CN116165548A

CN116165548A - Offline damage assessment method and system for lithium battery pack, electronic equipment and storage medium

Info

Publication number: CN116165548A
Application number: CN202310291179.XA
Authority: CN
Inventors: 王光; 邓鑫; 焦建芳; 谢家乐
Original assignee: North China Electric Power University
Current assignee: North China Electric Power University
Priority date: 2023-03-23
Filing date: 2023-03-23
Publication date: 2023-05-26

Abstract

The invention discloses a method, a system, electronic equipment and a storage medium for offline damage assessment of a lithium battery pack, which relate to the technical field of damage diagnosis and comprise the following steps: quantifying battery voltage synchronicity as an improved pearson correlation coefficient sequence using an improved pearson correlation coefficient curve of a lithium battery under the influence of damage from different severity levels (I, II, iii); calling a matlab tool box, and converting the phase relation sequence into a recursion diagram image in a recursion diagram conversion mode; constructing a training data set according to the recursion image corresponding to each damage severity level and the recursion image under the condition of no damage; training the M-RVM model by using a training data set; and inputting the recursion image corresponding to the lithium battery pack to be diagnosed into a trained model, and outputting a damage diagnosis result by using the trained model. The invention can reduce the calculated amount and improve the damage detection precision, the damage classification precision and the damage severity classification precision.

Description

Offline damage assessment method and system for lithium battery pack, electronic equipment and storage medium

Technical Field

The invention relates to the technical field of battery damage diagnosis, in particular to an offline damage assessment method and system for a lithium battery pack, electronic equipment and a storage medium.

Background

Due to long cycle life, high energy and power density, negligible self-discharge rate, lithium ion batteries have long been playing an irreplaceable role in energy storage applications on grid-oriented and grid-oriented power systems and electric vehicles. However, the inherent vulnerability of lithium batteries leads to the damage to components such as insulating separators, electrodes, and electrolytes, which affects the normal use of lithium batteries. If not maintained in time, the minor damage can be easily upgraded to system damage. Safety issues with lithium battery systems have attracted considerable attention in the industry and academia. Therefore, development of an effective diagnostic technique for various damages in lithium batteries is imperative.

The existing lithium battery damage diagnosis method can be classified into three modes of behavior mechanism modeling, priori knowledge accumulation and abnormal signal analysis according to different main technologies, and the characteristics of the method are as follows:

1. by integrating the observer or filter, the correlation between separator conductivity and shorting resistance is revealed using low frequency impedance characteristics based on a custom P2D model.

2. The method comprises the steps of exploring a pseudo-distributed battery model based on an electrochemical principle, obtaining damage evidence by modeling voltage, temperature and residual errors of a State of charge (SoC), and extracting damage characteristics, so that quantitative evaluation of an Internal Short Circuit (ISC) State is carried out by using a related vector machine.

3. To isolate the damage in the reconfigurable component, structural analysis is performed based on the electrothermal coupling model, and a minimum subset of sensors with optimal damage diagnostic reliability is pursued. An extended state space model is built that combines the conventional battery state and the short current state to detect ISC.

Existing schemes or custom P2D models (pseudo-two-dimensions) are based on revealing the correlation between separator conductivity and short-circuit resistance using low frequency impedance characteristics, which, although achieving a compelling performance, are computationally intensive due to the large number of partial derivatives employed to describe particle migration dynamics, resulting in significant computational overhead, or pseudo-distributed cell models or extended state space models, too dependent on model accuracy, which would be greatly reduced if reduced order or linear approximation were simplified, resulting in a greatly reduced performance of the damage diagnosis, and thus a greatly reduced damage detection accuracy (ability to distinguish damage from normal non-damage), damage classification accuracy, and damage severity classification accuracy.

In summary, how to improve the damage detection accuracy, the damage classification accuracy and the damage severity classification accuracy while reducing the calculation amount is a problem to be solved by those skilled in the art.

Disclosure of Invention

The invention aims to provide an offline damage assessment method, an offline damage assessment system, an offline damage assessment electronic device and an offline damage assessment storage medium for a lithium battery pack, wherein the damage detection precision, the damage classification precision and the damage severity classification precision can be improved while the calculated amount is reduced.

