Electric vehicle power battery state early warning method, system, equipment and medium
Through the battery state warning method based on the GRU neural network model, the battery inconsistency and undervoltage problems are predicted in advance, and the hysteresis and high cost problems in the existing technology are solved, and early warning of electric vehicle batteries and battery life extension are achieved.
Patent Information
- Application Number
- CN202410872646.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-01
- Publication Date
- 2024-11-08
AI Technical Summary
The existing power battery abnormality detection methods have hysteresis and high cost, making it difficult to achieve early warning of battery abnormalities, affecting the safe operation of electric vehicles and battery life.
The voltage difference and minimum single voltage prediction model based on the GRU neural network model are used to construct a prediction model through parameters such as total voltage, highest single voltage, battery SOC, etc., and data analysis is carried out in combination with attention mechanism to predict battery inconsistency and undervoltage in advance. Problem, and set thresholds according to battery type to provide early warning.
It realizes early warning of the battery status, reduces computing power requirements, is suitable for real-time applications, improves the accuracy and applicability of the early warning, and extends the service life of the battery.
Smart Images

Figure FDA0004922080080000021 
Figure FDA0004922080080000022 
Figure HDA0004922080100000011
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of electric vehicles, and in particular relates to a method, system, device and medium for early warning of the power battery status of an electric vehicle. Background Art
[0002] As the core component of new energy vehicles, the performance and safety of power batteries are directly related to the operating efficiency of the entire vehicle and the safety of passengers. With the widespread promotion of electric vehicles, the health status monitoring of power batteries has become particularly important. At present, the abnormal warning of power batteries mainly relies on the real-time monitoring of the battery management system, which evaluates the battery status by monitoring parameters such as voltage, current, and temperature. However, these traditional monitoring methods often only detect problems when the battery is obviously abnormal.
[0003] For early warning of battery abnormalities, current methods generally use complex machine learning algorithms such as LSTM for prediction, which require high computing power and long training time, resulting in low cost and adaptability. Therefore, there is an urgent need for a cost-effective and accurate power battery abnormality warning method to achieve early warning of battery abnormalities to ensure the safe operation of electric vehicles and extend the battery life. Summary of the invention
[0004] The purpose of the present invention is to provide an electric vehicle power battery status early warning method, system, device and medium to solve the technical problems of hysteresis and high cost in existing power battery abnormality detection methods.
[0005] In order to achieve the above object, the present invention adopts the following technical scheme:
[0006] An electric vehicle power battery status early warning method comprises the following steps:
[0007] A pressure difference prediction model is constructed with total voltage, maximum cell voltage, minimum cell voltage, and battery SOC as inputs and pressure difference as output. A minimum cell voltage prediction model is constructed with total voltage, maximum cell voltage, pressure difference, and battery SOC as inputs and minimum cell voltage as output.
[0008] Output prediction results using the trained voltage difference prediction model and the minimum cell voltage prediction model;
[0009] Compare the prediction results with the threshold and issue corresponding warnings when the threshold is exceeded;
[0010] The trained voltage difference prediction model and the minimum cell voltage prediction model both adopt a GRU-based neural network model.
[0011] Furthermore, the data sets for training and verifying the pressure difference prediction model and the minimum cell voltage prediction model come from the collection of power battery related data during vehicle driving.
[0012] Furthermore, the data set is preprocessed, and the specific process is to remove missing values and outliers and interpolate the missing values using the weighted moving average method while achieving noise reduction for the overall data.
[0013] Furthermore, the abnormal value is the data of the collection point that has a large deviation from the actual situation in the collected data.
[0014] Furthermore, the GRU-based neural network model consists of an input layer, an intermediate layer, and an output layer. The intermediate layer consists of 2 GRU layers, 2 attention layers, and 1 fully connected layer, and an attention mechanism is added to the intermediate layer.
