Battery production charging and discharging status monitoring and early warning method and system

Abstract

The present invention relates to the technical field of storage batteries, and more specifically, to a method and system for monitoring and warning the charge and discharge status of storage battery production. The method comprises: obtaining the operation data of the historical storage battery during charge and discharge, and setting a label for the charge and discharge status of the storage battery; obtaining the status score of the storage battery operation data and clustering it to obtain multiple clusters, screening the data pairs of the charge and discharge status that appear frequently and are difficult to distinguish during the charge and discharge of the storage battery in the cluster with the largest number of samples as the difficult-to-distinguish sample pairs, constructing the difficult-to-distinguish sample pairs and the standard data of the normal charge and discharge status of the storage battery into triplets, and obtaining multiple triplets; extracting features from the triplets based on the LSTM network, constructing a composite loss function according to the features, training a preset network model, and predicting the charge and discharge operation status of the storage battery production. The present invention screens out the difficult-to-distinguish sample pairs through cluster analysis, and constructs them into triplets with the standard data, thereby improving the accuracy of monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of storage batteries, and more specifically, to a method and system for monitoring and warning the charging and discharging status of storage batteries during production. Background Art

[0002] ‌With the development of industries such as telecommunications, electricity, and new energy vehicles, the use of batteries has increased dramatically. Therefore, it is increasingly important to accurately monitor and warn the charge and discharge status of batteries during production. Traditional battery monitoring methods, such as measuring electrolyte density and temperature, can no longer meet the needs of modern industry for high precision and high efficiency. Common battery monitoring methods include capacity discharge method and voltage inspection method, but these methods have problems such as long test time and unrealistic status. The internal resistance detection method is recognized by the industry as one of the best battery monitoring solutions because it can accurately and real-time reflect the performance and status of the battery.

[0003] The existing Chinese patent application document with publication number CN115051444A discloses a battery intelligent charging and discharging device and method, which relates to the field of battery technology. The specific steps are: classifying the batteries based on the charging curves of each battery in the battery pack; obtaining the ambient temperature, and dividing the charging stages of each battery according to the ambient temperature and the classification level; performing time-sharing staggered charging on batteries of each level in the same stage, and monitoring the charging data of each battery in real time; adjusting the charging plan and monitoring the status of each battery according to the charging data of each battery.

[0004] In the above scheme, the charging process of the battery is divided into different stages, the battery is monitored in real time by a detection device, and the charging state is automatically changed, so that the battery can be charged reasonably and quickly. At the same time, the damaged battery can be quickly isolated according to the isolation device, thereby ensuring the stability of the battery pack. However, in the current existing battery production charge and discharge monitoring process, a small number of abnormal mutations will appear in the operating data. The mutation of the operating data may be caused by external factors, such as unstable current and voltage, or it may be caused by the battery itself. Therefore, the charging and discharging status of these batteries is difficult to judge from the characteristics of the operating data, resulting in low accuracy of battery monitoring and early warning. Summary of the invention

[0005] In order to solve the problem that it is difficult to determine whether the abnormality in the operating data of the battery during the charging and discharging process is caused by external factors or the battery itself, resulting in low accuracy of battery monitoring and early warning, the present invention provides solutions in the following aspects.

[0006] In a first aspect, a method for monitoring and early warning of battery production charge and discharge status includes: obtaining historical battery charge and discharge operation data, and setting labels for the battery charge and discharge status according to the operation data; obtaining status scores of the battery operation data, clustering according to the status scores to obtain multiple clusters, screening data pairs with frequent and difficult-to-distinguish charge and discharge states during battery charge and discharge from the clusters with the largest number of samples as difficult-to-distinguish sample pairs, constructing the difficult-to-distinguish sample pairs and standard data with normal battery charge and discharge status labels into triplets to obtain multiple triplets; extracting features from the triplets based on an LSTM network, constructing a composite loss function according to the features, training a preset network model, and predicting the battery production charge and discharge operation status;

[0007] Among them, the composite loss function Satisfies the following relationship:

[0008] ;

[0009] In the formula, represents the logarithmic function, Indicates the first The labels of the data, Indicates the output of the preset network model. The probability that each data corresponds to a normal charging and discharging state is Indicates The exponential function with base , Indicates that the LSTM network extracts standard data in triples With normal data The distance between Indicates that the LSTM network extracts standard data in triples With abnormal data The distance between Indicates setting hyperparameters.

