Automatic identification method for battery material system

Abstract

The invention discloses an automatic identification method for a battery material system. The method comprises the following steps: acquiring voltage, current, time and SOC data; dividing into M groups according to SOC data or current values; obtaining Vmean / Vmain for each group of data; feature values Vmean (1), Vmean (M), Vmain (1), Vmain (M), detaVmean, detaVmain and the like are extracted; constructing a fuzzy neural network; and outputting a material system classification result. The method is based on a battery theoretical knowledge design process, combines the electrochemical characteristics of the battery material system to extract important characteristics, establishes a mapping relation between the characteristics and the battery material system according to a fuzzy neural network or a fuzzy rule knowledge base, realizes automation of confirmation of the battery material system, needs a small amount of calculation, is simple and convenient, and has high practical value. And intelligent technical support is provided for scenes requiring identification of battery material system types.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy vehicle battery applications, and more specifically to an automatic identification method for battery material systems. Background Art

[0002] The cloud diagnosis system for new energy vehicle batteries is one of the important guarantees for ensuring the healthy and safe operation of vehicles. This system conducts the excavation of battery hidden dangers based on data such as battery voltage, current, temperature, and SOC stored in the cloud. Currently, the main types of batteries installed in new energy vehicles are lithium iron phosphate batteries, ternary batteries, lithium manganate batteries, etc. Due to the large differences in the voltage characteristics of various batteries, the thresholds for overvoltage alarms, undervoltage alarms, and differential pressure alarms for batteries of different material systems are not the same, and different treatments are also required when making hidden danger warnings in the cloud to improve the accuracy and adaptability of the battery diagnosis system. Therefore, an important link in the process of cloud battery data processing is to classify the battery material system.

[0003] Currently, the commonly used classification methods are to classify according to the vehicle model configuration information or to manually sort out the battery material information table and look up the classification. Both belong to the classification based on the information collection method. However, with the continuous growth of the number of new energy vehicles, the workload of manual information statistics is increasing, and there is a certain probability of statistical errors, which misleads the subsequent battery cloud diagnosis. At the same time, for existing old vehicles that use new material system batteries after the original batteries are out of production, it is difficult to efficiently and accurately collect and update this part of information, thus giving rise to the need for automatic identification of battery material systems. Summary of the Invention

[0004] The present invention provides an automatic identification method for battery material systems, aiming to solve the disadvantages of the existing classification method for battery material systems based on the information collection method, such as large workload in the information collection process, high change and maintenance costs, and being not conducive to the timely update of battery information.

[0005] The present invention adopts the following technical solutions:

[0006] An automatic identification method for battery material systems includes the following steps:

[0007] Step 1: Obtain data including voltage, current, time, and SOC;

[0008] Step 2: Group according to SOC, and count the time serial numbers Tci when SOC is between A(k) and A(k - 1), with a total of M groups;

[0009] Step 3: Extract the corresponding voltage characteristics for each group of Tci, including the average value Vmean of the voltage value Vi corresponding to the Tci moment and the voltage value Vmain with the highest occurrence frequency extracted by statistically analyzing the probability density distribution of Vi;

[0010] Step 4: Calculate detaVmean\detaVmain from the Vmean\Vmain arrays calculated in Step 3 according to the following formulas (1) - (2);

[0011] detaVmean = Vmean(k) - Vmean(k - 1) (1)

[0012] detaVmain = Vmain(k) - Vmain(k - 1) (2)

[0013] Step 5: Extract the eigenvalue Vmean(1), Vmean(M), Vmain(1), Vmain(M), detaVmean, detaVmain, and build a fuzzy neural network to perform the training of a number of labeled data to achieve the automatic classification of the material system; or build a knowledge base according to the known material characteristics to achieve the automatic classification of the material system without labels and without training.

[0014] The method for grouping SOC in the above Step 2 is: set the array A of [0, a, 2*a, 3*a,..., n*a] according to the step size a 1*(n+1) .

[0015] The above step size a is set to 5 - 10, 5 < M ≤ n.

[0016] The method for extracting voltage characteristics in the above Step 3 is: calculate the voltage value Vi corresponding to the Tci moment, calculate the average value Vmean of Vi, and / or perform probability density distribution statistics on Vi, with the statistical step size being b, and extract the voltage value Vmain with the highest occurrence frequency; the statistical step size b is preferably less than 50 mV.

