A method and system for diagnosing discharge faults of a lithium-ion battery

Abstract

The present invention discloses a lithium-ion battery discharge fault diagnosis method and system, which includes diagnosis based on traditional threshold detection and diagnosis based on a voltage prediction model. Overdischarge is detected by comparing the battery voltage with the cutoff voltage. If the battery voltage is greater than the cutoff voltage, the fault diagnosis based on the voltage prediction model is entered to detect the previous overdischarge problem, and the lithium-ion battery overdischarge fault diagnosis is performed using the voltage prediction model. This diagnostic method performs a preliminary diagnosis of the lithium battery based on the battery voltage and the cutoff voltage, and determines whether the lithium battery requires a secondary diagnosis based on the results of the preliminary diagnosis. This simplifies the lithium battery fault diagnosis process and has the advantages of low computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnosis method, it can detect slight overdischarge problems of batteries in the battery pack.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent fault diagnosis of battery energy storage systems, and in particular to a method and system for diagnosing discharge faults of lithium-ion batteries. Background Art

[0002] In recent years, energy storage lithium batteries have been widely used in the field of electrochemical energy storage. At the same time, the safety performance of energy storage lithium batteries has also received widespread attention. Energy storage lithium batteries have high energy density and power density. Once a failure occurs, it will cause safety accidents such as thermal runaway, causing serious economic losses and social harm. Fault diagnosis, as one of the core functions of energy storage lithium battery management systems, is an important guarantee for the safe and reliable operation of battery systems. Studying the fault characteristics of energy storage lithium batteries and developing targeted early sign extraction and fault diagnosis methods are the basis for ensuring the safety of battery applications.

[0003] Existing fault diagnosis methods are primarily categorized as empirical methods based on experimental data, model-based methods, and data-driven methods. Traditional fault diagnosis methods based on experimental data fail to fully consider the complex operating conditions of energy storage batteries. Furthermore, both model-based and data-driven methods require a balance between accuracy and computational time, which impacts their overall diagnostic performance. Summary of the Invention

[0004] To promptly diagnose over-discharge faults in large-scale energy storage power plants, a lithium-ion battery discharge fault diagnosis method and system are proposed. In actual operation, once the discharge voltage drops below the cutoff threshold, the undervoltage alarm in the battery management system (BMS) is triggered. However, minor over-discharges cannot be detected promptly, and due to battery pack inconsistencies, some low-capacity cells may be prone to minor over-discharges. Therefore, a strategy is needed to promptly diagnose minor over-discharge issues based on traditional threshold-based over-discharge diagnosis methods.

[0005] To solve the above technical problems, this application provides the following technical solutions:

[0006] In a first aspect, the present application provides a lithium-ion battery discharge fault diagnosis method, the lithium-ion battery discharge fault diagnosis method comprising:

[0007] Determining whether the lithium-ion battery requires fault diagnosis based on the lithium-ion battery voltage and the cut-off voltage;

[0008] If the lithium-ion battery requires fault diagnosis, the discharge data of the lithium-ion battery is input into a voltage prediction model, and the voltage prediction model outputs a predicted discharge voltage of the lithium-ion battery;

[0009] A fault type of the lithium-ion battery is determined according to the predicted discharge voltage.

[0010] Preferably, the lithium-ion battery discharge fault diagnosis method further includes:

[0011] Determining parameters of the voltage prediction model according to the characteristics of the lithium-ion battery;

[0012] The voltage prediction model is trained using historical discharge data of lithium-ion batteries.

[0013] Preferably, determining the parameters of the voltage prediction model according to the characteristics of the lithium-ion battery includes:

[0014] Determining input characteristic parameters of the voltage prediction model according to characteristic parameters of the lithium-ion battery and the voltage of the lithium-ion battery;

[0015] Using a preset sliding time window to determine the training interval of the voltage prediction model;

[0016] The trained voltage prediction model is tested to optimize the parameters of the voltage prediction model.

[0017] Preferably, determining the input characteristic parameters of the voltage prediction model according to the characteristic parameters of the lithium-ion battery and the voltage of the lithium-ion battery includes:

[0018] Calculating, based on the characteristic parameters and the voltage of the lithium-ion battery, a correlation coefficient between each characteristic parameter and the voltage of the lithium-ion battery;

[0019] An input characteristic parameter of the voltage prediction model is determined according to the correlation coefficient.

[0020] Preferably, the adopting of a preset sliding time window to determine the training interval of the voltage prediction model includes:

[0021] forming an input data set for the voltage prediction model based on the lithium-ion battery discharge data;

[0022] When the length of the input data set is greater than the length of the sliding time window, the voltage prediction model performs discharge voltage prediction training.

[0023] Preferably, the testing of the trained voltage prediction model to optimize parameters of the voltage prediction model includes:

[0024] Obtaining a prediction accuracy of the voltage prediction model based on a discharge voltage test result of the voltage prediction model and a corresponding actual value of the discharge voltage of the lithium-ion battery;

[0025] The deviation value, variance value, and weight value of the voltage prediction model are adjusted according to the prediction accuracy.

[0026] Preferably, the characteristic parameters of the lithium-ion battery include: lithium-ion battery temperature, lithium-ion battery current and lithium-ion battery SOC.

