Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a lithium ion battery micro-short circuit diagnosis method and system based on clustering and charging voltage difference, and the detection and quantitative evaluation of the micro-short circuit of the energy storage lithium ion battery pack are realized based on an unsupervised clustering algorithm and the maximum charging voltage difference between adjacent cycles on the premise of no need of establishing a battery model and no need of setting a diagnosis threshold value.
In order to achieve the above object, the present invention provides the following solutions:
A lithium ion battery micro-short circuit diagnosis method based on clustering and charging voltage difference comprises the following steps:
Extracting an increment capacity IC from a constant current charging stage, and smoothing the increment capacity IC by utilizing moving average filtering to obtain a high-dimensional characteristic capable of effectively representing micro-short circuit faults of the lithium ion battery;
the PCA is utilized for main component analysis, the extracted high-dimensional features capable of effectively representing the micro-short circuit fault of the lithium ion battery are mapped to a two-dimensional plane, and the main component of the increment capacity is obtained;
constructing a fault detection model based on an unsupervised clustering algorithm, and detecting a monomer with micro short circuit fault in the lithium ion battery pack by taking the extracted main component of the increment capacity as input;
and quantitatively calculating the micro-short circuit current and micro-short circuit resistance of a single body with micro-short circuit fault in the lithium ion battery pack based on the maximum charging voltage difference between adjacent charging and discharging cycles.
Preferably, the method for extracting the incremental capacity IC includes:
where Q represents the battery charge capacity, V is the terminal voltage in the constant current charge mode, ICEVI represents the calculation IC based on the equal voltage interval Δv, and Q2-Q1 represents the charge capacity variation within the voltage interval Δv.
Preferably, the method for smoothing the incremental capacity IC by moving average filtering includes:
Wherein sr represents the value of the original signal s at the r-th moment,Is the corresponding filtered value, 2Np +1 is the window size of the sliding filter, Np is an integer, l is the time lag number, and sr-l is the value of the original signal at the r-l time.
Preferably, the method for mapping the extracted high-dimensional features capable of effectively representing the micro-short circuit fault of the lithium ion battery to the two-dimensional plane by using principal component analysis PCA comprises the following steps:
For all cells within the battery, the data matrix after incremental capacity extraction is expressed as:
Wherein,The method comprises the steps of representing an increment capacity curve extracted by an ith monomer in a jth cycle, i=1, 2, & n, j=1, 2, & d, n represents the number of the monomers in the battery pack, d represents the number of charge and discharge cycles, and m is the length of the increment capacity curve extracted by a corresponding monomer in a certain charge and discharge cycle;
The covariance matrix of the data matrix C is calculated as follows:
Wherein N is the number of rows of the matrix C, equal to N times d, and R is the covariance matrix;
Calculating eigenvalues and eigenvectors of the covariance matrix to determine a load matrix of the data matrix C, wherein the eigenvalue lambda= (lambda1,λ2,…,λm) is calculated, lambda1≥λ2≥...≥λm, the eigenvector matrix P= (P1,p2,...,pm) corresponding to the eigenvalue is the load matrix of the data matrix C, each column in P is called as a load vector, and the load vectors are mutually orthogonal;
based on the load matrix of the data matrix C, the data matrix C is decomposed into the following form:
Wherein,For the scoring matrix, each column vector Sh＝Cph of S represents the h-th column, essentially the projection of the data matrix C along the direction of the load vector,Representing the residual, q represents the number of principal components and satisfies q < m.
Preferably, the method for calculating the micro-short circuit current of the single body with the micro-short circuit fault in the lithium ion battery pack comprises the following steps:
Wherein, IMSC is micro short-circuit current, Qloss,j is estimated electric quantity loss at the j-th charging end of the micro short-circuit single body, Qloss,j-1 is estimated electric quantity loss at the j-1 th charging end of the micro short-circuit single body, Tj is the moment at the j-th charging end, and Tj-1 is the moment at the j-1 th charging end;
The method for calculating the micro-short circuit resistance of the single body with the micro-short circuit fault in the lithium ion battery pack comprises the following steps:
Wherein RMSC represents a micro-short circuit resistor, and UM is an average voltage between two adjacent cycle charging end times.