[0015] Furthermore, the calculation formula of the GRU layer is:
[0016] Reset gate: r t =σ(W r ·x t +U r ·h t-1 +b r )
[0017] Where W r For input x t The reset gate weight matrix, U r is the previous hidden state h t-1 The reset gate weight matrix, b z is the reset gate bias term;
[0018] Update gate: z t =σ(W z ·x t +U z ·h t-1 +b z )
[0019] Where σ is the sigmoid activation function, W z For input x t The reset gate weight matrix, U z is the previous hidden state h t-1 The reset gate weight matrix, b z To update the gate bias term;
[0020] Candidate hidden states:
[0021] Where tanh is the hyperbolic tangent activation function, W h For input x tCandidate hidden state weight matrix, U h is the previous hidden state h t-1 With reset gate r t The candidate hidden state weight matrix of the element-wise product, b h is the candidate hidden state bias term, ⊙ represents the element product;
[0022] Final hidden state:
[0023] In the formula, h t is the final hidden state at time step t, which is the previous hidden state h determined by the update gate t-1 and the current candidate hidden state The weighted sum of .
[0024] Furthermore, the threshold is set according to the difference in battery types.
[0025] An electric vehicle power battery status warning system includes a construction module, a prediction module, and a warning module, wherein:
[0026] Construction module: used to construct a pressure difference prediction model with total voltage, maximum cell voltage, minimum cell voltage, battery SOC as input and pressure difference as output, and to construct a minimum cell voltage prediction model with total voltage, maximum cell voltage, pressure difference, battery SOC as input and minimum cell voltage as output;
[0027] Prediction module: used to output prediction results using the trained voltage difference prediction model and the minimum cell voltage prediction model;
[0028] The early warning module is used to compare the prediction results with the threshold and issue corresponding early warnings when the threshold is exceeded;
[0029] The trained voltage difference prediction model and the minimum cell voltage prediction model both adopt a GRU-based neural network model.
[0030] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.
[0031] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.
[0032] Compared with the prior art, the present invention has the following beneficial technical effects: by real-time monitoring of the battery status and using a neural network algorithm for data analysis, it is able to predict battery inconsistencies and undervoltage problems earlier, thereby providing early warning; the advantage of the present invention is that the warning time is advanced and the requirement for computing power is relatively low, making it suitable for real-time application in electric vehicles.
[0033] Both the voltage difference prediction model and the minimum cell voltage prediction model use a GRU-based neural network model. GRU can effectively capture and utilize long-term dependencies, solving the problem that traditional RNNs tend to forget remote information when processing long sequence data. At the same time, compared with the long short-term memory network (LSTM), the GRU structure is simpler and has fewer parameters, which reduces the amount of calculation and improves the speed of training and reasoning.
[0034] Preferably, the data set for model training and verification of the present invention comes from the collection of power battery related data during vehicle driving, and the data set is preprocessed to ensure the accuracy of model training.
[0035] Preferably, the alarm threshold is set according to different battery types, so that the present invention is applicable to early warning of different types of power batteries.
[0036] Preferably, the data set is standardized before model training and verification to eliminate the dimension effect.
[0037] Preferably, in practical applications, the GRU model is optimized and an attention mechanism is introduced to further improve the prediction effect. The attention mechanism can help the model better focus on important time steps and improve the accuracy of early warning of voltage faults. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 The figure is a flow chart of a battery abnormal status early warning method based on big data in an embodiment of the present invention.
[0039] Figure 2 Schematic diagram of the neural network model in an embodiment of the present invention.
[0040] Figure 3 The present invention provides a schematic diagram of the overall process of constructing a prediction model for a battery abnormal state early warning method based on big data in an embodiment of the present invention.
[0041] Figure 4 Schematic diagram of comparison between predicted values and actual values for the monomer inconsistency problem in an embodiment of the present invention.
[0042] Figure 5 Schematic diagram of comparison between predicted value and actual value for undervoltage problem in an embodiment of the present invention. DETAILED DESCRIPTION
[0043] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0044] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.
