Battery fluid thermal fault early warning method, computer device and computer storage medium
By collecting operation data of power batteries and hydraulic system in real time, and using the XGBoost model to predict faults, the problem of low accuracy of fault warning in existing hydraulic system is solved, and higher prediction accuracy is achieved and the safety risks of power batteries are reduced.
Patent Information
- Application Number
- CN202510419574.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-03
- Publication Date
- 2025-07-11
AI Technical Summary
The existing hydraulic and thermal system fault warning depends on simple threshold judgment, and the accuracy is limited, so accurate warning cannot be achieved, resulting in the risk of degradation in power battery performance or safety accidents.
By collecting the operating data of the power battery and hydraulic system in real time, extracting features and using the XGBoost model for fault prediction, calculate the temperature increase rate of the highest temperature, the temperature increase rate of the lowest temperature, and the temperature difference of the battery cell are expanded, and combined with machine learning algorithms to conduct hydraulic thermal fault warning.
It improves the accuracy of hydraulic thermal failure prediction and effectively reduces the risk of power battery performance degradation and safety accidents.
Smart Images
Figure CN120287844A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fault prediction, and particularly to a method for warning of battery liquid heat faults, a computer device, and a computer storage medium. Background Art
[0002] With the rapid development of new energy vehicles, electric vehicles have become new energy vehicles with a wide range of applications in current society due to their relatively low cost and high safety. As the core component of an electric vehicle, the safety and stability of the power battery are crucial.
[0003] During the operation of the power battery, the ambient temperature where the power battery is located has a great influence on the battery output power. In a low-temperature environment, the battery may experience serious power loss, resulting in a reduction in the driving range of the electric vehicle. In existing electric vehicles, a liquid heat management system is used to ensure that the power battery operates within an appropriate temperature range. During actual operation, temperature changes and current fluctuations may both cause faults in the power battery liquid heat system, thereby leading to a decline in the performance of the power battery and even triggering safety accidents. In the prior art, warning of liquid heat system faults often relies on simple threshold judgment, and this threshold is generally set by professionals based on experience, with limited accuracy and unable to achieve precise warning. Summary of the Invention
[0004] The purpose of the present invention is to provide a method for warning of battery liquid heat faults, a computer device, and a computer storage medium to solve the problem that precise warning cannot be achieved for existing warning of liquid heat system faults.
[0005] The present invention provides a method for warning of battery liquid heat faults to solve the above technical problems. The steps include:
[0006] 1) Real-time collect the operation data of the power battery and the liquid heat system during driving or charging and extract features, and calculate the heating rate of the highest temperature of the battery cell, the heating rate of the lowest temperature of the battery cell, and the temperature difference expansion amount by using the extracted features; the extracted features include: the temperature of each battery cell at the start of heating, the temperature of each battery cell at the end of heating, the outlet water temperature at the start of heating, the return water temperature at the start of heating, the outlet water temperature at the end of heating, and the return water temperature at the end of heating;
[0007] 2) Input the heating rate of the highest temperature of the battery cell, the heating rate of the lowest temperature of the battery cell, the temperature difference expansion amount, and the extracted features into a trained machine learning model for battery liquid heat fault prediction. When training the machine learning model, the training data includes the operation data of the power battery and the liquid heat system corresponding to the existence of battery liquid heat faults and the operation data of the power battery and the liquid heat system corresponding to normal battery liquid heat.
[0008] Further, the SOC of the battery at the start of heating and the total current of the battery at the start of heating.
[0009] Further, the calculation method for the heating rate of the highest temperature of the battery cell is: subtracting the highest temperature of the battery cell at the end of heating from the highest temperature of the battery cell at the start of heating, and dividing the difference by the heating time to obtain the heating rate of the highest temperature of the battery cell; the calculation method for the heating rate of the lowest temperature of the battery cell is: subtracting the lowest temperature of the battery cell at the end of heating from the lowest temperature of the battery cell at the start of heating, and dividing the difference by the heating time.
[0010] Further, the amount of temperature difference expansion is the absolute value of the difference between the end temperature difference of the battery and the start temperature difference of the battery. The start temperature difference of the battery is the difference between the highest temperature and the lowest temperature of the battery cell at the start of heating, and the end temperature difference of the battery is the difference between the highest temperature and the lowest temperature of the battery cell at the end of heating.
[0011] Further, the machine learning model is an XGBoost model.
