the Taylor expansion at time k, since the remaining terms are very small, the first-order remainder terms are ignored. Rewritten as:

[0108]

[0109] In the second-order RC equivalent circuit model, the terminal voltage of the RC loop is the sum of the zero-input response and the zero-state response voltage. Therefore, the terminal voltage of the RC loop can be expressed as:

[0110]

[0111] Combining the above three formulas, we get:

[0112]

[0113] At time k+1 the current is:

[0114]

[0115] Since the terminal voltage of the lithium battery will be affected by the upper cut-off voltage U during use,max and lower cut-off voltage U min The restrictions are:

[0116]

[0117] Therefore, under the terminal voltage constraint, the peak current of the lithium battery is:

[0118]

[0119] Where, It represents the peak discharge current of the lithium battery in a sampling interval under OCV constraint at time k;

[0120] Since the above peak current is the peak discharge current within a sampling interval, and the peak current of a lithium battery in actual use cannot be maintained only within a single sampling interval, the long sampling interval is set to 𝐿×∆𝑡. At time k+2, the terminal voltage of the double RC circuit is:

[0121]

[0122] Assuming that the peak current can be maintained constant within the time L, then:

[0123]

[0124] right Taylor expansion at time k, and ignoring its first-order terms, the expression is as follows:

[0125]

[0126] Get the terminal voltage at time k+L:

[0127]

[0128] Arranging the above formula we can get:

[0129]

[0130] The current at the moment k+L can be obtained as:

[0131]

[0132] Due to the limitation of the terminal voltage in formula (5-8), the lithium battery should be within the normal operating terminal voltage range in practice, so its peak current from time k to k+L is:

[0133]

[0134] Where, represents the peak discharge current under OCV constraint from time k to k+L.

[0135] (3) Estimation of sustained peak power of lithium-ion batteries under multiple constraints

[0136] During the actual driving of electric vehicles, in order to ensure the safe and reliable use of power batteries, it is necessary to ensure that the terminal voltage does not exceed the cut-off voltage and the operating current is within the threshold specified by the battery; in addition, in order to ensure that the battery can provide sufficient power during acceleration, climbing, deceleration, and braking of electric vehicles, it is necessary to observe the SOC value in real time to avoid overcharging and discharging. Therefore, the continuous peak current under the SOC constraint and the continuous peak current under the OCV constraint are considered separately. In addition, the lithium battery itself will also have its own design current limitations. Therefore, the calculation of the peak power of the lithium battery needs to be carried out under the three constraints. When the lithium battery is discharging, the current is positive and the minimum value under the three constraints is taken; the expression of the continuous peak discharge current of the lithium battery is as follows:

[0137]

[0138] It indicates the continuous peak discharge current during this period of time.

[0139] Obtain the continuous discharge peak power of lithium batteries under multiple constraints:

[0140]

[0141] Where, represents the continuous peak discharge power under multiple constraints.

[0142] In the process of estimating the SOP of lithium batteries, the continuous peak current needs to be estimated first, because the continuous peak current will be affected by OCV, SOC and the upper and lower limits of the battery current. Among them, the accuracy of SOC has a great influence on the estimation of SOP. The results of SOC estimation in Section 3 are used to estimate the SOP of lithium batteries, and the SOP estimation and accuracy verification are carried out under dynamic DST conditions. The SOP estimation results are shown in Figure 2. Figure 3 As shown, the error between the estimated value and the true value is as follows Figure 4 shown.

[0143] S4. Take the charge and discharge current, terminal voltage, SOC estimation value, continuous peak current based on SOC constraint, continuous peak current based on OCV constraint, continuous peak current based on battery allowable current constraint and the preliminary SOP estimation value as input parameters, and the SOP obtained from actual test as output parameter, train the neural network model, and use the trained neural network model to optimize the preliminary SOP estimation value, and output the final SOP estimation value.

[0144] The process of BP neural network is mainly divided into two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer; the second stage is the back propagation of the error, from the output layer to the hidden layer, and finally to the input layer, adjusting the weights and bias from the hidden layer to the output layer, and the weights and bias from the input layer to the hidden layer in turn.

[0145] Each neuron in the input layer is responsible for receiving input information from the outside world and passing it to each neuron in the middle layer;

[0146] In this patent, the input layer is:

[0147] I: Battery charge and discharge current under DST conditions

[0148] U: Battery charge and discharge voltage under DST conditions

[0149] SOC: Use H under DST conditions ∞ Filtered state of charge estimate

[0150] :Continuous peak current estimation under DST condition based on real-time SOC constraint

[0151] :Continuous peak current estimation under DST condition based on real-time OCV constraint

[0152] : Estimated continuous peak current under the current limit constraint of the lithium battery itself under DST conditions

[0153] SOP: SOP estimation value obtained based on the model method under DST conditions

[0154] The middle layer is the internal information processing layer, responsible for information transformation. According to the requirements of information transformation capability, the middle layer is designed as a multi-hidden layer structure; the empirical formula for the number of nodes in the hidden layer is:

[0155]

[0156] Where S is the number of nodes in the hidden layer, m is the number of input nodes, and n is the number of output nodes.

