Battery voltage difference detection method, system, device and storage medium

Abstract

The application provides a battery voltage difference detection method, system, device and storage medium. The battery voltage difference detection method comprises the following steps: obtaining first detection data and second detection data of a battery; the first detection data comprises detection data in a first time period, and the second detection data comprises detection data in a second time period; the end time of the first time period and the end time of the second time period are both the current time; the length of the first time period is less than the length of the second time period; and the battery comprises a plurality of battery monomers; inputting the first detection data into a first detection model to obtain voltage anomaly detection results corresponding to the voltages of the battery monomers respectively; and inputting the voltage anomaly detection results, the first detection data and the second detection data into a second detection model to obtain a voltage difference detection result corresponding to the battery. Based on the above method, the accuracy of voltage difference detection can be improved.

Description

Technical Field

[0001] This application relates to the field of detection technology, and in particular to methods, systems, devices and storage media for detecting voltage range differences in batteries. Background Technology

[0002] With the development of new energy vehicles, the safety requirements for batteries that power these vehicles are becoming increasingly stringent. Typically, voltage difference detection is required to determine whether there are any abnormalities in the voltage difference between individual battery cells. If an abnormality is detected, appropriate early warning measures must be taken to reduce losses and minimize the possibility of safety accidents.

[0003] However, existing methods for detecting voltage range in batteries are inaccurate because the data they acquire is typically short-term, with a high degree of randomness and weak generalization. Summary of the Invention

[0004] The main technical problem addressed by this application is to provide a method, system, device, and storage medium for detecting voltage range in batteries, which can improve the accuracy of voltage range detection.

[0005] In a first aspect, this application provides a method for detecting voltage range of a battery, comprising: acquiring first detection data and second detection data of the battery; wherein the first detection data includes detection data within a first time period, the second detection data includes detection data within a second time period, the end time of the first time period and the end time of the second time period are both the current time, the duration of the first time period is shorter than the duration of the second time period, and the battery includes multiple battery cells; inputting the first detection data into a first detection model to obtain voltage anomaly detection results corresponding to the voltage of each battery cell; and inputting the voltage anomaly detection results, the first detection data and the second detection data into a second detection model to obtain voltage range detection results corresponding to the battery.

[0006] In the technical solution of this application embodiment, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model, so that the first detection model detects the probability of voltage anomalies in each battery cell using the first detection data to obtain a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model, so that the second detection model detects the probability of voltage range anomalies in each battery cell using the voltage anomaly detection result, the first detection data, and the second detection data to obtain a voltage range detection result. This achieves the goal of... The first detection model analyzes and processes the short-term condition of each battery cell in the battery based on the first detection data with a relatively short corresponding time span, and obtains the voltage anomaly detection result. Then, the second detection model combines the first detection data with a relatively short corresponding time span, the second detection data with a relatively long corresponding time span, and the voltage anomaly detection result to analyze and process the short-term condition, long-term condition, and voltage condition of each battery cell in the battery, and obtains the voltage range detection result. This enables a more comprehensive voltage range detection of the battery based on the detection data obtained from different time spans, reduces the possibility of distortion of detection results due to the strong randomness or weak generalization of the detection data, and improves the accuracy of voltage range detection.

[0007] In some embodiments, the loss function of the second detection model includes the sum of a loss term and a constraint penalty term; the constraint penalty term includes at least one of a first penalty term, a second penalty term, and a third penalty term; wherein the first penalty term is associated with the state of charge value of the battery in the training samples, the second penalty term is associated with the number of cycles of the battery in the training samples, and the third penalty term is associated with the current value of the battery in the training samples.

[0008] In the technical solution of this application embodiment, by setting a constraint penalty term in the loss function of the second detection model in addition to the loss term, and the constraint penalty term includes a first penalty term associated with the state of charge value of the battery in the training sample, a second penalty term associated with the number of cycles of the battery in the training sample, and a third penalty term associated with the current value of the battery in the training sample, the loss function of the second detection model can be adjusted accordingly for batteries in different states. Thus, after training with the help of this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0009] In some embodiments, the first penalty term is the average of the first values ​​corresponding to the batteries in all training samples; the first value is 0 when the second value is less than or equal to 0, and the second value is the second value when the second value is greater than 0; the second value is the difference obtained by subtracting the voltage extreme difference anomaly prediction probability of the corresponding battery from the first constraint boundary value; wherein the first constraint boundary value is associated with the state of charge value of the batteries in the training samples.

[0010] In the technical solution of this application embodiment, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge value of the battery in the training sample, a second penalty term associated with the number of cycles of the battery in the training sample, and a third penalty term associated with the current value of the battery in the training sample. The first penalty term, which reflects the state of charge value of each battery, can be obtained by calculating the average of the first values ​​corresponding to all batteries in the training sample. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with the help of this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0011] In some embodiments, the first constraint boundary value is the function value of the first Sigmoid function; the parameter value of the first Sigmoid function is the difference obtained by subtracting the first coefficient from the third value; the third value is the product obtained by multiplying the second coefficient by the fourth value; and the fourth value is the square of the difference between the state of charge value of the corresponding battery and 0.5.

[0012] In the technical solution of this application embodiment, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge value of the battery in the training sample, a second penalty term associated with the number of cycles of the battery in the training sample, and a third penalty term associated with the current value of the battery in the training sample. The average value of the first value corresponding to all batteries in the training sample is calculated. The first value is 0 when the second value is less than or equal to 0 and is the second value when the second value is greater than 0. The second value can be determined based on the first constraint boundary value obtained from the function value of the first Sigmoid function to perform boundary constraints. This yields the first penalty term that reflects the state of charge value of each battery. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0013] In some embodiments, the second penalty term is the average of the fifth values ​​corresponding to the batteries in all training samples; the fifth value is 0 when the sixth value is less than or equal to 0, and the sixth value is the sixth value when the sixth value is greater than 0; the sixth value is the difference obtained by subtracting the voltage extreme difference prediction probability of the corresponding battery from the second constraint boundary value; wherein the second constraint boundary value is associated with the number of cycles of the batteries in the training samples.

[0014] In the technical solution of this application embodiment, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge value of the battery in the training sample, a second penalty term associated with the number of cycles of the battery in the training sample, and a third penalty term associated with the current value of the battery in the training sample. The second penalty term, which reflects the number of cycles of each battery, can be obtained by calculating the average of the fifth values ​​corresponding to all batteries in the training sample. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with the loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0015] In some embodiments, the second constraint boundary value is the function value of the second Sigmoid function; the parameter value of the second Sigmoid function is the difference obtained by subtracting the third coefficient from the seventh value; the seventh value is the product obtained by multiplying the fourth coefficient by the eighth value; and the eighth value is the quotient obtained by dividing the cycle number of the corresponding battery by the maximum cycle number threshold.

