Battery connection abnormity detection method and device, electronic equipment and medium

Abstract

The invention provides a battery connection anomaly detection method and device, electronic equipment and a medium. The method comprises the steps of determining multiple pieces of operation data of a battery in a target period; determining a voltage distribution entropy of the battery in a target period according to the multiple pieces of operation data, forming a feature vector by the multiple pieces of operation data and the voltage distribution entropy, and inputting the feature vector into a detection model to obtain a connection state prediction probability of the battery in the target period, the detection model is constructed based on a corresponding relation between the feature vector of the battery in each period and the connection state of the internal battery core of the battery in the period; and according to the connection state prediction probability, calculating a connection abnormity score of the battery, and when the connection abnormity score is greater than an early warning threshold value corresponding to the target period, determining that the battery has connection abnormity. According to the invention, accurate detection of battery connection abnormity is realized.

Description

Technical Field

[0001] This application relates to the field of battery anomaly detection technology, and more specifically, to a method, apparatus, electronic device, and medium for detecting battery connection anomalies. Background Technology

[0002] With the booming development of new energy vehicles, battery safety remains a focal point for the industry and a pain point for consumers. Abnormal battery connection is a relatively common failure mode within battery systems. Although it rarely directly causes thermal runaway accidents, when abnormal connections occur in the battery system, it leads to increased contact resistance, increased heat during charging and discharging, accelerated battery aging, and consequently, shortened battery life and reduced efficiency. Therefore, early warning of abnormal power battery connections is essential. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method, apparatus, electronic device and medium for detecting abnormal battery connection, which aims to overcome at least one of the above-mentioned defects.

[0004] In a first aspect, this application provides a method for detecting battery connection anomalies. The method includes: determining multiple operating data of the battery under a target cycle; determining the voltage distribution entropy of the battery under the target cycle based on the multiple operating data, and inputting a feature vector composed of the multiple operating data and the voltage distribution entropy into a detection model to obtain a predicted probability of the connection state of the battery under the target cycle, wherein the detection model is constructed based on the correspondence between the feature vector of the battery under each cycle and the connection state between the internal cells of the battery under that cycle; calculating a connection anomaly score of the battery based on the predicted probability of the connection state, and determining that the battery has a connection anomaly when the connection anomaly score is greater than the warning threshold corresponding to the target cycle.

[0005] In one possible implementation, the detection model is constructed by: determining multiple running training data corresponding to the battery in multiple training cycles; for each training cycle, forming a training feature vector from the multiple running training data of that training cycle, and determining the connection state of the battery in that training cycle; and representing the training feature vector and the corresponding connection state of each training cycle through the detection model.

[0006] In one possible implementation, the multiple training data corresponding to each training cycle include average current value, maximum temperature value, current fluctuation standard deviation, state of charge value, and charge / discharge state. The training feature vector for each training cycle is composed as follows: a first feature interaction term is determined based on the voltage distribution entropy and the maximum temperature value of the training cycle; a second feature interaction term is determined based on the average current value and the state of charge value; and a third feature interaction term is determined based on the current fluctuation standard deviation and the maximum temperature value. The voltage distribution entropy, the charge / discharge state, the average current value, the maximum temperature value, the current fluctuation standard deviation, the state of charge value, the first feature interaction term, the second feature interaction term, and the third feature interaction term are combined to form the feature vector corresponding to the training cycle.

[0007] In one possible implementation, the voltage distribution entropy is determined by: for each cell of the battery, determining the proportion of the average voltage of that cell to the sum of the average voltages of all cells based on the average voltage of that cell during the target period; and determining the voltage distribution entropy based on all proportions.

[0008] In one possible implementation, the connection anomaly score of the battery is calculated by: determining the anomaly weight value and normal weight value corresponding to the target cycle; and calculating the connection anomaly score based on the anomaly weight value, the normal weight value, and the connection state prediction probability.

[0009] In one possible implementation, the warning threshold corresponding to the target cycle is determined by: determining the upper limit of the battery's temperature and the upper limit of its current; and determining the warning threshold based on the base threshold, the upper limit of the temperature, the upper limit of the current, and the battery's temperature, current, and state of charge values ​​during the target cycle.