[0027] In a second aspect, the present application provides a lithium-ion battery discharge fault diagnosis system, the lithium-ion battery discharge fault diagnosis system comprising:

[0028] Level 1 diagnostic module: determines whether the lithium-ion battery requires fault diagnosis based on the lithium-ion battery voltage and cut-off voltage;

[0029] Voltage prediction model: if the lithium-ion battery requires fault diagnosis, the discharge data of the lithium-ion battery is input into the voltage prediction model, and the voltage prediction model outputs the predicted discharge voltage of the lithium-ion battery;

[0030] A secondary diagnosis module is configured to determine a fault type of the lithium-ion battery according to the predicted discharge voltage and the cut-off voltage.

[0031] At the same time, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the above method is implemented when the processor executes the computer program.

[0032] At the same time, the present invention also provides a computer-readable storage medium, which stores a computer program for executing the above method.

[0033] As can be seen from the above technical solution, the present invention discloses a lithium-ion battery discharge fault diagnosis method and system, which includes diagnosis based on traditional threshold detection and diagnosis based on voltage prediction model. Over-discharge is detected by comparing the battery voltage with the cut-off voltage. If the battery voltage is greater than the cut-off voltage, the fault diagnosis based on the voltage prediction model is entered to detect the previous over-discharge problem, and the lithium-ion battery over-discharge fault diagnosis is performed through the voltage prediction model. This diagnostic method performs a preliminary diagnosis on the lithium battery based on the battery voltage and the cut-off voltage, and determines whether the lithium battery needs a secondary diagnosis based on the results of the preliminary diagnosis. It simplifies the fault diagnosis process of the lithium battery and has the advantages of low computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnosis method, it can detect slight over-discharge problems of batteries in the battery pack.

[0034] In order to make the above and other objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 Schematic diagram of a lithium-ion battery discharge fault diagnosis method in an embodiment of the present application.

[0037] Figure 2 This is a schematic diagram of the training set, test set, validation set in the embodiments of the present application and the sliding window in the present invention.

[0038] Figure 3 Schematic diagram of a parameter adjustment process in a lithium-ion battery discharge fault diagnosis method in an embodiment of the present application.

[0039] Figure 4 This is a structural diagram of a lithium-ion battery discharge fault diagnosis system in an embodiment of the present application.

[0040] Figure 5 Schematic diagram of the architecture of a lithium-ion battery discharge fault diagnosis system in an embodiment of the present application.

[0041] Figure 6 Schematic diagram of the structure of the electronic device in the application embodiment. DETAILED DESCRIPTION

[0042] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0043] Considering that slight overdischarge cannot be detected in a timely manner, and due to the inconsistency of battery packs, some low-capacity batteries may be prone to slight overdischarge, the present application provides a lithium-ion battery discharge fault diagnosis method, system, electronic device, and computer-readable storage medium. This method has the advantages of low computational complexity and high accuracy. Compared with traditional threshold-based overcharge diagnosis methods, it can detect slight overdischarge problems in batteries in a battery pack. This overdischarge diagnosis method detects overdischarge by comparing the battery voltage with a cutoff voltage. If the battery voltage is greater than the cutoff voltage, fault diagnosis based on a voltage prediction model is initiated to detect previous overdischarge problems.

[0044] Based on the above content, the present application also provides a lithium-ion battery discharge fault diagnosis device for implementing the lithium-ion battery discharge fault diagnosis method provided in one or more embodiments of the present application. The lithium-ion battery discharge fault diagnosis device can be communicatively connected with a user client device, and the user client terminal device can be provided with multiple, and the lithium-ion battery discharge fault diagnosis device can specifically access the client terminal device through an application server.

[0045] The lithium-ion battery discharge fault diagnosis device can receive a lithium-ion battery discharge fault diagnosis instruction from a client terminal device and obtain a lithium-ion battery discharge parameter from the lithium-ion battery discharge fault diagnosis instruction. When the lithium-ion battery discharge voltage is greater than the cut-off voltage, the lithium-ion battery discharge fault diagnosis device determines that the lithium-ion battery needs to be diagnosed. The lithium-ion battery discharge fault diagnosis device starts a voltage prediction model. The voltage prediction model predicts the discharge voltage of the lithium-ion battery based on the lithium-ion battery discharge data. The lithium-ion battery discharge fault diagnosis device determines that the lithium-ion battery has a slight over-discharge fault based on the predicted discharge voltage. Then, the lithium-ion battery discharge fault diagnosis device can send the lithium-ion battery fault diagnosis result to the client device for display, so that the user can perform battery maintenance operations based on the lithium-ion battery fault result.

[0046] It is understandable that the client device may include a smart phone, a tablet electronic device, a portable computer, a desktop computer, a personal digital assistant (PDA), etc.

[0047] In another practical application scenario, the portion of the lithium-ion battery discharge fault diagnosis process may be performed in the classification processing center as described above, or all operations may be performed in the client device. The specific selection may be based on the processing capabilities of the client device and the limitations of the user's usage scenario. This application does not impose any restrictions on this. If all operations are performed in the client device, the client device may further include a processor for performing specific processing for the lithium-ion battery discharge fault diagnosis process.

[0048] The above-mentioned client device may have a communication module (i.e., a communication unit) that can be connected to a remote server for communication and data transmission with the server. For example, the communication unit can send a lithium-ion battery discharge fault diagnosis instruction to a server in a classification processing center so that the server can perform lithium-ion battery discharge fault diagnosis processing according to the lithium-ion battery discharge fault diagnosis instruction. The communication unit can also receive the lithium-ion fault diagnosis result returned by the server. The server may include a server on the task scheduling center side, and other implementation scenarios may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or may include a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0049] The server and the client device may communicate using any suitable network protocol, including network protocols that have not yet been developed as of the filing date of this application. Examples of such network protocols include TCP / IP, UDP / IP, HTTP, and HTTPS. Furthermore, examples of such network protocols include RPC (Remote Procedure Call Protocol) and REST (Representational State Transfer) protocols, which are used on top of the aforementioned protocols.

