Lithium ion battery fault detection method and system based on operation data

Abstract

The invention discloses a lithium ion battery fault detection method and system based on operation data. The method comprises the following steps: step 1, preprocessing a lithium ion battery voltage signal acquired by a battery management system in an actual operation condition; 2, constructing a dimensionless feature set based on the preprocessed charging section battery voltage and discharging section battery voltage; 3, optimal dimensionless feature selection: calculating Laplacian scores of the constructed dimensionless features for the voltage abnormity of the charging section and the voltage abnormity of the discharging section of the battery, and defining the dimensionless feature with the highest score as the optimal dimensionless feature; 4, based on the charging voltage data of each battery in each charging process, extracting an extended average voltage feature, and further extracting a waveform factor from the extended average voltage feature by adopting a sliding window; and 5, performing standardized difference processing on the selected optimal dimensionless features and the extracted waveform factors, performing short-time scale real-time fault detection on the optimal dimensionless features after standardized difference by adopting a clustering algorithm, and performing long-time scale accumulative fault detection on the waveform factors after standardized difference by adopting a clustering algorithm.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithium ion battery fault diagnosis, and in particular to a lithium ion battery fault detection method and system based on operation data. Background Art

[0002] Lithium-ion batteries are widely used in transportation and grid energy storage due to their excellent energy density and long service life. However, in recent years, thermal runaway accidents have occurred frequently in high-proportion lithium-ion battery systems in electric vehicles and energy storage facilities. Most of these accidents can be traced back to the batteries themselves, and the cause may be defects in material distribution and structural design during battery manufacturing. Extreme conditions (such as thermal, electrical and mechanical abuse) can change the internal structure of the battery, trigger battery failure, and ultimately lead to thermal runaway, which greatly hinders the further development of lithium-ion batteries. There are many reasons for battery failure, including internal short circuit, external short circuit, overcharging and overheating. Different battery failures have different manifestations, so accurately identifying the cause of battery thermal runaway is crucial to ensure the safe and stable operation of lithium-ion battery systems.

[0003] Depending on whether the original signal is used to establish the battery physical model, the existing battery fault diagnosis methods can be mainly divided into two categories: model-based methods and data-driven methods. Model-based methods construct battery models, which include equivalent circuit models, electrochemical models, and thermal models, etc., and combine battery physical parameters (such as terminal voltage, resistance, and temperature) to simulate electrodynamics and thermodynamics. Battery system anomalies are determined by the residuals between the model estimated parameter values ​​and the observed parameter values. Different models are usually used to diagnose different faults. Although model-based methods can accurately describe the evolution of batteries from normal to faulty, the diagnostic accuracy of these methods depends largely on model accuracy and accurate residual thresholds. The complexity and disorder of actual operating conditions increase the difficulty of obtaining correct model parameters, which often leads to poor robustness of model-based methods.

[0004] Data-driven methods do not require a high-precision battery physical model and can directly analyze the battery signal in the real curve to achieve fault detection. These methods do not consider the complex fault mechanism of the battery, but identify the faulty battery by extracting fault-sensitive features from the voltage and temperature signals, which is only a mathematical processing of the original signal rather than modeling the internal mechanism of the battery. Popular data-driven methods can be divided into two categories: signal processing and machine learning. Some signal processing-based fault diagnosis methods use signal decomposition algorithms such as variation mode decomposition, empirical mode decomposition, and wavelet transform to decompose the voltage signal under dynamic conditions into components that are easy to extract fault features. Other methods directly extract fault features from the original voltage signal. The extracted fault features include similarity function entropy and improved entropy as well as distance function. The features extracted in the existing literature can detect one or several voltage abnormality patterns on their respective data sets, but the diagnosed patterns are not comprehensive. In terms of fault isolation, both threshold-based and clustering-based methods have been applied. However, most of these methods are based on fixed thresholds or manually set clustering algorithm parameters, resulting in weak generalization capabilities on various fault data sets. Machine learning methods, including neural networks, autoencoders, etc., are applied to achieve multi-fault identification through classification or sequence prediction. Machine learning methods require a large amount of fault data with accurate fault labels, but due to the lack of labeled fault data in real working conditions for sufficient training, these methods cannot be widely used in practice. In addition, these methods also have the problem of poor interpretability because they cannot explain the process of obtaining the detection results. For faulty batteries, the reason for the difference between the reconstructed voltage curve and the original voltage curve remains unclear, so the potential cause of the battery failure cannot be analyzed.