In order to achieve the above object, the present invention provides the following solutions:

a method of offline damage assessment of a lithium battery pack, the method comprising:

obtaining an improved pearson correlation coefficient curve of the constructed lithium battery pack under the influence of various injuries of each injury severity level; each improved Pelson correlation coefficient curve is obtained by applying off-line current excitation to the constructed lithium battery pack, simulating different types of damage, simulating different damage severity levels of each type of damage and detecting; the lithium battery pack comprises a plurality of brand-new lithium batteries connected in series; the improved pearson correlation coefficient is obtained by improving the pearson correlation coefficient by using recursive self-adaption, forgetting and noise suppression;

Quantifying voltage synchronicity between adjacent lithium cells in the lithium battery pack as an improved pearson correlation coefficient sequence using the improved pearson correlation coefficient curves for each of the improved pearson correlation coefficient curves;

calling a matlab tool box, and converting each improved pearson correlation coefficient sequence into a recursive graph image by adopting a recursive graph transformation mode aiming at each improved pearson correlation coefficient sequence;

constructing a training data set according to the recursion map image corresponding to various injuries of each injury severity level and the recursion map image corresponding to the non-injury; the training data set comprises a plurality of types of recursive image and damage diagnosis results corresponding to the recursive image; the damage diagnosis result comprises no damage, internal short-circuit damage, external short-circuit damage and overheat damage with different damage severity grades; the different lesion severity levels include micro-lesions, medium lesions and severe lesions;

training the M-RVM model by using the training data set to obtain a trained M-RVM model;

acquiring a recursion image corresponding to a lithium battery pack to be diagnosed;

And inputting the recursion diagram image corresponding to the lithium battery pack to be diagnosed into the trained M-RVM model, and outputting a damage diagnosis result by using the trained M-RVM model.

Optionally, the acquiring the recursion map image corresponding to the lithium battery pack to be diagnosed further includes:

obtaining an improved pearson correlation coefficient curve corresponding to a lithium battery pack to be diagnosed;

quantifying the voltage synchronism between adjacent lithium batteries in the lithium battery pack to be diagnosed into an improved pearson correlation coefficient sequence corresponding to the lithium battery pack to be diagnosed by utilizing the improved pearson correlation coefficient curve corresponding to the lithium battery pack to be diagnosed;

and converting the improved pearson correlation coefficient sequence corresponding to the lithium battery pack to be diagnosed into a recursion chart image corresponding to the lithium battery pack to be diagnosed in a recursion chart conversion mode.

Optionally, the recursive image corresponding to the damage-free mode is obtained by not simulating damage to the lithium battery pack, detecting an improved pearson correlation coefficient curve under the influence of the damage-free mode of the lithium battery pack, quantifying voltage synchronism between adjacent lithium batteries in the lithium battery pack into an improved pearson correlation coefficient sequence by using the improved pearson correlation coefficient curve, and converting the improved pearson correlation coefficient sequence into a recursive image by adopting a recursive image conversion mode.

The invention also provides the following scheme:

a lithium battery offline damage assessment system, the system comprising:

the improved pearson correlation coefficient curve acquisition module is used for acquiring an improved pearson correlation coefficient curve of the constructed lithium battery pack under the influence of various injuries of each injury severity level; each improved Pelson correlation coefficient curve is obtained by applying off-line current excitation to the constructed lithium battery pack, simulating different types of damage, simulating different damage severity levels of each type of damage and detecting; the lithium battery pack comprises a plurality of brand-new lithium batteries connected in series; the improved pearson correlation coefficient is obtained by improving the pearson correlation coefficient by using recursive self-adaption, forgetting and noise suppression;

a voltage synchronization quantization module for quantizing voltage synchronization between adjacent lithium batteries in the lithium battery pack to an improved pearson correlation coefficient sequence using the improved pearson correlation coefficient curves for each of the improved pearson correlation coefficient curves;

a recursive graph transformation module, configured to invoke a matlab toolbox, and for each of the modified pearson correlation coefficient sequences, convert the modified pearson correlation coefficient sequences into a recursive graph image in a form of recursive graph transformation;

The training data set construction module is used for constructing a training data set according to the recursion map image corresponding to various injuries of each injury severity level and the recursion map image corresponding to the non-injury; the training data set comprises a plurality of types of recursive image and damage diagnosis results corresponding to the recursive image; the damage diagnosis result comprises no damage, internal short-circuit damage, external short-circuit damage and overheat damage with different damage severity grades; the different lesion severity levels include micro-lesions, medium lesions and severe lesions;

the M-RVM model training module is used for training the M-RVM model by utilizing the training data set to obtain a trained M-RVM model;

the recursion image acquisition module is used for acquiring recursion images corresponding to the lithium battery pack to be diagnosed;

and the damage diagnosis module is used for inputting the recursion diagram image corresponding to the lithium battery pack to be diagnosed into the trained M-RVM model, and outputting a damage diagnosis result by using the trained M-RVM model.