The invention also provides a lithium ion battery micro-short circuit diagnosis system based on clustering and charging voltage difference, which comprises a processing module, a mapping module, a detection module and a calculation module;
The processing module is used for extracting the increment capacity IC from the constant current charging stage, and smoothing the increment capacity IC by utilizing moving average filtering to obtain high-dimensional characteristics capable of effectively representing the micro short circuit fault of the lithium ion battery;
The mapping module is used for utilizing principal component analysis PCA to map the extracted high-dimensional characteristics capable of effectively representing the micro-short circuit fault of the lithium ion battery to a two-dimensional plane so as to obtain principal components of increment capacity;
the detection module is used for constructing a fault detection model based on an unsupervised clustering algorithm, taking the extracted main components of the increment capacity as input, and detecting a monomer with micro short circuit fault in the lithium ion battery pack;
The calculation module is used for quantitatively calculating the micro-short circuit current and the micro-short circuit resistance of a single body with micro-short circuit faults in the lithium ion battery pack based on the maximum charging voltage difference between adjacent charging and discharging cycles.
Preferably, the process of extracting the delta capacity IC includes:
where Q represents the battery charge capacity, V is the terminal voltage in the constant current charge mode, ICEVI represents the calculation IC based on the equal voltage interval Δv, and Q2-Q1 represents the charge capacity variation within the voltage interval Δv.
Preferably, the process of smoothing the incremental capacity IC using moving average filtering includes:
Wherein sr represents the value of the original signal s at the r-th moment,Is the corresponding filtered value, 2Np +1 is the window size of the sliding filter, Np is an integer, l is the time lag number, and sr-l is the value of the original signal at the r-l time.
Preferably, the process of mapping the extracted high-dimensional features capable of effectively characterizing the micro-short circuit fault of the lithium ion battery to a two-dimensional plane by using principal component analysis PCA comprises the following steps:
For all cells within the battery, the data matrix after incremental capacity extraction is expressed as:
Wherein,The method comprises the steps of representing an increment capacity curve extracted by an ith monomer in a jth cycle, i=1, 2, & n, j=1, 2, & d, n represents the number of the monomers in the battery pack, d represents the number of charge and discharge cycles, and m is the length of the increment capacity curve extracted by a corresponding monomer in a certain charge and discharge cycle;
The covariance matrix of the data matrix C is calculated as follows:
Wherein N is the number of rows of the matrix C, equal to N times d, and R is the covariance matrix;
Calculating eigenvalues and eigenvectors of the covariance matrix to determine a load matrix of the data matrix C, wherein the eigenvalue lambda= (lambda1,λ2,…,λm) is calculated, lambda1≥λ2≥...≥λm, the eigenvector matrix P= (P1,p2,...,pm) corresponding to the eigenvalue is the load matrix of the data matrix C, each column in P is called as a load vector, and the load vectors are mutually orthogonal;
based on the load matrix of the data matrix C, the data matrix C is decomposed into the following form:
Wherein,For the scoring matrix, each column vector Sh＝Cph of S represents the h-th column, essentially the projection of the data matrix C along the direction of the load vector,Representing the residual, q represents the number of principal components and satisfies q < m.
Preferably, the process of calculating the micro-short circuit current of the single body with the micro-short circuit fault in the lithium ion battery pack comprises the following steps:
Wherein, IMSC is micro short-circuit current, Qloss,j is estimated electric quantity loss at the j-th charging end of the micro short-circuit single body, Qloss,j-1 is estimated electric quantity loss at the j-1 th charging end of the micro short-circuit single body, Tj is the moment at the j-th charging end, and Tj-1 is the moment at the j-1 th charging end;
The process for calculating the micro-short circuit resistance of the single body with the micro-short circuit fault in the lithium ion battery pack comprises the following steps:
Wherein RMSC represents a micro-short circuit resistor, and UM is an average voltage between two adjacent cycle charging end times.