[0016] In the technical solution of this application embodiment, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge value of the battery in the training sample, a second penalty term associated with the number of cycles of the battery in the training sample, and a third penalty term associated with the current value of the battery in the training sample. The average value of the fifth value corresponding to all batteries in the training sample is calculated. The fifth value is 0 when the sixth value is less than or equal to 0 and is the sixth value when the sixth value is greater than 0. The sixth value can be determined based on the second constraint boundary value obtained from the function value of the second Sigmoid function to perform boundary constraints. This yields the second penalty term that reflects the number of cycles of each battery. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0017] In some embodiments, the third penalty term is the average of the ninth values ​​corresponding to the batteries in all training samples; the ninth value is 0 when the tenth value is less than or equal to 0, and is the tenth value when the tenth value is greater than 0; the tenth value is the difference obtained by subtracting the voltage extreme difference anomaly prediction probability of the corresponding battery from the third constraint boundary value; wherein, the third constraint boundary value is associated with the current value of the batteries in the training samples.

[0018] In the technical solution of this application embodiment, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge value of the battery in the training sample, a second penalty term associated with the number of cycles of the battery in the training sample, and a third penalty term associated with the current value of the battery in the training sample. The third penalty term, which reflects the current status of each battery, can be obtained by calculating the average of the ninth values ​​corresponding to all batteries in the training sample. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with the help of this loss function, the accuracy of voltage range detection for batteries in different states can be improved.

[0019] In some embodiments, the third constraint boundary value is the function value of the third Sigmoid function; the parameter value of the third Sigmoid function is the difference obtained by subtracting the fifth coefficient from the eleventh value; the eleventh value is the product obtained by multiplying the sixth coefficient by the twelfth value; and the twelfth value is the quotient obtained by dividing the absolute value of the current of the corresponding battery by the maximum current threshold.

[0020] In the technical solution of this application embodiment, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge value of the battery in the training sample, a second penalty term associated with the number of cycles of the battery in the training sample, and a third penalty term associated with the current value of the battery in the training sample. The average value of the ninth value corresponding to all batteries in the training sample can be calculated. The ninth value is 0 when the tenth value is less than or equal to 0 and is the tenth value when the tenth value is greater than 0. The tenth value can be determined based on the third constraint boundary value obtained from the function value of the third Sigmoid function to perform boundary constraints. The third penalty term used to reflect the current status of each battery is obtained. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with the help of this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0021] In some embodiments, inputting the first detection data into the first detection model to obtain voltage anomaly detection results corresponding to the voltage of each battery cell includes: inputting the first detection data into the first detection model to obtain the anomaly probability of the voltage of each battery cell changing over time; processing the anomaly probability of the voltage of the battery cell changing over time to obtain voltage anomaly detection results; wherein the voltage anomaly detection results include at least one of the average, maximum, range, variance, skewness, kurtosis, and quantile of the anomaly probability of the voltage of the battery cell changing over time.

[0022] In the technical solution of this application embodiment, by acquiring first detection data of the battery within a relatively short period of time and second detection data of the battery within a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each battery cell, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each battery cell, thus obtaining a voltage range detection result. This achieves the processing of the first detection data within a relatively short period of time using the first detection model. The analysis and processing determine the abnormal probability of voltage changes of each cell in the battery over time. Based on the abnormal probability of each cell, various data reflecting the short-term condition of each cell are determined for voltage anomaly detection. A second detection model is then used to combine the relatively short-term first detection data, the relatively long-term second detection data, and the voltage anomaly detection results to analyze the short-term, long-term, and voltage conditions of each cell, obtaining voltage range detection results. This allows for more comprehensive voltage range detection of the battery using detection data obtained from different time spans, reducing the possibility of distortion due to strong randomness or weak generalization of the detection data and improving the accuracy of voltage range detection.

[0023] In some embodiments, the first detection model includes a first classifier and a second classifier, and the anomaly probability includes overvoltage probability and undervoltage probability. Inputting the first detection data into the first detection model to obtain the anomaly probability of the voltage change of each battery cell over time includes: inputting the first detection data into the first classifier to obtain the overvoltage probability of the voltage change of each battery cell over time, and inputting the first detection data into the second classifier to obtain the undervoltage probability of the voltage change of each battery cell over time; wherein the voltage anomaly detection result includes at least one of the following: average, maximum, range, variance, skewness, kurtosis, and quantile of the overvoltage probability, and at least one of the following: average, maximum, range, variance, skewness, kurtosis, and quantile of the undervoltage probability.

[0024] In the technical solution of this application embodiment, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model uses the first detection data to detect the probability of voltage anomalies in each battery cell, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each battery cell, thus obtaining a voltage range detection result. This achieves the processing of the first detection data over a relatively short period of time by the first classifier and the second classifier of the first detection model. The system performs analysis and processing to determine the undervoltage and overvoltage probabilities of each battery cell over time. Based on these probabilities, it determines various data points reflecting the short-term condition of each cell in the voltage anomaly detection results. A second detection model is then used to combine the relatively short-duration first detection data, the relatively long-duration second detection data, and the voltage anomaly detection results to analyze the short-term, long-term, and voltage conditions of each battery cell, yielding voltage range detection results. This allows for more comprehensive voltage range detection of the battery using detection data obtained from different time spans, reducing the possibility of data distortion due to strong randomness or weak generalization, and improving the accuracy of voltage range detection.

[0025] In some embodiments, the first detection data includes at least one of voltage detection data, cyclic detection data, and temperature detection data within a first time period; and / or, the second detection data includes at least one of voltage detection data, cyclic detection data, and temperature detection data within a second time period.

[0026] In the technical solution of this application embodiment, first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time are obtained. The first or second detection data includes at least one of voltage detection data, cycle detection data, and temperature detection data during the second period. The first detection data is then input into a first detection model, which uses the first detection data to detect the probability of voltage anomalies in each battery cell, thereby obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model, which uses the voltage anomaly detection result, the first detection data, and the second detection data to detect voltage range anomalies in each battery cell. The probability of voltage range detection is used to obtain voltage range detection results. A first detection model analyzes the short-term condition of each battery cell using relatively short-duration first detection data to obtain voltage anomaly detection results. A second detection model combines the relatively short-duration first detection data, the relatively long-duration second detection data, and the voltage anomaly detection results to analyze the short-term, long-term, and voltage conditions of each battery cell, obtaining voltage range detection results. This allows for more comprehensive voltage range detection of the battery using detection data obtained from different time spans, reducing the possibility of distortion in detection results due to strong randomness or weak generalization of the detection data, and improving the accuracy of voltage range detection.