[0010] In one possible implementation, the warning threshold is calculated using the following formula:

[0011] in, Here, t represents the warning threshold corresponding to the target period, and t represents the time corresponding to the target period. The basic threshold, This is the reference temperature value for the battery. The upper limit of the temperature is... The temperature value is... The current value This refers to the upper limit value of the current. The state of charge value, This is the adjustment coefficient.

[0012] Secondly, this application provides a battery connection anomaly detection device, the device comprising: a determining module, configured to determine multiple operating data of the battery under a target cycle; a predicting module, configured to determine the voltage distribution entropy of the battery under the target cycle based on the multiple operating data, and input a feature vector composed of the multiple operating data and the voltage distribution entropy into a detection model to obtain a predicted probability of the connection state of the battery under the target cycle, wherein the detection model is constructed based on the correspondence between the feature vector of the battery under each cycle and the connection state between the internal cells of the battery under that cycle; and a detection module, configured to calculate a connection anomaly score of the battery based on the predicted probability of the connection state, and determine that the battery has a connection anomaly when the connection anomaly score is greater than a warning threshold corresponding to the target cycle.

[0013] Thirdly, this application also provides an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and when the machine-readable instructions are executed by the processor, the steps of the method described above are performed.

[0014] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the method described above.

[0015] This application provides a method, apparatus, electronic device, and medium for detecting battery connection anomalies. The method includes: determining multiple operating data of the battery during a target cycle; determining the voltage distribution entropy of the battery during the target cycle based on the multiple operating data, and inputting a feature vector composed of the multiple operating data and the voltage distribution entropy into a detection model to obtain a predicted probability of the battery's connection state during the target cycle. The detection model is constructed based on the correspondence between the feature vector of the battery in each cycle and the connection state between the internal cells of the battery in that cycle; calculating a connection anomaly score for the battery based on the predicted connection state probability; and determining that the battery has a connection anomaly when the connection anomaly score is greater than a warning threshold corresponding to the target cycle. This application achieves accurate detection of battery connection anomalies.

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart of a battery connection anomaly detection method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating the construction of the detection model provided in an embodiment of this application; Figure 3 This is a schematic diagram of the battery connection anomaly detection device provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0020] First, the applicable scenarios for this application will be introduced. This application can be applied to battery anomaly detection technology.

[0021] Research has revealed that battery safety remains a key focus for the industry and a major pain point for consumers amidst the booming development of new energy vehicles. Battery connection anomalies are a relatively common failure mode within battery systems. While they rarely directly cause thermal runaway, they do lead to increased contact resistance, increased heat during charging and discharging, accelerated battery aging, and ultimately, shortened battery life and reduced efficiency. Therefore, early warning systems for power battery connection anomalies are crucial. In recent years, with the widespread adoption of cloud data storage and machine learning algorithms, research on battery connection anomaly warning and diagnostic technologies has shifted from short-cycle vehicle data to long-cycle cloud data analysis, from single-threshold judgments to multi-factor fusion, and from static models to dynamic adaptive algorithms. Despite these advancements in early warning and diagnosis, limitations remain: limited feature dimensions. Many existing battery connection anomaly warning and diagnostic technologies rely solely on a single physical quantity (such as pressure difference or voltage range). However, vehicles are affected by various factors during operation, such as temperature, operating conditions, and current, all of which are strongly correlated with connection anomalies. The current connection anomaly warning and diagnosis technologies suffer from several shortcomings. First, they lack flexibility. Key parameters (such as sliding window length, voltage threshold, current threshold, and characteristic parameters) require manual setting, lacking a dynamic optimization mechanism and failing to adapt to different battery types and their dynamic aging processes. Second, feature correlation analysis is insufficient. Existing methods only analyze the time-series changes of single signals, failing to construct cross-domain correlation models for parameters such as voltage, current, and temperature. This results in the inability to identify typical connection anomaly characteristics such as a sudden voltage drop in a cell during high-current discharge, or connection anomalies caused by the combined effects of vibration and high current. Third, real-time data processing capabilities are inadequate. Current technologies rely on multi-layer matrix operations and complex mathematical transformations to predict the sliding window operation, leading to high computational complexity and large computational costs, making it difficult to meet the needs of timely warnings.