[0010] The effect is: by combining the LSTM network and the BP neural network and using the composite loss function to train the model, it is possible to more accurately identify whether the battery charging and discharging status is normal or not. It is able to adapt to various operating conditions and environmental changes, and maintain a high accuracy rate even under the influence of external factors such as current and voltage instability. It eliminates battery monitoring anomalies caused by external factors, and combines subtle changes that may be caused by problems with the battery itself. The system can capture and warn in a timely manner, which helps to achieve early fault detection and preventive maintenance, and take maintenance measures in a timely manner to avoid overcharging or over-discharging of the battery, thereby extending the battery life.

[0011] Preferably, the operating data includes: temperature data, internal resistance data, voltage data, and current data during the battery charging and discharging process, wherein the labels are 1 and 0, 1 indicates normal and 0 indicates abnormal.

[0012] Preferably, the state scoring comprises the steps of:

[0013] The isolation forest algorithm is used for each group of data in the operating data to calculate the isolation score of each data point in each group of data, and the absolute value of the sum of the isolation scores of all data points in each group of data during each charging and discharging process is calculated as the state score.

[0014] The effect is: by combining the isolation forest algorithm to obtain the isolation score of each group of data, the isolation forest algorithm can handle data sets of different sizes and dimensions. Especially when the data set is large, it can effectively adapt to different data characteristics and anomaly detection needs through subsampling and adjustment of the number of trees. It can effectively improve the accuracy and efficiency of battery charge and discharge status monitoring, and help to timely discover and deal with potential battery problems.

[0015] Preferably, clustering is performed according to the state scores, including:

[0016] Use K-means clustering algorithm to determine the number of clusters based on the elbow method , calculate the intra-class sum of square differences of each cluster, build a rectangular coordinate system based on the number of clusters and the intra-class sum of square differences, obtain the inflection point in the rectangular coordinate system, and take the number of clusters corresponding to the inflection point as the number of cluster centers;

[0017] According to the number of cluster centers, different battery charge and discharge status scores are randomly selected as initial cluster centers, and the sum of square differences between each battery status score and the status score of each initial cluster center is calculated. The status scores are divided into clusters of the initial cluster center with the smallest sum of square differences, and several clusters are obtained, and the cluster centers are updated.

[0018] The effect is: by observing the relationship between the sum of squared differences within the class and the number of clusters, it helps to determine the optimal number of clusters, balances the compactness and quantity of clusters, avoids overfitting or underfitting, reflects the regular data in the actual operating data through the clustering results, and reduces the factors of environmental influence.

[0019] Preferably, the step of distinguishing difficult sample pairs comprises the steps of:

[0020] Get the cluster with the largest number of samples, select a preset number of state scores marked as normal state data in order from large to small, and each time a normal state is selected, select a state score of abnormal state data that is closest to the corresponding normal state, and record the battery operation data corresponding to these two scores as normal data and abnormal data to form a difficult-to-distinguish sample pair.

[0021] The effect is: by distinguishing difficult sample pairs, the model's ability to recognize abnormal situations can be enhanced, and the robustness of the model in practical applications can be improved. By combining abnormal state samples that are closest to the normal state, the model can learn more subtle feature differences, which helps to improve the model's ability to recognize data features. It can not only improve the accuracy of the monitoring and early warning system, but also provide deeper insights into battery health management, thereby achieving more effective maintenance and optimization.