[0050] The present application provides a lithium-ion battery discharge fault diagnosis method, system, electronic device and computer-readable storage medium, which includes a diagnosis based on traditional threshold detection and a diagnosis based on a voltage prediction model. Over-discharge is detected by comparing the battery voltage with the cut-off voltage. If the battery voltage is greater than the cut-off voltage, the fault diagnosis based on the voltage prediction model is entered to detect the previous over-discharge problem, and the lithium-ion battery over-discharge fault diagnosis is performed through the voltage prediction model. This diagnostic method performs a preliminary diagnosis of the lithium battery based on the battery voltage and the cut-off voltage, and determines whether the lithium battery requires a secondary diagnosis based on the results of the preliminary diagnosis. This simplifies the fault diagnosis process of the lithium battery and has the advantages of low computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnosis method, it can detect slight over-discharge problems of the batteries in the battery pack.

[0051] The details are described through the following multiple embodiments and application examples.

[0052] In order to solve the problem that slight over-discharge of lithium-ion batteries cannot be detected in time, and some batteries with low capacity may be prone to slight over-discharge due to the inconsistency of battery packs, the present application provides an embodiment of a lithium-ion battery discharge fault diagnosis method, see Figure 1 The lithium-ion battery discharge fault diagnosis method specifically includes the following contents:

[0053] Step 100: Determine whether the lithium-ion battery requires fault diagnosis based on the lithium-ion battery voltage and cut-off voltage;

[0054] It is understandable that the cut-off voltage is the termination voltage, which refers to the lowest operating voltage value at which the battery should no longer discharge when the voltage drops during discharge. Different battery types and different discharge conditions have different termination voltages. Since the lithium-ion battery discharge fault diagnosis process requires a lot of calculations, this application design divides lithium battery fault diagnosis into primary diagnosis and secondary diagnosis. First, the lithium battery voltage is diagnosed at the primary level. If the lithium-ion battery voltage is greater than the cut-off voltage, it means that the lithium-ion battery needs to be diagnosed at the secondary level.

[0055] Step 200: If the lithium-ion battery requires fault diagnosis, the discharge parameters of the lithium-ion battery are input into a voltage prediction model, and the voltage prediction model outputs a predicted discharge voltage of the lithium-ion battery;

[0056] It is understandable that the voltage prediction model of this application is the XGBoost model. XGBoost is a gradient boosting framework that has been widely used in the power industry. Compared with the gradient boosting decision tree (GDBT), the objective function of XGBoost has a regularizer, and the iterative effect and operability of each round of iteration will be improved. In addition, XGBoost supports parallel computing, which greatly reduces the calculation time. In the present invention, XGBoost is used to predict the discharge segment voltage of the energy storage battery.

[0057] Classification and Regression Tree (CART) is a typical binary decision tree in XGBoost that can be used for classification and regression processes. If the prediction result is discrete, CART will generate a classification decision tree. On the contrary, if the result is continuous, a regression decision tree is generated, whose output is the average of all predicted samples belonging to a certain leaf node. Gain_σ is selected to evaluate the splitting property of the decision tree. If the Gain_σ obtained by some attributes is small, there is almost no difference between the two subsamples obtained by dichotomy. At the same time, a certain attribute can be defined as a splitting attribute. The continuous prediction results, total variance and Gain_σ obtained from the sample set S can be calculated by the following formula:

[0058]

[0059]

[0060] Where μ and yk represent the average of the prediction results in the sample set S and the prediction result of sample k, respectively. Attribute A divides the sample set into two parts, and Gain_σA(S) is the Gain_σ obtained from attribute A.

[0061] The two parts calculated by attribute A continue to be divided by the above optimization method. Finally, as

[0062] For the sample set S, the minimum value of the optimal binary solution is selected. Compared with GDBT, XGBoost performs parallel computing in segmentation attribute optimization and has high training efficiency.

[0063] The objective optimization function of the XGBoost model is given by:

[0064]

[0065] Where l is the loss function, ft(xi) is the value of the i-th leaf node in the t-th regression tree, and Ω(ft) is the regularizer. The mean squared error (MSE) is used as the loss function to ensure global optimization.

[0066] The discharge voltage prediction based on the XGBoost regression algorithm can be expressed as:

[0067]

[0068] Among them, f t (X i ) is the t-th regression tree, T is the total number of regression trees, Y i is the i-th predicted voltage.

[0069] Step 300: Determine the fault type of the lithium-ion battery according to the predicted discharge voltage;

[0070] It is understandable that when the predicted discharge voltage drops to the cut-off threshold, it indicates that the lithium-ion battery has a slight over-discharge fault.

[0071] As can be seen from the above description, the present invention discloses a lithium-ion battery discharge fault diagnosis method and system, which includes a primary diagnosis based on traditional threshold detection and a secondary diagnosis based on a voltage prediction model. The primary diagnosis detects over-discharge by comparing the battery voltage with the cut-off voltage. If the battery voltage is greater than the cut-off voltage, the secondary diagnosis based on the voltage prediction model is entered to detect the previous over-discharge problem, and the secondary over-discharge fault diagnosis of the lithium-ion battery is performed through the voltage prediction model. This diagnostic method performs a preliminary diagnosis on the lithium battery through the primary diagnosis, and determines whether the lithium battery requires a secondary diagnosis based on the result of the primary diagnosis, which simplifies the fault diagnosis process of the lithium battery and has the advantages of low computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnosis method, it can detect slight over-discharge problems of batteries in the battery pack.

