Battery failure determination method, device, non-volatile storage medium and electronic device

Abstract

The present application discloses a battery fault determination method, device, non-volatile storage medium and electronic device. The method includes: obtaining current state data of a target battery and multiple historical state data; determining the time series characteristics of the target battery based on the time sequence between the multiple historical state data; determining the current expected voltage of the target battery based on the multiple historical state data, the current state data, and the time series characteristics; determining the fault identification result of the target battery based on the current expected voltage. The present application solves the technical problem that the accuracy of the battery fault determination result in the related art is not ideal.

Description

Technical Field

[0001] The present application relates to the technical field of lithium battery failure warning, and in particular to a battery failure determination method, device, non-volatile storage medium and electronic device. Background Art

[0002] In recent years, with the gradual popularization of new energy vehicles, electric bicycles, drones, mobile smart devices, portable electronic devices and other products, the demand for high-performance, high-safety and long-life lithium-ion batteries has been growing. However, lithium batteries may fail during use, such as internal short circuits, abnormal self-discharge, etc. These failures may cause the performance of lithium batteries to deteriorate, shorten their lifespan, and even cause safety accidents such as fires and explosions. Battery voltage, as an important indicator of battery status, mainly reflects the health status and performance of the battery. In related technologies, fault diagnosis of lithium-ion batteries is mainly based on parameter detection and operating condition detection. Since parameter detection and operating condition detection only rely on parameters or operating conditions outside the battery, it is impossible to accurately evaluate the chemical reaction process inside the battery. As a result, it is impossible to accurately identify the fault state inside the battery, and there is a problem of unsatisfactory accuracy in the battery fault determination results.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide a battery fault determination method, device, non-volatile storage medium and electronic device to at least solve the technical problem of unsatisfactory accuracy of battery fault determination results existing in the related art.

[0005] According to one aspect of the embodiment of the present application, a battery fault determination method is provided, including: obtaining the current state data of the target battery and multiple historical state data; determining the time series characteristics of the target battery based on the time sequence between the multiple historical state data; determining the current expected voltage of the target battery based on the multiple historical state data, the current state data, and the time series characteristics; determining the fault identification result of the target battery based on the current expected voltage. By comparing the voltage value predicted by the model with the actual measured voltage value, the voltage abnormality can be quickly identified, so that the possible fault of the battery can be discovered in time, and measures can be taken in advance, such as adjusting the charging and discharging strategy, replacing or repairing the battery unit, to avoid the deterioration of the fault.

[0006] Optionally, based on multiple historical status data, current status data, and time series characteristics, the current expected voltage of the target battery is determined, including: detecting the target battery to obtain the current actual voltage; determining the voltage difference between the current expected voltage and the current actual voltage; and determining the fault identification result based on the voltage difference and a predetermined fault threshold. By detecting the current voltage of the battery in real time and comparing it with the voltage predicted by the model, voltage anomalies can be discovered in time, so that an early warning can be issued at the early stage of the fault to prevent the occurrence of battery safety accidents.

[0007] Optionally, the method further includes: determining a first coefficient representing the current health state of the target battery; determining a second coefficient representing the operating frequency of the target battery and the temperature of the region in which it is located; and determining a predetermined fault threshold of the target battery based on the first coefficient and the second coefficient. By dynamically adjusting the predetermined fault threshold, the predetermined fault threshold can be optimized according to the actual health state and operating conditions of the battery, thereby more accurately identifying fault signals and reducing false positives and false negatives.

[0008] Optionally, the fault identification result is determined based on the voltage difference and the predetermined fault threshold, including: determining the current detection result of the target battery based on the voltage difference and the predetermined fault threshold; obtaining the historical detection result of the target battery when the current detection result indicates that the target battery has a fault; and determining the fault identification result as a fault state when the historical detection result indicates that the target battery has a fault. By detecting the voltage difference in real time and combining the historical detection results, the reliability of fault identification is further verified, avoiding misjudgment caused by a single abnormal detection, filtering out voltage fluctuations caused by accidental factors, reducing false alarms, and improving the robustness and stability of the system, as well as the timeliness and accuracy of fault response.

[0009] Optionally, the current state data includes multiple parameters, and the multiple parameters describe the state of the target battery from different dimensions. Based on multiple historical state data, current state data, and time series features, the current expected voltage of the target battery is determined, including: determining the spatial sequence features of the target battery based on the current state data, wherein the spatial sequence features are used to represent the correlation between multiple parameters at the same time; and determining the current expected voltage by performing feature fusion based on the time series features and the spatial sequence features. The introduction of spatial sequence features can capture the mutual influence between multiple parameters, ensuring that the prediction model can comprehensively consider the time evolution and the correlation between multiple parameters, so that the model takes into account more comprehensive information when predicting the voltage, thereby improving the accuracy and stability of the prediction.

[0010] Optionally, feature fusion is performed based on time series features and space series features to determine the current expected voltage, including: obtaining temperature data and pressure data of the area where the target battery is located; extracting features based on the temperature data and pressure data to determine the environmental features of the target battery; and performing feature fusion based on environmental features, time series features, and space series features to determine the current expected voltage. The introduction and fusion of environmental features can more comprehensively reflect the actual operating conditions of the battery, enable the early warning system to better adapt to the actual operating environment of the battery, improve the accuracy of the model in predicting battery voltage, and enhance the generalization ability of the model.