[0045] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0046] The present invention is further described in detail below in conjunction with the accompanying drawings:
[0047] like Figure 1 As shown, a method for early warning of the power battery status of an electric vehicle comprises the following steps:
[0048] Step 1: Take the total voltage, the highest cell voltage, the lowest cell voltage, and the battery SOC as inputs and the pressure difference as output to build a pressure difference prediction model; take the total voltage, the highest cell voltage, the pressure difference, and the battery SOC as inputs and the lowest cell voltage as output to build a lowest cell voltage prediction model;
[0049] Step 2: Output prediction results using the trained voltage difference prediction model and the minimum cell voltage prediction model;
[0050] Step three: compare the prediction results with the threshold and issue corresponding warnings when the threshold is exceeded.
[0051] Example
[0052] The present invention provides an electric vehicle power battery status early warning method, comprising the following steps:
[0053] S1, collecting relevant data of the power battery during vehicle driving and preprocessing the data;
[0054] S2, for the problem of battery cell inconsistency and undervoltage, the processed data is made into two sets of data sets, with the training set and the validation set divided into 8:2 respectively;
[0055] S3, respectively build GRU-based neural network models, use two types of training sets for model training, use validation sets as validation, and output the models whose validation results meet the standards;
[0056] S4, in the actual prediction scenario, obtain and process the battery data within the time period according to the method of S1, make a data set and implement the prediction in the model trained in S3;
[0057] S5, comparing the prediction result with the threshold value, and issuing corresponding warnings for situations exceeding the threshold value.
[0058] Specifically, data collection and processing and data set preparation and division are as follows: Figure 3 As shown;
[0059] S1 specifically includes the following steps:
[0060] S101, the test vehicle is a pure electric commercial vehicle equipped with a ternary lithium battery. The battery management system is used to collect battery data through various sensors. The specific collected indicators include total voltage, SOC, maximum single cell voltage, minimum single cell voltage, etc.
[0061] S102, after data collection, outliers beyond the specified range are deleted, and then missing data are processed. When there are continuous missing data segments, the missing data segments are directly deleted. When there are a small number of missing data segments, the missing values are interpolated by the weighted sliding average method; at the same time, the sliding average method is used to reduce the noise of each data.
[0062] In S101, the battery data collection frequency in this case is once every 30 seconds.
[0063] In S102 , the total voltage is normally within the range of 0V to 1000V, the SOC is between 0% and 100%, the highest and lowest values of the single cell voltage are within the range of 0V to 15V, and the data outside this range is discarded.
[0064] S2 specifically includes the following steps:
[0065] S201, create a 15-minute sliding window to traverse the battery data to produce a data set. To address the problem of battery cell inconsistency, data set A uses total voltage, maximum cell voltage, minimum cell voltage, and battery SOC as inputs, and uses differential pressure as output; to address the undervoltage problem, data set B uses total voltage, maximum cell voltage, differential pressure, and battery SOC as inputs, and uses differential pressure as output;
[0066] S202, dividing the data set A and the data set B into a training set and a validation set according to an 8:2 ratio;
[0067] In S201, since the battery collection frequency in this case is once every 30 seconds, a total of 30 sets of data points are collected as input in the 15-minute sliding window, and the relevant indicators at the next time point after 30 seconds are output. It should be noted that the 15-minute sliding window is only a specific case in this implementation case, and the time window size can be changed according to factors such as the application scenario in actual applications.
[0068] In S202, the data partitioning method is randomly selected.
[0069] like Figure 3 As shown, S3 specifically includes the following steps:
[0070] S301, inputting two training sets A and B into the model and performing training;
[0071] S302, using the corresponding validation set to calculate the prediction accuracy of the training model;
[0072] S303, retraining the models whose prediction accuracy does not meet the standard, and outputting the models whose accuracy meets the standard;
[0073] In S301, the model structure is as follows Figure 2 As shown in the figure, it consists of an input layer, an intermediate layer, and an output layer. The intermediate layer consists of 2 GRU layers, 2 attention layers, and 1 fully connected layer. After adding the attention mechanism, the model can better capture the key parts of the input data. The number of neurons in the GRU layer is set to 256, the model learning rate is set to 0.01, the optimizer is Adam, the number of iterations is 20, the prediction step is 1, the batch size is 32, and the sliding window size is 30.