[0072] In one embodiment of a lithium-ion battery discharge fault diagnosis method provided in the present application, the lithium-ion battery discharge fault diagnosis method further includes:

[0073] Determining parameters of the voltage prediction model according to the characteristics of the lithium-ion battery;

[0074] The voltage prediction model is trained using historical discharge data of lithium-ion batteries.

[0075] In this embodiment, the input parameters of the voltage prediction model are determined based on the characteristics of the lithium-ion battery. Furthermore, a sliding window is used to predict voltage in a timely manner using newly input operating data. Finally, the regression trees are gradually optimized, and the final predicted voltage is the sum of the results of all the trees.

[0076] In one embodiment of a lithium-ion battery discharge fault diagnosis method provided by the present application, determining the parameters of the voltage prediction model according to the characteristics of the lithium-ion battery includes:

[0077] Determining input characteristic parameters of the voltage prediction model according to characteristic parameters of the lithium-ion battery and the voltage of the lithium-ion battery;

[0078] Using a preset sliding time window to determine the training interval of the voltage prediction model;

[0079] The trained voltage prediction model is tested to optimize the parameters of the voltage prediction model.

[0080] In this embodiment, based on the characteristic parameters and the lithium-ion battery voltage, a correlation coefficient between each characteristic parameter and the lithium-ion battery voltage is calculated; based on the correlation coefficient, the input characteristic parameters of the voltage prediction model are determined. The Pearson correlation coefficient is used to determine the correlation between voltage and different input characteristics, such as temperature, current, SOC, etc., and parameters with high correlation coefficients (greater than a certain threshold) are selected as input characteristics of the battery voltage prediction model in data-driven. The threshold value of the Pearson correlation coefficient is determined through experiments.

[0081] When new time series data is available, the online voltage prediction value should be updated promptly. Therefore, a sliding time window is used to extract the training interval, and the battery characteristics of this calculation interval are input into the trained voltage prediction model for voltage prediction. Furthermore, to minimize the error between the predicted and actual values, the voltage prediction model parameters need to be optimized.

[0082] In one embodiment of a lithium-ion battery discharge fault diagnosis method provided by the present application, the method of using a preset sliding time window to determine the training interval of the voltage prediction model includes:

[0083] forming an input data set for the voltage prediction model based on the lithium-ion battery discharge data;

[0084] When the length of the input data set is greater than the length of the sliding time window, the voltage prediction model performs discharge voltage prediction training.

[0085] In this embodiment, see Figure 2 , with a sliding window of training set, test set and validation set such as Figure 2 As shown. The sample set Vprediction = {(Xk, Yk)} for time series prediction is constructed using a sliding window with a length of w. Xi = {Sk, Sk+1,…, Sk+w-1} is the input data of XGBoost. Sk includes all variables with a Pearson correlation coefficient with voltage higher than a certain threshold, Yk = {Sk+w} is the output data after prediction, and k is the serial number of the sample set. When the data length does not reach the size of the sliding window w, it will be saved and accumulated. Only when the data length is greater than w, the data of the previous w frames will be assimilated and used to predict the next frame. Therefore, the voltage prediction model is given by the following formula:

[0086]

[0087] The size of the training data has a certain impact on XGBoost's prediction accuracy. Large amounts of training data may increase XGBoost's training complexity and training time. However, if the amount of training data is too small, XGBoost will not be able to learn voltage-related features, resulting in inaccurate prediction results. Therefore, the length of the sliding time window in the present invention is determined through experiments.

[0088] In one embodiment of a lithium-ion battery discharge fault diagnosis method provided by the present application, the testing of the trained voltage prediction model and the optimization of the parameters of the voltage prediction model include:

[0089] Obtaining a prediction accuracy of the voltage prediction model based on a discharge voltage test result of the voltage prediction model and a corresponding actual value of the discharge voltage of the lithium-ion battery;

[0090] The deviation value, variance value, and weight value of the voltage prediction model are adjusted according to the prediction accuracy.

[0091] In this embodiment, the prediction model is trained using discharge data from a normal battery. MSE represents the prediction accuracy, and the residual between the predicted value and the actual value is given by the following formula:

[0092]

[0093]

[0094] The flow chart of parameter adjustment is as follows Figure 3 As shown:

[0095] (1) Adjust the number of estimators to optimize the accuracy of XGBoost while fixing other parameters. The accuracy of the model (also known as generalization error) is determined by variance, bias, and noise, which are given by the following formula:

[0096] E = bias 2 +var+ε 2 (2)

[0097] Where E is the generalization error, bias is the deviation, var is the variance, and ε is the noise. By adjusting bias and var, we get the estimator that minimizes the generalization error.

[0098] (2) Adjust the maximum depth and minimum child weight of the tree (affecting the conservatism or complexity of the algorithm).

[0099] (3) Adjust Gamma to make the model more conservative.

[0100] (4) Perform different combinations of subsample and colsample bytree to prevent overfitting.

[0101] (5) Adjust the learning rate. Generally, the learning rate should not be too high to prevent overfitting, nor too low to prevent falling into a local optimal solution.

[0102] The present invention uses MSE to evaluate the prediction results of the test data during each cross-validation. In order to make the model more conservative, ten-fold cross-validation is used to obtain the optimized adjustment results.