Compared with the prior art, the invention has the beneficial effects that:
The invention discloses a lithium ion battery micro-short circuit diagnosis method and system based on clustering and charging voltage difference, which are characterized in that increment capacity IC is extracted from a constant current charging stage, the increment capacity IC is smoothed by moving average filtering to obtain high-dimensional characteristics capable of effectively representing lithium ion battery micro-short circuit faults, PCA is analyzed by using a main component, the extracted high-dimensional characteristics capable of effectively representing lithium ion battery micro-short circuit faults are mapped to a two-dimensional plane to obtain main components of increment capacity, a fault detection model is constructed based on an unsupervised clustering algorithm, the main components of the extracted increment capacity are used as input to detect single bodies with micro-short circuit faults in the lithium ion battery, and micro-short circuit currents and micro-short circuit resistances of the single bodies with micro-short circuit faults in the lithium ion battery are quantitatively calculated based on the maximum charging voltage difference between adjacent charging and discharging cycles. The invention does not need to establish a battery model and set a diagnosis threshold value in the process of detecting micro-short circuit monomers in the battery pack and calculating the short circuit resistance of the micro-short circuit monomers, thereby avoiding the difficulties that an accurate battery model is difficult to establish and an accurate diagnosis threshold value is difficult to determine.
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
As shown in fig. 4, the invention provides a lithium ion battery micro-short circuit diagnosis method based on clustering and charging voltage difference, so as to realize detection and quantitative evaluation of the micro-short circuit of the energy storage lithium ion battery pack on the premise of not establishing a battery model and setting a diagnosis threshold value. Firstly, the incremental capacity (INCREMENTAL CAPACITY, abbreviated as IC) is extracted from the constant current charging stage and smoothed by moving average filtering to obtain the characteristic capable of effectively representing the micro short circuit fault of the lithium ion battery. Secondly, the extracted high-dimensional feature space is mapped to a two-dimensional plane by utilizing principal component analysis (PRINCIPAL COMPONENT ANALYSIS, abbreviated as PCA) so as to facilitate fault detection and result visualization. Then, a fault detection model is constructed based on an unsupervised clustering algorithm, and a single body with micro short circuit faults in the lithium ion battery pack is detected by taking the extracted main component of the increment capacity as input. And finally, quantitatively calculating the micro-short circuit current and the micro-short circuit resistance of the detected micro-short circuit monomer based on the maximum charging voltage difference between adjacent charging and discharging cycles. The specific implementation process is as follows:
In this embodiment, an IC curve is obtained by using a numerical differentiation method, and the calculation formula is as follows:
where Q represents the battery charge capacity, V is the terminal voltage in the constant current charge mode, ICEVI represents the calculation IC based on the equal voltage interval Δv, and Q2-Q1 represents the charge capacity variation within the voltage interval Δv.
After that, the IC curve is smoothed by moving average filtering. Moving average filtering smoothes the signal by calculating the average of the signal over a predetermined window. Given a time-varying signal s contaminated with noise, the moving average filter may be designed to:
Wherein sr represents the value of the original signal s at the r-th moment,Is the corresponding filtered value, 2Np +1 is the window size of the sliding filter, Np is an integer, l is the time lag number, and sr-l is the value of the original signal at the r-l time.
In this embodiment, in order to facilitate fault detection and visualization, the present invention maps a high-dimensional feature space with linear correlation of features into a two-dimensional feature space by using PCA, and the following describes the process of PCA calculation.
The invention regards the incremental capacity curve extracted from each monomer in one charge-discharge cycle as one sample. For all cells in the battery, the data matrix after incremental capacity extraction is expressed as:
Wherein the method comprises the steps ofThe incremental capacity curve for the i-th monomer extracted at the j-th cycle, i=1, 2. n represents the number of cells in the battery, d represents the number of charge and discharge cycles, and m is the length of the incremental capacity curve extracted for a certain cell during a certain charge and discharge cycle.