Optionally, the system further comprises:

the to-be-diagnosed Pelson correlation coefficient curve acquisition module is used for acquiring an improved Pelson correlation coefficient curve corresponding to the to-be-diagnosed lithium battery pack;

The to-be-diagnosed voltage synchronism quantifying module is used for quantifying the voltage synchronism between adjacent lithium batteries in the to-be-diagnosed lithium battery pack into an improved pearson correlation coefficient sequence corresponding to the to-be-diagnosed lithium battery pack by utilizing an improved pearson correlation coefficient curve corresponding to the to-be-diagnosed lithium battery pack;

and the pearson correlation coefficient sequence conversion module to be diagnosed is used for converting the improved pearson correlation coefficient sequence corresponding to the lithium battery pack to be diagnosed into a recursion chart image corresponding to the lithium battery pack to be diagnosed in a recursion chart conversion mode.

The invention also provides the following scheme:

an electronic device comprising a memory for storing a computer program and a processor running the computer program to cause the electronic device to perform the lithium battery pack offline damage assessment method.

The invention also provides the following scheme:

a computer readable storage medium storing a computer program which when executed by a processor implements the method of offline damage assessment of a lithium battery pack.

According to the specific embodiment provided by the invention, the invention discloses the following technical effects:

according to the offline damage assessment method, system, electronic equipment and storage medium of the lithium battery pack, voltage synchronicity among batteries is quantized into an IPCC (improved pearson correlation coefficient sequence) sequence, wherein damage-independent deduction such as load dynamics and noise can be effectively shielded, and key information is further reserved; invoking a matlab tool box, adopting an RP (recursive graph) transformation mode to convert a time sequence (an improved pearson correlation coefficient sequence) into an RP image (a recursive graph image), intuitively visualizing hidden cross-time autocorrelation characteristics into image textures, avoiding a large amount of useless data calculation, adopting matlab to realize the conversion of the time sequence into the RP image without adopting a large amount of partial differentiation, and thus reducing the calculation amount; the RP image contains rich texture details, can be used as fault evidence, RP images of different faults have unique textures, can be effectively distinguished by using an M-RVM model, is based on a real high-precision damage data set constructed by the RP image, can be used for obtaining a model with higher precision by training the M-RVM model, can accurately identify no damage and damage by using the trained M-RVM model, and accurately diagnose damage types and damage severity levels, thereby improving damage detection precision, damage classification precision and damage severity classification precision.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flowchart of a method for offline damage assessment of a lithium battery pack according to an embodiment of the present invention;

FIG. 2 is a flowchart showing an off-line damage assessment method for a lithium battery pack based on RP images;

FIG. 3 is a physical view and schematic of an experimental set-up of the present invention;

FIG. 4 is a schematic diagram of a cyclic load curve of the present invention;

FIG. 5 is a comparative schematic diagram of battery voltage during a damage period;

FIG. 6 is a schematic graph of IPCC curves affected by different types and classes of lesions;

FIG. 7 is a schematic diagram of an RP image that is affected by different types and grades of damage;

FIG. 8 is a schematic diagram of a diagnostic model of injury according to the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.

Example 1

Fig. 1 is a flowchart of an embodiment of an off-line damage assessment method for a lithium battery pack according to the present invention. As shown in fig. 1, the embodiment provides an offline damage assessment method for a lithium battery pack, which includes the following steps:

step 101: obtaining an improved pearson correlation coefficient curve of the constructed lithium battery pack under the influence of various injuries of each injury severity level; each improved Pelson correlation coefficient curve is obtained by applying off-line current excitation to a constructed lithium battery pack, simulating different types of damage, simulating different damage severity levels of each type of damage, and detecting; the lithium battery pack comprises a plurality of brand-new lithium batteries connected in series; the improved pearson correlation coefficient is obtained by improving the pearson correlation coefficient using recursive adaptation, forgetting and noise suppression.

Step 102: for each modified pearson correlation coefficient curve, the modified pearson correlation coefficient curve is utilized to quantify the voltage synchronicity between adjacent lithium cells in the lithium battery pack as a modified pearson correlation coefficient sequence.

Step 103: calling a matlab kit, and converting the improved pearson correlation coefficient sequences into a recursive graph image by adopting a recursive graph transformation mode aiming at each improved pearson correlation coefficient sequence.

Step 104: constructing a training data set according to the recursion map image corresponding to various injuries of each injury severity level and the recursion map image corresponding to the non-injury; the training data set comprises a plurality of recursive image images and damage diagnosis results corresponding to the recursive image images; the damage diagnosis result comprises no damage, internal short-circuit damage, external short-circuit damage and overheat damage with different damage severity grades; different lesion severity levels include micro-lesions, medium lesions, and severe lesions.