[0005] In summary, the current battery fault detection methods have the problems of insufficient sensitivity of fault characteristics resulting in low fault identification accuracy, and artificially set fault thresholds resulting in a high false alarm rate, making it difficult to accurately locate the faulty cell. Summary of the invention

[0006] In view of the problems and defects described in the above-mentioned prior art, the present invention provides a lithium-ion battery fault detection method and system based on operating data, which analyzes the historical operating data of the battery system within a certain period of time without interfering with the operation of the lithium-ion battery system, constructs fault characteristic parameters and performs feature screening, and finally identifies the faulty cells in the battery pack based on the clustering algorithm, promptly discovers the single cells with potential abnormalities, and provides theoretical guidance for the maintenance of the battery system.

[0007] In a first aspect, the present invention adopts the following technical solution: a lithium-ion battery fault detection method based on operation data, comprising: Preprocessing the voltage signal of the lithium-ion battery in actual operating conditions; wherein the voltage signal of the lithium-ion battery can be collected through a battery management system or other means; Based on the preprocessed battery voltage in the charging stage and the battery voltage in the discharging stage, a dimensionless feature set is constructed; Select the optimal dimensionless feature: Calculate the Laplace scores of the constructed dimensionless feature for the battery charging stage voltage anomaly and the battery discharging stage voltage anomaly, and define the dimensionless feature with the highest score as the optimal dimensionless feature; Based on the charging voltage data of each battery in each charging process, the extended average voltage feature is extracted, and the waveform factor is further extracted from the extended average voltage feature using a sliding window; The selected optimal dimensionless features and the extracted waveform factors are subjected to standardized difference processing, and a clustering algorithm is used for the optimal dimensionless features after standardized difference to realize real-time fault detection of batteries on a short time scale, and a clustering algorithm is used for the waveform factors after standardized difference to realize cumulative fault detection of batteries on a long time scale.

[0008] On the basis of the above technical solution, the preprocessing of the lithium-ion battery voltage signal in the actual operating condition includes data cleaning and charge and discharge segment division, and the method is as follows: Data cleaning: delete missing data in the collected battery voltage signal; Charge and discharge segment division: Based on the battery system charging status flag, the charging segment voltage data is obtained, and the other data is used as the discharge segment voltage data.

[0009] On the basis of the above technical solution, the method for constructing the dimensionless feature is as follows: Set the window width to w , for each battery in the window w The dimensionless characteristic is constructed by using the voltage value. The calculation formula of the dimensionless characteristic of battery voltage is shown in formula (1): ; in, z ( v ) represents voltage v The function of p ( v )represent v The probability density distribution of and m is a positive integer; In the formula, voltage v Function z ( v )include: ; in, μ For each battery in the windoww The average value of the voltage.

[0010] Based on the above technical solution, the Laplace score is calculated as follows: use w Dimensionless feature samples to build a nearest neighbor graph G , No. Nodes correspond to ,like and is connected, then construct a weighted matrix S : ; In the formula, is the weight matrix S Each element in t is a constant, is the Euclidean distance; For r Features, definition , D is a diagonal matrix; In order to avoid the phenomenon that some dimensional data have great differences and dominate the construction of the nearest neighbor graph, each feature is de-meaned, and formula (4) is obtained: ; Definition r The Laplace score of features: ; The smaller the Laplace fraction, the more sensitive this feature is to faults.

[0011] Based on the above technical solution, the calculation method of the extended average voltage characteristic is as follows: Calculate the voltage of num cells in the battery pack at time during each charging process t The average value of: ; In the formula, For the moment t The average battery pack voltage, is the battery cell c at time t Voltage; Select each charging process The time corresponding to the fixed interval in t 1 and t 2. Calculation t 1 to t The extended average voltage of each monomer at time 2 is calculated as shown in formula (7): ; Among them, EAV is the extended average voltage of a single cell during a charging process. v t For a single t The voltage at the moment, For the time interval.