First, a covariance matrix of the data matrix C is calculated as shown in the following formula:
Wherein N is the number of rows of the matrix C, equal to N times d, and R is the covariance matrix;
Further, the eigenvalues and eigenvectors of the covariance matrix are obtained by solving the equation (5), so as to determine the load matrix:
R-λIm|=0 (5)
Wherein, Im is a unit array of m dimensions. The eigenvalue λ= (λ1,λ2,…,λm), where λ1≥λ2≥...≥λm, is calculated. The eigenvector matrix p= (P1,p2,...,pm) corresponding to the eigenvalue is the load matrix of C, each column in P is called a load vector, and they are mutually orthogonal.
Thus, the data matrix C can be decomposed into the following forms:
Wherein,For the scoring matrix, each column vector Sh＝Cph of S represents the h-th column, essentially the projection of the data matrix C along the direction of the load vector,Representing the residual, q represents the number of principal components and satisfies q < m. The invention selects the first two principal components in the principal component space to represent the original data matrix.
In this embodiment, a density-based clustering algorithm (DBSCAN for short) is a common and efficient clustering method, and compared with other clustering algorithms, the algorithm is characterized in that the number of super parameters to be set is small, the robustness to cluster shapes is high, the cluster shape is insensitive to noise, and the method is very suitable for density clustering tasks with irregular shapes and without knowing the number of target clusters in advance. Since DBSCAN identifies clusters based on the distribution density of data in feature space, points in lower density regions are typically marked as outliers, which enables the clustering algorithm to identify outliers in the dataset. Therefore, the invention constructs a fault detection model based on the DBSCAN algorithm to realize accurate detection of the micro short circuit monomer in the lithium ion battery pack.
The core idea of DBSCAN is to divide regions with sufficient density into one class. The algorithm has two input parameters, a neighborhood radius epsilon and a density threshold MinPts. All points in the sample set X are divided into core points, density reachable points, density connecting points and extra-local points, and the specific definition is as follows:
If a point k includes itself at least MinPts points within the epsilon range of distance, then that point is referred to as the core point, and points within the epsilon range are considered to be reachable by the direct density of k.
If there is a series of points k1,k2,···,kn, and each ki+1 is directly density reachable by ki, then kn is said to be density reachable by k1.
Given the objects a, k, b e X, if both k and b are reachable from a with respect to epsilon and MinPts densities, then k and b are said to be connected with respect to epsilon and MinPts densities.
All points not reachable by any point density are then called outliers.
The principle of the DBSCAN algorithm is shown in fig. 1. Wherein minpts=5, the black dot is a core point, and there are at least five samples within the range of the neighborhood radius epsilon. All points within the range of the core point epsilon are said to be reachable with the core point direct density. The core points that are not within the same hypersphere are density reachable. A series of core points connected by solid black arrows constitute a sequence of reachable density. All sample points within epsilon of all density reachable sequences are density-connected. The black square is not within epsilon of any core point, i.e. is not reachable by any point density, and is therefore an extra-local point.
The basic idea of the algorithm is that from a certain core point, the algorithm is continuously expanded to a region with reachable density, so that a maximized region containing the core point and the boundary point is obtained, and any two points in the region are connected in density. The method comprises the following specific steps:
(1) Defining a neighborhood radius epsilon and a density threshold MinPts;
(2) Randomly selecting an unaccessed point k, counting the number n (including the point itself) of points with a distance less than epsilon, if n is more than or equal to MinPts, the point k is a core point, creating a new cluster V, placing the points in the neighborhood radius epsilon into the cluster, and marking the points as accessed;
(3) Traversing the points in the k adjacent domains, if the point b in the adjacent domains is a core point, dividing the points in the adjacent domains of the point b into a cluster V, and the like until the V is not expanded any more;
(4) And repeating the step 2 and the step3 until all the points are accessed.
After the fault detection model based on the unsupervised clustering algorithm is constructed, the single body with micro short circuit fault in the lithium ion battery pack is detected by taking the extracted main component of the increment capacity as input. Specifically, normal monomers and micro-short circuit monomers in the lithium ion battery pack are detected and distinguished through observing the clustering result. In the clustering result, the monomer corresponding to the sample point in the high-density area is the normal monomer. In contrast, the cell corresponding to the sample point farther from the high density region represents a micro-short cell. In fig. 1, core points and non-core points located in circles represent samples corresponding to normal single bodies, and on the contrary, the outlier points represent sample points corresponding to micro-short circuit single bodies.