In step 104, the recursive image corresponding to the damage-free is obtained by not simulating the damage to the lithium battery pack, detecting and obtaining an improved pearson correlation coefficient curve under the influence of the damage-free of the lithium battery pack, quantifying the voltage synchronism between adjacent lithium batteries in the lithium battery pack into an improved pearson correlation coefficient sequence by using the improved pearson correlation coefficient curve, and converting the improved pearson correlation coefficient sequence into the recursive image by adopting a recursive image conversion mode.

Step 105: and training the M-RVM model by using the training data set to obtain a trained M-RVM model.

Step 106: and acquiring a recursion image corresponding to the lithium battery pack to be diagnosed.

This step 106 is preceded by:

and obtaining an improved Pelson correlation coefficient curve corresponding to the lithium battery pack to be diagnosed.

And quantifying the voltage synchronism between adjacent lithium batteries in the lithium battery pack to be diagnosed into an improved pearson correlation coefficient sequence corresponding to the lithium battery pack to be diagnosed by utilizing the improved pearson correlation coefficient curve corresponding to the lithium battery pack to be diagnosed.

Step 107: and inputting the recursion image corresponding to the lithium battery pack to be diagnosed into a trained M-RVM model, and outputting a damage diagnosis result by using the trained M-RVM model.

The technical scheme of the invention is described in the following by a specific embodiment:

the invention relates to an off-line damage assessment method of a lithium battery pack based on RP images, which is shown in a specific flow chart in figure 2 and comprises the following steps:

And step one, connecting a plurality of lithium batteries (lithium ion batteries) in series to construct a lithium ion battery pack (lithium battery pack).

Fig. 3 is a physical view and schematic of the experimental setup of the present invention. As shown in fig. 3, four new lithium ion batteries (NCM, 3.7v,4 ah) are assembled into a lithium ion battery pack (Li-ion battery pack) by a serial assembly, and the cascade voltage of the batteries is synchronously measured by using a self-designed circuit board (voltage acquisition unit), and a special chip LTC6811 is used as a core processor. The lithium ion battery is a ternary lithium battery (NCM NI-CO-Mnbattery). An electronic load, a controllable dc power supply and several relays work together to simulate the load excitation on the assembly. The STM 32-based module (i.e. the local control unit) switches charge and discharge, uploads terminal data and triggers damage. The RS-232 bus is used for inter-device data and command transmission. LabView-based applications running on the host are designed for global control and data management. The signal sampling frequency was set to 5Hz. FIG. 4 is a schematic diagram of a cyclic load curve of the present invention, with the multi-stage load condition shown in FIG. 4 as a load-on-pump cycle.

And step two, artificially simulating the behavior of damaging the battery.

The damage simulation is performed because the battery itself has no problem in the first step, and the damage is required to be simulated manually when the damage is to be measured. Three types of damage shown in table 1 were induced by performing multiple abuse operations on the cells, and simulations were performed on three levels of each type (i.e., damage severity levels). Three types are external short (External short circuit, ESC), overheating (OHT), internal short-circuit (ISC). Three classes, I: microdamage, II: moderate injury, III: severe damage. Wherein, grade I is mild (Minor), grade II is moderate (Medium), and grade III is severe (series). The lesion configurations are shown in table 1.

TABLE 1 injury configuration

The internal resistance in table 1 refers to the internal resistance measured after the battery was overcharged, and the distance refers to the distance from the spray gun nozzle to the battery surface.

The simulation of ISC damage is to change the state of charge (SOC) percentage and resistance of the internal resistance (Internal resistance), the simulation of ESC damage is to change the resistance of the Short resistance (Short resistance) and the simulation of OHT damage is to change the heating temperature (Heating temperature) and distance.

The measured voltage is a representation of the damage because it is a difference between the voltage of the damaged cell and the normal cell. Lithium battery damage can affect ion intercalation and deintercalation rates, solid electrolyte phase interface (solid electrolyte interphase, SEI) thickness, and electrolyte conductivity, macroscopically manifesting as changes in overpotential and internal resistance, so monitoring electrical synchrony (voltage synchrony) between batteries can provide informative evidence for damage Diagnosis (Fault Diagnosis).

The pearson correlation coefficient may reflect a relationship between voltages. The pearson correlation coefficient (Pearson correlation coefficient, PCC), i.e. pearson correlation coefficient, is a linear correlation coefficient, and is the most commonly used one cov (X, Y) to reflect the degree of linear correlation of two variables X and Y, the larger the absolute value of which indicates a stronger correlation, which is used here to observe the correlation of voltages between battery cells for correlation analysis (corelationalysis).