[0012] On the basis of the above technical solution, the method of further extracting the waveform factor from the extended average voltage feature using a sliding window is as follows: choose w One charging segment is used as a window, and each battery in the window is w 1 Expanded average voltage value to calculate the waveform factor. The waveform factor calculation formula is shown in (8): ; Among them, EVA i For a single unit i The extended average voltage of the charging segment.

[0013] On the basis of the above technical solution, the standardized difference processing method of the optimal dimensionless characteristics and waveform factors is as follows: The optimal dimensionless characteristics and waveform factors are differentiated, and the calculation formula is shown in formula (9): ; in, is the optimal dimensionless characteristic or waveform factor after differentiation, is the optimal dimensionless characteristic or shape factor of the current window, is the optimal dimensionless characteristic or shape factor of the previous window; Afterwards, Z The score normalization method is calculated as shown in formula (10): ; in, is the optimal dimensionless characteristic or waveform factor after normalized difference, is the optimal dimensionless characteristic or waveform factor after differentiation, is the average value of the differential optimal dimensionless characteristics or form factors of all cells in the battery pack in the same window, It is the standard deviation of the differential optimal dimensionless characteristic or form factor of all cells in the battery pack in the same window.

[0014] On the basis of the above technical solution, the clustering algorithm is an adaptive DBSCAN clustering algorithm, and the specific steps are as follows: Divide the grid based on the boundary of the standardized differential feature data set, and obtain the peak density points of the grid by statistics; Calculate the distance from each feature point to the peak density point ； Calculate the distance matrix from each feature point to other feature points Dist , the longest distance in the distance matrix The two feature points are marked as x and y, and the maximum distance from x to a point other than y is defined as ;like Greater than , then the parameter neighborhood radius equal ,otherwise equal The maximum value in ; Set the parameter MinPts to 1 and find each feature The number of points in the neighborhood. If the number of points is greater than or equal to MinPts, then the point is a core point; Start from the core point. All points in the neighborhood are classified into the same cluster, and points that do not belong to any cluster are marked as feature points of the faulty monomer, thus realizing lithium-ion battery fault detection.

[0015] In a second aspect, a lithium-ion battery fault detection system based on operation data is provided, comprising: A preprocessing module is used to preprocess the voltage signal of the lithium-ion battery in actual operating conditions; A feature set construction module constructs a dimensionless feature set based on the preprocessed battery voltage in the charging stage and the battery voltage in the discharging stage; Feature selection module, selects the optimal dimensionless feature: calculates the Laplace scores of the constructed dimensionless features for the battery charging stage voltage anomaly and the discharge stage voltage anomaly, and defines the dimensionless feature with the highest score as the optimal dimensionless feature; The waveform factor extraction module extracts the extended average voltage feature based on the charging voltage data of each battery in each charging process, and further extracts the waveform factor from the extended average voltage feature using a sliding window; The fault detection module performs standardized difference processing on the selected optimal dimensionless features and the extracted waveform factors, and uses a clustering algorithm to realize real-time fault detection of batteries on a short time scale for the optimal dimensionless features after standardized difference, and uses a clustering algorithm to realize cumulative fault detection of batteries on a long time scale for the waveform factors after standardized difference.

[0016] On the basis of the above technical solution, the preprocessing module preprocesses the voltage signal of the lithium-ion battery in the actual operating condition, including data cleaning and charge and discharge segment division, and the method is as follows: Data cleaning: delete missing data in the collected battery voltage signal; Charge and discharge segment division: Based on the battery system charging status flag, the charging segment voltage data is obtained, and the other data is used as the discharge segment voltage data.

[0017] On the basis of the above technical solution, the method for constructing dimensionless features by the feature set construction module is as follows: Set the window width to w , for each battery in the window w The dimensionless characteristic is constructed by using the voltage value. The calculation formula of the dimensionless characteristic of battery voltage is shown in formula (1): ; in, z ( v ) represents voltage v The function of p ( v )represent v The probability density distribution of and m is a positive integer; In the formula, voltage v Function z ( v )include: ; in, μ For each battery in the window w The average value of the voltage.