[0011] Optionally, obtaining the current state data of the target battery includes: determining a target clustering algorithm based on the data characteristics of the current state data; and using the target clustering algorithm to perform outlier elimination processing on the current state data to obtain the eliminated current state data. By eliminating outliers, noise and outliers in the data can be removed, the purity of the current state data set can be improved, and a more accurate data basis can be provided for subsequent fault warning model training. At the same time, the robustness of the system to these abnormal situations can be enhanced, false alarms and missed alarms can be reduced, and the stability of fault warnings can be improved.

[0012] According to another aspect of an embodiment of the present application, a battery fault determination device is provided, including: a data acquisition module, used to acquire current status data of a target battery, and multiple historical status data; a time series feature determination module, used to determine the time series characteristics of the target battery based on the time sequence between multiple historical status data; a current expected voltage determination module, used to determine the current expected voltage of the target battery based on multiple historical status data, current status data, and time series characteristics; a fault identification result determination module, used to determine the fault identification result of the target battery based on the current expected voltage.

[0013] According to another aspect of an embodiment of the present application, a non-volatile storage medium is provided, wherein the non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing any one of the battery fault determination methods.

[0014] According to another aspect of an embodiment of the present application, an electronic device is provided, comprising: one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement any one of the battery fault determination methods.

[0015] In the embodiment of the present application, by acquiring the current state data of the target battery and multiple historical state data; determining the time series characteristics of the target battery based on the time sequence between the multiple historical state data; determining the current expected voltage of the target battery based on the multiple historical state data, the current state data, and the time series characteristics; and determining the fault identification result of the target battery based on the current expected voltage. The purpose of determining the expected voltage of the battery through the historical state data and the current state data of the battery, and determining the fault identification result of the battery based on the expected voltage is achieved, and the technical effect of improving the accuracy of the battery fault identification result is achieved, thereby solving the technical problem of the unsatisfactory accuracy of the battery fault determination result existing in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 is a flowchart of an optional battery failure determination method provided according to an embodiment of the present application;

[0018] Figure 2 is a schematic diagram of an optional battery fault determination method provided according to an embodiment of the present application;

[0019] Figure 3 is a first model diagram of an optional battery fault determination method provided according to an embodiment of the present application;

[0020] Figure 4 is a second model diagram of an optional battery fault determination method provided according to an embodiment of the present application;

[0021] Figure 5 is a fault identification diagram of an optional battery fault determination method provided in an embodiment of the present application;

[0022] Figure 6 It is a schematic diagram of an optional battery fault determination device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0026] Thermal runaway refers to a self-accelerating process of a series of exothermic reactions occurring inside the battery, which may cause the battery temperature to rise uncontrollably, thereby triggering thermal runaway events, including fire and explosion.

[0027] SOC (State of Charge) is the ratio of the remaining battery power to the power when the battery is fully charged, usually expressed as a percentage.

[0028] BMS (Battery Management System) is used in application scenarios such as electric vehicles, energy storage systems, portable electronic devices, etc. that require efficient, safe and reliable management and monitoring of battery performance. It has functions such as battery status detection, balancing management, fault diagnosis and early warning, and charge and discharge control.

[0029] CNN (Convolutional Neural Network) is a deep learning model suitable for processing data with grid structures, such as images, videos, and sequence data. When processing these data, CNN can effectively identify and extract local features and combine these features into more complex representations through a hierarchical structure.

[0030] TCN (Temporal Convolutional Network) is a neural network architecture for processing sequence data. It uses one-dimensional convolutional layers to capture the temporal dependencies in sequences. It has the characteristics of parallel computing capabilities, long-distance dependencies, and high stability.

[0031] DNN (Deep Neural Network) is a machine learning model. The DNN network structure contains multiple hidden layers and can learn complex and abstract features in data. It is widely used in image recognition, speech recognition, natural language processing, time series prediction and other fields.

[0032] The related art provides a lithium battery voltage prediction method based on the principle of time series analysis, which uses a recurrent neural network combined with a learning rate decay algorithm to predict the voltage time series. However, it does not consider the impact of other factors such as current, temperature, SOC, etc. on the voltage, so the accuracy of the voltage prediction result is poor.

[0033] Related technologies provide a lithium battery thermal runaway warning method and a battery thermal runaway warning model training method. Fault detection and warning are performed by integrating multiple abnormality recognition methods and balanced training samples. However, the deep feature analysis of data, such as the correlation between elements and the extraction of associated elements, is not well studied.

[0034] The related art provides a lithium battery health status prediction method based on an optimized time domain convolutional network and a lithium battery remaining life status assessment method based on an improved TCN. By optimizing and improving the TCN network, the time series is predicted, thereby realizing the battery health status assessment. However, the above method does not improve the defect that TCN has a poor effect on long-term dependency problems.