[0022] Preferably, the triplet includes: a battery charging and discharging standard data with a marked state of normal , a battery with normal charging and discharging data marked as normal , a battery charging and discharging abnormal data with abnormal status ;

[0023] From the scores of all normal state data, the normal battery charging and discharging data that are non-repetitive and less than the marked state are randomly selected as standard data. The number of selected data is the same as the number of difficult-to-distinguish sample pairs. Each difficult-to-distinguish sample pair and a standard data are constructed into a triplet.

[0024] Preferably, the hyperparameters are set according to normal data With standard data Difference in charge and discharge status scores and abnormal data With standard data Difference in charge and discharge scores To obtain;

[0025] The setting of the hyperparameters satisfies the following relationship:

[0026] ;

[0027] In the formula, Indicates setting hyperparameters, Indicates The exponential function with base , Represents normal data in a triple With standard data The difference between the charge and discharge status scores, Indicates abnormal data in triples With standard data The difference between charge and discharge scores, , Indicates a preset constant.

[0028] The effect is that the hyperparameters are dynamically adjusted for each triple , the model can better distinguish between normal and abnormal states, especially when the score difference is small and the labels are different, which helps the model capture the characteristics of the data more accurately.

[0029] Preferably, the preset network model is: BP neural network, wherein the network structure is: input layer: for receiving raw data input, hidden layer, output layer: for outputting network prediction results;

[0030] The collected battery operation data is input and the corresponding probability of the battery charge and discharge state being normal is output; if the probability is greater than the set value, the battery charge and discharge state is normal; otherwise, the state is abnormal and an early warning is issued.

[0031] In a second aspect, a battery production charge and discharge status monitoring and early warning system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned battery production charge and discharge status monitoring and early warning method is implemented.

[0032] The present invention has the following effects:

[0033] 1. The present invention selects difficult-to-distinguish sample pairs through cluster analysis, and constructs the difficult-to-distinguish sample pairs and standard data in normal state into triplets, which is conducive to the network model to better learn and distinguish whether the battery charging and discharging operation data is abnormal due to external factors, and further obtain the abnormal state of the battery itself, thereby improving the robustness of the model and the accuracy of the monitoring effect.

[0034] 2. The present invention sets the calculation method of hyperparameters so that the model can dynamically adjust the value of the set hyperparameters according to the score difference between normal data, abnormal data and standard data, which is conducive to the model to better adapt to different data distributions and improve the adaptability of the early warning system and the accuracy of the early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0036] Figure 1It is a method flow chart of steps S1 to S3 in the method for monitoring and early warning of the battery production charging and discharging status in an embodiment of the present invention.

[0037] Figure 2 It is a structural block diagram of a battery production charge and discharge status monitoring and early warning system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0039] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0040] Reference Figure 1 The battery production charge and discharge status monitoring and early warning method includes steps S1 to S3, which are as follows:

[0041] S1: Obtain historical battery charging and discharging operation data, and set labels for the battery charging and discharging status according to the operation data.

[0042] The operation data includes: temperature data, internal resistance data, voltage data, and current data during the battery charging and discharging process, where the labels are 1 and 0, 1 indicates normal and 0 indicates abnormal.

[0043] Exemplarily, the battery charging and discharging time and charging and discharging conditions are set according to the battery performance standard, so that the charging and discharging time and conditions of each battery are consistent, and the temperature, internal resistance, voltage, and current data of the battery charging and discharging process are collected at set time intervals using sensors. When the battery charging process is completed, the battery is discharged, and the same type of data in the charging process and the discharging process are sorted and recorded in time series to obtain four sets of data of temperature, internal resistance, voltage, and current in the battery charging and discharging process. The data of the four dimensions of each battery charging and discharging in history are obtained; the personnel in this field mark the battery charging and discharging status according to the four-dimensional data of the battery charging and discharging.

[0044] S2: Obtain the status score of the battery operation data, perform clustering according to the status score, obtain multiple clusters, select data pairs with frequent and difficult-to-distinguish charging and discharging states during battery charging and discharging from the cluster with the largest number of samples as difficult-to-distinguish sample pairs, construct the difficult-to-distinguish sample pairs and standard data with normal battery charging and discharging status labels into triplets, and obtain multiple triplets.