In order to perceive system anomalies online, it is necessary to introduce recursive and noise suppression mechanisms to the PCC. Furthermore, revealing damage-related information from lengthy non-damaged data, rather than being overwhelmed by lengthy non-damaged data, requires adding a forgotten window to the PCC to balance damage sensitivity and data retention, i.e., improving pearson correlation coefficient (Impoving Pearson correlation coefficient, IPCC), quantifying electrical synchronicity between units into pearson correlation coefficient sequences by IPCC (improved pearson correlation coefficient) curves, where damage-independent inferences such as load dynamics and noise are effectively masked.

Improved pearson correlation coefficient:

PCC is the most common index for measuring the correlation between two variables, and for discrete variable X, Y, the PCC expression is:

at the same time, in order to detect system anomalies online, it is necessary to introduce recursive and noise suppression mechanisms. In addition, to reveal information about the damage, a forgetting window is required to balance the damage sensitivity and the data retention. The PCC formula is modified as:

wherein:

the P, Q, R, S and T are simply referred to herein, are convenient for writing, and have no meaning per se, and are specifically shown in the following form:

Wherein the method comprises the steps of

v _x,i And v _y,i Is the original signal (e.g. battery voltage), +.>

Is an auxiliary square wave that eliminates disordered oscillations caused by measurement errors and noise in steady state, w is the forgetting window length, and a recursive calculation is performed. In the present invention, w coincides with the length and width of an image described later, +.>

Set to 6 sampling intervals. IPCC can effectively mitigate false alarms caused by load fluctuations while preserving critical information.

The above formulas are all existing formulas, and the meaning of each parameter involved in the formulas is described in the existing publications, so that the description is omitted.

The pearson correlation coefficient is an upgrade of the euclidean distance (i.e. it provides a processing step that varies over a range of variable values, the differences in the dimensions of the different variables being removed during the calculation), and an improvement in cosine similarity in the absence of dimensional values.

In order to observe the damage to the battery, it is necessary to apply a damage operation by hand, and table 1 shows a specific operation of the damage arrangement.

And thirdly, obtaining an IPCC curve, wherein the subsequent application is the image acquisition source of the fourth step.

In the previous step two, multiple groups of abuse operations are performed on the battery units to induce OHT and ISC damage, and voltages of adjacent batteries are collected by the voltage collecting unit as shown in fig. 5, fig. 5 is a comparative schematic diagram of battery voltages during damage, part (a) in fig. 5 shows voltages of healthy batteries and OHT batteries, and part (b) in fig. 5 shows voltages of healthy batteries and ISC. Taking the example of both OHT and ISC damage, the terminal voltages of two adjacent cells (one damage free, the other) are depicted in fig. 5, wherein the dashed box area represents the duration of time that the damage is activated. It can be seen that in both cases, the voltage of the damaged battery gradually deviates from the voltage of the healthy battery. However, using conventional threshold-based approaches does not provide enough evidence to give damage details, as voltage differences are easily overwhelmed by high dynamic loads, temperature drift of the sensor, and EMI distortion. Since these adverse factors often interfere with all voltmeters at the same time, they can be accurately suppressed using IPCC. From an electrochemical perspective, OHT generally has a direct impact on electrolyte, lithiation and delithiation, thereby affecting particle transport resistance; ISCs caused by separator rupture typically reduce static internal resistance without significantly changing dynamic internal resistance. These unique impedance changes, which manifest as distinct responses in the frequency domain, can be preserved by IPCC and then materialized as image texture. IPCC generally shows both sensitivity and reliability in lesion detection. FIG. 6 is a schematic of IPCC curves affected by different types and classes of lesions. FIG. 6 shows the IPCC curves detected under the influence of different types and classes of lesions. In fig. 6, (a) shows an IPCC curve detected under the influence of no damage, fig. 6, (b) shows an IPCC curve detected under the influence of ESC damage, fig. 6, (c) shows an IPCC curve detected under the influence of ISC damage, and fig. 6, (d) shows an IPCC curve detected under the influence of OHT damage.

Step four: the temporal sequence is converted into RP images using RP-transforms.

RP is a method of restorative visualization proposed on the basis of poincare theorem. The method obtains priori knowledge from the internal structure of the time sequence, interprets the similarity and information quantity of the time sequence, can perform predictability analysis on signals, and is an important method for analyzing the periodicity, chaos and non-stationarity of the time sequence. The nature of RP is a time-to-time signal processing method used to encode cyclic behavior occurring in a time series into two-dimensional images for representation. Time series data can be classified using unique repetitive behaviors such as periodic and irregular periodic aspects.