[0035] According to an embodiment of the present application, a method embodiment of a battery fault determination method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] Figure 1 is a flowchart of an optional battery failure determination method provided according to an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0037] Step S102, obtaining current status data of the target battery and a plurality of historical status data;

[0038] It can be understood that the current state data of the target battery is obtained, such as current, temperature, SOC and insulation, etc. Multiple historical state data of the target battery are obtained, such as voltage, average voltage, etc. By obtaining the current state data and historical state data of the target battery, a comprehensive data basis is provided for voltage prediction and fault warning of the target battery.

[0039] In an optional embodiment, obtaining the current state data of the target battery includes: determining a target clustering algorithm based on data characteristics of the current state data; and using the target clustering algorithm to perform outlier elimination processing on the current state data to obtain the current state data after elimination processing.

[0040] It can be understood that based on the data characteristics of the current state data of the target battery, such as distribution characteristics, dimension, noise level, dynamics, etc., a suitable clustering algorithm is selected. For example, as a target clustering algorithm, the current state data is clustered to remove the outliers in the previous state data, thereby obtaining the current state data after the removal process. By removing outliers, the noise and outliers in the data can be removed, the purity of the current state data set can be improved, and a more accurate data basis can be provided for the subsequent fault warning model training. At the same time, the robustness of the system to these abnormal situations can be enhanced, false alarms and missed alarms can be reduced, and the stability of fault warnings can be improved.

[0041] Step S104, determining the time series characteristics of the target battery based on the time sequence between the multiple historical status data;

[0042] It can be understood that the time series characteristics of the historical status data are determined according to the time sequence between the historical status data of the target battery, and the above time series characteristics contain the laws and trends of battery changes over time. By extracting the time series characteristics of the historical status data, the potential pattern of battery status changes is revealed, providing a data-driven basis for subsequent voltage anomaly detection and fault warning, reducing the model's overreaction to accidental or short-term fluctuations, thereby reducing the false alarm rate in voltage anomaly detection and improving the reliability and accuracy of battery fault warning.

[0043] Optionally, an improved TCN (temporal convolutional network) network model can be used to extract time series features of the target battery. Due to the local feature extraction of the convolution kernel, the traditional TCN network model ignores the long-term dependency of the historical state data of the target battery on the time series (that is, the current voltage may be correlated with the voltage a long time ago). At the same time, the traditional TCN network model does not distinguish the importance of different time periods and ignores the impact of different time periods on the current expected voltage. Therefore, the prediction results for voltage prediction problems with long dependencies are less accurate. The above problems can be solved by adding an Attention module to the residual structure of the traditional TCN network model. On the one hand, the matrix operation of the Attention module can be used to obtain the sequence features from a long time ago to solve the long dependency problem; on the other hand, the self-attention mechanism can be used to give different attention (that is, weights) to the historical time series. Specifically, the Attention module first uses positional encoding to add temporal features (i.e., time series features), and then uses the Multi-Head Attention module to perform matrix operations to obtain the importance (i.e., weights) of different time periods, and calculates the correlation and output values ​​with the long time series, and then optimizes the model through the residual and regularization modules, and finally outputs the results of the Attention module.

[0044] Step S106, determining a current estimated voltage of the target battery based on a plurality of historical state data, current state data, and time series characteristics;

[0045] It can be understood that based on the historical state data, current state data of the target battery, and the time series characteristics based on the historical state data, the prediction model is used to determine the current expected voltage of the target battery, that is, the expected voltage value at the current moment. By integrating and utilizing the historical and current state data, as well as the time series characteristics, the accuracy of battery voltage prediction can be significantly improved, the generalization ability of the prediction model can be enhanced, and real-time fault warning of the battery can be achieved. At the same time, false alarms and missed alarms caused by incomplete understanding of the data can be reduced, and the battery management strategy can be optimized.

[0046] Optionally, the above prediction model can use a hybrid network model of improved TCN to predict the voltage of the target battery. The hybrid network model of improved TCN can effectively utilize the spatiotemporal fusion features, and provide a guarantee for accurate voltage fault warning by accurately predicting the voltage time series. First, the improved TCN (time convolutional network) network structure is used to process the characteristic time series features and extract the correlation between the voltage in the time series. Then, CNN (convolutional neural network) is used to process the spatial sequence features and extract the correlation between voltage and current, SOC, temperature and other features. Next, the dimension of the features can be transformed through the FeedForward module (feedforward neural network) to facilitate the fusion of time series features and spatial series features. Then, the Fusion module (fusion module) is used to fuse the spatiotemporal features. Finally, a DNN network is used to analyze the spatiotemporal features to obtain the current expected voltage.

[0047] In an optional embodiment, the current expected voltage of the target battery is determined based on multiple historical status data, current status data, and time series characteristics, including: detecting the target battery to obtain the current actual voltage; determining the voltage difference between the current expected voltage and the current actual voltage; and determining the fault identification result based on the voltage difference and a predetermined fault threshold.