[0012] Further, the adopted XGBoost model is a model with qualified accuracy. When the accuracy of the XGBoost model does not meet the requirements, the model parameters are optimized until the accuracy meets the requirements.
[0013] Further, the operation data of the power battery and the liquid heating system during driving or charging when feature extraction is performed are data after data cleaning and denoising processing.
[0014] Further, the battery liquid heating fault warning method further includes: giving an alarm when a liquid heating fault is detected by the machine learning model.
[0015] A computer device, characterized in that the computer device includes a processor to implement the battery liquid heating fault warning method as described above.
[0016] A computer storage medium, in which a computer program is stored to implement the battery liquid heating fault warning method as described above.
[0017] The beneficial effects of the present invention are as follows: As an improved invention, the present invention extracts corresponding characteristic data from the operation data by collecting the operation data of the power battery and the liquid heating system in real time, and calculates using the characteristic data. The characteristic data, the heating rate of the highest temperature of the battery cell, the heating rate of the lowest temperature of the battery cell, and the temperature difference expansion amount are used as the input data of the machine learning model. The machine learning model is used to learn the relationship between these data and the liquid heating fault, and the liquid heating fault is predicted based on the learned relationship. The characteristics extracted from the operation data by the present invention, as well as the heating rate of the highest temperature of the battery cell, the heating rate of the lowest temperature of the battery cell, and the temperature difference expansion amount calculated, all affect the liquid heating management system. The present invention uses machine learning algorithms and more comprehensive data to predict the faults of the liquid heating system, effectively improving the accuracy of liquid heating fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 is the flow chart for constructing and evaluating the power battery liquid heating fault system of the present invention;
[0019] Figure 2 is the data source of the power battery in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0020] The following further describes the specific embodiments of the present invention with reference to the accompanying drawings.
[0021] The present application proposes a method for warning liquid heating faults of batteries, which uses machine learning algorithms and more comprehensive data to predict liquid heating fault data, improving the accuracy of liquid heating fault prediction.
[0022] Embodiment of the method for warning liquid heating faults of batteries
[0023] As Figure 1 shown, the method for warning liquid heating faults of batteries based on XGBoost proposed by the present invention includes the following steps:
[0024] 1) Data collection and preprocessing.
[0025] Collect the operation data of the power battery and the liquid heating system in real time during driving or charging, extract the characteristics, and preprocess the collected data. The preprocessing process specifically includes data cleaning, noise removal, and data integrity evaluation. The data collection source is as Figure 2 shown. Subsequently, feature extraction is performed on the processed data. The extracted features include: the temperature of each battery cell at the start of heating, the temperature of each battery cell at the end of heating, the outlet water temperature at the start of heating, the return water temperature at the start of heating, the outlet water temperature at the end of heating, the return water temperature at the end of heating, the SOC of the battery at the start of heating, and the total current of the battery at the start of heating.
[0026] Subsequently, the heating rate of the highest temperature of the battery cell, the heating rate of the lowest temperature, and the temperature difference expansion amount are calculated using the extracted features.
[0027] Specifically, the calculation method for the heating rate of the highest temperature of the battery cell is as follows: subtract the highest temperature of the battery cell at the end of heating from the highest temperature of the battery cell at the start of heating, and divide the difference by the liquid thermal heating time.
[0028] The calculation formula for the heating rate ratio1 of the highest temperature of the battery cell is:
[0029]
[0030] where heat bgn_maxt is the highest temperature of the battery cell at the start of heating, heat end_maxt is the highest temperature of the battery cell at the end of heating, and time during is the liquid thermal heating time.
[0031] The calculation method for the heating rate of the lowest temperature of the battery cell is: subtract the lowest temperature of the battery cell at the end of heating from the lowest temperature of the battery cell at the start of heating, and divide the difference by the heating time. The calculation formula for the heating rate ratio2 of the lowest temperature of the battery cell is:
[0032]
[0033] where heat end_mint is the lowest temperature of the battery cell at the start of heating, and heat bgn_mint is the lowest temperature of the battery cell at the end of heating.
[0034] The temperature difference expansion amount is the absolute value of the difference between the end temperature difference and the starting temperature difference of the battery. The calculation method for the end temperature difference of the battery is: calculate the difference between the highest temperature and the lowest temperature of the battery cell at the start of heating as the starting temperature difference of the battery, and calculate the difference between the highest temperature and the lowest temperature of the battery cell at the end of heating as the end temperature difference of the battery.