[0017] The present invention can also adopt the following technical solution: a method for automatically identifying a battery material system, including the following steps:

[0018] Step 1: Obtain data including voltage, current, time, and SOC;

[0019] Step 2: Group according to the current, screen the time serial numbers with the absolute value of the current value less than the threshold c, and mark them as Tci', with a total of M' groups;

[0020] Step 3: Extract the corresponding voltage characteristics for each group of Tci', including the average value Vmean' of the voltage value Vi' corresponding to the Tci' moment and performing probability density distribution statistics on Vi' and extracting the voltage value Vmain' with the highest occurrence frequency;

[0021] Step 4: Calculate the detaVmean' and detaVmain' arrays obtained in Step 3 according to the following formulas (3) - (4);

[0022] detaVmean' = Vmean′(k) - Vmean'(k - 1) (3)

[0023] detaVmain' = Vmain'(k) - Vmain'(k - 1) (4)

[0024] Step 5: Extract the eigenvalue Vmean'(1), Vmean'(M’), Vmain'(1), Vmain'(M’), detaVmean', and detaVmain’, and build a fuzzy neural network to implement the training of several labeled data to achieve the automatic classification of the material system; or build a knowledge base according to the known material characteristics, and achieve the automatic classification of the material system without labels and without training.

[0025] In the above Step 3, the statistical step size of the probability density distribution of Vi' is b, and b is less than 50 mV.

[0026] From the above description of the present invention, compared with the prior art, the present invention has the following advantages:

[0027] The present invention designs a process based on battery theory knowledge, extracts important features in combination with the electrochemical characteristics of the battery material system, establishes a mapping relationship between the features and the battery material system according to the fuzzy neural network or the fuzzy rule knowledge base, automates the confirmation of the battery material system, avoids relying on manual information collection, improves the work efficiency of the battery material system confirmation link, reduces subsequent diagnostic errors caused by information collection errors, has a small amount of calculation required by the method, is simple and convenient, and provides intelligent technical support for scenarios that need to identify the types of battery material systems. Brief Description of the Drawings

[0028] Figure 1 It is a schematic flow chart of Embodiment 1 of the present invention.

[0029] Figure 2 It is a schematic flow chart of Embodiment 2 of the present invention. Detailed Description of the Invention

[0030] The following refers to Figure 1 to illustrate the specific implementation manners of the present invention. In order to fully understand the present invention, many details are described below, but for those skilled in the art, the present invention can be implemented without these details. For well-known components, methods, and processes, no further detailed description will be given below.

[0031] Embodiment 1

[0032] This embodiment provides a method for automatically identifying a battery material system. Refer to Figure 1 , and specifically includes the following steps:

[0033] Step 1: Obtain data such as voltage, current, time, and SOC of a new energy vehicle battery.

[0034] Step 2: Perform grouping according to SOC. The specific grouping method is as follows: Set a series of arrays A = [0, a, 2*a, 3*a,..., n*a] according to a certain step size a 1*(n+1) , and count the time serial numbers where SOC is distributed between A(k) and A(k - 1), marked as Tci, with a total of M groups. Among them, the step size a is set to 5 - 10, M ≤ n, and M is preferably greater than 5.

[0035] Step 3: Extract the corresponding voltage features for each group of Tci. The voltage features in this embodiment are extracted using the following method:

[0036] Calculate the voltage value Vi corresponding to the Tci moment, and calculate the average value Vmean of Vi, and / or perform probability density distribution statistics on Vi, with a statistical step size of b, and extract the voltage value Vmain with the highest occurrence frequency, where b is preferably less than 50 mV.

[0037] Step 4: Calculate detaVmean\detaVmain by performing calculations on the Vmean\Vmain arrays obtained in Step 3 according to the following formulas (1) - (2).

[0038] detaVmean = Vmean(k) - Vmean(k - 1) (1)

[0039] detaVmain = Vmain(k) - Vmain(k - 1) (2)

[0040] Step 5: Extract eigenvalue Vmean(1), Vmean(M), Vmain(1), Vmain(M), detaVmean, detaVmain, etc., and build a fuzzy neural network to perform training on a number of labeled data to achieve automatic classification of the material system; or build a knowledge base according to known material characteristics to achieve automatic classification of the material system without labels and without training.

[0041] Embodiment 2

[0042] The main difference between the method for automatically identifying a battery material system in this embodiment and that in Embodiment 1 lies in the differences in Step 2 and Step 3, and the remaining steps are the same. Refer to Figure 2 , and specifically includes the following steps:

[0043] Step 1: Obtain data such as the voltage, current, time, and SOC of the new energy vehicle battery.

[0044] Step 2: Implement grouping based on the current value. The specific grouping method is as follows: Set arrays A = [0, a, 2*a, 3*a,..., n*a] according to a certain step size a. 1*(n+1) , Screen the time serial numbers with the absolute value of the current value less than the threshold c, mark them as Tci’, and there are M’ groups in total. The step size a is set to 5 - 10, M’ ≤ n, and M’ is preferably greater than 5.