[0018] Based on the above technical solution, the feature selection module calculates the Laplace score as follows: use w Dimensionless feature samples to build a nearest neighbor graph G , No. Nodes correspond to x i ,like x i and x j is connected, then construct a weighted matrix S : ; In the formula, is the weight matrix S Each element in t is a constant, is the Euclidean distance; For r Features, definition , D is a diagonal matrix; In order to avoid the phenomenon that some dimensional data have great differences and dominate the construction of the nearest neighbor graph, each feature is de-meaned, and formula (4) is obtained: ; Definition r The Laplace score of features: ; The smaller the Laplace fraction, the more sensitive this feature is to faults.

[0019] On the basis of the above technical solution, the calculation method of the extended average voltage feature by the waveform factor extraction module is as follows: Calculate the voltage of num cells in the battery pack at time during each charging process t The average value of: ; In the formula, For the moment t The average battery pack voltage, is the battery cell c at time t Voltage; Select each charging process The time corresponding to the fixed interval in t 1 and t 2. Calculation t 1 to t The extended average voltage of each monomer at time 2 is calculated as shown in formula (7): ; Among them, EAV is the extended average voltage of a single cell during a charging process. v t For a single t The voltage at the moment, For the time interval.

[0020] On the basis of the above technical solution, the waveform factor extraction module uses a sliding window to further extract the waveform factor from the extended average voltage feature as follows: choose w One charging segment is used as a window, and each battery in the window is w 1 Expanded average voltage value to calculate the waveform factor. The waveform factor calculation formula is shown in (8): ; in, For a single unit i The extended average voltage of the charging segment.

[0021] On the basis of the above technical solution, the fault detection module performs standardized differential processing on the optimal dimensionless features and waveform factors as follows: The optimal dimensionless characteristics and waveform factors are differentiated, and the calculation formula is shown in formula (9): ; in, is the optimal dimensionless characteristic or waveform factor after differentiation, The optimal dimensionless characteristic or shape factor for the current window, is the optimal dimensionless characteristic or shape factor of the previous window; Afterwards, Z The score normalization method is calculated as shown in formula (10): ; in, is the optimal dimensionless characteristic or waveform factor after normalized difference, is the optimal dimensionless characteristic or waveform factor after differentiation, is the average value of the differential optimal dimensionless characteristics or form factors of all cells in the battery pack in the same window, It is the standard deviation of the differential optimal dimensionless characteristic or form factor of all cells in the battery pack in the same window.

[0022] On the basis of the above technical solution, the clustering algorithm adopted by the fault detection module is an adaptive DBSCAN clustering algorithm, and the specific steps are as follows: Divide the grid based on the boundary of the standardized differential feature data set, and obtain the peak density points of the grid by statistics; Calculate the distance from each feature point to the peak density point ； Calculate the distance matrix from each feature point to other feature points Dist , the longest distance in the distance matrix The two feature points are marked as x and y, and the maximum distance from x to a point other than y is defined as ;like Greater than , then the parameter neighborhood radius equal ,otherwise equal The maximum value in ; Set the parameter MinPts to 1 and find each feature ε The number of points in the neighborhood. If the number of points is greater than or equal to MinPts, then the point is a core point; Start from the core point. εAll points in the neighborhood are classified into the same cluster, and points that do not belong to any cluster are marked as feature points of the faulty monomer, thus realizing lithium-ion battery fault detection.

[0023] In a third aspect, an electronic device is provided, comprising a memory, a processor, a computer program stored in the memory and executable on the processor, and a database, wherein when the processor executes the program, any one of the lithium-ion battery fault detection methods based on operating data is implemented.