[0103] From the above description, it can be seen that the present application provides a lithium-ion battery discharge fault diagnosis method, which includes a second-level diagnosis based on traditional threshold detection and XGBoost. The second-level diagnostic model construction process based on XGBoost includes correlation analysis of battery characteristic parameters, determining the XGBoost sliding window length and the hyperparameters of the model, and finally training the XGBoost prediction model. After the prediction model is trained, the lithium-ion battery secondary over-discharge fault diagnosis is performed. This diagnostic method has the advantages of small computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnostic method, it can detect slight over-discharge problems of batteries in the battery pack.

[0104] Secondly, in order to solve the problem that slight over-discharge of lithium-ion batteries cannot be detected in time, and some batteries with low capacity may tend to be slightly over-discharged due to the inconsistency of battery packs, the present application provides an embodiment of a lithium-ion battery discharge fault diagnosis system, see Figure 4 The lithium-ion battery discharge fault diagnosis system specifically includes the following contents:

[0105] Level 1 diagnostic module 01: determines whether the lithium-ion battery requires fault diagnosis based on the lithium-ion battery voltage and cut-off voltage;

[0106] Voltage prediction model 02: If the lithium-ion battery requires fault diagnosis, the discharge data of the lithium-ion battery is input into the voltage prediction model, and the voltage prediction model outputs the predicted discharge voltage of the lithium-ion battery;

[0107] Secondary diagnosis module 03: determining the fault type of the lithium-ion battery according to the predicted discharge voltage and the cut-off voltage.

[0108] In this embodiment, the primary diagnosis module 01 determines the relationship between the discharge voltage and the cut-off voltage of the lithium-ion battery. If the lithium-ion battery voltage is greater than the cut-off voltage, it indicates that the lithium-ion battery needs to undergo a secondary diagnosis.

[0109] The voltage prediction model 02 starts the voltage prediction model, which predicts the discharge voltage of the lithium-ion battery based on the lithium-ion battery discharge data and transmits the predicted discharge voltage to the secondary diagnosis module 03.

[0110] The secondary diagnosis module 03 determines the relationship between the predicted discharge voltage and the voltage cut-off threshold. If the predicted discharge voltage drops to the cut-off threshold, it indicates that the lithium-ion battery has a slight over-discharge fault.

[0111] As can be seen from the above description, an embodiment of the present application provides a lithium-ion battery discharge fault diagnosis system, which includes a first-level diagnosis based on traditional threshold detection and a second-level diagnosis based on a voltage prediction model. The first-level diagnosis detects over-discharge by comparing the battery voltage with the cut-off voltage. If the battery voltage is greater than the cut-off voltage, the second-level diagnosis based on the voltage prediction model is entered to detect the previous over-discharge problem, and the second-level over-discharge fault diagnosis of the lithium-ion battery is performed through the voltage prediction model. This diagnostic method has the advantages of low computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnosis method, it can detect slight over-discharge problems of batteries in the battery pack.

[0112] The following combination Figure 5 , the overall strategy architecture of a lithium-ion battery discharge fault diagnosis system provided in an embodiment of the present application is specifically described:

[0113] Step 1: Correlation analysis of battery characteristic parameters

[0114] The Pearson correlation coefficient is used to measure the correlation between voltage and various input characteristics, such as temperature, current, and SOC. Parameters with high correlation coefficients (greater than a certain threshold) are selected as input characteristics for the data-driven XGBoost model for battery voltage prediction. The threshold for the Pearson correlation coefficient is determined experimentally.

[0115] Step 2: Determine the XGBoost sliding window length

[0116] In addition, in order to timely predict the voltage using the newly input operating data, a sliding window is used for prediction. Finally, the regression tree is gradually optimized, and the final predicted voltage is the sum of the results of all trees.

[0117] When new time series data is available, the online voltage prediction value should be updated in time. Therefore, a sliding time window is used to extract the training interval, and the battery characteristics of the calculation interval are input into the trained XGBoost for voltage prediction. The training set, test set, and validation set with sliding window are as follows: Figure 2 shown.

[0118] like Figure 2As shown, the sample set Vprediction = {(Xk, Yk)} for time series prediction is constructed using a sliding window with a length of w. Xi = {Sk, Sk+1, ..., Sk+w-1} is the input data of XGBoost. Sk includes all variables with a Pearson correlation coefficient with voltage higher than a certain threshold, Yk = {Sk+w} is the output data after prediction, and k is the sequence number of the sample set. When the data length does not reach the size of the sliding window w, it will be saved and accumulated. Only when the data length is greater than w will the data of the previous w frames be assimilated and used to predict the next frame. Therefore, the voltage prediction model is given by the following formula:

[0119]

[0120] The size of the training data has a certain impact on XGBoost's prediction accuracy. Large amounts of training data may increase XGBoost's training complexity and training time. However, if the amount of training data is too small, XGBoost will not be able to learn voltage-related features, resulting in inaccurate prediction results. Therefore, the length of the sliding time window in the present invention is determined through experiments.

[0121] Step 3: Determine the hyperparameters of the XGBoost model

[0122] In order to minimize the error between the predicted value and the actual value, the XGBoost model parameters are optimized. The flow chart of parameter adjustment is as follows Figure 3 As shown:

[0123] (1) First, adjust the number of estimators to optimize the accuracy of XGBoost while fixing other parameters. The accuracy of the model (also known as generalization error) is determined by variance, bias, and noise, which are given by the following formula:

[0124] E = bias 2 +var+ε 2 (2)

[0125] Where E is the generalization error, bias is the deviation, var is the variance, and ε is the noise. By adjusting bias and var, we get the estimator that minimizes the generalization error.

[0126] (2) Adjust the maximum depth and minimum child weight of the tree (affecting the conservatism or complexity of the algorithm).