[0045] Status scoring includes the following steps:

[0046] The isolation forest algorithm is used for each group of data in the operating data to calculate the isolation score of each data point in each group of data, and the absolute value of the sum of the isolation scores of all data points in each group of data during each charging and discharging process is calculated as the state score.

[0047] Specifically, the status score satisfies the following relationship:

[0048] ;

[0049] In the formula, Indicates the status score, Indicates Group data, Indicates During the first charge and discharge process The first The isolation score of a data point, Represents the total number of data points.

[0050] That is to say, the inner summation is to accumulate the isolated scores of all data points in each charging and discharging process, and the outer summation is to accumulate the four-dimensional data.

[0051] In this embodiment, the isolation score measures the degree of abnormality of a data point in the data set. In the isolation forest algorithm, the closer the isolation score of a data point is to 0, the more normal the data point is in the data set. Conversely, the more it deviates from 0, the more likely the data point is an abnormal data point. It should be noted that the larger the state score value, the more abnormal the charge and discharge state of the battery.

[0052] Further explanation, in response to the status score Small, indicating that the isolated scores of data points are generally close to 0, that is, the data changes smoothly, without obvious abnormal mutations, so it can be considered that the charging and discharging state of the battery is normal. A larger value indicates that there are multiple isolated scores in the data points far away from 0, that is, there are more mutations in the data changes, so it can be considered that the charging and discharging state of the battery is abnormal.

[0053] In addition, the difference between the charging and discharging operation data detected at each moment and the set operation data can be calculated and normalized, and the difference of each dimension data can be summed up as the battery charging and discharging status score. At the same time, the larger the status score, the more abnormal it is.

[0054] To further illustrate, clustering based on state scores includes the following steps:

[0055] Use K-means clustering algorithm to determine the number of clusters based on the elbow method , calculate the intra-class sum of square differences of each cluster, build a rectangular coordinate system based on the number of clusters and the intra-class sum of square differences, obtain the inflection point in the rectangular coordinate system, and take the number of clusters corresponding to the inflection point as the number of cluster centers;

[0056] For example, the number of clusters The value is set to , the corresponding intra-class sum of squared differences Value is ,when As the value increases from 1 to 4, The value dropped significantly, but when When the value increases from 4 to 5, The decrease in value is significantly reduced, so you can choose as the total number of clusters.

[0057] According to the number of cluster centers, different battery charge and discharge status scores are randomly selected as initial cluster centers, and the sum of squared differences between each battery status score and the status score of each initial cluster center is calculated. The status scores are divided into clusters of the initial cluster center with the smallest sum of squared differences, and several clusters are obtained. The cluster centers are updated, and the mean of all data points in each cluster is used as the new cluster center. Clustering is performed again and the sum of squared differences is calculated until the set number of times is reached or the sum of squared differences no longer changes significantly, then the update is stopped.

[0058] Specifically, the sum of squared differences satisfies the following relationship:

[0059] ;

[0060] In the formula, represents the sum of squares of intra-class differences, Represents the total number of clusters. represents the total number of data points in the cluster. Indicates The first data points, Indicates The cluster centers of the clusters.

[0061] It should be noted that the elbow method is used to determine Value, from Start to calculate the different The sum of squared differences within each cluster in the clustering result corresponding to the value ;

[0062] In addition, the state scores can also be clustered by using a hierarchical clustering algorithm and a DBSCAN algorithm, which will not be described in detail.

[0063] Obtaining difficult-to-distinguish sample pairs includes the following steps:

[0064] Get the cluster with the largest number of samples, select a preset number of state scores marked as normal state data in order from large to small, and each time a normal state is selected, select a state score of abnormal state data that is closest to the corresponding normal state, and record the battery operation data corresponding to these two scores as normal data and abnormal data to form a difficult-to-distinguish sample pair.

[0065] Further explanation is given by obtaining the state score of the abnormal state data closest to the corresponding normal state, wherein the closest value range is between 0.5, so as to ensure that enough difficult-to-distinguish sample pairs are obtained.