[0048] It can be understood that at the current moment, the target battery is detected to obtain the current actual voltage value. The current actual voltage value is taken from the current expected voltage to obtain a voltage difference. The above voltage difference is compared with the predetermined fault threshold of the target battery at the current moment, such as 150mV (millivolts), to determine the fault identification result of the target battery. If the voltage difference is greater than the predetermined fault threshold, the fault identification result of the target battery is a fault; if the voltage difference is less than or equal to the predetermined fault threshold, the fault identification result of the target battery is normal. By detecting the current voltage of the battery in real time and comparing it with the voltage predicted by the model, voltage anomalies can be discovered in time, so that an early warning can be issued at the early stage of the fault to prevent the occurrence of battery safety accidents. At the same time, by reasonably setting the fault threshold, false alarms and missed alarms can be effectively reduced, and the accuracy of fault identification can be improved.

[0049] In an optional embodiment, the method also includes: determining a first coefficient representing the current health status of the target battery; determining a second coefficient representing the operating frequency of the target battery and the temperature of the area in which it is located; and determining a predetermined fault threshold of the target battery based on the first coefficient and the second coefficient.

[0050] It can be understood that the first coefficient and the second coefficient of the target battery are determined, wherein the first coefficient is used to reflect the current health status of the target battery, and the second coefficient is used to reflect the operating frequency of the target battery and the temperature of the area in which it is located. Based on the first coefficient and the second coefficient, the predetermined fault threshold of the target battery at the current moment is determined. By dynamically adjusting the predetermined fault threshold, the predetermined fault threshold can be optimized according to the actual health status and operating conditions of the battery, so as to more accurately identify fault signals and reduce false alarms and missed alarms. At the same time, by considering factors such as health status, operating frequency and temperature, personalized early warning for each battery can be achieved, thereby improving the pertinence and effectiveness of the early warning.

[0051] Optionally, the first coefficient is used to reflect the current state of health of the target battery. This is usually calculated based on the SOH (State of Health) of the battery. SOH can be estimated by a variety of methods, such as using an electrochemical model, an equivalent circuit model, or a machine learning model based on historical data. The lower the state of health, the larger the first coefficient may be, indicating that the battery is closer to a fault state. The second coefficient is used to reflect the operating frequency of the target battery and the temperature of the area in which it is located. The operating frequency refers to the number or frequency of charge and discharge of the battery. Frequent charging and discharging will accelerate battery aging; temperature is an important factor affecting battery performance and life. Too high or too low temperature will have a negative impact on the battery. The second coefficient can be calculated by statistical methods, such as mapping the operating frequency and temperature to a standardized coefficient range, or using a machine learning model to learn from historical data the degree of influence of these two on battery performance.

[0052] In an optional embodiment, a fault identification result is determined based on a voltage difference and a predetermined fault threshold, including: determining a current detection result of a target battery based on a voltage difference and a predetermined fault threshold; when the current detection result indicates that a fault exists in the target battery, obtaining a historical detection result of the target battery; when the historical detection result indicates that a fault exists in the target battery, determining that the fault identification result is a fault state.

[0053] It can be understood that the current voltage difference is obtained based on the difference between the current actual voltage value and the current expected voltage, and the current voltage difference is compared with the current predetermined fault threshold to determine the current detection result of the target battery. If the current detection result shows that the target battery is faulty, it is necessary to determine the state of the target battery in combination with the historical detection results of the target battery. Determine the historical detection results of the target battery, wherein the historical detection results are obtained based on the historical actual voltage values ​​and the historical expected voltages. If the historical detection results also indicate that the target battery is faulty, it means that the fault identification result of the target battery is a fault state, that is, the target battery is faulty. By detecting the voltage difference in real time and combining the historical detection results, the reliability of fault identification is further verified, avoiding misjudgments caused by single detection abnormalities, filtering out voltage fluctuations caused by accidental factors, reducing false alarms, and improving the robustness and stability of the system, as well as the timeliness and accuracy of fault response.

[0054] Optionally, the target battery's current predicted voltage is predicted using an improved TCN hybrid network model, and the current predicted voltage is compared with the true voltage (i.e., the current actual voltage) to determine the target battery's fault identification result. The predicted voltage (i.e., the current predicted voltage) is determined based on the historical state data of the voltage's historical characteristic law and the current state data reflecting the target battery's current condition. If the predicted voltage differs greatly from the true voltage, it means that the target battery may have a voltage fault. The predetermined fault thresholds for the predicted voltage and the true voltage can be set to 150mV. At the same time, in order to cope with the complex working conditions of the power battery during operation, a predetermined number of frames, such as 3 frames, generally appear continuously. Only when the above abnormalities are finally determined to be voltage faults, will an early warning be issued.

[0055] In an optional embodiment, the current state data includes multiple parameters, and the multiple parameters describe the state of the target battery from different dimensions respectively. Based on multiple historical state data, current state data, and time series characteristics, the current expected voltage of the target battery is determined, including: based on the current state data, determining the spatial sequence characteristics of the target battery, wherein the spatial sequence characteristics are used to represent the correlation between multiple parameters at the same time; and performing feature fusion based on the time series characteristics and the spatial sequence characteristics to determine the current expected voltage.