[0027] In some embodiments, the first detection data includes at least one of the voltage range of each cell in the battery, the voltage change rate of the battery, and the average voltage of each cell in the battery during a first time period; and / or, the second detection data includes at least one of the voltage range of each cell in the battery, the average voltage of each cell in the battery, the average number of daily charge cycles of the battery, the average number of daily discharge cycles of the battery, and the maximum value of the temperature of each cell in the battery during charging during a second time period.

[0028] In the technical solution of this application embodiment, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first or second detection data includes at least one of the following: voltage range of each battery cell, average voltage of each battery cell, average daily charge cycle count, average daily discharge cycle count, and maximum temperature of each battery cell during charging. The first detection data is then input into a first detection model, which uses the first detection data to detect the probability of voltage anomalies in each battery cell, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model, which uses the voltage anomaly detection result and the first detection data to obtain a voltage anomaly detection result. The first detection model uses first detection data (corresponding to a relatively short duration) to analyze the short-term condition of each battery cell and obtain voltage range detection results. The second detection model combines the short-term first detection data, the long-term second detection data, and the voltage range detection results to analyze the short-term, long-term, and voltage conditions of each battery cell and obtain voltage range detection results. This allows for more comprehensive voltage range detection of the battery using detection data obtained from different time spans, reducing the possibility of distortion due to strong randomness or weak generalization of the detection data and improving the accuracy of voltage range detection.

[0029] Secondly, this application provides a battery management system, including: a processing module, the processing module being used to execute the above-described method.

[0030] Thirdly, this application provides a battery system including the aforementioned battery management system.

[0031] Fourthly, this application provides an electrical device including the aforementioned battery system.

[0032] Fifthly, this application provides a computer-readable storage medium storing program instructions that, when executed by a processor, implement the above-described method.

[0033] It is understood that the beneficial effects of the second, third, fourth and fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0034] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A schematic flowchart of one or more embodiments of the battery voltage range detection method provided in this application;

[0037] Figure 2 A second schematic flowchart of one or more embodiments of the battery voltage difference detection method provided in this application;

[0038] Figure 3 This is a schematic diagram of the structure of one or more embodiments of the first detection model and the second detection model provided in this application;

[0039] Figure 4 A schematic diagram of the structure of one or more embodiments of the battery management system provided in this application;

[0040] Figure 5 A schematic diagram of the structure of one or more embodiments of the battery system provided in this application;

[0041] Figure 6 A schematic diagram of the structure of one or more embodiments of the electrical device provided in this application;

[0042] Figure 7 A schematic diagram of one or more embodiments of the computer-readable storage medium provided in this application. Detailed Implementation

[0043] 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. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0045] In the description of the embodiments of this application, the technical terms "first", "second", etc. are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features.

[0046] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0047] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0048] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0049] In the description of the embodiments of this application, unless otherwise expressly specified and limited, the technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0050] With the development of new energy vehicles, the safety requirements for batteries that power these vehicles are becoming increasingly stringent. Typically, voltage difference detection is required to determine whether there are any abnormalities in the voltage difference between individual battery cells. If an abnormality is detected, appropriate early warning measures must be taken to reduce losses and minimize the possibility of safety accidents.

[0051] However, existing methods for detecting voltage range in batteries are inaccurate because the data they acquire is typically short-term, with a high degree of randomness and weak generalization.

[0052] Existing battery voltage range detection methods typically rely on short-term charge / discharge data of the batteries in new energy vehicles to be tested, or directly detect the current voltage status of the battery. After analysis, the battery voltage range is determined, enabling corresponding risk assessment and early warning. However, it's important to note that because the battery detection data collected by existing methods is relatively uniform in duration and is short-term, the randomness of the data is amplified, resulting in poor generalization. This makes the detection results highly susceptible to variations in battery operating conditions. For example, a battery corresponding to a set of detection data might be considered normal under one operating condition but abnormal under another. Due to the strong randomness and weak generalization of the data obtained by existing methods, the trained detection model may fail to accurately determine the risk of abnormal voltage range, thus failing to obtain accurate voltage range detection results.

[0053] Based on the above considerations, this application provides a method, system, device, and storage medium for detecting voltage range in a battery. The voltage range detection method includes: acquiring first detection data and second detection data of the battery; the first detection data includes detection data within a first time period, and the second detection data includes detection data within a second time period; the end time of both the first and second time periods is the current time, and the duration of the first time period is shorter than the duration of the second time period; the battery includes multiple individual battery cells; inputting the first detection data into a first detection model to obtain voltage anomaly detection results corresponding to the voltage of each individual battery cell; inputting the voltage anomaly detection results, the first detection data, and the second detection data into a second detection model to obtain voltage range detection results corresponding to the battery. Based on the above method, the accuracy of voltage range detection can be improved.

[0054] See Figure 1 , Figure 1 This is a schematic flowchart of one or more embodiments of the battery voltage difference detection method provided in this application.

[0055] like Figure 1 As shown, the battery voltage range detection method includes:

[0056] Step S11: Obtain the first and second detection data of the battery.

[0057] The first detection data includes detection data within a first time period, the second detection data includes detection data within a second time period, the end time of the first time period and the end time of the second time period are both the current time, the duration of the first time period is shorter than the duration of the second time period, and the battery includes multiple battery cells.

[0058] For example, the first time period can be a period of several seconds, minutes, or hours before the current time, so that the first detection data can reflect the short-term characteristics of the battery.

[0059] The second time period can be a period of several months or years prior to the current moment, so that the second detection data can reflect the medium-term and / or long-term characteristics of the battery.

[0060] The first and second time periods can also be combinations of time periods with different durations, as in other examples; this is not limited here.

[0061] After the battery acquires the test data, it can be uploaded to a cloud server for network storage, or it can be transported to the device where the battery is located (such as a new energy vehicle or other device equipped with a battery) for local storage. There is no limitation here.

[0062] The first and second detection data may specifically include data detected by various types of sensors installed in the battery. The detection data at different time periods can reflect the battery's operating conditions in the short and long term, respectively, so that subsequent detection models can perform corresponding analysis and processing. The specific data types of detection can be determined according to the sensors installed, and are not limited here.

[0063] The acquired detection data can be preprocessed accordingly. For example, each detection data can be processed according to the requirements of GB / T32960, and / or, noise removal processing, and / or, outlier removal processing, and / or, missing value removal processing, and / or, other types of preprocessing, which are not limited here.

[0064] Step S12: Input the first detection data into the first detection model to obtain the voltage anomaly detection results corresponding to the voltage of each battery cell.

[0065] Specifically, the first detection model can be a random forest model or other types of detection models, which can output the voltage status of each battery cell based on the first detection data, and obtain the corresponding voltage anomaly detection results.

[0066] Step S13: Input the voltage anomaly detection result, the first detection data, and the second detection data into the second detection model to obtain the voltage range detection result corresponding to the battery.

