A method for online identification of abnormal battery self-discharge rate

Abstract

This invention provides a method for online identification of abnormal battery self-discharge rates, relating to the field of new energy battery technology. Specifically, it utilizes discharge data, by collecting data on a specific date (D). i Battery-related parameters are input into the constructed battery model, and the open-circuit voltage at each time step is obtained using an identification algorithm. This voltage is then input into the MQ solver module to obtain the corresponding MQ. i Value, when [D i MQ i ] n×2 When the matrix has 2 rows, the SDR is calculated using a formula. When the matrix dimension is greater than 2, the DBSCAN algorithm is used to remove outliers from the matrix, and the remaining points are linearly fitted to obtain the SDR. This invention improves the accuracy of internal short circuit diagnosis and the efficiency of detecting minute internal short circuits. The required pre-reserved parameters only involve the SOC-OCV data of the new battery, without any other offline testing or calibration. It is not limited by operating conditions, does not rely on charging data, and does not require waiting for the battery to discharge to a certain depth or meet certain resting conditions. The algorithm increases the calculable vehicle coverage and has strong versatility.

Description

Technical Field

[0001] This invention relates to the field of new energy battery application technology, and more specifically to a method for online identification of abnormal battery self-discharge rates. Background Technology

[0002] After a power battery system is installed in a vehicle, some individual battery cells may experience internal short circuits due to misuse or manufacturing defects. Short circuits are one of the main causes of battery thermal runaway. To ensure battery safety, it is necessary to monitor battery short circuits and issue timely warnings in the early stages of even minor short circuits. Currently, the common method for detecting internal short circuits in batteries is to leave them unattended for an extended period and monitor changes in battery voltage, with the storage time being one week or more. Due to the limitation of the storage time, this method cannot be used to monitor short circuits in batteries after installation in a vehicle.

[0003] To achieve online monitoring of vehicle batteries, cloud-based remote diagnostic technology has emerged in related fields. This technology analyzes the battery's short-circuit status based on battery data uploaded from the vehicle to the cloud. Currently, various methods for monitoring internal short circuits in batteries exist. For example, application publication number CN 111929602A discloses a quantitative diagnostic method for single-cell battery leakage or micro-short circuits based on capacity estimation, including the following steps: S1 acquiring charge and discharge data of individual battery cells; S2 estimating the battery charging capacity C using traditional capacity estimation methods. C and discharge capacity C D S3 calculates the ratio of discharge capacity to charging capacity, and determines that a leakage fault has occurred when the ratio is less than a threshold. However, since the charging capacity of a real vehicle is greatly affected by temperature and current ratio, and the discharge capacity of a real vehicle is also greatly affected by ambient temperature and vehicle operating conditions, this method is obviously not applicable to real vehicles.

[0004] For example, Chinese patent application CN111208439A discloses a quantitative detection method for micro short-circuit faults in series lithium-ion battery packs. Based on an improved dual Kalman filter (DEKF), it estimates the battery's OCV; the battery's SOC is calculated by interpolation, thereby calculating the SOC difference between batteries. However, for lithium iron phosphate batteries, the SOC obtained from OCV has a large error, resulting in low algorithm accuracy and making it difficult to identify micro short circuits.

[0005] For example, Chinese patent application CN 113848495A discloses a method for diagnosing internal micro-short circuit faults based on charging curves. This patent mainly extracts aging characteristics from the IC curve obtained from the battery charging curve to understand the current aging state of the battery. However, in actual use of new energy vehicles, many vehicles struggle to achieve constant current charging, and many vehicles have charging currents greater than 1C. Charging with a current greater than 1C cannot reflect the charging characteristics of the two platforms, resulting in low algorithm versatility.

[0006] Therefore, we provide a method for online identification of abnormal battery self-discharge rates. Summary of the Invention

[0007] This invention provides a method for online identification of abnormal battery self-discharge rates, overcoming the shortcomings of existing methods such as low accuracy and low versatility in identifying abnormal battery self-discharge rates.