[0056] It can be understood that the current state data of the target battery includes multiple parameters, such as current, temperature, SOC, etc., each of which describes the state of the battery from different dimensions. Based on the current state data of the target battery, the spatial sequence characteristics of the target battery are determined, and the extraction of spatial sequence characteristics aims to describe the correlation between these parameters at the same time. Based on the time series characteristics of the historical state data of the target battery and the spatial sequence characteristics of the current state data, the prediction model is used to perform feature fusion processing and obtain the current estimated voltage of the target battery. The introduction of spatial sequence features can capture the mutual influence between multiple parameters, ensuring that the prediction model can comprehensively consider the time evolution and the correlation between multiple parameters, so that the model takes into account more comprehensive information when predicting voltage, thereby improving the accuracy and stability of the prediction.

[0057] Optionally, voltage time series prediction refers to predicting future voltage using voltage characteristics of historical time. In order to predict the lowest voltage (i.e., expected voltage) at n time points after the current moment, the time series characteristics of historical state data are first extracted, such as the lowest voltage and average voltage characteristics at m time points before the current moment, which is called feature 1, as is the case with common time series prediction algorithms. In addition to being associated with historical voltage, voltage is also related to current temperature, current, SOC and other characteristics. To accurately predict future voltage, it is also necessary to extract future current, SOC, maximum temperature, insulation and other characteristics, which are called feature 2, i.e., the correlation between different variables. The process of extracting and fusing the extracted features 1 and 2 is called the extraction and fusion of spatiotemporal features.

[0058] In an optional embodiment, feature fusion is performed based on time series features and space series features to determine the current expected voltage, including: acquiring temperature data and pressure data of the area where the target battery is located; performing feature extraction based on the temperature data and pressure data to determine the environmental characteristics of the target battery; and performing feature fusion based on environmental characteristics, time series features, and space series features to determine the current expected voltage.

[0059] It can be understood that the voltage of the target battery will be affected by temperature and pressure. The temperature data and pressure data of the area where the target battery is located are obtained, and the features of the above temperature data and pressure data are extracted to generate the environmental characteristics of the target battery. The environmental characteristics, time series characteristics, and spatial series characteristics of the target battery are fused to predict the voltage of the target battery and obtain the current expected voltage. The introduction and fusion of environmental characteristics can more comprehensively reflect the actual conditions of battery operation, enable the early warning system to better adapt to the actual environment of battery operation, improve the accuracy of the model's prediction of battery voltage, and enhance the generalization ability of the model.

[0060] Step S108, determining a fault identification result of the target battery based on the current estimated voltage.

[0061] It can be understood that the fault identification result of the target battery is determined based on the current predicted voltage of the target battery, combined with the current actual voltage and the predetermined fault threshold. By comparing the voltage value predicted by the model with the actual measured voltage value, the voltage anomaly can be quickly identified, so that the possible fault of the battery can be discovered in time, and measures can be taken in advance, such as adjusting the charging and discharging strategy, replacing or repairing the battery unit, to avoid the deterioration of the fault.

[0062] Through the above step S102, the current state data of the target battery and multiple historical state data are obtained; step S104, based on the time sequence between multiple historical state data, the time series characteristics of the target battery are determined; step S106, based on multiple historical state data, current state data, and time series characteristics, the current expected voltage of the target battery is determined; step S108, based on the current expected voltage, the fault identification result of the target battery is determined. The purpose of determining the expected voltage of the battery through the historical state data and current state data of the battery, and determining the fault identification result of the battery based on the expected voltage can be achieved, achieving the technical effect of improving the accuracy of the battery fault identification result, thereby solving the technical problem of unsatisfactory accuracy of the battery fault determination result existing in the related art.

[0063] Based on the above embodiments and optional embodiments, the present application proposes an optional implementation method for fault identification of lithium batteries. Lithium-ion batteries (i.e., lithium batteries) have the advantages of high energy density and long service life, which make them widely used in new energy vehicles and energy storage systems. However, lithium batteries may fail during use, such as internal short circuits, abnormal self-discharge, SOC jumps, etc. These failures may cause the performance of lithium batteries to deteriorate, shorten their lifespan, and even cause safety accidents such as fires and explosions. Therefore, it is usually necessary to provide thermal runaway warnings for lithium batteries. Thermal runaway warning technology is a technology focused on the safety management of lithium-ion batteries. The technology aims to identify and prevent safety risks caused by overheating of lithium batteries.

[0064] In the actual use of lithium batteries, there are some irreversible electrochemical reactions inside the lithium batteries, which cause the performance of the lithium batteries to decline and fail, and even cause safety accidents. The voltage value is the main indicator for diagnosing whether the working state of the lithium battery is normal. Using the trend prediction scheme of the lithium battery voltage, the voltage value in the future is predicted through the voltage value at the historical moment, and the residual analysis of the actual value and the predicted value is performed to provide a quantitative basis for the fault diagnosis and thermal runaway warning of lithium batteries.

[0065] The voltage prediction problem of lithium batteries can be regarded as a time series analysis and prediction problem. As a widely used machine learning method, neural networks have been widely used in the fields of time series prediction and fault warning, and have achieved certain results.