[0024] The present invention has the following beneficial technical effects: The present invention can perform fault detection on lithium-ion battery packs based on operating data. According to the method of the present invention, a dimensionless feature set for battery voltage signals in the charging and discharging stages is analyzed, and the dimensionless feature set is screened using Laplace fractions. Then, the extended average voltage is extracted from the battery charging voltage data, and the waveform factor features are further extracted from the extended average voltage using a sliding window. The dimensionless features and waveform factor features after screening are subjected to standardized differential processing, and a clustering algorithm is used to identify faulty battery cells, providing an important theoretical basis for the maintenance and replacement of battery systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the process of the method of the present invention; Figure 2 Schematic diagram of dimensionless characteristic Laplace fractions of the charging section and the discharging section of an embodiment of the present invention; Figure 3 This is a schematic diagram of an extended average voltage curve in a voltage range of 3.9-4.1V according to an embodiment of the present invention; Figure 4 Schematic diagram of a waveform factor curve with a window width of 6 according to an embodiment of the present invention; Figure 5 It is a schematic diagram of the standardized differential optimal dimensionless characteristic curve of the charging stage according to an embodiment of the present invention; Figure 6 It is a schematic diagram of the standardized differential optimal dimensionless characteristic curve of the discharge section according to an embodiment of the present invention; Figure 7 FIG. 4 is a schematic diagram of a normalized differential form factor curve according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0027] like Figure 1 As shown, a lithium-ion battery fault detection method based on operating data and a system flow chart are shown, which are performed in the following steps: Step 1: Preprocess the lithium-ion battery voltage signal collected by the battery management system in actual operating conditions.

[0028] Delete the missing data in the collected battery voltage signal, obtain the charging segment voltage data based on the battery system charging status flag, and use the other data as the discharging segment voltage data.

[0029] Step 2: Construct a dimensionless feature set based on the preprocessed charging segment battery voltage and discharging segment battery voltage.

[0030] Set the window width to w , for each battery in the window w The dimensionless characteristic is constructed by using the voltage value. The calculation formula of the dimensionless characteristic of battery voltage is shown in formula (1): ; in, z ( v ) represents voltage v The function of p ( v )represent v The probability density distribution of and m is a positive integer. In the formula, the voltage v Function z ( v )include: ; in, μ For each battery in the window w The average value of the voltage.

[0031] In this embodiment, the window width w is 30, for the function z ( v ) = v and z ( v ) = | v – μ |, yes l and m Assign all 10 positive integer values ​​in the interval [1, 10] to obtain 200 eigenvalues; for the function z ( v ) = v – μ ,right l Assign all 10 positive integer values ​​in the interval [1, 10] to m Assigning five values ​​{2, 4, 6, 8, 10}, 50 dimensionless features can be obtained, and these 250 features are constructed for the battery charging voltage and discharging voltage respectively.p ( v ) = 1 / 30.

[0032] Step 3: Optimal dimensionless feature selection: Calculate the Laplace scores of the constructed dimensionless features for the battery charging voltage anomaly and the discharging voltage anomaly, and define the dimensionless feature with the highest score as the optimal dimensionless feature.

[0033] use w Dimensionless feature samples to build a nearest neighbor graph G , No. i Nodes correspond to x i ,like x i and x j is connected, then construct a weighted matrix S : ; In the formula, S ij is the weight matrix S Each element in t is a constant, is the Euclidean distance.

[0034] For r Features, definition , D is a diagonal matrix. In order to avoid the phenomenon that some dimensional data have great differences and dominate the construction of the neighbor graph, each feature is de-meaned, and formula (4) is obtained. ; Definition r The Laplace score of features: ; The smaller the Laplace fraction, the more sensitive this feature is to faults.

[0035] In this embodiment, w = 30. Figure 2 The Laplace score results of the dimensionless characteristics of the charging and discharging stages are shown. The optimal dimensionless characteristics of the battery charging stage are selected as ; The optimal dimensionless characteristic of the battery discharge segment is selected as ; Step 4: Based on the charging voltage data of each battery in each charging process, the extended average voltage feature is extracted, and the waveform factor is further extracted from the extended average voltage feature using a sliding window.

[0036] Calculate the voltage of num cells in the battery pack at time during each charging process t The average value of: ; In the formula, For the moment t The average battery pack voltage, is the battery cell c at time t voltage.

[0037] Select each charging process The time corresponding to the fixed interval in t 1 and t 2. Calculation t 1 to t The extended average voltage of each monomer at time 2 is calculated as shown in formula (9): ; Among them, EAV is the extended average voltage of a single cell during a charging process. v t For a single t The voltage at the moment, For the time interval.