[0067] The second detection model can be a gradient boosting decision tree model (i.e., the LightGBM model) or other types of detection models.

[0068] The voltage anomaly detection results, which characterize the voltage status of each individual cell in the battery, the first detection data, which characterizes the battery's operating condition over a relatively short period of time, and the second detection data, which characterizes the battery's operating condition over a relatively long period of time, can be input into the second detection model. This allows the second detection model to reliably analyze the current status or operating condition of the battery from three dimensions based on the voltage anomaly detection results, the first detection data, and the second detection data. This reduces the possibility of voltage range detection results being distorted due to the randomness of the detection data and improves the generalization of the detection data, thereby improving the accuracy of voltage range detection.

[0069] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each individual cell of the battery, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each individual cell of the battery, thus obtaining a voltage range detection result. This achieves the goal of passing the first detection... The model analyzes and processes the short-term condition of each battery cell in the battery based on the first detection data with a relatively short corresponding time span, and obtains the voltage anomaly detection result. Then, through a second detection model, it combines the first detection data with a relatively short corresponding time span, the second detection data with a relatively long corresponding time span, and the voltage anomaly detection result to analyze and process the short-term condition, long-term condition, and voltage condition of each battery cell in the battery, and obtains the voltage range detection result. This enables a more comprehensive voltage range detection of the battery based on the detection data obtained from different time spans, reduces the possibility of distortion of detection results due to strong randomness or weak generalization of detection data, and improves the accuracy of voltage range detection.

[0070] In some embodiments, the loss function of the second detection model includes the sum of a loss term and a constraint penalty term.

[0071] The constraint penalty term includes at least one of a first penalty term, a second penalty term, and a third penalty term. Specifically, the first penalty term is associated with the state of charge (SOC) value of the battery in the training samples, the second penalty term is associated with the number of cycles of the battery in the training samples, and the third penalty term is associated with the current value of the battery in the training samples.

[0072] Specifically, the second detection model can be the LightGBM model or other types of detection models, which are not limited here.

[0073] The loss function of the second detection model can be specifically shown below:

[0074]

[0075]

[0076] in, The loss function of the second detection model is... For loss items, To constrain penalties, The weighting coefficients for the constraint penalty term, It can typically be between 0.01 and 0.1.

[0077] The first penalty term is used to apply physical constraints on voltage polarity anomalies based on the battery's state of charge (SOC). This is the second penalty term, used to apply appropriate aging physical constraints based on the number of battery cycles. This is the third penalty term, used to apply appropriate physical constraints on the current based on the battery current. The weighting coefficient for the first penalty term. The weighting coefficient for the second penalty term. This is the weighting coefficient for the third penalty term.

[0078] The weight coefficients corresponding to the first, second, and third penalty terms can be adaptively set according to the type of battery to improve the reliability of the loss function.

[0079] For example, the loss term is the cross-entropy loss term. If the loss function is constructed using the Balanced Binary Cross Entropy (Balanced BCE) function, it can be as follows:

[0080]

[0081] in, For loss items, This represents the total number of training samples input into the second detection model during training. Each training sample includes first and second detection data for one battery. The sample weights are typically the quotient of the number of positive samples divided by the number of negative samples in all training samples. Positive samples are training samples of batteries with normal voltage ranges, while negative samples are training samples of batteries with abnormal voltage ranges. The true label of the i-th training sample among all training samples (e.g., normal voltage range or abnormal voltage range). This represents the predicted probability of a normal voltage range obtained by the second detection model when predicting the voltage range for the i-th training sample. The base can be e or other preset values, which are not limited here.

[0082] The above are just examples. In other examples, the loss term can be constructed by other formulas, or the loss term can be replaced by other types of loss terms. This is not limited here.

[0083] In this application, by setting a constraint penalty term in the loss function of the second detection model in addition to the loss term, and setting a first penalty term related to the state of charge value of the battery in the training sample, a second penalty term related to the number of cycles of the battery in the training sample, and a third penalty term related to the current value of the battery in the training sample, the loss function of the second detection model can be adjusted accordingly for batteries in different states. Thus, after training with the help of this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0084] Optionally, the first penalty term is the average of the first values ​​corresponding to the batteries in all training samples.

[0085] The first value is 0 when the second value is less than or equal to 0, and the second value is greater than 0.

[0086] The second value is the difference between the first constraint boundary value and the corresponding battery voltage extreme difference anomaly prediction probability.

[0087] The first constraint boundary value is associated with the state of charge value of the battery in the training samples.

[0088] Specifically, the first constraint boundary value can be the function value of the first Sigmoid function.

[0089] The parameter value of the first sigmoid function is the difference between the third value and the first coefficient.

[0090] The third value is the product of the second coefficient and the fourth value.

[0091] The fourth value is the square of the difference between the corresponding battery's state of charge value and 0.5.

[0092] In one example, the first penalty term could be as follows:

[0093]

[0094]

[0095] in, As the first penalty item, This represents the total number of training samples input into the second detection model during training. The voltage range anomaly prediction probability is obtained by the second detection model from predicting the voltage range of each cell in the corresponding training sample. (x) is the Sigmoid function (that is, the Sigmoid function can be...). ,For example, For function values, (for parameter values) This is the first Sigmoid function. As the first coefficient, As the second coefficient, This represents the state of charge (SOC) value of the battery corresponding to the training sample.

[0096] The remaining energy level of a battery can be reflected by its state of charge (SOC) value.

[0097] The first coefficient can be used to adjust the physical sensitivity, and is usually 2.0-3.0.

[0098] By using the above method of determining the first value, when the risk value predicted by the thermal runaway detection data model is lower than the actual value, a corresponding penalty mechanism can be set to constrain this underestimation behavior, thereby improving the reliability of the second detection model through the loss function.

[0099] In this application, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge (SOC) value of the batteries in the training samples, a second penalty term associated with the number of cycles of the batteries in the training samples, and a third penalty term associated with the current value of the batteries in the training samples. The average value of the first value corresponding to all batteries in the training samples is calculated. The first value is 0 when the second value is less than or equal to 0 and is the second value when the second value is greater than 0. The second value can be determined based on the first constraint boundary value obtained from the function value of the first Sigmoid function to perform boundary constraints. This yields the first penalty term that reflects the SOC value of each battery. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0100] Optionally, the second penalty term is the average of the fifth values ​​corresponding to the batteries in all training samples.

[0101] The fifth value is 0 when the sixth value is less than or equal to 0, and the sixth value is greater than 0.

[0102] The sixth value is the difference between the second constraint boundary value and the corresponding battery voltage extreme difference anomaly prediction probability.

[0103] The second constraint boundary value is related to the number of cycles of the batteries in the training samples.

[0104] Specifically, the second constraint boundary value can be the function value of the second Sigmoid function.