[0008] The present invention adopts the following technical solution:

[0009] A method for online identification of abnormal battery self-discharge rates includes the following steps:

[0010] (1) Collect relevant battery parameters during the operation of vehicle Di on a certain date, including voltage V, current I, and time t;

[0011] (2) The current I and time t arrays collected in step (1) are processed to obtain the capacity value Q(k) at each moment. Then, by building a model and identifying parameters, D is obtained. i The open-circuit voltage data uoc(k) corresponding to each voltage data point for the day;

[0012] (3) Input the open-circuit voltage data uoc(k) from step (2) into the MQ solver module to obtain the corresponding MQ. i value;

[0013] (4) Select another date and repeat steps (1) to (3) to obtain [D] i MQ i ] n×2 A matrix, where n is the number of repetitions;

[0014] (5) When [D i MQ i ] n×2 When n=2 in the matrix, the formula SDR=[MQ2-MQ1-(MQ2)] is used. 0 -MQ1 0 The SDR is calculated by ] / (D2-D1) / Q0, where MQ2 0 With MQ1 0 The corresponding MQ values ​​are obtained by repeating steps (1) to (3) on the standard voltage curve for days D2 and D1, respectively; when n>2, first calculate the MQ values. i After correction, the DBSCAN algorithm is used to delete outliers in the matrix. The remaining points are linearly fitted to obtain the slope K. Then, the self-discharge rate of the battery is calculated using the formula SDR = K / Q0.

[0015] In a preferred embodiment, the battery system is required to be fully charged (SOC≥99%) at the start time of the above step (1) of collecting battery-related parameter data, and the data collection ends at the time before the vehicle stops running and starts charging.

[0016] In a preferred embodiment, the capacity value Q(k) at each moment in step (2) above is obtained by processing according to the following two formulas: Q(1)=Q0-(1), Q(k)=Q(k-1)-I(k)×[t(k)-t(k-1)] / 3600-(2); where k is the sequence number from 1 to N, N is the total number of time t arrays, and Q0 is the rated capacity of the battery system.

[0017] In a preferred embodiment, the specific steps of the MQ solving module in step (3) above include:

[0018] S1. First, initialize the three parameters Mov_Q, Mov_V, and Comp, and substitute them into the following formulas (3)-(5) to obtain the objective function J:

[0019] x(k)=(Q(k)-Mov_Q)×Comp / Q0- (3);

[0020] y(k)=fi(x(k))- (4);

[0021]

[0022] Wherein, Mov_Q is a design parameter with the same dimension as the capacity Q, with a minimum of 0 and a maximum of Q0; Mov_V is a design parameter with the same dimension as uoc, with a minimum of 0 and a maximum of 0.1; Comp is a dimensionless design parameter, with a minimum of 1 and a maximum of 10; x(k) is the corresponding state of charge transition value; f(x) is the relationship function between the battery's state of charge SOC and open-circuit voltage uoc data; y(k) is the open-circuit voltage conversion value obtained by converting x(k) through the relationship function;

[0023] S2. Use the parameter optimization method to find the optimal Mov_Q, Mov_V, and Comp, so that J is minimized;

[0024] S3. Using the Mov_Q obtained in S2, calculate MQ according to formula (6):

[0025] MQ = f 1 (y0)*Q0 / Comp+Mov_Q (6)

[0026] Among them, f 1 (y) is the inverse function of fi(x), where y takes the value y0, and y0 is a fixed open-circuit voltage value. iThe corresponding MQ tag is MQ i .

[0027] In a preferred embodiment, in the specific step S3 of the above MQ solving module, y0 is taken as the open-circuit voltage value corresponding to the obvious characteristics of the state of charge, such as the SOC value corresponding to SOC=1, SOC=0, or the SOC-uoc curve showing an obvious inflection point; when taking the open-circuit voltage value when SOC=0, f 1 (y0) = 0, MQ = Mov_Q.

[0028] In a preferred embodiment, the above-mentioned MQ can also be solved by finding the sequence number based on the correspondence between Q(k) and y(k).

[0029] In a preferred embodiment, the optimization method in step S2 of the MQ solving module can be any one of the following: traversal algorithm, genetic algorithm, particle swarm optimization algorithm, or intelligent machine learning optimization algorithm.

[0030] In a preferred embodiment, in the specific step S3 of the above MQ solution module, f(x) is the relationship function between the battery's state of charge (SOC) and open-circuit voltage (uoc) data. Specifically, it is a function obtained by formula fitting, or a smoothing function, or an interpolation function, or a two-column array. When it is a two-column array, x(k) obtained by formula (4) is rounded, rounded, or interpolated to correspond with the SOC array.

[0031] In a preferred embodiment, the voltage data in step (1) above is the lowest single-cell voltage value at any time in the battery system, or the voltage value corresponding to the lowest state of charge of a single cell, or the voltage of a specific single cell.

[0032] In a preferred embodiment, step (5) above [D i MQ i ] n×2 When n=2 in the matrix, and the battery discharge SOC is too high to meet the requirement of dropping to y0, MQ2 0 With MQ1 0 The value is omitted and replaced with 0.