[0157] The information transmitted from the last hidden layer to each neuron in the output layer is further processed to complete a forward propagation process of learning, and the output layer outputs the information processing results to the outside world. In this patent, the output layer is: the true value of SOP.

[0158] When the actual output does not match the expected output, the error back propagation phase begins. The error passes through the output layer, and the weights of each layer are corrected according to the error gradient descent method, and then propagated back to the hidden layer and input layer layer by layer. The repeated forward propagation of information and back propagation of errors is the process of continuous adjustment of the weights of each layer, and the process of neural network learning and training. This process continues until the error of the network output is reduced to the expected level, or the pre-set number of learning times. The final SOP estimation optimization value result is as follows: Figure 5 As shown, the error value results are as follows Figure 6 As shown in the figure, it can be concluded that using the traditional model method combined with the neural network to estimate SOP can obtain a more accurate SOP estimation value.

[0159] In order to verify its universality, this patent uses the traditional model method combined with the RBF neural network to estimate the SOP. Using the same input and output values, the SOP estimation optimization value results are as follows: Figure 7 As shown, the error value results are as follows Figure 8 As shown in Figure 2, it can be seen that the use of the traditional model method combined with the RBF neural network to estimate SOP can also improve its accuracy.

[0160] It should be noted that the above embodiments can be freely combined as needed. The above are only preferred embodiments of the present invention. It should be pointed out that ordinary relevant personnel in this technical field can make several improvements and modifications without departing from the principles of the present invention. Such improvements and modifications should also be considered as the scope of protection of the present invention.

[0161] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

Claims

1. A lithium battery SOP estimation optimization method, characterized in that, The following steps are involved: S1. Establish an equivalent circuit model of the lithium battery and perform online identification of the model parameters; S2. Estimate the state of charge (SOC) of the lithium battery based on the model parameters after online identification; S3. Calculate a preliminary estimate of the lithium battery's SOP based on an equivalent circuit model with identified parameters, subject to constraints on SOC, open-circuit voltage (OCV), and battery current. S4. Input the parameters related to SOP estimation and the estimation results of the model method into the neural network model, and output the SOP optimization estimation results.

2. The method according to claim 1, wherein: The equivalent circuit model in step S1 is a dual-polarization second-order RC network model, which consists of a voltage source, an ohmic internal resistance, a first resistance-capacitance branch, and a second resistance-capacitance branch.

3. The method according to claim 1, characterized in that The online identification of the model parameters in step S2 adopts the least square method with a forgetting factor, which specifically includes the following steps: S11. Setting the initial parameter vector estimate and the initial covariance matrix; S12. At each sampling moment, collect battery terminal current and terminal voltage data and calculate the estimated error at the current moment; S13. Calculate the algorithm gain, where the algorithm gain is determined by the current covariance matrix, the system input matrix, and the forgetting factor; S14. Update the current covariance matrix, wherein the covariance matrix is ​​weighted recursively calculated according to the forgetting factor; S15. Update the estimated value of the parameter vector, and correct the parameter vector using the algorithm gain and the estimated error; S16. According to the estimated value of the parameter vector, the resistance and capacitance parameters in the dual-polarization model are identified in real time.

4. The method according to claim 3, characterized in that : The value range of the forgetting factor is 0-1.

5. The method according to claim 1, characterized in that In step S2, the SOC estimation uses the H∞ filtering algorithm. Based on the model structure and the parameters obtained by online identification, the battery current and voltage data are input into the state space equation to achieve robust estimation of the SOC. The specific steps include: S21. Construct a discrete linear system equation based on a dual-polarization equivalent circuit model; S22. Set the filter initial state and initial covariance matrix; S23. Execute the time update step to predict the system state and covariance at the current moment; S24. Execute the measurement update step, use the measurement value to correct the system state and covariance, and obtain the SOC estimation value.

6. The method according to claim 1, characterized in that In step S3, the initial SOP estimate is obtained as follows: S31. Calculate the corresponding continuous peak current based on the SOC constraint, OCV constraint, and the maximum / minimum allowable operating current constraint of the battery; S32. Take the minimum value of the continuous peak current under the above three constraints and calculate the preliminary SOP estimate based on the battery terminal voltage.

7. The method according to claim 1, wherein: The neural network in step S4 is a BP neural network or an RBF neural network, and the input parameters include charge and discharge current, terminal voltage, SOC estimation value, continuous peak current under various constraints and model method SOP estimation value, and the output is the optimized SOP estimation value.

Images