[0105] The parameter value of the second Sigmoid function is the difference between the seventh value and the third coefficient.

[0106] The seventh value is the product of the fourth coefficient and the eighth value.

[0107] The eighth value is the quotient obtained by dividing the number of cycles of the corresponding battery by the maximum number of cycles threshold.

[0108] In one example, the second penalty term could be as follows:

[0109]

[0110]

[0111] in, As the second penalty item, This represents the total number of training samples input into the second detection model during training. The voltage range anomaly prediction probability is obtained by the second detection model from predicting the voltage range of each cell in the corresponding training sample. (x) is the Sigmoid function (that is, the Sigmoid function can be...). ,For example, For function values, (for parameter values) This is the second Sigmoid function. The third coefficient, It is the fourth coefficient. Let be the number of cycles for the battery corresponding to the i-th training sample. This is the maximum cycle count threshold for the battery.

[0112] The number of battery cycles can reflect the degree of battery aging. Aging is an irreversible process. The higher the degree of aging, the lower the safety and the higher the possibility of voltage difference abnormalities.

[0113] The third coefficient can be used to adjust the physical sensitivity, also known as the bias term, and is usually 2.0-3.0.

[0114] The fourth factor, also known as the aging factor, is usually between 2.0 and 4.0.

[0115] By using the above-mentioned method of determining the fifth value, when the risk value predicted by the thermal runaway detection data model is lower than the actual value, a corresponding penalty mechanism can be set to constrain this underestimation behavior, thereby improving the reliability of the second detection model through the loss function.

[0116] In this application, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term associated with the state of charge value of the batteries in the training samples, a second penalty term associated with the number of cycles of the batteries in the training samples, and a third penalty term associated with the current value of the batteries in the training samples. The average value of the fifth value corresponding to all batteries in the training samples is calculated. The fifth value is 0 when the sixth value is less than or equal to 0 and is the sixth value when the sixth value is greater than 0. The sixth value can be determined based on the second constraint boundary value obtained from the function value of the second sigmoid function to perform boundary constraints. This yields the second penalty term that reflects the number of cycles of each battery. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0117] Optionally, the third penalty term is the average of the ninth values ​​corresponding to the batteries in all training samples.

[0118] The ninth value is 0 when the tenth value is less than or equal to 0, and the tenth value is greater than 0.

[0119] The tenth value is the difference between the third constraint boundary value and the corresponding battery voltage extreme difference anomaly prediction probability.

[0120] The third constraint boundary value is related to the current value of the battery in the training samples.

[0121] Specifically, the third constraint boundary value is the function value of the third Sigmoid function.

[0122] The parameter value of the third Sigmoid function is the difference between the eleventh value and the fifth coefficient.

[0123] The eleventh value is the product of the sixth coefficient and the twelfth value.

[0124] The twelfth value is the quotient obtained by dividing the absolute value of the current of the corresponding battery by the maximum current threshold.

[0125] In one example, the third penalty term could be as follows:

[0126]

[0127]

[0128] in, As the third penalty item, This represents the total number of training samples input into the second detection model during training. The voltage range anomaly prediction probability is obtained by the second detection model from predicting the voltage range of each cell in the corresponding training sample. (x) is the Sigmoid function (that is, the Sigmoid function can be...). ,For example, For function values, (for parameter values) This is the third Sigmoid function. It is the fifth coefficient. It is the sixth coefficient. Let be the current value of the battery corresponding to the i-th training sample. This is the maximum current threshold of the battery.

[0129] The battery's current value can reflect the battery's ohmic voltage drop and thermal effect. Under high current conditions, the battery's internal resistance difference is amplified, and the possibility of abnormal voltage range is higher.

[0130] The fifth coefficient is usually between 2.0 and 4.0.

[0131] The sixth coefficient, also known as the current coefficient, is usually 2.5-5.0.

[0132] By using the above method of determining the value of the ninth value, when the risk value predicted by the thermal runaway detection data model is lower than the actual value, a corresponding penalty mechanism can be set to constrain this underestimation behavior, thereby improving the reliability of the second detection model through the loss function.

[0133] In this application, in addition to the loss term, a constraint penalty term is set in the loss function of the second detection model. The constraint penalty term includes a first penalty term related to the state of charge value of the battery in the training sample, a second penalty term related to the number of cycles of the battery in the training sample, and a third penalty term related to the current value of the battery in the training sample. The average of the ninth value corresponding to all batteries in the training sample is calculated. The ninth value is 0 when the tenth value is less than or equal to 0 and is the tenth value when the tenth value is greater than 0. The tenth value can be determined based on the third constraint boundary value obtained from the function value of the third sigmoid function to perform boundary constraints. The third penalty term used to reflect the current status of each battery is obtained. This allows the loss function of the second detection model to be adjusted accordingly for batteries in different states. As a result, after training with this loss function, the second detection model can improve the accuracy of voltage range detection for batteries in different states.

[0134] In some embodiments, see Figure 2 , Figure 2 This is a second schematic flowchart of one or more embodiments of the battery voltage difference detection method provided in this application.

[0135] like Figure 2 As shown, step S12 may specifically include:

[0136] Step S121: Input the first detection data into the first detection model to obtain the abnormal probability of the voltage change of each battery cell over time.

[0137] Step S122: Process the abnormal probability of the voltage change of the battery cell over time to obtain the voltage abnormality detection result.

[0138] Among them, the voltage anomaly detection results include at least one of the following: average, maximum, range, variance, skewness, kurtosis, and quantile of the abnormal probability of voltage change of individual cells over time.

[0139] Specifically, the functions of the first detection model may include:

[0140]

[0141] in, Let be the abnormal probability of the voltage change of the j-th cell in the corresponding battery over time (the value is between 0 and 1). For any time, Let be the feature vector of the first detection data corresponding to the j-th battery cell. Let be a function of the first detection model. If the first detection model is a random forest model, then this function is a function of the random forest classifier.

[0142] Subsequently, based on the abnormal probability of the voltage change of individual battery cells over time, data statistics can be performed to obtain:

[0143] average value:

[0144] Maximum value:

[0145] Voltage range:

[0146] variance:

[0147] Skewness:

[0148] Kuroshi:

[0149] Quantiles: at least one of the 10th quantile, 20th quantile, 30th quantile, 90th quantile, and other quantiles.

[0150] For example, the 90th percentile:

[0151]

[0152] in, This represents the total number of individual battery cells in the corresponding battery. To find the maximum value function, To find the minimum value function, This is a function for calculating quantiles.

[0153] Based on the above method, the abnormal probability of voltage change of each cell in the corresponding battery can be calculated according to the characteristics of the first detection data, and at least one of the average, maximum, range, variance, skewness, kurtosis and quantile of the abnormal probability of each cell can be obtained by further statistical analysis, forming a voltage abnormality detection result for use by the second detection model.