[0033] In a preferred embodiment, the MQ of step (5) above i The correction method includes calculating the corresponding MQ obtained using the standard voltage curve. i 0 The MQ value is obtained by changing with the date, thus showing how it is affected by factors such as temperature, normal lifespan aging, charge rate, and SOC error. The corrected MQ value is obtained after deducting the interference caused by these factors. i value.

[0034] As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following advantages:

[0035] 1. This invention provides a method for online identification of abnormal battery self-discharge rates. It utilizes vehicle driving discharge condition data, employing points with distinct open-circuit voltage characteristics as the basis for MQ calculation, and within [D...] i MQ i ] n×2 When the matrix has 2 rows, the SDR is calculated using a formula; when the matrix dimension is greater than 2, the DBSCAN algorithm is used to remove outliers from the matrix, and the remaining points are linearly fitted to obtain the SDR. This invention improves the accuracy of internal short circuit diagnosis and the efficiency of detecting minute internal short circuits, solving the problems of low accuracy and low versatility in existing methods for identifying abnormal battery self-discharge rates.

[0036] 2. The parameters required for advance storage in this invention only involve the SOC-OCV data of the new battery. No other offline testing and calibration are required. It is not limited by operating conditions, does not depend on charging data, does not require waiting for the battery to discharge to a certain depth, or to meet certain static conditions, does not require disassembling the battery box, and does not require long-term static storage. The algorithm in the whole method improves the calculable vehicle coverage and has strong versatility. Attached Figure Description

[0037] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0038] The following reference Figure 1 Specific embodiments of the present invention will be described below. Many details are described below to provide a comprehensive understanding of the invention; however, those skilled in the art can implement the invention without these details. Well-known components, methods, and processes will not be described in detail hereafter.

[0039] Example 1

[0040] This embodiment provides a method for online identification of abnormal battery self-discharge rates, including the following steps:

[0041] Step 1: Select a date D i Collect D i Battery-related parameters during vehicle operation on that day included the highest single-cell voltage Vmax, the lowest single-cell voltage Vmin, current I, and time t. Data collection began when the battery system was fully charged (SOC ≥ 99%), and ended when the vehicle finished operating and before recharging began.

[0042] Step 2: Process the current I and time t arrays in the data collected in Step 1 according to formulas (1)-(2) to obtain the capacity value Q(k) at each moment, where: k is the sequence number from 1 to N, N is the total number of time t arrays, and Q0 is the rated capacity of the battery system.

[0043] Q(1)=Q0 (1)

[0044] Q(k)=Q(k-1)-I(k)×[t(k)-t(k-1)] / 3600 (2)

[0045] Step 3: Using the data obtained in Steps 1 and 2, build a battery model and identify parameters to obtain D. i The open-circuit voltage data corresponding to each Vmin of the day is denoted as uoc(k).

[0046] Step 4: Solve MQ. In the MQ solving module, first initialize the three parameters Mov_Q, Mov_V, and Comp, and substitute them into formulas (3)-(5) to obtain the objective function J;

[0047] x(k)=(Q(k)-Mov_Q)×Comp / Q0 (3)

[0048] y(k)=fi(x(k)) (4)

[0049]

[0050] Wherein, Mov_Q is a design parameter with the same dimension as the capacity Q, and can take a minimum of 0 and a maximum of Q0; Mov_V is a design parameter with the same dimension as uoc, and can take a minimum of 0 and a maximum of 0.1; Comp is a dimensionless design parameter, and can take a minimum of 1 and a maximum of 10; x(k) is the corresponding state of charge transition value; f(x) is the relationship function between the battery's state of charge SOC and open circuit voltage uoc data; y(k) is the open circuit voltage conversion value obtained by converting x(k) through the relationship function.

[0051] Step 5: Use the parameter optimization method to find the optimal Mov_Q, Mov_V, and Comp, so that J is minimized.

[0052] Step 6: Using the Mov_Q obtained in Step 5, calculate MQ according to formula (6):

[0053] MQ = f 1 (y0)*Q0 / Comp+Mov_Q (6)

[0054] Among them, f 1(y) is the inverse function of f(x), where y takes the value y0, which is a fixed open-circuit voltage value. This value can be chosen when the state of charge exhibits distinct characteristics, such as the SOC value corresponding to SOC = 1, SOC = 0, or the SOC-uoc curve showing a clear inflection point. When the open-circuit voltage value at SOC = 0 is taken, f... 1 (y0) = 0, MQ = Mov_Q; D i The corresponding MQ tag is MQ i .