Fig. 7 is a schematic diagram of an RP image affected by different types and levels of damage. Part (a) in fig. 7 shows an RP image affected by ISC damage, part (b) in fig. 7 shows an RP image affected by ESC damage, and part (c) in fig. 7 shows an RP image affected by OHT damage. To compare texture differences between images, nine sets of RP images collected according to vertical type and horizontal level are shown in fig. 7. Images within the same beam have the same lesion type and grade, but are affected by different loads. After load calibration, images of the same location of the different beams are derived. From the observation, the RP images contain abundant texture details and can be used as fault evidence, which is why RP transformation is adopted instead of direct judgment by using IPCC curves.

The first to fourth steps are to obtain a real injury dataset for training the mRVM model (M-RVM model). The invention obtains a relatively real training data set, namely a real high-precision damage data set, based on the steps one to four. The invention has high precision because the more real training data set is obtained through the first step to the fourth step, and the model with higher precision can be obtained by training the M-RVM model by adopting the data set, thereby realizing the accurate classification of the damage.

Training the M-RVM model to obtain a damage diagnosis (damage detection) result.

M-RVM (multi-class relevance vectormachine), a multi-class relevance vector machine; the RVM algorithm is a Bayesian framework-based machine learning model, which is trained by maximizing marginal likelihood to obtain correlation vectors (RVs) and weights, and then machine learning by the M-RVM algorithm (M-RVM model). During training iterations, the posterior distribution of most parameters tends to zero, and samples corresponding to non-negligible parameters are called RVs, which reflect the primary information of all samples. The number of RVs undergoes an increasing-plane-decreasing three-phase change, which means that a highly sparse model is obtained. The model can diagnose which damage class of the battery in the battery pack is, such as No fault and Minor, medium, serious, and classifies damage types of the battery into ISC, ESC, OHT, the flow is shown in FIG. 8, and FIG. 8 is a schematic diagram of the damage diagnosis model of the invention. A total of 10000 signal samples, which cover ten states (including one non-invasive state and nine invasive states shown in table 1), are input into the M-RVM algorithm for machine learning model training. Using these data, 10000 RPs were generated, respectively, as shown in table 2. The image sets were then segmented at 8:2 ratio for training and testing, respectively, as shown in Table 2.

Table 2 image samples and labels for different lesion configurations for model training and testing

Label (Label)	Injury state	Training	Testing
				A	No damage	800	200
B	External short-circuit damageI)	800	200
				C	External short-circuit damage (II)	800	200
D	External short-circuit damage (III)	800	200
				E	Internal short-circuit damage (I)	800	200
F	Internal short-circuit damage (II)	800	200
				G	Internal short-circuit damage (III)	800	200
H	Overheat damage (I)	800	200
				I	Overheat damage (II)	800	200
J	Overheat damage (III)	800	200

The training data set is three-fourths of 10000 data samples for model training, and the rest for method test, input characteristic samples and output damage detection and separation matrix.

And (3) damage detection:

800 samples (including a) from the "test" dataset in table 2 and another 600 samples from the three lesion types were mixed together. Only two ISC samples on RP are misjudged as ESC and nondestructive, the detection precision is perfect, and all nondestructive samples can be correctly identified. This means that the RP images of different faults have unique textures, which can be effectively distinguished using the mvm algorithm (M-RVM model).

Injury grading:

and verifying the damage grading performance. Referring to the test data of table 2, three data sets for each lesion type were mixed with the non-lesion data sets, namely "b+c+d+a", "e+f+g+a" and "h+i+j+a", so each test set contained 800 samples. Table 3 summarizes the results and it can be seen that a significant portion of the lesion samples were misdiagnosed. 61% on ESC sample RP. In fact, this result is predicted from the mechanism of RP, i.e., the uniformly partitioned form confuses the IPCC data of ESCs, which are mostly concentrated near the minimum, with little fluctuation. For the ISC case, the average scoring accuracy of RP was 63.3%, with an accuracy of RP slightly higher than 51.9% for class II ("F" row), which can be explained by the considerable randomness of the damaging effects of high temperature on cell structure and materials, and the overheating consequences are not fully manifested when a defective cell is subjected to dynamic loading. The total accuracy of the injury classification of RP is 58.7% respectively, which is basically consistent with training. Table 3 is a damage grading matrix for RP with reference to the labels in table 2.

TABLE 3 injury ranking matrix on RP

/>

The bolded data in Table 3 represents successful classification (i.e., the type of lesion successfully diagnosed), the remaining data, except bolded data, data 0, and accuracy data (including 61.0%, 63.3%, 51.9%, and 58.7%), represent unsuccessful diagnosis of the lesion type, and zero represents no sample.