[0154] This enables the second detection model to not only analyze short-term and medium-term / long-term data based on the first and second detection data, but also to assess the probability of anomalies in each individual battery cell based on the voltage anomaly detection results. This allows for a more comprehensive analysis and processing of the battery, resulting in more accurate voltage range detection results and improving the accuracy of voltage range detection.

[0155] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each individual cell of the battery, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each individual cell of the battery, thus obtaining a voltage range detection result. This allows the first detection model to analyze and process the corresponding relatively short period of first detection data. The method determines the abnormal probability of voltage changes of each cell in the battery over time. Then, based on the abnormal probability corresponding to each cell, it determines various data to reflect the short-term condition of each cell in the voltage anomaly detection result. By combining the first detection data with a relatively short corresponding time period, the second detection data with a relatively long corresponding time period, and the voltage anomaly detection result through a second detection model, the short-term condition, long-term condition, and voltage condition of each cell in the battery are analyzed and processed to obtain the voltage range detection result. This enables a more comprehensive voltage range detection of the battery based on the detection data obtained from different time spans, reduces the possibility of distortion of detection results due to strong randomness or weak generalization of detection data, and improves the accuracy of voltage range detection.

[0156] Optionally, the first detection model includes a first classifier and a second classifier, and the anomaly probability includes overvoltage probability and undervoltage probability.

[0157] Step S121 may include:

[0158] The first detection data is input into the first classifier to obtain the overvoltage probability of each battery cell's voltage change over time. The first detection data is then input into the second classifier to obtain the undervoltage probability of each battery cell's voltage change over time.

[0159] The voltage anomaly detection results include at least one of the following: average, maximum, range, variance, skewness, kurtosis, and quantile of the overvoltage probability; and at least one of the following: average, maximum, range, variance, skewness, kurtosis, and quantile of the undervoltage probability.

[0160] Specifically, in one example, see Figure 3 , Figure 3 This is a schematic diagram of the structure of one or more embodiments of the first detection model and the second detection model provided in this application.

[0161] The original battery data can be obtained from the cloud server and / or the local storage of the device where the battery is located. Then, the first detection data can be extracted based on the first data extractor (for example, the first detection data can be obtained by processing the data of the first time period in the original data with a sliding time window algorithm). The second detection data can be extracted based on the second data extractor (for example, the second detection data can be obtained by performing working condition behavior data analysis on the data of the second time period in the original data). The first data extractor and the second data extractor can be the same processing device or different processing devices, which is not limited here.

[0162] After inputting the first detection data into the first detection model, assuming the first detection model is a random forest model, the first classifier is the first random forest classifier, and the second classifier is the second random forest classifier. The first classifier can output the overvoltage probability of each battery cell's voltage change over time, and the second classifier can output the undervoltage probability of each battery cell's voltage change over time.

[0163] In the first classifier, the overvoltage probability of each battery cell's voltage change over time when the anomaly is overvoltage can be calculated according to the anomaly probability calculation method mentioned above. In the second classifier, the undervoltage probability of each battery cell's voltage change over time can be calculated according to the anomaly probability calculation method mentioned above.

[0164] Furthermore, at least one of the following is obtained for each overvoltage probability: average value, maximum value, range, variance, skewness, kurtosis, and quantile; and at least one of the following is obtained for each undervoltage probability: average value, maximum value, range, variance, skewness, kurtosis, and quantile, forming a voltage anomaly detection result, which is then input into the second detection model.

[0165] The first and second detection data are also input into the second detection model so that the second detection model can determine whether there are problems such as excessive voltage difference or overcharging / over-discharging in the battery based on the specific operating conditions of the battery (such as determining the current charge of the battery by the state of charge of each battery, or determining the degree of aging by the number of cycles of each battery, or determining whether there is overcharging or over-discharging by the current of each battery). This allows the second detection model to assess the possibility of thermal runaway and other problems. The relatively short period of the first detection data can improve the timeliness of the final voltage difference detection, while the relatively long period of the second detection data can improve the accuracy of the final voltage difference detection. This enables the battery to perform voltage difference detection from multiple dimensions with high timeliness, accuracy and reliability, and obtain voltage difference detection results to effectively assess the possibility of voltage difference anomalies in the battery, thereby improving the safety of the battery and the equipment in which the battery is located.

[0166] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each individual cell of the battery, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses these data to detect the probability of voltage range anomalies in each individual cell of the battery, thus obtaining a voltage range detection result. This allows the first and second classifiers of the first detection model to analyze the corresponding relatively short period of first detection data. The method determines the undervoltage and overvoltage probabilities of each cell's voltage over time. Based on these probabilities, it determines various data points reflecting the short-term condition of each cell in the battery. A second detection model combines the relatively short-duration first detection data, the relatively long-duration second detection data, and the voltage anomaly detection results to analyze the short-term, long-term, and voltage conditions of each cell, yielding voltage range detection results. This allows for comprehensive voltage range detection of the battery using data obtained from different time spans, reducing the possibility of data distortion due to strong randomness or weak generalization, and improving the accuracy of voltage range detection.

[0167] In some embodiments, the first detection data includes at least one of voltage detection data, cycle detection data, and temperature detection data within a first time period.

[0168] And / or,

[0169] The second detection data includes at least one of voltage detection data, cycle detection data, and temperature detection data during the second time period.

[0170] Specifically, the first detection data includes the voltage range of each cell in the battery during the corresponding time period, the voltage change rate of the battery, and the average voltage of each cell in the battery.

[0171] And / or,

[0172] The second set of test data includes the voltage range of each cell in the battery during the corresponding time period, the average voltage of each cell in the battery, the average number of daily charging cycles of the battery, the average number of daily discharging cycles of the battery, and the maximum temperature of each cell in the battery during charging.

[0173] Specifically, the first and second detection data may contain data of the same type but with different durations in the corresponding time periods, or they may not contain data of the same type. The specific choice depends on the actual needs and is not limited here.

[0174] In one example, the first detection data or the second detection data includes at least one of voltage detection data, cycle detection data and temperature detection data during the second time period. For example, the first detection data or the second detection data includes at least one of the voltage range of each cell of the battery, the average voltage of each cell of the battery, the average number of daily charging cycles of the battery, the average number of daily discharging cycles of the battery, and the maximum value of the temperature of each cell of the battery during charging during the corresponding time period.

[0175] The calculation methods for each type of detection data are as follows:

[0176] 1. Voltage range of each cell in the battery:

[0177]

[0178] Where D represents the voltage range of each individual cell in the battery. Let be the voltage value of the i-th cell in the battery.