In this embodiment, after detecting the micro-short circuit unit in the battery pack, the invention further estimates the short circuit current and the short circuit resistance thereof to quantitatively evaluate the severity and the evolution stage of the micro-short circuit, which is helpful for the battery management system to take targeted countermeasures. Taking a series battery pack subjected to Constant Current (CC) charge dynamic stress test (DYNAMIC STRESS TEST, abbreviated as DST) discharge as an example, a voltage curve diagram of the series battery pack in two consecutive charge-discharge cycles is shown in fig. 2 (a). Wherein CCV represents a charge cutoff voltage, a solid black line represents a terminal voltage curve of a normal cell in the battery pack, and a broken black line represents a terminal voltage curve of a cell in which a micro short circuit occurs. As shown in fig. 2 (a), the micro-short cell needs to discharge the short resistor, i.e., the short resistor continuously consumes the energy of the micro-short cell, in both the charging phase and the discharging phase, which results in the terminal voltage of the micro-short cell being lower than that of the normal cell.
Further, the charge voltage curves of the micro short circuit monomers in the adjacent two charge-discharge cycles were individually observed, as shown in fig. 2 (b). As can be seen from the figure, the maximum charge voltage (MCV 2) of the micro-short circuit cell in the last charge-discharge Cycle (i.e. the 2 nd charge-discharge Cycle, denoted as Cycle 2) is significantly smaller than the maximum charge voltage (MCV 1) in the previous charge-discharge Cycle (i.e. the 1 st charge-discharge Cycle, denoted as Cycle 1). This maximum charge voltage difference between adjacent cycles is also due to the additional consumption of energy by the micro-shorting cells by the shorting resistors. It is assumed that at the end of the 2 nd charge of the battery pack, the micro-short cell is taken out from the battery pack and charged alone so that its maximum charge voltage is the same as that at the end of the first charge cycle, as shown by the solid black line in fig. 2 (b), and the charge time Δt required at this time is defined as the remaining charge time of the micro-short cell.
In the actual operation process of the energy storage lithium ion battery pack, the micro-short circuit monomer cannot be taken out independently to be charged so as to obtain the residual charging time delta t. The remaining charging time Δt is therefore theoretically unknown. However, it can be noted that if the maximum charge voltage MCV 2 of the next cycle is interpolated on the charge voltage curve of the previous cycle and the corresponding time is noted as t, the time interval from the time t to the charge end time in the previous cycle is equal to the remaining charge time Δt. Therefore, the remaining charging time Δt of the micro-short circuit monomer in the current cycle can be obtained by interpolating the maximum charging voltage of the current cycle on the charging voltage curve of the previous cycle.
After the remaining charging time delta t of the micro-short circuit monomer in the current charging and discharging cycle is obtained, the electric quantity loss of the micro-short circuit monomer in the current charging and discharging cycle can be further obtained by integrating the residual charging time delta t with the charging current of the battery pack, wherein the electric quantity loss is caused by a short circuit resistor, namely:
Qloss,i=I·Δt (7)
Wherein, Qloss,i is the power loss of the micro-short circuit monomer at the end of I times of charging, I represents the charging current. The change rate of the electric quantity loss along with time is the short-circuit current of the micro-short-circuit monomer, so that the micro-short-circuit monomer has
Wherein IMSC is micro-short circuit current, Qloss,j is estimated electric quantity loss at the j-th charging end of the micro-short circuit single body, Qloss,j-1 is estimated electric quantity loss at the j-1 th charging end of the micro-short circuit single body, Tj is time at the j-th charging end, Tj-1 is time at the j-1 th charging end, and based on the obtained micro-short circuit current, the micro-short circuit resistance value can be further obtained by the following formula. The smaller the short-circuit resistance, the greater the battery short-circuit heat generation power and the greater the likelihood of thermal runaway.