[0035] The calculation formula for the temperature difference expansion amount temp diff is:
[0036] temp diff = |(heat end_maxt - heat end_mint ) - (heat bgn_maxt - heat bgn_mint )|.
[0037] The finally obtained data for fault warning is shown in Table 1. Using this data, the fault prediction of the power battery can be carried out.
[0038] Table 1 Classification and Explanation of Evaluation Indicators
[0039] Serial Number Field Name Valid Value Range Unit 1 Battery Management System Status Driving or Charging / 2 Highest Temperature of Single Cell at Heating Start 0-100 ℃ 3 Lowest Temperature of Single Cell at Heating End 0-100 ℃ 4 SOC at Heating Start 0-100 ％ 5 Total Current at Heating Start / A 6 Outlet Water Temperature at Heating Start 0-100 ℃ 7 Return Water Temperature at Heating Start 0-100 ℃ 8 Highest Temperature of Single Cell at Heating End 0-100 ℃ 9 Lowest Temperature of Single Cell at Heating End 0-100 ℃ 10 Outlet Water Temperature at Heating End 0-100 ℃ 11 Return Water Temperature at Heating End 0-100 ℃ 12 Temperature Rise Rate of the Highest Temperature / ℃ / min 13 Temperature Rise Rate of the Lowest Temperature / ℃ / min
[0040] 2) Use the trained machine learning model to predict the thermal failure of the battery liquid.
[0041] The present invention adopts the XGBoost (Extreme Gradient Boosting Decision Tree) extreme gradient boosting classification algorithm to perform parallel calculations on the characteristic data of the power battery and the liquid thermal system, and determines whether there is a risk of failure of the power battery.
[0042] In order to use the XGBoost model to predict the thermal failure of the battery liquid, it is necessary to train the XGBoost model. First, initialize the parameters of the XGBoost model, and determine the number, depth, and learning rate of the decision trees. Subsequently, train the initialized XGBoost model. In the XGBoost model, new trees are split based on features. Each time a new tree is added to the model, it is to learn a new function f(x) to fit the residual of the previous prediction. When the training is completed, k trees are obtained in the XGBoost model, and there is a corresponding score at each leaf node of each tree. Different features of a sample will fall to the corresponding leaf nodes on the corresponding trees, and the score corresponding to the leaf node is obtained. The scores obtained by all features of a sample are added together to obtain the sample score, and it is judged whether the battery corresponding to the sample fails according to the sample score.
[0043] Specifically, the XGBoost model is trained using the training dataset. The training dataset includes N samples, and each sample includes m features. According to the data types provided in Table 1, m is 13 in this embodiment. The expression of the training dataset D is:
[0044] D = {(x i , y i ): x i ∈R m , y i ∈R, |D| = N}
[0045] Among them, x i represents the i-th feature, y i is the predicted value, and R is the sample set.
[0046] The XGBoost model contains CART (Classification And Regression Tree classification and regression tree). The set of classification and regression trees is denoted as:
[0047] F = {fk (x i ) = w(x i ): k ∈ R m → T}
[0048] where k is the number of CART trees, T is all the leaf nodes of a tree, w(x i ) is the rule for mapping the input samples to a definite leaf node, w is the score corresponding to this leaf node, and f k (x i ) is the objective function.
[0049] Use the XGBoost algorithm based on CART trees to predict y i , and the expression of the predicted value is:
[0050]
[0051] where the regularization term Ω(f t ) is a function representing the complexity of the tree. The smaller the value, the lower the complexity and the stronger the generalization ability. Before training the XGBoost model, set the splitting gain threshold and the maximum number of splits. During the training process, when the column-wise gain is lower than the splitting gain threshold or the number of splits has reached the maximum number of splits, stop the training to obtain the trained XGBoost model.
[0052] To ensure the good accuracy of the trained XGBoost model, in this embodiment, use the evaluation dataset to optimize the XGBoost model parameters by methods such as cross-validation and grid search. To ensure the accuracy of the evaluation results, the total number of vehicles collected in the evaluation dataset should not be less than 1000, and the data should include the basic faults of the existing liquid heat systems in the market, such as large temperature difference under heating and no temperature rise during heating. Whether the vehicles corresponding to the data in the evaluation dataset are faulty should be known. The ratio of the number of faulty vehicles to normal vehicles in the evaluation dataset can be set by oneself.
[0053] Input the evaluation dataset into the trained XGBoost model for fault prediction, and use the confusion matrix shown in Table 2 to evaluate the results output by the XGBoost model.