[0179] 2. Battery voltage change rate:

[0180]

[0181] in, This represents the rate of change of battery voltage. The interval duration refers to the duration of the interval; different time periods correspond to different interval durations. This represents the battery voltage value after a certain time interval (e.g., the battery voltage value at the current moment). This is the battery voltage value before the interval (e.g., the battery voltage value before the current interval).

[0182] 3. Average voltage of each individual cell in the battery:

[0183]

[0184] in, This represents the average voltage of each individual cell in the battery. Let be the voltage value of the i-th cell in the battery. This represents the total number of individual battery cells in the battery.

[0185] 4. The average number of charging cycles per day for a battery is the value obtained by dividing the total number of charging cycles within the corresponding time period by the total number of days within the corresponding time period.

[0186] 5. The average number of discharge cycles per day for a battery is the value obtained by dividing the total number of discharge cycles in the corresponding period by the total number of days in the corresponding period.

[0187] 6. The maximum temperature of each individual cell during charging is the maximum value among all the maximum temperatures reached by the entire battery cell during charging.

[0188] In another example, the acquired detection data (such as any of the above-mentioned voltage range, voltage change rate, average voltage, average daily charge cycle count, average daily discharge cycle count, and the maximum temperature of each battery cell during charging) can be normalized, and then input into the second detection model after normalization. The normalization formula is shown below:

[0189]

[0190] in, The value obtained after normalizing the corresponding type of detection data. For the corresponding type of detection data, This represents the average value of the corresponding type of detection data. This represents the standard deviation of the corresponding type of test data.

[0191] In yet another example, the above normalization process is performed only on the data input to the second detection model.

[0192] Based on the above method, the accuracy of voltage range detection in the second detection model can be improved.

[0193] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first or second detection data includes at least one of the following: voltage range of each battery cell, average voltage of each battery cell, average daily charge cycle count, average daily discharge cycle count, and maximum temperature of each battery cell during charging. The first detection data is then input into a first detection model, which uses the first detection data to detect the probability of voltage anomalies in each battery cell, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model, which uses the voltage anomaly detection result, the first detection data, and the second detection data to obtain a voltage anomaly detection result. The two detection models detect the probability of abnormal voltage ranges in each cell of the battery, obtaining voltage range detection results. The first detection model analyzes the short-term condition of each cell using relatively short-duration first detection data to obtain voltage anomaly detection results. The second detection model combines the relatively short-duration first detection data, the relatively long-duration second detection data, and the voltage anomaly detection results to analyze the short-term, long-term, and voltage conditions of each cell, obtaining voltage range detection results. This allows for more comprehensive voltage range detection of the battery using detection data obtained from different time spans, reducing the possibility of distortion caused by strong randomness or weak generalization of the detection data, and improving the accuracy of voltage range detection.

[0194] Please refer to Figure 4 , Figure 4 This is a schematic diagram of one or more embodiments of the battery management system provided in this application.

[0195] like Figure 4 As shown, the battery management system 20 includes a processing module 21, which is used to execute the voltage range detection method described in any of the preceding embodiments.

[0196] Specifically, a battery management system can refer to a BMS (BATTERY MANAGEMENT SYSTEM). A battery management system can be used to regulate and monitor the individual batteries in a battery system, reduce the occurrence of overcharging or over-discharging or other malfunctions, extend the life of the battery system, and improve the reliability of the battery system.

[0197] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each individual cell of the battery, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each individual cell of the battery, thus obtaining a voltage range detection result. This achieves the goal of passing the first detection... The model analyzes and processes the short-term condition of each battery cell in the battery based on the first detection data with a relatively short corresponding time span, and obtains the voltage anomaly detection result. Then, through a second detection model, it combines the first detection data with a relatively short corresponding time span, the second detection data with a relatively long corresponding time span, and the voltage anomaly detection result to analyze and process the short-term condition, long-term condition, and voltage condition of each battery cell in the battery, and obtains the voltage range detection result. This enables a more comprehensive voltage range detection of the battery based on the detection data obtained from different time spans, reduces the possibility of distortion of detection results due to strong randomness or weak generalization of detection data, and improves the accuracy of voltage range detection.

[0198] Please refer to Figure 5 , Figure 5 This is a schematic diagram of one or more embodiments of the battery system provided in this application.

[0199] like Figure 5 As shown, the battery system 30 includes a battery management system 20.

[0200] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each individual cell of the battery, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each individual cell of the battery, thus obtaining a voltage range detection result. This achieves the goal of passing the first detection... The model analyzes and processes the short-term condition of each battery cell in the battery based on the first detection data with a relatively short corresponding time span, and obtains the voltage anomaly detection result. Then, through a second detection model, it combines the first detection data with a relatively short corresponding time span, the second detection data with a relatively long corresponding time span, and the voltage anomaly detection result to analyze and process the short-term condition, long-term condition, and voltage condition of each battery cell in the battery, and obtains the voltage range detection result. This enables a more comprehensive voltage range detection of the battery based on the detection data obtained from different time spans, reduces the possibility of distortion of detection results due to strong randomness or weak generalization of detection data, and improves the accuracy of voltage range detection.

[0201] Please refer to Figure 6 , Figure 6 This is a schematic diagram of one or more embodiments of the electrical device provided in this application.

[0202] like Figure 6 As shown, the electrical device 40 includes a battery system 30.

[0203] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each individual cell of the battery, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each individual cell of the battery, thus obtaining a voltage range detection result. This achieves the goal of passing the first detection... The model analyzes and processes the short-term condition of each battery cell in the battery based on the first detection data with a relatively short corresponding time span, and obtains the voltage anomaly detection result. Then, through a second detection model, it combines the first detection data with a relatively short corresponding time span, the second detection data with a relatively long corresponding time span, and the voltage anomaly detection result to analyze and process the short-term condition, long-term condition, and voltage condition of each battery cell in the battery, and obtains the voltage range detection result. This enables a more comprehensive voltage range detection of the battery based on the detection data obtained from different time spans, reduces the possibility of distortion of detection results due to strong randomness or weak generalization of detection data, and improves the accuracy of voltage range detection.

[0204] See Figure 7 , Figure 7 A schematic diagram of one or more embodiments of the computer-readable storage medium provided in this application.

[0205] like Figure 7 As shown, the computer-readable storage medium 50 stores program instructions 51 thereon, which, when executed by a processor (not shown), implement the voltage range detection method in the above embodiments.

[0206] In this embodiment, the computer-readable storage medium 50 may be, but is not limited to, a USB flash drive, SD card, PD optical drive, portable hard drive, high-capacity floppy drive, flash memory, multimedia memory card, storage unit in a server, FPGA, or ASIC, etc.