[0022] Based on this, embodiments of this application provide a battery connection anomaly detection method, apparatus, electronic device, and medium, aiming to overcome at least one of the above-mentioned defects. Please see Figure 1 , Figure 1 This is a flowchart illustrating a battery connection anomaly detection method provided in an embodiment of this application. Figure 1 As shown in the figure, the battery connection anomaly detection method provided in this application includes: S101. Determine multiple operating data of the battery under the target cycle.

[0023] Here, multiple operational data are included, such as vehicle status, charging status, voltage value, current value, temperature value, and state of charge (SOC). The length of one cycle can be set to 1 minute.

[0024] Specifically, the vehicle's BMS system can collect the aforementioned operational data and perform necessary data cleaning and preprocessing, including outlier cleaning and normalization.

[0025] S102. Based on multiple operating data, determine the voltage distribution entropy of the battery under the target cycle, and input the feature vector composed of multiple operating data and voltage distribution entropy into the detection model to obtain the predicted probability of the battery's connection state under the target cycle.

[0026] Specifically, the average cell voltage within a 1-minute target period is taken to obtain the voltage sequence for that period; the voltage of the [number]th cell is calculated. The proportion of the average voltage of each cell to the total average voltage of the battery pack The voltage distribution entropy of the battery under the target cycle can be obtained using the following formula:

[0027] here, Indicates the first The average voltage of each cell This represents the total number of battery cells in the battery pack, and then the proportion. Substituting into the Shannon entropy formula:

[0028] This yields the voltage distribution entropy. When a cell has an abnormal connection, its voltage percentage deviates from the mean and increases significantly.

[0029] Next, a feature vector composed of multiple operating data and voltage distribution entropy is input into the detection model to obtain the predicted probability of the battery's connection state in the target cycle. The detection model is constructed based on the correspondence between the battery's feature vector in each cycle and the connection state between the internal cells of the battery in that cycle.

[0030] The following is through Figure 2 This section describes the specific process of building the detection model.

[0031] Figure 2 This is a flowchart illustrating the construction of the detection model provided in an embodiment of this application.

[0032] S201. Determine the multiple running training data corresponding to the battery in multiple training cycles.

[0033] Here, multiple operational training data points corresponding to various periods over the past 6 months are collected, including normal operation data and known battery connection anomaly data, for feature construction. Specifically, the multiple operational training data points corresponding to each training period include average current values. Maximum temperature value Standard deviation of current fluctuation SOC and state of charge / discharge The charging state includes the discharging state, the charging state, and the resting state. The discharging state is marked as 0, the charging state is marked as 1, and the resting state is marked as 2.

[0034] Specifically, the standard deviation of current fluctuation This is used to measure the stability of the current. A larger standard deviation indicates more severe current fluctuations, which may indicate problems such as abnormal connections or unstable loads. The average current value for each cycle can be calculated, reflecting the battery's charge and discharge intensity within that cycle. The standard deviation of current fluctuation can then be calculated.

[0035] here, For the standard deviation of current fluctuation, This is the average current over one cycle.

[0036] S202. For each training cycle, the multiple running training data of that training cycle are combined into a training feature vector, and the connection status of the internal cells of the battery is determined in that training cycle.

[0037] Specifically, the voltage distribution entropy for each training cycle is first calculated based on the training data for each run. The specific calculation process is shown in step S102, which will not be elaborated here.

[0038] Next, the voltage distribution entropy and the corresponding training data are combined to form the training feature vector for the corresponding training period, specifically represented as follows:

[0039] All feature vectors are stacked row-wise to construct an input feature matrix with dimensions m×6, where m is the number of samples. Simultaneously, based on the actual fault conditions, the corresponding state is labeled for each feature vector's period, constructing a label vector y, y∈{0,1}, where 0 represents a normal state and 1 represents an abnormal state.

[0040] Then, all feature vectors are preprocessed to determine the charge / discharge state. One-hot encoding is performed, and after encoding, the feature vector is expanded to 8 dimensions:

[0041] The encoding function Defined as:

[0042]

[0043] Next, the first characteristic interaction term is determined based on the voltage distribution entropy and the maximum temperature value, the second characteristic interaction term is determined based on the average current value and the state of charge value, and the third characteristic interaction term is determined based on the current fluctuation standard deviation and the maximum temperature value.