[0127] (3) Adjust Gamma to make the model more conservative.

[0128] (4) Perform different combinations of subsample and colsample bytree to prevent overfitting.

[0129] (5) Adjust the learning rate. Generally, the learning rate should not be too high to prevent overfitting, nor too low to prevent falling into a local optimal solution.

[0130] The present invention uses MSE to evaluate the prediction results of the test data during each cross-validation. In order to make the model more conservative, ten-fold cross-validation is used to obtain the optimal adjustment results. The data used to train the XGBoost model is divided into training data and test data in a ratio of 7:3, such as Figure 2 shown.

[0131] Step 4: Train the XGBoost prediction model

[0132] XGBoost is a gradient boosting framework that has been widely used in the power industry. Compared to the gradient boosted decision tree (GDBT), XGBoost's objective function has a regularizer, which improves the iteration effect and operability of each round. In addition, XGBoost supports parallel computing, which greatly reduces calculation time. In this paper, XGBoost is used to predict the discharge voltage of energy storage batteries.

[0133] Classification and Regression Tree (CART) is a typical binary decision tree in XGBoost that can be used for classification and regression processes. If the prediction result is discrete, CART will generate a classification decision tree. On the contrary, if the result is continuous, a regression decision tree is generated, whose output is the average of all predicted samples belonging to a certain leaf node. Gain_σ is selected to evaluate the splitting property of the decision tree. If the Gain_σ obtained by some attributes is small, there is almost no difference between the two subsamples obtained by dichotomy. At the same time, a certain attribute can be defined as a splitting attribute. The continuous prediction results, total variance and Gain_σ obtained from the sample set S can be calculated by the following formula:

[0134]

[0135]

[0136] Where μ and yk represent the average of the prediction results in the sample set S and the prediction result of sample k, respectively. Attribute A divides the sample set into two parts, and Gain_σA(S) is the Gain_σ obtained from attribute A.

[0137] The two parts calculated by attribute A continue to be divided by the above optimization method. Finally, as

[0138] For the sample set S, the minimum value of the optimal binary solution is selected. Compared with GDBT, XGBoost performs parallel computing in segmentation attribute optimization and has high training efficiency.

[0139] The objective optimization function of the XGBoost model is given by:

[0140]

[0141] Where l is the loss function, ft(xi) is the value of the i-th leaf node in the t-th regression tree, and Ω(ft) is the regularizer. The mean squared error (MSE) is used as the loss function to ensure global optimization.

[0142] The discharge voltage prediction based on the XGBoost regression algorithm can be expressed as:

[0143]

[0144] where ft(Xi) is the t-th regression tree, T is the total number of regression trees, and Yi is the i-th predicted voltage.

[0145] The prediction model is trained using the discharge data of a normal battery. The MSE represents the prediction accuracy, and the residual between the predicted value and the actual value is given by the following formula:

[0146]

[0147]

[0148] Step 5: Perform secondary over-discharge fault diagnosis

[0149] After the XGBoost prediction model is trained, a second level of over-discharge fault diagnosis is performed. The first level uses a traditional threshold-based approach to detect over-discharge by comparing the battery voltage with the cutoff voltage. If the battery voltage is greater than the cutoff voltage, the second level of diagnosis based on XGBoost is performed to detect the previous over-discharge problem.

[0150] From a hardware perspective, in order to address the current problem that slight over-discharge of lithium-ion batteries cannot be detected in a timely manner, and due to the inconsistency of battery packs, some low-capacity batteries may be prone to slight over-discharge, the present application provides an embodiment of an electronic device that includes all or part of the content of a lithium-ion battery discharge fault diagnosis method, and the electronic device specifically includes the following content:

[0151] Figure 6 Schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. Figure 6 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that the Figure 6 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0152] In one embodiment, the lithium-ion battery discharge fault diagnosis function can be integrated into the central processing unit. The central processing unit can be configured to perform the following control:

[0153] Step 100: Performing a primary diagnosis on the lithium-ion battery according to the lithium-ion battery voltage and the cut-off voltage, and then determining whether the lithium-ion battery requires a secondary diagnosis;

[0154] The cutoff voltage is the final voltage, which refers to the minimum operating voltage at which the battery can no longer discharge during discharge. The cutoff voltage varies depending on the battery type and discharge conditions. If the voltage of a lithium-ion battery is greater than the cutoff voltage, it indicates that the battery requires secondary diagnosis.

[0155] Step 200: If the lithium-ion battery requires secondary diagnosis, the discharge parameters of the lithium-ion battery are input into a voltage prediction model, and the voltage prediction model outputs a predicted discharge voltage of the lithium-ion battery;

[0156] It is understandable that the voltage prediction model of this application is the XGBoost model. XGBoost is a gradient boosting framework that has been widely used in the power industry. Compared with the gradient boosting decision tree (GDBT), the objective function of XGBoost has a regularizer, and the iterative effect and operability of each round of iteration will be improved. In addition, XGBoost supports parallel computing, which greatly reduces the calculation time. In the present invention, XGBoost is used to predict the discharge segment voltage of the energy storage battery.

[0157] Classification and Regression Tree (CART) is a typical binary decision tree in XGBoost that can be used for classification and regression processes. If the prediction result is discrete, CART will generate a classification decision tree. On the contrary, if the result is continuous, a regression decision tree is generated, whose output is the average of all predicted samples belonging to a certain leaf node. Gain_σ is selected to evaluate the splitting property of the decision tree. If the Gain_σ obtained by some attributes is small, there is almost no difference between the two subsamples obtained by dichotomy. At the same time, a certain attribute can be defined as a splitting attribute. The continuous prediction results, total variance and Gain_σ obtained from the sample set S can be calculated by the following formula:

[0158]

[0159]

[0160] Where μ and yk represent the average of the prediction results in the sample set S and the prediction result of sample k, respectively. Attribute A divides the sample set into two parts, and Gain_σA(S) is the Gain_σ obtained from attribute A.