[0207] In this application, by acquiring first detection data of the battery over a relatively short period of time and second detection data of the battery over a relatively long period of time, the first detection data is input into a first detection model. The first detection model then uses the first detection data to detect the probability of voltage anomalies in each individual cell of the battery, thus obtaining a voltage anomaly detection result. The voltage anomaly detection result, the first detection data, and the second detection data are then input into a second detection model. The second detection model then uses the voltage anomaly detection result, the first detection data, and the second detection data to detect the probability of voltage range anomalies in each individual cell of the battery, thus obtaining a voltage range detection result. This achieves the goal of passing the first detection... The model analyzes and processes the short-term condition of each battery cell in the battery based on the first detection data with a relatively short corresponding time span, and obtains the voltage anomaly detection result. Then, through a second detection model, it combines the first detection data with a relatively short corresponding time span, the second detection data with a relatively long corresponding time span, and the voltage anomaly detection result to analyze and process the short-term condition, long-term condition, and voltage condition of each battery cell in the battery, and obtains the voltage range detection result. This enables a more comprehensive voltage range detection of the battery based on the detection data obtained from different time spans, reduces the possibility of distortion of detection results due to strong randomness or weak generalization of detection data, and improves the accuracy of voltage range detection.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application. In particular, as long as there is no structural conflict, the various technical features mentioned in the embodiments can be combined in any way. This application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for detecting the voltage range of a battery, characterized in that, include: Acquire first and second detection data of the battery; wherein the first detection data includes voltage detection data within a first time period, the second detection data includes voltage detection data within a second time period, the end time of the first time period and the end time of the second time period are both the current time, the duration of the first time period is less than the duration of the second time period, and the battery includes multiple battery cells; The first detection data is input into the first detection model to obtain the abnormal probability of the voltage change of each battery cell over time; wherein, the abnormal probability includes overvoltage probability and undervoltage probability; The abnormal probability of the voltage change of the battery cell over time is processed to obtain the voltage abnormality detection result; The voltage anomaly detection result, the first detection data, and the second detection data are input into the second detection model, so that the second detection model detects the probability of an abnormal voltage range of each battery cell in the battery through the voltage anomaly detection result, the first detection data, and the second detection data, and obtains the voltage range detection result corresponding to the battery.

2. The voltage range detection method according to claim 1, characterized in that, The loss function of the second detection model includes the sum of a loss term and a constraint penalty term; The constraint penalty term includes at least one of a first penalty term, a second penalty term, and a third penalty term; wherein the first penalty term is associated with the state of charge value of the battery in the training sample, the second penalty term is associated with the number of cycles of the battery in the training sample, and the third penalty term is associated with the current value of the battery in the training sample.

3. The voltage range detection method according to claim 2, characterized in that, The first penalty term is the average of the first values ​​corresponding to the batteries in all the training samples; The first value is 0 when the second value is less than or equal to 0, and is the second value when the second value is greater than 0; The second value is the difference between the first constraint boundary value and the corresponding voltage extreme difference anomaly prediction probability of the battery. The first constraint boundary value is associated with the state of charge value of the battery in the training sample.

4. The voltage range detection method according to claim 3, characterized in that, The first constraint boundary value is the function value of the first Sigmoid function; The parameter value of the first Sigmoid function is the difference between the third value and the first coefficient; The third value is the product of the second coefficient and the fourth value; The fourth value is the square of the difference between the state of charge value of the corresponding battery and 0.

5.

5. The voltage range detection method according to claim 2, characterized in that, The second penalty term is the average of the fifth values ​​corresponding to the batteries in all the training samples; The fifth value is 0 when the sixth value is less than or equal to 0, and is the sixth value when the sixth value is greater than 0; The sixth value is the difference between the second constraint boundary value and the corresponding voltage extreme difference anomaly prediction probability of the battery. The second constraint boundary value is associated with the number of cycles of the battery in the training samples.

6. The voltage range detection method according to claim 5, characterized in that, The second constraint boundary value is the function value of the second Sigmoid function; The parameter value of the second Sigmoid function is the difference between the seventh value and the third coefficient; The seventh value is the product of the fourth coefficient and the eighth value. The eighth value is the quotient obtained by dividing the number of cycles of the corresponding battery by the maximum number of cycles threshold.

7. The voltage range detection method according to claim 2, characterized in that, The third penalty term is the average of the ninth values ​​corresponding to the batteries in all the training samples; The ninth value is 0 when the tenth value is less than or equal to 0, and is the tenth value when the tenth value is greater than 0; The tenth value is the difference between the third constraint boundary value and the corresponding voltage extreme difference anomaly prediction probability of the battery. The third constraint boundary value is associated with the current value of the battery in the training sample.

8. The voltage range detection method according to claim 7, characterized in that, The third constraint boundary value is the function value of the third Sigmoid function; The parameter value of the third Sigmoid function is the difference between the eleventh value and the fifth coefficient; The eleventh value is the product of the sixth coefficient and the twelfth value; The twelfth value is the quotient obtained by dividing the absolute value of the current of the corresponding battery by the maximum current threshold.

9. The voltage range detection method according to any one of claims 1 to 5, characterized in that, The voltage anomaly detection results include at least one of the following: average, maximum, range, variance, skewness, kurtosis, and quantile, representing the anomaly probability of the voltage change of the individual battery cell over time.

10. The voltage range detection method according to claim 9, characterized in that, The first detection model includes a first classifier and a second classifier; The step of inputting the first detection data into the first detection model to obtain the abnormal probability of the voltage change of each battery cell over time includes: The first detection data is input into the first classifier to obtain the overvoltage probability of the voltage of each battery cell changing over time. The first detection data is also input into the second classifier to obtain the undervoltage probability of the voltage of each battery cell changing over time. The voltage anomaly detection results include at least one of the following: the average value, maximum value, range, variance, skewness, kurtosis, and quantile of the overvoltage probability; and at least one of the following: the average value, maximum value, range, variance, skewness, kurtosis, and quantile of the undervoltage probability.

11. The voltage range detection method according to any one of claims 1 to 5, characterized in that, The first detection data also includes at least one of the cyclic detection data and temperature detection data within the first time period; And / or, The second detection data also includes at least one of the cyclic detection data and temperature detection data within the second time period.

12. The voltage range detection method according to claim 11, characterized in that, The first detection data includes at least one of the following during the first time period: the voltage range of each of the battery cells, the voltage change rate of the battery, and the average voltage of each of the battery cells. And / or, The second detection data includes at least one of the following during the second time period: the voltage range of each battery cell of the battery, the average voltage of each battery cell of the battery, the average number of daily charging cycles of the battery, the average number of daily discharging cycles of the battery, and the maximum value of the temperature of each battery cell of the battery during charging.

13. A battery management system, characterized in that, include: A processing module, the processing module being configured to perform the method as described in any one of claims 1 to 12.

14. A battery system, characterized in that, Including the battery management system as described in claim 13.

15. An electrical appliance, characterized in that, Includes the battery system as described in claim 14.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 12.

Images