[0044] Specifically, a first feature interaction term is constructed that interacts between voltage distribution entropy and the maximum temperature value. It captures the positive feedback effect of "temperature increase → contact resistance increase → voltage distribution intensification"; and constructs a second characteristic interaction term that interacts with the average current value and the state of charge (SOC). The study identifies the pattern of "low SOC high current discharge → increased cell polarization → increased risk of connection abnormalities"; and constructs a third feature interaction term that reflects the interaction between the standard deviation of current fluctuation and the maximum temperature value. The composite anomaly of "vibration-induced contact resistance fluctuations + high-temperature accelerated oxidation" was identified, resulting in a new 11-dimensional feature vector.

[0045] S203. Represent the training feature vector and the corresponding connection state of each training cycle through the detection model.

[0046] Here, the detection model preferred in this application is the LightGBM model.

[0047] Specifically, the dataset is first divided into 11-dimensional feature vectors. Stacking is performed to obtain a new feature matrix, and each feature in the preprocessed feature matrix is ​​then used to... and corresponding label vector according to The proportions are divided into training and test sets using stratified sampling to ensure that the ratios of abnormal and normal samples in the training and test sets are similar, thus avoiding the impact of data imbalance on model training.

[0048] Next, the core parameters of the model were configured. `objective = 'bunary'` was set to explicitly define the model's binary classification task. `metric = 'auc'` was set as the evaluation metric. To control tree complexity, `num_leaves` was set within a reasonable range based on the dataset size and feature complexity. A learning rate decay strategy was used, gradually reducing the learning rate as training progresses to balance training speed and model performance. The minimum number of samples per leaf node (`min_child_samples`) was set based on the data scale. Specifically, binary logistic loss and AUC were chosen as evaluation metrics. Based on the dataset size and feature complexity, 5-fold cross-validation was used to determine the optimal `num_leaves` value within the pre-defined parameter search space, thereby controlling tree complexity. A learning rate decay strategy was employed, with an initial `learning_rate` set to 0.1. `min_child_samples` was set based on the data scale to ensure that the number of samples per leaf node was not less than a certain amount. `subsample` and `colsample_bytree` were set, and their optimal values ​​were determined through cross-validation to improve the model's generalization ability.

[0049] The model training process is based on the principle of gradient boosting, which iteratively builds multiple decision trees, reaching the [number]th [tree]. In each iteration, the model will adjust its predictions based on the current results. Calculate the gradient for each sample. Hessian matrix In the model, a binary logistic loss function is selected. For optimization, the gradient calculation formula is as follows:

[0050] The formula for calculating the Hessian matrix is:

[0051] Based on gradient and Hessian matrix information, a new decision tree is constructed by using a histogram algorithm to find the optimal split point for each feature. The histogram algorithm discretizes continuous eigenvalues ​​into several intervals, calculates the gradient sum and Hessian matrix sum of samples within each interval, thus reducing the computational cost of finding the optimal split point. After finding the optimal split point, the node is split into left and right child nodes, and the output value of the node is calculated. For leaf nodes... Its output value The calculation formula is:

[0052] in, Indicates belonging to a leaf node The sample set, This is a regularization parameter used to prevent overfitting and update the model's predictions.

[0053] in, The learning rate is used to iterate through the above process until the evaluation metric no longer improves, thus completing the model training.

[0054] After training, the model is evaluated using accuracy, precision, recall, F1 score, and AUC-ROC. Based on the evaluation results, the model is optimized by adjusting parameters, optimizing feature engineering, and adjusting the settings of core parameters.

[0055] The LightGBM model outputs the predicted probability of anomalous states. To transform the probability into a classification result of normal or abnormal, the Youden index is used. Determine the classification threshold Youden Index The calculation formula is:

[0056] Where TPR represents recall rate and FPR represents false positive rate, the formulas for calculating TPR and FPR are as follows:

[0057]

[0058] Where TP (True Positive Examples) represents the number of samples correctly predicted by the model as battery connection abnormality, TN (True Negative Examples) represents the number of samples correctly predicted by the model as battery normality, FP (False Positive Examples) represents the number of samples incorrectly predicted as abnormality but actually normality, and FN (False Negative Examples) represents the number of samples incorrectly predicted as normality but actually abnormality. This is achieved by iterating through different thresholds. Calculate the corresponding TPR and FPR, and then obtain the corresponding... Value, making Threshold for the value to reach its maximum value The optimal classification threshold is the threshold that ensures a high true positive rate while minimizing the false positive rate, thereby optimizing the model's classification performance and balancing the false negative and false positive rates in anomaly detection. The optimal classification threshold determines whether the battery connection is normal or abnormal.