[0038] choose w One charging segment is used as a window, and each battery in the window is w 1 Expanded average voltage value to calculate the waveform factor. The waveform factor calculation formula is shown in (10): ; in, For a single unit i The extended average voltage of the charging segment.

[0039] In this embodiment, the selected The fixed voltage range is 3.9V-4.1V, and the result is as follows Figure 3 The expanded average voltage results are shown. w 1 = 6, so Figure 4 The form factor results are shown.

[0040] Step 5: Perform standardized difference processing on the selected optimal dimensionless features and the extracted waveform factors, and use a clustering algorithm for the optimal dimensionless features after standardized difference to realize real-time fault detection of batteries on a short time scale, and use a clustering algorithm for the waveform factors after standardized difference to realize cumulative fault detection of batteries on a long time scale.

[0041] The optimal dimensionless characteristics and waveform factors are differentiated, and the calculation formula is shown in formula (11): ; in, is the optimal dimensionless characteristic or waveform factor after differentiation, is the optimal dimensionless characteristic or shape factor of the current window, is the optimal dimensionless characteristic or shape factor of the previous window.

[0042] Afterwards, Z The score normalization method is calculated as shown in formula (12): ; in, is the optimal dimensionless characteristic or waveform factor after normalized difference, is the optimal dimensionless characteristic or waveform factor after differentiation, is the average value of the differential optimal dimensionless characteristics or form factors of all cells in the battery pack in the same window, It is the standard deviation of the differential optimal dimensionless characteristic or form factor of all cells in the battery pack in the same window.

[0043] Divide the grid based on the boundary of the standardized differential feature data set, obtain the peak density points of the grid by statistics, and calculate the distance from each feature point to the peak density point , calculate the distance matrix from each feature point to other feature points Dist , the longest distance in the distance matrix The two feature points are marked as x and y, and the maximum distance from x to a point other than y is defined as ,like Greater than , then the parameter neighborhood radius ε equal , otherwise ε is equal to The value of the parameter MinPts is 1, and each feature is found ε The number of points in the neighborhood. If the number of points is greater than or equal to MinPts, then the point is a core point. Starting from the core point, ε All points in the neighborhood are classified into the same cluster, and those points that do not belong to any cluster are marked as feature points of the faulty cell, thus realizing lithium-ion battery fault detection.

[0044] In this embodiment, the optimal dimensionless characteristics of the charging stage, the optimal dimensionless characteristics of the discharging stage, and the waveform factor after standardized difference are as follows: Figure 5 , Figure 6 and Figure 7 shown.

[0045] In this embodiment, adaptive DBSCAN clustering is performed on the standardized differential optimal dimensionless features with a window width of 30, and battery No. 81 is detected as a real-time fault in the short-time scale charging segment, and battery No. 27 is detected as a real-time fault in the short-time scale discharging segment.

[0046] In this embodiment, adaptive DBSCAN clustering is performed on the standardized differential waveform factor with a voltage range of 3.9-4.1V and a window width of 6, and the No. 52 battery is detected as a long-time scale cumulative fault.

[0047] It should be pointed out that those skilled in the art can understand that other clustering algorithms (such as K-means, etc.) can be used in step 5 to achieve the purpose of the invention.

[0048] In addition, the lithium-ion battery fault detection system based on operation data provided by the present invention includes: A preprocessing module is used to preprocess the voltage signal of the lithium-ion battery in actual operating conditions; A feature set construction module constructs a dimensionless feature set based on the preprocessed battery voltage in the charging stage and the battery voltage in the discharging stage; Feature selection module, selects the optimal dimensionless feature: calculates the Laplace scores of the constructed dimensionless features for the battery charging stage voltage anomaly and the discharge stage voltage anomaly, and defines the dimensionless feature with the highest score as the optimal dimensionless feature; The waveform factor extraction module extracts the extended average voltage feature based on the charging voltage data of each battery in each charging process, and further extracts the waveform factor from the extended average voltage feature using a sliding window; The fault detection module performs standardized differential processing on the selected optimal dimensionless features and the extracted waveform factors, uses a clustering algorithm for the standardized differential optimal dimensionless features to realize real-time fault detection of the battery in a short time scale, and uses a clustering algorithm for the standardized differential waveform factors to realize cumulative fault detection of the battery in a long time scale. The lithium-ion battery fault detection system based on the knowledge graph described below and the lithium-ion battery fault detection method based on the knowledge graph described above can be referred to each other.