[0054] Table 2 Confusion Matrix of the Prediction Results of the Liquid Heat Fault Warning of Power Batteries
[0055]
[0056] The confusion matrix classifies the prediction results output by the XGBoost model into four categories, including normal vehicles predicted as normal vehicles, faulty vehicles predicted as normal vehicles, normal vehicles predicted as faulty vehicles, and faulty vehicles predicted as faulty vehicles. The XGBoost model is evaluated based on the proportion of these four situations in all prediction results. The specific classification and description of the evaluation indicators are shown in Table 3.
[0057] Table 3 Classification and Description of Evaluation Indicators
[0058]
[0059]
[0060] Among them, the recall rate refers to the proportion of all actually normal vehicles that are determined to be normal vehicles by the XGBoost model. The calculation formula is:
[0061] Precision = TP / (TP + FP)
[0062] Among them, Precision is the recall rate, with the unit of %, TP is the number of samples of actually normal vehicles determined to be normal vehicles, and FP is the number of samples of actually faulty vehicles determined to be faulty vehicles.
[0063] The precision rate refers to the proportion of all samples correctly determined by the XGBoost model in the total number of samples. The calculation formula is:
[0064] Recall = TP / (TP + FN)
[0065] Among them, Recall is the precision rate, with the unit of %, and FN is the number of samples of actually normal vehicles determined to be faulty vehicles.
[0066] The accuracy rate refers to the proportion of correctly determined vehicles among all vehicles determined to be normal by the XGBoost model. The calculation formula is:
[0067] Accuracy = (TP + TN) / (TP + TN + FP + FN)
[0068] Among them, Accuracy is the accuracy rate, with the unit of %, and TN is the number of samples of actually faulty vehicles determined to be faulty vehicles.
[0069] The F1 score refers to the harmonic mean of the recall rate, precision rate, and accuracy rate. The F1 score is mainly used to measure the precision and accuracy of the model. The calculation formula is:
[0070] F1 = 2 * (Precision * Recall) / (Precision + Recall)
[0071] Among them, F1 is the F1 score, with the unit of %, and the closer the F1 score is to 1, the better the performance of the evaluated model. In the present invention, the evaluated model is the XGBoost model.
[0072] The XGBoost model is evaluated based on recall, precision, accuracy, and the F1 score. The specific evaluation rules are shown in Table 4.
[0073] Table 4 Evaluation level determination rules
[0074] Evaluation Index Evaluation Grade Grade Characteristics Accuracy Rate ≥ 90%, Precision Rate ≥ 90%, Recall Rate ≥ 90%, F1 Score ≥ 90% A Demonstration Leadership Accuracy Rate ≥ 85%, Precision Rate ≥ 85%, Recall Rate ≥ 85%, F1 Score ≥ 85% B Outstanding Performance Accuracy Rate ≥ 75%, Precision Rate ≥ 75%, Recall Rate ≥ 75%, F1 Score ≥ 75% C Effective Development Accuracy Rate ≥ 70%, Precision Rate ≥ 70%, Recall Rate ≥ 70%, F1 Score ≥ 70% D Initial Scale Accuracy Rate ≥ 60%, Precision Rate ≥ 60%, Recall Rate ≥ 60%, F1 Score ≥ 60% E Qualified Others F Unqualified
[0075] Combined with the evaluation level result and the precision required for the XGBoost model in actual use, it is determined whether the evaluated XGBoost model needs to be optimized. If the precision of the evaluated XGBoost model cannot meet the usage requirements, the model parameters are optimized through methods such as cross-validation and grid search. Subsequently, the XGBoost model with optimized parameters is evaluated again until the precision of the XGBoost model meets the actual requirements.
[0076] In this embodiment, 4442 pieces of power battery and liquid thermal system data collected from December 1, 2023 to December 31, 2023 are allocated according to 4:1. 4 / 5 of the data is used as the training dataset for training the XGBoost model, and the remaining 1 / 5 of the data is used as the evaluation dataset for testing the precision of the XGBoost model. The test results are shown in Table 5.
[0077] Table 5 Test results of power battery liquid thermal fault warning based on XGBOOST
[0078] Faulty Vehicle Predicted Correctly Faulty Vehicle Predicted Incorrectly Non-Faulty Vehicle Predicted Incorrectly Number of Vehicles 114 10 5
[0079] Using the method proposed in the present invention to predict the power battery liquid thermal fault, the accuracy rate of the XGBoost model on the test data is 99.2%. The expected accuracy rate is 95%, which actually meets the project's expected requirements, that is, the safety warning algorithm model can meet the standards.