In summary, the offline damage assessment method of the lithium battery pack provided by the invention has obvious advantages in damage type detection (damage detection). Although the performance of OHT is significantly worse in terms of further lesion severity grading, weak texture differences between different lesion severity are difficult to identify, while less satisfactory, showing the prospect of further optimization.

The invention discloses an offline damage assessment method of a lithium battery pack, which is a damage diagnosis method of the lithium battery pack and aims to construct a novel serial battery pack damage diagnosis framework, namely an intelligent diagnosis framework: firstly, the synchronism between adjacent batteries is quantified in real time through an electrical system to indicate that the lithium battery is abnormal. The modified PCC (IPCC) sequence is then converted into a Recursive Plot (RP) by RP transformation. Then, using a Recursive Plot (RP), the phase relation sequence is converted into an RP image (using matlab toolbox), and the texture of the RP image reflects the information details of the system state. Finally, the occurrence of damage is detected by utilizing an M-RVM algorithm and integrating characteristic evidence, and the type and the severity of the damage are judged, so that the effects of improving the overall damage detection precision (the capability of distinguishing damaged fragments from normal nondestructive fragments) and the damage classification precision, namely the damage severity classification precision (I, II and III classification) in the damage diagnosis of the large battery pack are achieved.

The intelligent diagnosis framework is constructed aiming at common damage of the lithium battery ESC, the ISC and the OHT, firstly, the Pearson Correlation Coefficient (PCC) is improved through recursive self-adaption, forgetting and noise suppression, then the synchronicity between adjacent batteries is quantified in real time through an electric system to indicate that the lithium battery is abnormal, and then the Improved PCC (IPCC) sequence is converted into an RP image through RP conversion, so that time line fluctuation representing different damage can be regarded as image texture, finally, the M-RVM algorithm is utilized to synthesize feature evidence to detect the occurrence of the damage, and the type and the severity of the damage are judged, so that the overall damage detection precision and the damage grading precision in the damage diagnosis of the large battery are improved.

The key points of the invention are as follows:

1. IPCC datasets and images under different types and levels of fault influence.

2. Structural difference data and images of RP in different fault states.

3. The sparse M-RVM model (M-RVM model) was used to diagnose the type of injury (detection and injury grade (evaluation)).

Compared with the prior art, the invention has the advantages that:

1. the electrical synchronicity between the batteries is quantified as a sequence of correlation coefficients (IPCC), wherein damage-independent inferences, such as load dynamics and noise, can be effectively masked.

2. By using RP transformation as an example, the time sequence is converted into RP images, so that hidden cross-time autocorrelation characteristics can be visualized into image textures intuitively, and a large amount of useless data calculation (after conversion into the image textures, a matlab tool box can be directly called for identification) is avoided. In the process of converting the time sequence into the RP image, the invention is realized by matlab, and because a large amount of partial differentiation is not adopted, the calculated amount is small, and the calculated amount is greatly reduced.

3. At present, a method for estimating internal parameters of a battery by combining behavior mechanism modeling with priori knowledge accumulation is used for performing damage estimation on a battery pack, the adopted battery model is inaccurate and insufficient in data, and an electrochemical reaction mechanism in the battery is not considered, but the actual process is more complicated than that of the estimated battery. The invention uses the real-time measurement data of the battery pack, avoids the error of battery model estimation, and the data is completely obtained by the actual measurement in real time, which is sufficient, the obtained data completely contains the electrochemical reaction mechanism inside the battery, and the accuracy is higher.

4. By applying different physical damages to the battery, including thermal and electrical abuse, to induce damage consequences and obtain a truly high-precision damage dataset.

In addition, other more complex experimental algorithms can obtain better damage identification effect to a certain extent, but the model and algorithm are not suitable for engineering realization of the existing damage identification due to the complexity of the model and algorithm, so the invention belongs to a relatively preferable scheme.

Example two

In order to perform a corresponding method of the above embodiment to achieve the corresponding functions and technical effects, the following provides an offline damage assessment system for a lithium battery pack, which includes:

the improved pearson correlation coefficient curve acquisition module is used for acquiring an improved pearson correlation coefficient curve of the constructed lithium battery pack under the influence of various injuries of each injury severity level; each improved Pelson correlation coefficient curve is obtained by applying off-line current excitation to a constructed lithium battery pack, simulating different types of damage, simulating different damage severity levels of each type of damage, and detecting; the lithium battery pack comprises a plurality of brand-new lithium batteries connected in series; the improved pearson correlation coefficient is obtained by improving the pearson correlation coefficient using recursive adaptation, forgetting and noise suppression.

And the voltage synchronism quantifying module is used for quantifying the voltage synchronism between adjacent lithium batteries in the lithium battery pack into an improved pearson correlation coefficient sequence by utilizing the improved pearson correlation coefficient curve aiming at each improved pearson correlation coefficient curve.