[0066] In actual use scenarios, lithium battery failure samples are very scarce, which brings great challenges to lithium battery failure warning. In related technologies, lithium battery failure warning is mainly divided into theoretical modeling and data-driven methods. Theoretical modeling is to build models based on different theories, including electrochemical models and equivalent circuit models. However, the model construction of theoretical modeling is difficult, and it is difficult to obtain high prediction accuracy, and it is difficult to obtain good model generalization due to the large differences in battery cells. Data-driven methods include statistical analysis, machine learning, deep learning and other methods. Statistical analysis and machine learning are sensitive to the selection of thresholds and parameters, and have poor generalization in different scenarios; the deep neural network model in the deep learning method can describe the strong nonlinear relationship between input and output, and can be trained or learned to mine the internal potential characteristics of lithium batteries from big data. However, the voltage prediction of lithium batteries may be related to the voltage state and operating conditions a long time ago, and the degree of influence in different time periods is also different. It is necessary to propose a more complex and more suitable network for time series prediction. The deep neural network model is still insufficient for long-dependent time series problems. Moreover, in the related art, the voltage prediction of lithium batteries usually only considers the characteristics of the time series, and the correlation between the characteristics is not effectively mined and utilized.

[0067] Figure 2 is a schematic diagram of an optional battery fault determination method provided according to an embodiment of the present application, such as Figure 2 As shown, the steps for voltage fault identification and early warning for lithium batteries (i.e., target batteries) are explained. S1. Collect time series data of BMS such as voltage and current; S2. Extract time dimension features of time series, and extract related features between different feature variables, that is, extract and fuse spatiotemporal features; S3. Train and predict voltage based on the hybrid network model of improved TCN; S4. Use the difference analysis between the predicted voltage (i.e., the current estimated voltage) and the real voltage (i.e., the current actual voltage) to identify anomalies and issue fault warnings. The above steps will be explained in detail below.

[0068] S1. Collect the time series data of BMS such as voltage and current.

[0069] Obtain the current status data of the target lithium battery, such as current, temperature, SOC, insulation, etc. Obtain multiple historical status data of the target battery, such as voltage, average voltage, etc. By obtaining the current status data and historical status data of the target battery, a comprehensive data foundation is provided for voltage prediction and fault warning of the target battery.

[0070] S2. Extract the time dimension features of the time series, and extract the correlation features between different feature variables, that is, extract and fuse the spatiotemporal features.

[0071] Voltage time series prediction refers to the use of historical voltage characteristics to predict future voltage. The extraction and fusion of time series features and space series features are shown in Table 1. In order to predict the lowest voltage (i.e., expected voltage) at n time points after the current moment, the time series features of historical state data are first extracted, such as the lowest voltage and average voltage features at m time points before the current moment, which is called feature 1, as is the case with common time series prediction algorithms. In addition to being associated with historical voltage, voltage is also related to current temperature, current, SOC and other features. To accurately predict future voltage, it is also necessary to extract future current, SOC, maximum temperature, insulation and other features, which are called feature 2, i.e., the correlation between different variables. The process of extracting and fusing the extracted features 1 and 2 is called the extraction and fusion of spatiotemporal features.

[0072] Table 1 Extraction and fusion of time series features and spatial series features

[0073]

[0074] S3. Train and predict voltage based on the hybrid network model of improved TCN.

[0075] The voltage of the target battery is predicted using a hybrid network model of improved TCN. Figure 3 is a first model diagram of an optional battery fault determination method provided in an embodiment of the present application, and the process of predicting the current expected voltage using the hybrid network model of the improved TCN is as follows Figure 3As shown in the figure. The hybrid network model of the improved TCN can effectively utilize the spatiotemporal fusion features, and provide guarantee for accurate voltage fault warning by accurately predicting the voltage time series. First, the improved TCN (time convolutional network) network structure is used to process the characteristic time series features and extract the correlation between the voltage in the time series. Then, the CNN (convolutional neural network) is used to process the spatial sequence features and extract the correlation between the voltage and the current, SOC, temperature and other features. Next, the dimension of the features can be transformed through the Feed Forward module (feedforward neural network) to facilitate the fusion of the time series features and the spatial series features. Then, the Fusion module (fusion module) is used to fuse the spatiotemporal features. Finally, a DNN network is used to analyze the spatiotemporal features, and the MSE Loss is used as the loss function to calibrate and correct the model, and then the voltage is predicted to obtain the current expected voltage.

[0076] An improved TCN (temporal convolutional network) network model is used to extract time series from lithium batteries. Figure 4 is a second model diagram of an optional battery fault determination method provided in an embodiment of the present application, such as Figure 4 As shown in the figure, the traditional TCN (i.e., original TCN) network model ignores the long-term dependency of the historical state data of the target battery on the time series due to the local feature extraction of the convolution kernel (i.e., the current voltage may be associated with the voltage a long time ago). At the same time, the traditional TCN network model does not distinguish the importance of different time periods and ignores the impact of different time periods on the current expected voltage. Therefore, the prediction results for the voltage prediction problem with long-term dependency are less accurate. The above problems can be solved by adding an Attention module to the residual structure of the traditional TCN network model. On the one hand, the matrix operation of the Attention module can be used to obtain the sequence features of a long time ago to solve the long dependency problem; on the other hand, different attention (i.e., weights) can be given to the historical time series through the self-attention mechanism. Specifically, the Attention module first uses positional encoding to add time series features (i.e., time series features), and then uses the Multi-Head Attention module to perform matrix operations to obtain the importance of different time periods (i.e., weights), and calculates the correlation and output value with the long time series, and then optimizes the model through the residual and regularization modules, and finally outputs the results of the Attention module.