Wherein RMSC represents a micro-short circuit resistor, and UM is an average voltage between two adjacent cycle charging end times.
Example two
The invention also provides a lithium ion battery micro-short circuit diagnosis system based on clustering and charging voltage difference, which comprises a processing module, a mapping module, a detection module and a calculation module;
The processing module is used for extracting the increment capacity IC from the constant current charging stage, and smoothing the increment capacity IC by utilizing moving average filtering to obtain high-dimensional characteristics capable of effectively representing the micro short circuit fault of the lithium ion battery;
The mapping module is used for utilizing principal component analysis PCA to map the extracted high-dimensional characteristics capable of effectively representing the micro-short circuit fault of the lithium ion battery to a two-dimensional plane so as to obtain principal components of increment capacity;
The detection module is used for constructing a fault detection model based on an unsupervised clustering algorithm, taking the extracted main components of the increment capacity as input, and detecting a monomer with micro short circuit fault in the lithium ion battery pack;
The calculation module is used for quantitatively calculating the micro-short circuit current and the micro-short circuit resistance of the single body with the micro-short circuit fault in the lithium ion battery pack based on the maximum charging voltage difference between the adjacent charging and discharging cycles.
In the present embodiment, the process of extracting the incremental capacity IC includes:
where Q represents the battery charge capacity, V is the terminal voltage in the constant current charge mode, ICEVI represents the calculation IC based on the equal voltage interval Δv, and Q2-Q1 represents the charge capacity variation within the voltage interval Δv.
In the present embodiment, the process of smoothing the delta capacity IC using moving average filtering includes:
Wherein sr represents the value of the original signal s at the r-th moment,Is the corresponding filtered value, 2Np +1 is the window size of the sliding filter, Np is an integer, l is the time lag number, and sr-l is the value of the original signal at the r-l time.
In this embodiment, the process of mapping the extracted high-dimensional features capable of effectively characterizing the micro-short fault of the lithium ion battery to the two-dimensional plane by using the principal component analysis PCA includes:
For all cells within the battery, the data matrix after incremental capacity extraction is expressed as:
Wherein,The method comprises the steps of representing an increment capacity curve extracted by an ith monomer in a jth cycle, i=1, 2, & n, j=1, 2, & d, n represents the number of the monomers in the battery pack, d represents the number of charge and discharge cycles, and m is the length of the increment capacity curve extracted by a corresponding monomer in a certain charge and discharge cycle;
The covariance matrix of the data matrix C is calculated as follows:
Wherein N is the number of rows of the matrix C, equal to N times d, and R is the covariance matrix;
Calculating eigenvalues and eigenvectors of the covariance matrix to determine a load matrix of the data matrix C, wherein the eigenvalue lambda= (lambda1,λ2,…,λm) is calculated, lambda1≥λ2≥...≥λm, the eigenvector matrix P= (P1,p2,...,pm) corresponding to the eigenvalue is the load matrix of the data matrix C, each column in P is called as a load vector, and the load vectors are mutually orthogonal;
based on the load matrix of the data matrix C, the data matrix C is decomposed into the following form:
Wherein,For the scoring matrix, each column vector Sh＝Cph of S represents the h-th column, essentially the projection of the data matrix C along the direction of the load vector,Representing the residual, q represents the number of principal components and satisfies q < m.
In this embodiment, the process of calculating the micro-short circuit current of the cell having the micro-short circuit fault in the lithium ion battery pack includes:
Wherein, IMSC is micro short-circuit current, Qloss,j is estimated electric quantity loss at the j-th charging end of the micro short-circuit single body, Qloss,j-1 is estimated electric quantity loss at the j-1 th charging end of the micro short-circuit single body, Tj is the moment at the j-th charging end, and Tj-1 is the moment at the j-1 th charging end;
The process for calculating the micro-short circuit resistance of the single body with the micro-short circuit fault in the lithium ion battery pack comprises the following steps:
Wherein RMSC represents a micro-short circuit resistor, and UM is an average voltage between two adjacent cycle charging end times.
Example III
The invention tests a battery pack formed by connecting 8 cylindrical lithium ion battery cells in series, and verifies the effectiveness of the technology. The monomer specification parameters are shown in table 1.