[0161] The two parts calculated by attribute A continue to be divided by the above optimization method. Finally, as

[0162] For the sample set S, the minimum value of the optimal binary solution is selected. Compared with GDBT, XGBoost performs parallel computing in segmentation attribute optimization and has high training efficiency.

[0163] The objective optimization function of the XGBoost model is given by:

[0164]

[0165] Where l is the loss function, ft(xi) is the value of the i-th leaf node in the t-th regression tree, and Ω(ft) is the regularizer. The mean squared error (MSE) is used as the loss function to ensure global optimization.

[0166] The discharge voltage prediction based on the XGBoost regression algorithm can be expressed as:

[0167]

[0168] Among them, f t (X i ) is the t-th regression tree, T is the total number of regression trees, Y i is the i-th predicted voltage.

[0169] Step 300: Determine the fault type of the lithium-ion battery according to the predicted discharge voltage;

[0170] It is understandable that when the predicted discharge voltage drops to the cut-off threshold, it indicates that the lithium-ion battery has a slight over-discharge fault.

[0171] As can be seen from the above description, an electronic device provided by an embodiment of the present application includes a primary diagnosis based on traditional threshold detection and a secondary diagnosis based on a voltage prediction model. The primary diagnosis detects over-discharge by comparing the battery voltage with the cut-off voltage. If the battery voltage is greater than the cut-off voltage, the secondary diagnosis based on the voltage prediction model is entered to detect the previous over-discharge problem, and the secondary over-discharge fault diagnosis of the lithium-ion battery is performed through the voltage prediction model. This diagnostic method has the advantages of low computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnostic method, it can detect slight over-discharge problems of batteries in the battery pack.

[0172] In another embodiment, the lithium-ion battery discharge fault diagnosis device can be configured separately from the central processing unit 9100. For example, the lithium-ion battery discharge fault diagnosis device can be configured as a chip connected to the central processing unit 9100, and the lithium-ion battery discharge fault diagnosis function can be realized through the control of the central processing unit.

[0173] like Figure 6 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily have to include Figure 6 In addition, the electronic device 9600 may also include all components shown in Figure 6 For components not shown, reference may be made to the prior art.

[0174] like Figure 6 As shown, the central processing unit 9100 is sometimes also referred to as a controller or operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.

[0175] Memory 9140 can be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It can store the aforementioned failure-related information and also store programs that execute the relevant information. The CPU 9100 can execute the programs stored in memory 9140 to implement information storage or processing.

[0176] The input unit 9120 provides input to the central processing unit 9100. The input unit 9120 may be, for example, a keypad or touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display objects such as images and text. The display may be, for example, an LCD display, but is not limited thereto.

[0177] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), or a SIM card. Alternatively, it may be a memory that retains information even when power is off, can be selectively erased, and is provided with more data. Examples of such memory are sometimes referred to as EPROMs. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 by the central processing unit 9100.

[0178] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various driver programs for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0179] The communication module 9110 is a transmitter / receiver 9110 that transmits and receives signals via an antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as in a conventional mobile communication terminal.

[0180] Based on different communication technologies, multiple communication modules 9110 can be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby implementing common telecommunication functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Furthermore, the audio processor 9130 is also coupled to the central processing unit 9100, enabling local recording via the microphone 9132 and playback of stored audio via the speaker 9131.

[0181] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all steps of the lithium-ion battery discharge fault diagnosis method in the above-mentioned embodiment. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, all steps of the lithium-ion battery discharge fault diagnosis method in the above-mentioned embodiment, in which the execution subject is a server or a client, are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0182] Step 100: Performing a primary diagnosis on the lithium-ion battery according to the lithium-ion battery voltage and the cut-off voltage, and then determining whether the lithium-ion battery requires a secondary diagnosis;

[0183] The cutoff voltage is the final voltage, which refers to the minimum operating voltage at which the battery can no longer discharge during discharge. The cutoff voltage varies depending on the battery type and discharge conditions. If the voltage of a lithium-ion battery is greater than the cutoff voltage, it indicates that the battery requires secondary diagnosis.

[0184] Step 200: If the lithium-ion battery requires secondary diagnosis, the discharge parameters of the lithium-ion battery are input into a voltage prediction model, and the voltage prediction model outputs a predicted discharge voltage of the lithium-ion battery;

[0185] It is understandable that the voltage prediction model of this application is the XGBoost model. XGBoost is a gradient boosting framework that has been widely used in the power industry. Compared with the gradient boosting decision tree (GDBT), the objective function of XGBoost has a regularizer, and the iterative effect and operability of each round of iteration will be improved. In addition, XGBoost supports parallel computing, which greatly reduces the calculation time. In the present invention, XGBoost is used to predict the discharge segment voltage of the energy storage battery.