[0059] return Figure 1S103. Calculate the battery connection anomaly score based on the connection status prediction probability. When the connection anomaly score is greater than the warning threshold corresponding to the target cycle, it is determined that the battery has a connection anomaly.

[0060] Here, the LightGBM model outputs whether the battery connection is normal or abnormal. However, for more refined anomaly identification, this application also needs to define a connection anomaly score based on the model's predicted probability. i This allows for a more detailed assessment of the severity of the anomaly, resulting in a connection anomaly score. i To determine the battery's condition, a connection anomaly score can be calculated using the following formula. i :

[0061] in, For the model, the feature vector samples corresponding to the target period The probability of connection state prediction. and These are the abnormal weight value and the normal weight value, respectively. The weight setting needs to be adjusted according to the actual application scenario and requirements. (Due to connection anomaly scoring...) The range of values ​​and and It is relevant that it can quantify the degree of anomalies and provide rich information for subsequent early warning decisions.

[0062] Next, the connection anomaly score needs to be calculated. i It is often compared with the warning threshold corresponding to the target cycle to determine whether there is a connection abnormality in the battery.

[0063] Specifically, in order for the early warning mechanism to adapt to different operating conditions, it is necessary to set dynamic thresholds for each cycle. The calculation method is as follows:

[0064] here, Here, t represents the warning threshold corresponding to the target period, and t represents the time corresponding to the target period. This is a preliminary threshold set based on historical data and experience. These are adjustment coefficients, all determined through grid search, with values ​​ranging from [value range missing]. That is, traversal Determine the optimal combination among all possible combinations, with the goal of maximizing the model's F1 score on the validation set or minimizing the sum of the false positive and false negative rates. for Battery temperature at any given time. This is the reference temperature value for the battery. This refers to the upper limit of the battery's temperature. for The current value at time [time]. This represents the upper limit of the battery's current. for The state of charge value at time t, where, I(t), SOC(t) and The calculation uses data from the same time window, such as the 1-minute average value corresponding to the target period.

[0065] Among them, when A timely warning is issued when a vehicle is identified as having a risk of abnormal battery connection. At that time, the vehicle was in normal condition.

[0066] Compared to existing technologies, this application utilizes voltage distribution entropy to quantify the degree of unevenness in cell voltage distribution, and constructs feature vectors by combining multi-dimensional features and feature interaction terms, providing rich input information for the model. Simultaneously, it employs the LightGBM model, combined with parameter settings and optimization strategies, and the Youden index to determine the optimal classification threshold, improving the model's ability to identify battery connection anomalies. Regarding adaptability and flexibility, this invention constructs a dynamic threshold and anomaly scoring mechanism, enabling the early warning strategy to be closely integrated with battery operating data for adaptive adjustment. The model can cover various vehicle models and operating conditions. In terms of operational efficiency, this invention uses the LightGBM model, which is based on histogram algorithms and gradient boosting principles, resulting in fast training speed and low computational complexity. This model also possesses technological foresight, providing new ideas for industry development and ensuring the safe operation of new energy vehicle batteries in terms of social benefits, reducing threats caused by battery safety issues.

[0067] Based on the same inventive concept, this application also provides a battery connection anomaly detection device corresponding to the battery connection anomaly detection method. Since the principle of the device in this application is similar to the battery connection anomaly detection method described above in this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0068] Please see Figure 3 , Figure 3 This is a schematic diagram of the battery connection anomaly detection device provided in an embodiment of this application. Figure 3 As shown, the battery connection anomaly detection device 300 includes: The determination module 301 is used to determine multiple operating data of the battery under the target cycle.

[0069] The prediction module 302 is used to determine the voltage distribution entropy of the battery under the target cycle based on the multiple operating data, and input the multiple operating data and the voltage distribution entropy into the detection model to obtain the predicted probability of the connection state of the battery under the target cycle. The detection model is constructed based on the correspondence between the feature vector of the battery under each cycle and the connection state of the internal cells of the battery under that cycle.