[0066] The triplet includes: a battery charging and discharging standard data with a normal state , a battery with normal charging and discharging data marked as normal , a battery charging and discharging abnormal data with abnormal status ;

[0067] From the scores of all normal state data, the normal battery charging and discharging data that are non-repetitive and less than the marked state are randomly selected as standard data. The number of selected data is the same as the number of difficult-to-distinguish sample pairs. Each difficult-to-distinguish sample pair and a standard data are constructed into a triplet.

[0068] S3: Extract features from the triples based on the LSTM network, build a composite loss function based on the features, train the preset network model, and predict the battery production charging and discharging operation status.

[0069] Specifically, the composite loss function Satisfies the following relationship:

[0070] ;

[0071] In the formula, represents the logarithmic function, Indicates the first The labels of the data, Indicates the output of the preset network model. The probability that each data corresponds to a normal charging and discharging state is Indicates The exponential function with base , Indicates that the LSTM network extracts standard data in triples With normal data The distance between Indicates that the LSTM network extracts standard data in triples With abnormal data The distance between Indicates setting hyperparameters.

[0072] That is to say, It represents the penalty coefficient for the loss of each group of triple samples. For a group of triple samples, the closer the probability that the model predicts that the battery charging and discharging status is normal is to 0.5, the more difficult it is to correctly classify the triple sample, and the larger the corresponding coefficient will be, and the greater the loss will be.

[0073] Further explanation: Based on the extracted features, a composite loss function is constructed. The composite loss function combines the cross entropy loss and the triplet loss as well as sets the hyperparameters. , set the hyperparameters The strength of the distance penalty term is adaptively adjusted based on the triplet. One triplet corresponds to a set hyperparameter. The classification accuracy of the model for normal and abnormal states is improved by minimizing the cross entropy loss, thereby improving the classification accuracy of the model for normal and abnormal states through the hyperparameter. Allows to adjust the sensitivity of the distance penalty term to adapt to different data distributions and classification tasks.

[0074] This further illustrates that setting hyperparameters based on normal data With standard data Difference in charge and discharge status scores and abnormal data With standard data Difference in charge and discharge scores To obtain;

[0075] The hyperparameters are set to satisfy the following relationship:

[0076] ;

[0077] In the formula, Indicates setting hyperparameters, Indicates The exponential function with base , Represents normal data in a triple With standard data The difference between the charge and discharge status scores, Indicates abnormal data in triples With standard data The difference between charge and discharge scores, , Indicates a preset constant.

[0078] That is to say, The closer it is to 1, the closer the scores of normal data and abnormal data in the triplet are, which means that the samples of the triplet are more difficult to classify. The corresponding setting hyperparameters The bigger it is, the When equal to 1, set the hyperparameter Maximum and equal to the preset constant .

[0079] In this embodiment, , .

[0080] The preset network model is: BP neural network, where the network structure is: input layer: used to receive raw data input, hidden layer, output layer: used to output network prediction results;

[0081] Input the real-time collected battery charging and discharging operation data, and output the corresponding probability that the battery charging and discharging state is normal; if the probability is greater than the set value, the battery charging and discharging state is normal; otherwise, the state is abnormal and an early warning is issued.

[0082] To further illustrate, in this embodiment, the set value is 0.9.

[0083] The present invention also provides a battery production charge and discharge status monitoring and early warning system. Figure 2 As shown, the system includes a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the battery production charge and discharge status monitoring and early warning method according to the first aspect of the present invention is implemented.