[0049] Those skilled in the art can understand that the lithium-ion battery fault detection system based on operation data and the lithium-ion battery fault detection method based on operation data can refer to each other.

[0050] It should be noted that any process or method description in the embodiments may be understood as representing a module, fragment or portion of a code including one or more executable instructions for implementing steps of a specific logical function or process, and that the scope of the preferred embodiments of the present invention includes alternative implementations in which the functions may not be performed in the order shown or discussed, including performing the functions in a substantially simultaneous manner or in the reverse order according to the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0051] It should be noted that the logic and / or steps in the embodiments, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or equipment (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or equipment and execute instructions), or in combination with these instruction execution systems, devices or equipment. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or equipment, or in combination with these instruction execution systems, devices or equipment. More specific examples of computer-readable media (non-exhaustive list) include the following: an electrical connection with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0052] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0053] A person skilled in the art may understand that all or part of the steps of implementing the above-mentioned embodiment method may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one of the steps of the method embodiment or a combination thereof.

[0054] In addition, each functional module in each embodiment of the present invention may be integrated into a processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0055] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0056] The above embodiments describe the technical solutions of the present invention in detail. Obviously, the present invention is not limited to the described embodiments. Based on the embodiments of the present invention, people familiar with the technical field can also make various changes accordingly, but any changes that are equivalent or similar to the present invention belong to the scope of protection of the present invention.

[0057] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

Claims

1. A lithium-ion battery fault detection method based on operating data, characterized in that: The steps include: Preprocessing of lithium-ion battery voltage signals in actual operating conditions; Based on the preprocessed battery voltage in the charging stage and the battery voltage in the discharging stage, a dimensionless feature set is constructed; Select the optimal dimensionless feature: Calculate the Laplace scores of the constructed dimensionless feature for the battery charging stage voltage anomaly and the battery discharging stage voltage anomaly, and define the dimensionless feature with the highest score as the optimal dimensionless feature; Based on the charging voltage data of each battery in each charging process, the extended average voltage feature is extracted, and the waveform factor is further extracted from the extended average voltage feature using a sliding window; The selected optimal dimensionless features and the extracted waveform factors are subjected to standardized difference processing, and a clustering algorithm is used for the optimal dimensionless features after standardized difference to realize real-time fault detection of batteries on a short time scale, and a clustering algorithm is used for the waveform factors after standardized difference to realize cumulative fault detection of batteries on a long time scale.

2. The lithium-ion battery fault detection method based on operating data according to claim 1, characterized in that: The preprocessing of the lithium-ion battery voltage signal in the actual operating condition includes data cleaning and charge and discharge segment division, and the method is as follows: Data cleaning: delete missing data in the collected battery voltage signal; Charge and discharge segment division: Based on the battery system charging status flag, the charging segment voltage data is obtained, and the other data is used as the discharge segment voltage data.

3. The lithium-ion battery fault detection method based on operating data according to claim 1, characterized in that: The method for constructing the dimensionless feature is as follows: Set the window width to w , for each battery in the window w The dimensionless characteristic is constructed by using the voltage value. The calculation formula of the dimensionless characteristic of battery voltage is shown in formula (1): ; in, z ( v ) represents voltage v The function of p ( v )represent v The probability density distribution of and m is a positive integer; In the formula, voltage v Function z ( v )include: ; in, μ For each battery in the window w The average value of the voltage.