TABLE 1 specification parameters of lithium ion batteries
And carrying out constant-current charging of 0.5C on the battery pack, and stopping charging the battery pack when the maximum single terminal voltage reaches the charging cut-off voltage of 4.2V so as to prevent the single battery from being overcharged. The battery is then discharged for dynamic stress testing. When the minimum cell terminal voltage reaches a discharge cut-off voltage of 2.75V, the battery pack stops discharging to prevent overdischarge of the cells. A total of 15 charge and discharge cycles were performed on the battery pack.
The invention adopts a mode of connecting external resistors in parallel at two ends of the battery to simulate micro short circuit faults. The method can well control the triggering time and the triggering position of the micro short circuit, simulate the evolution process of the micro short circuit, and has good controllability and repeatability.
Specifically, in order to simulate the evolution process of micro-short circuit, short-circuit resistors with different magnitudes and corresponding manual switches are connected in series and then connected to two ends of the single body 4 and the single body 8 in parallel. By closing different switches, short-circuit resistors with different magnitudes can be connected in parallel to two ends of the battery so as to simulate the occurrence and evolution of micro-short circuit. The smaller the short-circuit resistance, the larger the short-circuit current, indicating that the short-circuit is more severe. The parallel short-circuit resistances corresponding to the respective charge and discharge cycles are shown in table 2.
TABLE 2 short-circuit resistance corresponding to different charge-discharge cycles
The micro-short detection result based on the PCA and DBSCAN clustering algorithm is shown in fig. 3, wherein black dots represent normal samples, and black cross symbols represent detected fault samples. As described above, the present invention regards the features extracted from each monomer in one charge-discharge cycle as one sample. Thus, for the battery used for the experiment, 8 cells output 120 samples in total for 15 charge-discharge cycles. Each sample is numbered v_i_j, corresponding to the characteristics of the ith cell at the jth charge-discharge cycle, where i=1, 2. Meanwhile, in order to more clearly show the fault detection result, fig. 3 only marks the sample numbers of the detected fault samples, and does not show the sample numbers of the normal samples with a larger number.
As shown in fig. 3, the normal samples are all concentrated in one high density region. More importantly, at the 4 th to 15 th charge-discharge cycles, most of the samples from cell 4 and cell 8 were diagnosed as faulty samples, and as the equivalent short-circuit resistance across the battery decreases, the distance between the faulty samples and the high-density region where the normal samples are located becomes longer and longer. In general, the detection accuracy (the number of correctly detected samples divided by the total number of samples) of the micro-short circuit detection method provided by the invention is up to 99.17%. In a word, the fault detection result is basically consistent with the fault setting, which proves that the method provided by the invention can accurately detect the micro-short circuit monomer in the lithium ion battery pack, and proves the effectiveness of the method.
After detecting the micro-short circuit unit in the battery pack, the method provided by the invention is further utilized to estimate the short circuit resistance of the battery pack so as to quantify the severity and evolution stage of the micro-short circuit. The actual values, estimates and relative errors of the shorting resistances of the micro-shorting cells 4 and 8 are shown in Table 3. As shown in table 3, the maximum relative error of the short circuit resistance estimation result was 3.85% for the micro short circuit cell 4, and 4.26% for the micro short circuit cell 8. Furthermore, it can be seen from table 3 that the relative error of the estimation results shows a decreasing trend with decreasing real short-circuit resistance across the battery, because the fault signature caused by micro-short-circuits becomes more and more pronounced with decreasing short-circuit resistance. The estimation error of the short circuit resistance is in an acceptable range, and the effectiveness of the quantitative evaluation method of the micro-short circuit provided by the invention is proved.
TABLE 3 estimation results of micro-shorting resistance
The above embodiments are merely illustrative of the preferred embodiments of the present invention, and the scope of the present invention is not limited thereto, but various modifications and improvements made by those skilled in the art to which the present invention pertains are made without departing from the spirit of the present invention, and all modifications and improvements fall within the scope of the present invention as defined in the appended claims.