[0186] Classification and Regression Tree (CART) is a typical binary decision tree in XGBoost that can be used for classification and regression processes. If the prediction result is discrete, CART will generate a classification decision tree. On the contrary, if the result is continuous, a regression decision tree is generated, whose output is the average of all predicted samples belonging to a certain leaf node. Gain_σ is selected to evaluate the splitting property of the decision tree. If the Gain_σ obtained by some attributes is small, there is almost no difference between the two subsamples obtained by dichotomy. At the same time, a certain attribute can be defined as a splitting attribute. The continuous prediction results, total variance and Gain_σ obtained from the sample set S can be calculated by the following formula:

[0187]

[0188]

[0189] Where μ and yk represent the average of the prediction results in the sample set S and the prediction result of sample k, respectively. Attribute A divides the sample set into two parts, and Gain_σA(S) is the Gain_σ obtained from attribute A.

[0190] The two parts calculated by attribute A continue to be divided by the above optimization method. Finally, as

[0191] For the sample set S, the minimum value of the optimal binary solution is selected. Compared with GDBT, XGBoost performs parallel computing in segmentation attribute optimization and has high training efficiency.

[0192] The objective optimization function of the XGBoost model is given by:

[0193]

[0194] Where l is the loss function, ft(xi) is the value of the i-th leaf node in the t-th regression tree, and Ω(ft) is the regularizer. The mean squared error (MSE) is used as the loss function to ensure global optimization.

[0195] The discharge voltage prediction based on the XGBoost regression algorithm can be expressed as:

[0196]

[0197] Among them, f t (X i ) is the t-th regression tree, T is the total number of regression trees, Y i is the i-th predicted voltage.

[0198] Step 300: Determine the fault type of the lithium-ion battery according to the predicted discharge voltage;

[0199] It is understandable that when the predicted discharge voltage drops to the cut-off threshold, it indicates that the lithium-ion battery has a slight over-discharge fault.

[0200] As can be seen from the above description, an embodiment of the present application provides a computer-readable storage medium, which includes a primary diagnosis based on traditional threshold detection and a secondary diagnosis based on a voltage prediction model. The primary diagnosis detects over-discharge by comparing the battery voltage with the cut-off voltage. If the battery voltage is greater than the cut-off voltage, the secondary diagnosis based on the voltage prediction model is entered to detect the previous over-discharge problem, and the secondary over-discharge fault diagnosis of the lithium-ion battery is performed through the voltage prediction model. This diagnostic method has the advantages of low computational complexity and high accuracy. Compared with the traditional threshold-based overcharge diagnostic method, it can detect slight over-discharge problems of batteries in the battery pack.

[0201] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, apparatus, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0202] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks 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 produce 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.

[0203] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0204] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0205] Specific embodiments are used in the present invention to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A method for diagnosing discharge faults of lithium-ion batteries, characterized in that: include: Determining whether the lithium-ion battery requires fault diagnosis based on the lithium-ion battery voltage and the cut-off voltage; If the lithium-ion battery requires fault diagnosis, the discharge parameters of the lithium-ion battery are input into a voltage prediction model, and the voltage prediction model outputs a predicted discharge voltage of the lithium-ion battery; determining a fault type of the lithium-ion battery according to the predicted discharge voltage; The lithium-ion battery discharge fault diagnosis method further includes: Determining parameters of the voltage prediction model according to the characteristics of the lithium-ion battery; Using historical discharge data of lithium-ion batteries to train the voltage prediction model; Determining the parameters of the voltage prediction model according to the characteristics of the lithium-ion battery includes: Determining input characteristic parameters of the voltage prediction model according to characteristic parameters of the lithium-ion battery and the voltage of the lithium-ion battery; Using a preset sliding time window to determine the training interval of the voltage prediction model; Testing the trained voltage prediction model to optimize parameters of the voltage prediction model; The step of determining the input characteristic parameters of the voltage prediction model according to the characteristic parameters of the lithium-ion battery and the voltage of the lithium-ion battery includes: Calculating, based on the characteristic parameters and the voltage of the lithium-ion battery, a correlation coefficient between each characteristic parameter and the voltage of the lithium-ion battery; Determining input characteristic parameters of the voltage prediction model according to the correlation coefficient; The method of using a preset sliding time window to determine the training interval of the voltage prediction model includes: forming an input data set for the voltage prediction model based on the lithium-ion battery discharge data; When the length of the input data set is greater than the length of the sliding time window, the voltage prediction model performs discharge voltage prediction training; The testing of the trained voltage prediction model to optimize parameters of the voltage prediction model includes: Obtaining a prediction accuracy of the voltage prediction model based on a discharge voltage test result of the voltage prediction model and a corresponding actual value of the discharge voltage of the lithium-ion battery; The deviation value, variance value, and weight value of the voltage prediction model are adjusted according to the prediction accuracy.

2. The lithium-ion battery discharge fault diagnosis method according to claim 1, characterized in that: The characteristic parameters of the lithium-ion battery include: lithium-ion battery temperature, lithium-ion battery current and lithium-ion battery SOC.

3. A lithium-ion battery discharge fault diagnosis system, based on the lithium-ion battery discharge fault diagnosis method according to any one of claims 1 to 2, characterized in that: include: A primary diagnosis module: performs a primary diagnosis on the lithium-ion battery according to the lithium-ion battery voltage and cut-off voltage, and then determines whether the lithium-ion battery requires a secondary diagnosis; Voltage prediction model: if the lithium-ion battery requires secondary diagnosis, the discharge data of the lithium-ion battery is input into the voltage prediction model, and the voltage prediction model outputs the predicted discharge voltage of the lithium-ion battery; A secondary diagnosis module is configured to determine a fault type of the lithium-ion battery according to the predicted discharge voltage and the cut-off voltage.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the lithium-ion battery discharge fault diagnosis method according to any one of claims 1 to 2 is implemented.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the lithium-ion battery discharge fault diagnosis method according to any one of claims 1 to 2 is implemented.

Images