[0055] Step 7: Select another date and repeat steps 1 through 6 to obtain [D] i MQ i ] n×2 A matrix, where n is the number of repetitions;

[0056] Step 8: When n = 2, calculate the battery's self-discharge rate SDR according to formula (7), where MQ2 0 With MQ1 0 These are the corresponding MQ values ​​calculated using steps one through six on days D2 and D1, respectively, based on the standard voltage curve. When the battery discharge SOC is too high to meet the requirement of dropping to y0, MQ2... 0 With MQ1 0 It can be omitted as 0; when n>2, first check MQ. i Correction was performed, and then the DBSCAN algorithm was used to [D] i MQ i ] n×2 Outliers in the matrix are removed, and the remaining points are linearly fitted to obtain the slope K. The self-discharge rate of the battery is calculated by formula (8).

[0057] SDR = [MQ2 - MQ1 - (MQ2)] 0 -MQ1 0 )] / (D2-D1) / Q0(n=2) (7)

[0058] SDR=K / Q0 (n>2) (8)

[0059] Step 9: Set the corresponding short-circuit warning level according to the self-discharge rate to achieve short-circuit warning.

[0060] The battery system involved in this invention can be a new energy vehicle battery system or an energy storage system.

[0061] The voltage data in step one above can be the lowest single-cell voltage value at any given moment in the battery system, or the voltage value corresponding to the single cell with the lowest state of charge, or the voltage of a specific single cell. The current value I in step one is defined as negative for charging and positive for discharging; of course, the current value I can also be defined as positive for charging and negative for discharging, then formula (1) is adjusted accordingly to Q(1)=-Q0.

[0062] In step two above, the initial value of Q is set to the rated capacity value, or it can be set to 0 or other values, and the associated formulas are modified accordingly. For example, when the initial value of Q is 0, all Q values ​​in the above formulas are replaced with (Q0-Q).

[0063] The battery model in step three above can be an equivalent circuit model or an electrochemical model. The identification algorithm can be any algorithm capable of identifying the OCV, such as the least squares identification algorithm, Kalman filter algorithm, H-infinity algorithm, or intelligent machine learning optimization algorithm.

[0064] In step four above, f(x) is the relationship function between the battery's state of charge (SOC) and open-circuit voltage (uoc) data. It can be a function obtained by fitting the formula, a smoothing function, an interpolation function, or a two-column array. When it is an array, x(k) obtained by formula (4) can be matched with the SOC array by rounding, rounding, interpolation, etc.

[0065] The parameter optimization solution method in step five above can be any method that can solve the problem, such as traversal algorithm, genetic algorithm, particle swarm optimization algorithm, intelligent machine learning optimization algorithm, etc.

[0066] The standard voltage curve in step eight above can be the average voltage of the battery system, the highest single-cell voltage of the battery system, or the voltage of a single cell that has been marked as normal. The MQ in step eight... i The correction method includes calculating the corresponding MQ obtained using the standard voltage curve. i 0 The MQ value is obtained by changing with the date, thus showing how it is affected by factors such as temperature, normal lifespan aging, charge rate, and SOC error. The corrected MQ value is obtained after deducting the interference caused by these factors. i value.

[0067] Example 2

[0068] The method for online identification of abnormal battery self-discharge rate in this embodiment is basically the same as the steps in Embodiment 1, with the main difference being step six.

[0069] Step six in this embodiment involves calculating MQ using the Mov_Q obtained in step five according to the formula:

[0070] MQ=Q(i), i=find(y(k)=y0)

[0071] Where y0 is the open-circuit voltage value corresponding to a distinct state of charge, such as the open-circuit voltage value corresponding to SOC = 0.62; D i The corresponding MQ tag is MQ i .

[0072] The above are merely specific embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing upon the protection scope of the present invention.