4. The lithium-ion battery fault detection method based on operating data according to claim 1, characterized in that: The Laplace score is calculated as follows: use w Dimensionless feature samples to build a nearest neighbor graph G , No. Nodes correspond to ,like and is connected, then construct a weighted matrix S : ； In the formula, is the weight matrix S Each element in t is a constant, is the Euclidean distance; For r Features, definition D is a diagonal matrix; After de-averaging each feature, we get formula (4): ； Definition r The Laplace score of features: 。 5. The lithium-ion battery fault detection method based on operating data according to claim 1, characterized in that: The calculation method of the extended average voltage characteristic is as follows: Calculate the voltage of num cells in the battery pack at time during each charging process t The average value of: ； In the formula, For the moment t The average battery pack voltage, is the battery cell c at time t Voltage; Select each charging process The time corresponding to the fixed interval in t 1 and t 2. Calculation t 1 to t The extended average voltage of each monomer at time 2 is calculated as shown in formula (7): ； Among them, EAV is the extended average voltage of a single cell during a charging process. v t For a single t The voltage at the moment, is the time interval.

6. The lithium-ion battery fault detection method based on operating data according to claim 1, characterized in that: The method of further extracting the waveform factor from the extended average voltage feature using a sliding window is as follows: choose w One charging segment is used as a window, and each battery in the window is w 1 Expanded average voltage value to calculate the waveform factor. The waveform factor calculation formula is shown in (8): ； in, For a single unit The extended average voltage of the charging segment.

7. The lithium-ion battery fault detection method based on operating data according to claim 1, characterized in that: The standardized difference processing method of the optimal dimensionless characteristics and waveform factors is as follows: The optimal dimensionless characteristics and waveform factors are differentiated, and the calculation formula is shown in formula (9): ； in, is the optimal dimensionless characteristic or waveform factor after differentiation, is the optimal dimensionless characteristic or shape factor of the current window, is the optimal dimensionless characteristic or shape factor of the previous window; Afterwards, Z The score normalization method is calculated as shown in formula (10): ； in, is the optimal dimensionless characteristic or waveform factor after normalized difference, is the optimal dimensionless characteristic or waveform factor after differentiation, is the average value of the differential optimal dimensionless characteristics or form factors of all cells in the battery pack in the same window, It is the standard deviation of the differential optimal dimensionless characteristic or form factor of all cells in the battery pack in the same window.

8. The lithium-ion battery fault detection method based on operating data according to claim 1, characterized in that: The clustering algorithm is an adaptive DBSCAN clustering algorithm, and the specific steps are: Divide the grid based on the boundary of the standardized differential feature data set, and obtain the peak density points of the grid by statistics; Calculate the distance from each feature point to the peak density point ； Calculate the distance matrix from each feature point to other feature points Dist , the longest distance in the distance matrix The two feature points are marked as x and y, and the maximum distance from x to a point other than y is defined as ;like Greater than , then the parameter neighborhood radius equal ,otherwise equal The maximum value in ; Set the parameter MinPts to 1 and find each feature The number of points in the neighborhood, if the number of points is greater than or equal to MinPts, then this point is the core point; Start from the core point. All points in the neighborhood are classified into the same cluster, and points that do not belong to any cluster are marked as feature points of the faulty monomer, thus realizing lithium-ion battery fault detection.

9. A lithium-ion battery fault detection system based on operating data, characterized in that: include: A preprocessing module is used to preprocess the voltage signal of the lithium-ion battery in actual operating conditions; A feature set construction module constructs a dimensionless feature set based on the preprocessed battery voltage in the charging stage and the battery voltage in the discharging stage; Feature selection module, selects the optimal dimensionless feature: calculates the Laplace scores of the constructed dimensionless features for the battery charging stage voltage anomaly and the discharge stage voltage anomaly, and defines the dimensionless feature with the highest score as the optimal dimensionless feature; The waveform factor extraction module extracts the extended average voltage feature based on the charging voltage data of each battery in each charging process, and further extracts the waveform factor from the extended average voltage feature using a sliding window; The fault detection module performs standardized difference processing on the selected optimal dimensionless features and the extracted waveform factors, and uses a clustering algorithm to realize real-time fault detection of batteries on a short time scale for the optimal dimensionless features after standardized difference, and uses a clustering algorithm to realize cumulative fault detection of batteries on a long time scale for the waveform factors after standardized difference.

10. An electronic device comprising a memory, a processor, a computer program stored in the memory and executable on the processor, and a database, wherein: When the processor executes the program, the lithium-ion battery fault detection method based on operating data as described in any one of claims 1 to 8 is implemented.

Images