And the recursive graph transformation module is used for calling the matlab tool box and converting the improved pearson correlation coefficient sequences into recursive graph images by adopting a recursive graph transformation mode aiming at each improved pearson correlation coefficient sequence.

The training data set construction module is used for constructing a training data set according to the recursion map images corresponding to various injuries of each injury severity level and the recursion map images corresponding to the non-injury injuries; the training data set comprises a plurality of recursive image images and damage diagnosis results corresponding to the recursive image images; the damage diagnosis result comprises no damage, internal short-circuit damage, external short-circuit damage and overheat damage with different damage severity grades; different lesion severity levels include micro-lesions, medium lesions, and severe lesions.

The method comprises the steps of obtaining a recursive graph image corresponding to a non-damage state, detecting and obtaining an improved pearson correlation coefficient curve of the lithium battery pack under the non-damage effect, quantifying voltage synchronism between adjacent lithium batteries in the lithium battery pack into an improved pearson correlation coefficient sequence by using the improved pearson correlation coefficient curve, and converting the improved pearson correlation coefficient sequence into the recursive graph image by adopting a recursive graph conversion mode.

And the M-RVM model training module is used for training the M-RVM model by utilizing the training data set to obtain a trained M-RVM model.

And the recursion image acquisition module is used for acquiring recursion images corresponding to the lithium battery pack to be diagnosed.

The damage diagnosis module is used for inputting the recursion diagram image corresponding to the lithium battery pack to be diagnosed into the trained M-RVM model, and outputting a damage diagnosis result by using the trained M-RVM model.

The offline damage assessment system of the lithium battery pack further comprises:

and the Pelson correlation coefficient curve obtaining module to be diagnosed is used for obtaining an improved Pelson correlation coefficient curve corresponding to the lithium battery pack to be diagnosed.

And the to-be-diagnosed voltage synchronism quantifying module is used for quantifying the voltage synchronism between adjacent lithium batteries in the to-be-diagnosed lithium battery pack into an improved pearson correlation coefficient sequence corresponding to the to-be-diagnosed lithium battery pack by utilizing an improved pearson correlation coefficient curve corresponding to the to-be-diagnosed lithium battery pack.

The offline damage assessment system of the lithium battery pack is based on RP images, and can improve damage detection precision, damage classification precision and damage severity classification precision while reducing calculated amount.

Example III

An electronic device according to a third embodiment of the present invention includes a memory and a processor, where the memory is configured to store a computer program, and the processor is configured to run the computer program to cause the electronic device to execute the method for offline damage assessment of a lithium battery pack according to the first embodiment.

The electronic device may be a server.

Example IV

The fourth embodiment of the present invention provides a computer readable storage medium storing a computer program, which when executed by a processor, implements the method for offline damage assessment of a lithium battery pack according to the first embodiment.

In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.

The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.

Claims

1. An offline damage assessment method for a lithium battery pack, comprising the steps of:

2. The method for offline damage assessment of a lithium battery pack according to claim 1, wherein the acquiring the recursion map image corresponding to the lithium battery pack to be diagnosed further comprises:

3. The method for offline damage assessment of a lithium battery pack according to claim 1, wherein the recursive image corresponding to the damage-free image is obtained by detecting an improved pearson correlation coefficient curve of the lithium battery pack under the influence of the damage-free image, quantifying voltage synchronism between adjacent lithium batteries in the lithium battery pack into an improved pearson correlation coefficient sequence by using the improved pearson correlation coefficient curve, and converting the improved pearson correlation coefficient sequence into a recursive image by adopting a recursive image conversion mode.

4. An offline lithium battery pack damage assessment system, the system comprising:

5. The lithium battery pack offline damage assessment system of claim 4, wherein the system further comprises:

6. The lithium battery pack offline damage assessment system according to claim 4, wherein the recursive image corresponding to the damage-free image is obtained by detecting an improved pearson correlation coefficient curve under the influence of the damage-free image of the lithium battery pack, quantifying voltage synchronism between adjacent lithium batteries in the lithium battery pack into an improved pearson correlation coefficient sequence by using the improved pearson correlation coefficient curve, and converting the improved pearson correlation coefficient sequence into a recursive image by adopting a recursive image conversion mode.

7. An electronic device comprising a memory for storing a computer program and a processor that runs the computer program to cause the electronic device to perform the lithium battery pack offline damage assessment method of any one of claims 1-3.

8. A computer readable storage medium, characterized in that it stores a computer program, which when executed by a processor implements the lithium battery pack offline damage assessment method according to any one of claims 1-3.