[0077] In such Figure 4In the traditional TCN (i.e., original TCN) network model shown in the figure, Dropout is used for regularization. Dropout is a regularization technique commonly used in training deep neural networks. It is mainly used to prevent model overfitting, enhance the generalization ability of the model, and accelerate the training process to a certain extent. Dilated Causal Convolution is a convolution technique commonly used in the field of deep learning, especially in processing time series data. It is used to capture and model long-range dependencies in time series data. WeightNorm is used to optimize the weights of the model. WeightNorm is a normalization method used in the field of deep learning to optimize the weights of neural networks. It is used to simplify the initialization and update process of weights, thereby accelerating the training of the network and improving the performance of the model. ReLU (Rectified Linear Unit) is used as the activation function of the model. ReLU is an activation function widely used in deep learning. It is used to help the model learn nonlinear feature relationships and improve the expression ability and prediction accuracy of the model. 1*1Conv(optional) is a special type of convolutional layer that can linearly combine input features without changing the spatial size of the input feature map. It is mainly used for feature mapping and dimension conversion.

[0078] In such Figure 4 In the improved TCN (temporal convolutional network) network model shown, Add&Norm combines residual connections and layer normalization to solve the gradient vanishing problem in the deep network training process, improve the stability and expressiveness of the model, and accelerate the training process. The Input Encoding method is used to encode the data input to the model. Input Encoding refers to converting the original input data into a form that the model can understand and process, usually including steps such as data preprocessing, embedding, and feature encoding.

[0079] S4. Utilize the difference analysis between the predicted voltage (i.e., the current estimated voltage) and the real voltage (i.e., the current actual voltage) to identify abnormalities and issue fault warnings.

[0080] The improved TCN hybrid network model is used to predict the current predicted voltage of the lithium battery, and the current predicted voltage is compared with the real voltage (i.e., the current actual voltage) to determine the fault identification result of the target battery. The predicted voltage (i.e., the current predicted voltage) is determined based on the historical state data of the voltage history characteristic law and the current state data reflecting the current environmental conditions of the target battery. If the predicted voltage and the real voltage differ greatly, it means that the target battery may have a voltage fault. The predetermined fault threshold of the predicted voltage and the real voltage can be set to 150mV. At the same time, in order to cope with the complex working conditions of the power battery during operation, a predetermined number of frames, such as 3 frames, generally appear continuously. Only when the above abnormalities are finally determined as voltage faults, an early warning is issued. Figure 5 is a fault identification diagram of an optional battery fault determination method provided in an embodiment of the present application, Figure 5 The horizontal axis is time and the vertical axis is voltage. Figure 5 As shown, at position 1, although the voltage difference is greater than the predetermined fault threshold of 150mV, only one frame is abnormal, so the final fault identification result is a normal state and no warning is issued; at position 1, there is a situation where the voltage difference is greater than the predetermined fault threshold of 150mV for more than 3 consecutive frames, so the final fault identification result is a fault state and a warning is required.

[0081] The above optional implementation methods achieve at least the following effects: a hybrid network model based on improved TCN is proposed, through which time and space features can be effectively extracted and feature fusion can be performed, thereby improving the accuracy of lithium battery voltage prediction results, reducing false positives and omissions caused by incomplete understanding of data, and optimizing battery management strategies; in order to solve the long-term dependency problem of voltage time series, an Attention module is added to the TCN network structure to extract long-term dependency features and weights of different time periods, further improving the accuracy of voltage prediction results, enhancing the generalization ability of the prediction model, and realizing real-time fault warning of the battery.

[0082] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0083] In this embodiment, a battery fault determination device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the terms "module" and "device" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0084] According to an embodiment of the present application, a device embodiment for implementing a battery failure determination method is also provided. Figure 6 is a schematic diagram of a battery fault determination device according to an embodiment of the present application, such as Figure 6 As shown, the battery fault determination device includes a data acquisition module 602, a time series feature determination module 604, a current estimated voltage determination module 606, and a fault identification result determination module 608. The device is described below.

[0085] A data acquisition module 602 is used to acquire current state data of a target battery and a plurality of historical state data;

[0086] A time series characteristic determination module 604, connected to the data acquisition module 602, is used to determine the time series characteristics of the target battery based on the time sequence between the multiple historical state data;

[0087] A current estimated voltage determination module 606, connected to the time series feature determination module 604, is used to determine the current estimated voltage of the target battery based on a plurality of historical state data, current state data, and time series features;

[0088] The fault identification result determination module 608 is connected to the current estimated voltage determination module 606 and is used to determine the fault identification result of the target battery based on the current estimated voltage.