[0084] The system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0085] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0086] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0087] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for monitoring and early warning of battery production charge and discharge status, characterized in that: include: Acquire historical battery charging and discharging operation data, and set a label for the battery charging and discharging state according to the operation data; Acquire the state score of the battery operation data, perform clustering according to the state score to obtain multiple clusters, select data pairs that frequently appear and are difficult to distinguish the charge and discharge states of the battery from the cluster with the largest number of samples as difficult-to-distinguish sample pairs, and construct the difficult-to-distinguish sample pairs and standard data with a normal battery charge and discharge state label into triplets to obtain multiple triplets; Extract features from the triplet based on the LSTM network, construct a composite loss function based on the features, train a preset network model, and predict the battery production charge and discharge operation status; Among them, the composite loss function Satisfies the following relationship: ； In the formula, represents the logarithmic function, Indicates the first The labels of the data, Indicates the output of the preset network model. The probability that each data corresponds to a normal charging and discharging state is is the natural exponential function, Indicates that the LSTM network extracts standard data in triples With normal data The distance between Indicates that the LSTM network extracts standard data in triples With abnormal data The distance between Indicates setting hyperparameters; The calculation formula for setting the hyperparameters is: ； In the formula, is the natural exponential function, Represents normal data in a triple With standard data The difference between the charge and discharge status scores, Indicates abnormal data in triples With standard data The difference between charge and discharge scores, , Indicates a preset constant.

2. The battery production charge and discharge status monitoring and early warning method according to claim 1 is characterized in that: The operation data includes: temperature data, internal resistance data, voltage data, and current data during the battery charging and discharging process, wherein the labels are 1 and 0, 1 indicates normal and 0 indicates abnormal.

3. The battery production charge and discharge status monitoring and early warning method according to claim 1 is characterized in that: The state scoring comprises the steps of: The isolation forest algorithm is used for each group of data in the operating data to calculate the isolation score of each data point in each group of data, and the absolute value of the sum of the isolation scores of all data points in each group of data during each charging and discharging process is calculated as the state score.

4. The battery production charge and discharge status monitoring and early warning method according to claim 1 is characterized in that: Clustering is performed according to the status scores, including: Use K-means clustering algorithm to determine the number of clusters based on the elbow method , calculate the intra-class sum of square differences of each cluster, build a rectangular coordinate system based on the number of clusters and the intra-class sum of square differences, obtain the inflection point in the rectangular coordinate system, and take the number of clusters corresponding to the inflection point as the number of cluster centers; According to the number of cluster centers, different battery charge and discharge status scores are randomly selected as initial cluster centers, and the sum of square differences between each battery status score and the status score of each initial cluster center is calculated. The status scores are divided into clusters of the initial cluster center with the smallest sum of square differences, and several clusters are obtained, and the cluster centers are updated.

5. The battery production charge and discharge status monitoring and early warning method according to claim 1 is characterized in that: The method of distinguishing difficult sample pairs comprises the steps of: Get the cluster with the largest number of samples, select a preset number of state scores marked as normal state data in order from large to small, and each time a normal state is selected, select a state score of abnormal state data that is closest to the corresponding normal state, and record the battery operation data corresponding to these two scores as normal data and abnormal data to form a difficult-to-distinguish sample pair.

6. The battery production charge and discharge status monitoring and early warning method according to claim 1 is characterized in that: in, The triplet includes: a battery charging and discharging standard data with a normal state , a battery with normal charging and discharging data marked as normal , a battery charging and discharging abnormal data with abnormal status ; From the scores of all normal state data, the normal battery charging and discharging data that are non-repetitive and less than the marked state are randomly selected as standard data. The number of selected data is the same as the number of difficult-to-distinguish sample pairs. Each difficult-to-distinguish sample pair and a standard data are constructed into a triplet.

7. The battery production charge and discharge status monitoring and early warning method according to claim 1 is characterized in that: The preset network model is: BP neural network, wherein the network structure is: input layer: for receiving raw data input, hidden layer, output layer: for outputting network prediction results; Input the collected battery operation data, and output the probability that the corresponding battery charge and discharge state is normal; if the probability is greater than the set value, the battery charge and discharge state is normal; Otherwise, the status is abnormal and an early warning is issued.

8. Battery production charge and discharge status monitoring and early warning system, characterized by: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the battery production charge and discharge status monitoring and early warning method according to any one of claims 1 to 7 is implemented.

Images