[0045] Step 3: Extract the corresponding voltage features for each group of Tci'. The voltage features in this embodiment are extracted using the following method:

[0046] Calculate the average value Vmean’ of the voltage value Vi’ corresponding to the moment of Tci’, and / or perform probability density distribution statistics on Vi', with a statistical step size of b, and extract the voltage value Vmain’ with the highest occurrence frequency. b is preferably less than 50mV.

[0047] Step 4: Calculate detaVmean’ / detaVmain’ by performing calculations on the Vmean’ / Vmain’ arrays obtained in Step 3 according to the following formulas (3) - (4);

[0048] detaVmean’ = Vmean’(k) - Vmean’(k - 1) (3)

[0049] detaVmain’ = Vmain’(k) - Vmain’(k - 1) (4)

[0050] Step 5: Extract the eigenvalue Vmean’(1), Vmean’(M’), Vmain’(1), Vmain’(M’), detaVmean’, detaVmain’, build a fuzzy neural network to implement the training of several labeled data to achieve the automatic classification of the material system; or build a knowledge base according to the known material features to achieve the automatic classification of the material system without labels and without training.

[0051] It should be noted that the voltage in Step 1 of the above two embodiments can be the highest single - cell voltage of the battery system, the lowest single - cell voltage of the battery system, the voltage of a certain battery cell in the battery system, or the average single - cell voltage of the entire battery system.

[0052] The above is only the specific implementation manner of the present invention, but the design concept of the present invention is not limited thereto. Any non - substantial modification made to the present invention using this concept shall fall within the scope of infringement of the protection of the present invention.

Claims

1. A method for automatically identifying a battery material system, characterized in that: The following steps are involved: Step 1: Obtain data including voltage, current, time and SOC; Step 2: Grouping is performed based on SOC, and the time series numbers Tci of SOC distribution between A(k) and A(k-1) are counted, with a total of M groups; Step 3: extract the corresponding voltage features for each group Tci, including the average value Vmean of the voltage value Vi corresponding to the Tci moment and the voltage value Vmain with the highest frequency of occurrence extracted by performing probability density distribution statistics on Vi; Step 4: According to the following formulas (1) to (2), the Vmean\Vmain array calculated in step 3 is used to calculate detaVmean\detaVmain; detaVmean=Vmean(k)-Vmean(k-1) (1) detaVmain=Vmain(k)-Vmain(k-1) (2) Step 5. Extract the characteristic values ​​Vmean(1), Vmean(M), Vmain(1), Vmain(M), detaVmean, detaVmain, build a fuzzy neural network, implement training of several labeled data to realize automatic classification of material system; or build a knowledge base according to known material characteristics to realize automatic classification of material system without labels and training.

2. A method for automatically identifying a battery material system according to claim 1, characterized in that: The method of grouping SOC in step 2 is: according to the step length a, an array A of [0, a, 2*a, 3*a, ..., n*a] is set. 1*(n+1) .

3. A method for automatically identifying a battery material system according to claim 2, characterized in that: The step length a is set to 5-10, 5<M≤n.

4. A method for automatically identifying a battery material system according to claim 2, characterized in that: The method for extracting the voltage feature in step three is: calculating the voltage value Vi corresponding to the time Tci, calculating the average value Vmean of Vi, and / or performing probability density distribution statistics on Vi with a statistical step length of b, and extracting the voltage value Vmain with the highest occurrence frequency.

5. A method for automatically identifying a battery material system according to claim 2, characterized in that: The statistical step length b is less than 50 mV.

6. A method for automatically identifying a battery material system, characterized in that: The following steps are involved: Step 1: Obtain data including voltage, current, time and SOC; Step 2: Grouping based on current, screening the time series numbers whose absolute current values ​​are less than the threshold c, marking them as Tci', with a total of M' groups; Step 3: extract the corresponding voltage features for each group Tci', including the average value Vmean' of the voltage value Vi' corresponding to the time Tci' and perform probability density distribution statistics on Vi' and extract the voltage value Vmain' with the highest occurrence frequency; Step 4: According to the following formulas (3) to (4), the Vmean'\Vmain' array calculated in step 3 is used to calculate detaVmean'\detaVmain'; detaVmean'=Vmean'(k)-Vmean'(k-1) (3) detaVmain'=Vmain'(k)-Vmain'(k-1) (4) Step 5. Extract the characteristic values ​​Vmean'(1), Vmean'(M'), Vmain'(1), Vmain'(M'), detaVmean', detaVmain', build a fuzzy neural network, implement training of several labeled data to realize automatic classification of material systems; or build a knowledge base according to known material characteristics to realize automatic classification of material systems without labels or training.

7. A method for automatically identifying a battery material system according to claim 6, characterized in that: In the step 3, the statistical step length of the probability density distribution of Vi' is b, and b is less than 50 mV.

Images