[0089] In a battery fault determination device provided by an embodiment of the present application, a data acquisition module 602 is set to acquire the current state data of the target battery and multiple historical state data; a time series feature determination module 604 is connected to the data acquisition module 602 and is used to determine the time series features of the target battery based on the time sequence between multiple historical state data; a current expected voltage determination module 606 is connected to the time series feature determination module 604 and is used to determine the current expected voltage of the target battery based on multiple historical state data, current state data, and time series features; a fault identification result determination module 608 is connected to the current expected voltage determination module 606 and is used to determine the fault identification result of the target battery based on the current expected voltage. The purpose of determining the expected voltage of the battery through the historical state data and current state data of the battery and determining the fault identification result of the battery based on the expected voltage is achieved, and the technical effect of improving the accuracy of the battery fault identification result is achieved, thereby solving the technical problem of the unsatisfactory accuracy of the battery fault determination result existing in the related art.

[0090] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0091] It should be noted that the above data acquisition module 602, time series feature determination module 604, current estimated voltage determination module 606, and fault identification result determination module 608 correspond to steps S102 to S108 in the embodiment, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments. It should be noted that the above modules as part of the device can be run in a computer terminal.

[0092] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0093] The above-mentioned battery fault determination device may also include a processor and a memory. The data acquisition module 602, the time series feature determination module 604, the current expected voltage determination module 606, the fault identification result determination module 608, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0094] The processor includes a kernel, which retrieves the corresponding program unit from the memory. There can be one or more kernels. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0095] An embodiment of the present application provides a non-volatile storage medium on which a program is stored. When the program is executed by a processor, a battery failure determination method is implemented.

[0096] The embodiment of the present application provides an electronic device, which includes a processor, a memory, and a program stored in the memory and can be run on the processor. When the processor executes the program, the following steps are implemented: obtaining the current state data of the target battery and multiple historical state data; determining the time series characteristics of the target battery based on the time sequence between the multiple historical state data; determining the current expected voltage of the target battery based on the multiple historical state data, the current state data, and the time series characteristics; determining the fault identification result of the target battery based on the current expected voltage. The device in this article can be a server, a PC, etc.

[0097] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: obtaining current status data of a target battery, and multiple historical status data; determining a time series characteristic of the target battery based on a time sequence between multiple historical status data; determining a current expected voltage of the target battery based on multiple historical status data, current status data, and time series characteristics; and determining a fault identification result of the target battery based on the current expected voltage.

[0098] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0099] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0101] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0102] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0103] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0104] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0105] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0106] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0107] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for determining a battery failure, characterized in that: include: Obtain current status data of the target battery, as well as multiple historical status data; Determining a time series feature of the target battery based on a time sequence between the plurality of historical status data; Determining a current estimated voltage of the target battery based on the plurality of historical state data, the current state data, and the time series characteristics; Based on the current estimated voltage, a fault identification result of the target battery is determined.

2. The method according to claim 1, characterized in that The determining the current estimated voltage of the target battery based on the plurality of historical status data, the current status data, and the time series characteristics comprises: Detecting the target battery to obtain the current actual voltage; determining a voltage difference between the current expected voltage and the current actual voltage; The fault identification result is determined based on the voltage difference and a predetermined fault threshold.

3. The method according to claim 2, characterized in that The method further comprises: Determining a first coefficient representing a current health state of the target battery; Determine a second coefficient representing the target battery operation frequency and the temperature of the area where the target battery is located; Based on the first coefficient and the second coefficient, a predetermined failure threshold of the target battery is determined.

4. The method according to claim 2, characterized in that: The determining the fault identification result based on the voltage difference and a predetermined fault threshold comprises: Determining a current detection result of the target battery based on the voltage difference and the predetermined fault threshold; When the current detection result indicates that the target battery is faulty, obtaining a historical detection result of the target battery; In a case where the historical detection result indicates that the target battery has a fault, the fault identification result is determined to be a fault state.

5. The method according to claim 1, characterized in that: The current state data includes multiple parameters, and the multiple parameters describe the state of the target battery from different dimensions respectively. The current estimated voltage of the target battery is determined based on the multiple historical state data, the current state data, and the time series characteristics, including: Based on the current state data, determining a spatial sequence feature of the target battery, wherein the spatial sequence feature is used to represent an association relationship between the multiple parameters at the same time; Feature fusion is performed based on the time series features and the space series features to determine the current estimated voltage.

6. The method according to claim 5, characterized in that The step of performing feature fusion based on the time series feature and the space series feature to determine the current estimated voltage includes: Acquiring temperature data and pressure data of the area where the target battery is located; Perform feature extraction based on the temperature data and the pressure data to determine the environmental characteristics of the target battery; Feature fusion is performed based on the environmental features, the time series features, and the space series features to determine the current expected voltage.

7. The method according to claim 1, characterized in that The obtaining of current status data of the target battery includes: Determining a target clustering algorithm based on data characteristics of the current state data; The target clustering algorithm is used to perform outlier elimination processing on the current state data to obtain the current state data after the elimination processing.

8. A battery failure determination device, characterized in that: include: A data acquisition module, used to acquire current status data of a target battery and multiple historical status data; A time series feature determination module, used to determine the time series feature of the target battery based on the time sequence between the plurality of historical state data; a current estimated voltage determination module, configured to determine a current estimated voltage of the target battery based on the plurality of historical state data, the current state data, and the time series characteristics; A fault identification result determination module is used to determine a fault identification result of the target battery based on the current estimated voltage.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the battery fault determination method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: One or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the battery fault determination method described in any one of claims 1 to 7.

Images