Claims

1. A method for online identification of abnormal battery self-discharge rate, characterized in that, Includes the following steps: (1) Collect relevant battery parameters during the operation of vehicle Di on a certain date, including voltage V, current I, and time t; (2) The current I and time t arrays collected in step (1) are processed to obtain the capacity value Q(k) at each moment. Then, by building a model and identifying parameters, D is obtained. i The open-circuit voltage data uoc(k) corresponding to each voltage data of the day; the capacity value Q(k) at each moment is obtained by processing according to the following two formulas: Q(1)=Q0-(1), Q(k)=Q(k-1)-I(k)×[t(k)-t(k-1)] / 3600-(2); where k is the sequence number from 1 to N, I(k) is the kth current value, t(k) is the kth time value, and Q0 is the rated capacity of the battery system; (3) Input the open-circuit voltage data uoc(k) from step (2) into the MQ solver module to obtain the corresponding MQ. i Value; specific steps include: S1. First, initialize the three parameters Mov_Q, Mov_V, and Comp, and substitute them into the following formulas (3)-(5) to obtain the objective function J: x(k)=(Q(k)-Mov_Q)×Comp / Q0-(3); y(k)=f(x(k))-(4; -（5）； Where N is the total number of elements in the time t array; Mov_Q is a design parameter with the same dimensions as the capacity Q, with a minimum of 0 and a maximum of Q0; Mov_V is a design parameter with the same dimensions as uoc, with a minimum of 0 and a maximum of 0.1; Comp is a dimensionless design parameter, with a minimum of 1 and a maximum of 10; x(k) is the corresponding state of charge transition value; f(x) is the relationship function between the battery's state of charge SOC and open-circuit voltage uoc data; y(k) is the open-circuit voltage conversion value obtained by converting x(k) through the relationship function. S2. Use the parameter optimization method to find the optimal Mov_Q, Mov_V, and Comp, so that J is minimized; S3. Using the Mov_Q obtained in S2, calculate MQ according to formula (6): MQ=f -1 (y0)*Q0 / Comp+Mov_Q (6) Among them, f -1 (y) is the inverse function of f(x), where y takes the value y0, and y0 is a fixed open-circuit voltage value. i The corresponding MQ tag is MQ i ; (4) Select another date and repeat steps (1) to (3) to obtain [D] i MQ i ] n×2 A matrix, where n is the number of repetitions; (5) When [D i MQ i ] n×2 When n=2 in the matrix, the formula SDR=[MQ2 -MQ1-(MQ2)] is used. 0 -MQ1 0 The SDR is calculated by ] / (D2-D1) / Q0, where SDR is the battery self-discharge rate, and MQ2 is the metric. 0 With MQ1 0 The corresponding MQ values ​​are obtained by repeating steps (1) to (3) on the standard voltage curve for days D2 and D1, respectively; when n>2, first calculate the MQ values. i Correction is performed, including calculating the corresponding MQ obtained using the standard voltage curve. i 0 The MQ value is obtained by changing with the date, thus showing how it is affected by factors such as temperature, normal lifespan aging, charge rate, and SOC error. The corrected MQ value is obtained after deducting the interference caused by these factors. i The value is then used; the DBSCAN algorithm is then used to remove outliers in the matrix, and the remaining points are linearly fitted to obtain the slope K. The self-discharge rate of the battery is then calculated using the formula SDR=K / Q0.

2. The method for online identification of abnormal battery self-discharge rate as described in claim 1, characterized in that: The start time of step (1) collecting battery-related parameter data requires the battery system to be fully charged, i.e., SOC≥99%, and the end time of data collection is the time before the vehicle stops running and starts charging.

3. The method for online identification of abnormal battery self-discharge rate as described in claim 1, characterized in that: In the specific step S3 of the MQ solution module, the open-circuit voltage value corresponding to the obvious characteristics of the y0 charge state includes the SOC value corresponding to SOC=1, SOC=0 or the SOC-uoc curve showing an obvious inflection point. When the open-circuit voltage value at SOC=0 is taken, f -1 (y0)=0, MQ=Mov_Q.

4. The method for online identification of abnormal battery self-discharge rate as described in claim 1, characterized in that: The solution for MQ is obtained by using the index-based search method based on the correspondence between Q(k) and y(k).

5. The method for online identification of abnormal battery self-discharge rate as described in claim 1, characterized in that: The optimization method in step S2 of the MQ solving module can be any one of the following: traversal algorithm, genetic algorithm, particle swarm algorithm, or intelligent machine learning optimization algorithm.

6. The method for online identification of abnormal battery self-discharge rate as described in claim 4, characterized in that: In the specific steps S3 of the MQ solution module, f(x) is the relationship function between the battery's state of charge (SOC) and open-circuit voltage (uoc) data. Specifically, it is a function obtained by formula fitting, a smoothing function, an interpolation function, or a two-column array. When it is a two-column array, x(k) obtained by formula (4) is rounded, rounded, or interpolated to correspond with the SOC array.

7. The method for online identification of abnormal battery self-discharge rate as described in claim 1, characterized in that: The voltage data in step (1) is the lowest single-cell voltage value in the battery system at any given moment, or the voltage value corresponding to the lowest state of charge of a single cell, or the voltage of a specific single cell.

8. The method for online identification of abnormal battery self-discharge rate as described in claim 1, characterized in that: The [D] in step (5) i MQ i ] n×2 When n=2 in the matrix, and the battery discharge SOC is too high to meet the requirement of dropping to y0, MQ2 0 With MQ1 0 The value is omitted and replaced with 0.

Images