[0080] Embodiment of the computer device
[0081] The present invention discloses a computer device, which includes a processor to implement the above-mentioned battery liquid thermal fault warning method based on XGBoost.
[0082] Specifically, the computer program includes a data acquisition module, a data processing module, a data calculation module, and an XGBoost model. Among them, the acquisition module is mainly used to collect the operation data of the power battery and the liquid thermal system in real time. The data processing module is mainly used to clean, denoise, and extract features from the data. The data calculation module is mainly used to calculate the heating rate and the temperature difference expansion amount of the highest temperature and the lowest temperature of the battery cells according to the feature extraction results. The XGBoost model is mainly used to determine the liquid thermal fault of the power battery according to the highest temperature of the battery cells, the heating rate and the temperature difference expansion amount of the lowest temperature of the battery cells, and the extracted features.
[0083] The computer device embodiments described above are only illustrative. The division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.
[0084] The specific implementation process has been described in detail in the method embodiments and will not be repeated here.
[0085] Computer storage medium
[0086] The present invention provides a computer storage medium in which a computer program is stored to implement the battery liquid thermal fault warning method as described above.
[0087] The specific implementation process has been described in detail in the method embodiments and will not be repeated here.
Claims
1. A method for warning of thermal faults in a battery liquid, characterized in that the steps Including: 1) Real-time collect the operation data of the power battery and the liquid heating system during driving or charging, extract features, and calculate the heating rate of the highest temperature of the battery cell, the heating rate of the lowest temperature, and the enlarged temperature difference using the extracted features; the extracted features include: the temperature of each battery cell at the start of heating, the temperature of each battery cell at the end of heating, the outlet water temperature at the start of heating, the return water temperature at the start of heating, the outlet water temperature at the end of heating, and the return water temperature at the end of heating; 2) Input the heating rate of the highest temperature of the battery cell, the heating rate of the lowest temperature of the battery cell, the enlarged temperature difference, and the extracted features into a trained machine learning model for battery liquid heating fault prediction. When training the machine learning model, the training data includes the operation data of the power battery and the liquid heating system corresponding to the existence of battery liquid heating faults and the operation data of the power battery and the liquid heating system corresponding to normal battery liquid heating.
2. The battery liquid thermal fault warning method according to claim 1, wherein The extracted features also include: the SOC of the battery at the start of heating and the total current of the battery at the start of heating.
3. The battery liquid thermal fault warning method according to claim 1, wherein The calculation method of the heating rate of the highest temperature of the battery cell is: subtract the highest temperature of the battery cell at the end of heating from the highest temperature of the battery cell at the start of heating, and divide the difference by the heating time to obtain the heating rate of the highest temperature of the battery cell; the calculation method of the heating rate of the lowest temperature of the battery cell is: subtract the lowest temperature of the battery cell at the end of heating from the lowest temperature of the battery cell at the start of heating, and divide the difference by the heating time.
4. The battery liquid thermal fault warning method according to claim 1, wherein, The enlarged temperature difference is the absolute value of the difference between the end temperature difference of the battery and the starting temperature difference of the battery. The starting temperature difference of the battery is the difference between the highest temperature and the lowest temperature of the battery cell at the start of heating, and the end temperature difference of the battery is the difference between the highest temperature and the lowest temperature of the battery cell at the end of heating.
5. The battery liquid thermal fault warning method according to claim 1, wherein The machine learning model is an XGBoost model.
6. The battery liquid thermal fault warning method according to claim 5, wherein The adopted XGBoost model is a model with precision meeting the requirements. When the precision of the XGBoost model does not meet the requirements, optimize the model parameters until the precision meets the requirements.
7. The battery liquid thermal fault warning method according to claim 1, wherein The operation data of the power battery and the liquid heating system during driving or charging when extracting features are data after data cleaning and denoising processing.
8. The battery liquid thermal fault warning method according to claim 1, wherein The battery liquid heating fault warning method further includes: alarm when the liquid heating fault detected by the machine learning model occurs.
9. A computer device, characterized in that, The computer device includes a processor to implement the battery liquid heating fault warning method according to any one of claims 1-8 above.
10. A computer storage medium, characterized in that, The computer storage medium stores a computer program to implement the battery liquid heating fault warning method according to any one of claims 1-8 above.