[0070] The detection module 303 is used to calculate the connection anomaly score of the battery based on the connection status prediction probability, and determine that the battery has a connection anomaly when the connection anomaly score is greater than the warning threshold corresponding to the target period.

[0071] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 400 includes a processor 410, a memory 420, and a bus 430.

[0072] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, they can perform the operations described above. Figure 1 as well as Figure 2 The steps of the battery connection anomaly detection method in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0073] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 as well as Figure 2 The steps of the battery connection anomaly detection method in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0074] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0075] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0076] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0077] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0078] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0079] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting abnormal battery connection, characterized in that, The method includes: Determine multiple operational data points for the battery at the target cycle; Based on the multiple operating data, the voltage distribution entropy of the battery under the target cycle is determined, and the multiple operating data and the voltage distribution entropy are combined to form a feature vector and input into the detection model to obtain the predicted probability of the connection state of the battery under the target cycle. The detection model is constructed based on the correspondence between the feature vector of the battery under each cycle and the connection state between the internal cells of the battery under that cycle. Based on the predicted connection status probability, a connection anomaly score for the battery is calculated. When the connection anomaly score is greater than the warning threshold corresponding to the target period, it is determined that the battery has a connection anomaly.

2. The method according to claim 1, characterized in that, The detection model is constructed using the following method: Determine the multiple running training data corresponding to each battery in multiple training cycles; For each training cycle, the multiple running training data of that training cycle are combined into a training feature vector, and the connection state of the battery in that training cycle is determined. The training feature vector and the corresponding connection state for each training cycle are represented by the detection model.

3. The method according to claim 2, characterized in that, Each training cycle corresponds to multiple training data points, including average current value, maximum temperature value, current fluctuation standard deviation, state of charge value, and charge / discharge state. The training feature vector for each training cycle is constructed in the following manner: The first feature interaction term is determined based on the voltage distribution entropy of the training period and the maximum temperature value; the second feature interaction term is determined based on the average current value and the state of charge value; and the third feature interaction term is determined based on the current fluctuation standard deviation and the maximum temperature value. The voltage distribution entropy, the charge / discharge state, the average current value, the maximum temperature value, the current fluctuation standard deviation, the state of charge value, the first feature interaction term, the second feature interaction term, and the third feature interaction term are used to form the feature vector corresponding to this training period.

4. The method according to claim 1, characterized in that, The voltage distribution entropy is determined in the following manner: For each cell of the battery, the proportion of the average voltage of that cell to the sum of the average voltages of all cells is determined based on the average voltage of that cell during the target period. The voltage distribution entropy is determined based on all the proportions.

5. The method according to claim 1, characterized in that, The battery connection anomaly score is calculated using the following method: Determine the abnormal weight value and normal weight value corresponding to the target period; The connection anomaly score is calculated based on the anomaly weight value, the normal weight value, and the connection state prediction probability.

6. The method according to claim 1, characterized in that, The warning threshold corresponding to the target period is determined using the following method: Determine the upper limit values ​​for the temperature and current of the battery; The warning threshold is determined based on the basic threshold, the upper limit of temperature, the upper limit of current, and the temperature, current, and state of charge values ​​of the battery during the target cycle.

7. The method according to claim 6, characterized in that, The warning threshold is calculated using the following formula: in, Here, t represents the warning threshold corresponding to the target period, and t represents the time corresponding to the target period. The basic threshold, This is the reference temperature value for the battery. The upper limit of the temperature is... The temperature value is... The current value This refers to the upper limit value of the current. The state of charge value, This is the adjustment coefficient.

8. A battery connection anomaly detection device, characterized in that, The device includes: The determination module is used to determine multiple operating data of the battery under the target cycle; The prediction module is used to determine the voltage distribution entropy of the battery under the target cycle based on the multiple operating data, and input the multiple operating data and the voltage distribution entropy into the detection model to obtain the predicted probability of the connection state of the battery under the target cycle. The detection model is constructed based on the correspondence between the feature vector of the battery under each cycle and the connection state between the internal cells of the battery under that cycle. The detection module is used to calculate the connection anomaly score of the battery based on the probability prediction of the connection status, and to determine that the battery has a connection anomaly when the connection anomaly score is greater than the warning threshold corresponding to the target period.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is in operation, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method as described in any one of claims 1 to 7.

Images