[0074] In S303, the accuracy judgment standard is that the MSE error is less than 0.005, where the calculation formula of the MSE error is as follows:
[0075] MSE error:
[0076] where y i and are the true value and the predicted value respectively;
[0077] Step 4 specifically includes the following steps:
[0078] S401, according to the two situations of battery cell inconsistency and undervoltage, a data set is prepared according to the input specifications of the two models;
[0079] S402, standardizing the data set;
[0080] S403, inputting the standardized data set into the two models respectively and predicting the voltage difference and the minimum cell voltage after 30 seconds;
[0081] In S402, the data standardization method needs to be performed according to the standardization specifications of the initial training set.
[0082] S5 specifically includes the following steps:
[0083] S501, comparing the voltage difference with the lowest voltage of the monomer and the set threshold, and issuing an early warning if the result exceeds the threshold;
[0084] S502, as new data is generated, the data set is re-created and predicted to achieve continuous abnormal warning;
[0085] In S501, according to the vehicle and battery characteristics, the voltage difference threshold is selected as 0.3V, and the minimum voltage threshold is selected as 2.9V;
[0086] Application Verification:
[0087] In this embodiment, the collected data comes from a new energy vehicle monitoring platform. The prediction results for the voltage difference (battery cell inconsistency problem) and the minimum cell voltage (undervoltage problem) are as follows: Figure 4 and Figure 5 As shown, Figure 4 The figure shows the comparison between the pressure difference prediction result and the true value, where the solid line represents the predicted data and the dotted line represents the actual data. The overall MSE error is 0.0014. It can be seen from the figure that the two curves are very close at most time points. In the section where the pressure difference decreases, the predicted value is basically consistent with the true value. When the pressure difference rises sharply, the model prediction result has only a small deviation. Figure 5 The comparison between the predicted result of the lowest cell voltage and the true value shows that the overall MSE error is 0.0032. In most time periods, the predicted value can follow the trend of the true value and the prediction result is relatively accurate.
[0088] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, 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 disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0089] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), 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 the 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.
[0090] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims of the invention.
Claims
1. A method for early warning of the power battery status of an electric vehicle, characterized in that: The following steps are involved: A pressure difference prediction model is constructed with total voltage, maximum cell voltage, minimum cell voltage, and battery SOC as inputs and pressure difference as output. A minimum cell voltage prediction model is constructed with total voltage, maximum cell voltage, pressure difference, and battery SOC as inputs and minimum cell voltage as output. Output prediction results using the trained voltage difference prediction model and the minimum cell voltage prediction model; Compare the prediction results with the threshold and issue corresponding warnings when the threshold is exceeded; The trained voltage difference prediction model and the minimum cell voltage prediction model both adopt a GRU-based neural network model.
2. The electric vehicle power battery status early warning method according to claim 1, characterized in that: The data sets for training and verifying the pressure difference prediction model and the minimum cell voltage prediction model come from the collection of power battery related data during vehicle driving.
3. The electric vehicle power battery status early warning method according to claim 2, characterized in that: The data set is preprocessed, and the specific process is to remove missing values and outliers and use the weighted moving average method to interpolate the missing values while achieving noise reduction for the overall data.
4. The electric vehicle power battery status early warning method according to claim 3 is characterized in that: The abnormal value is the data of the collected point that deviates from the actual situation in the collected data.
5. The electric vehicle power battery status early warning method according to claim 1, characterized in that: The GRU-based neural network model consists of an input layer, an intermediate layer, and an output layer. The intermediate layer consists of 2 GRU layers, 2 attention layers, and 1 fully connected layer, and an attention mechanism is added to the intermediate layer.
6. The electric vehicle power battery status early warning method according to claim 5, characterized in that: The calculation formula of the GRU layer is: Reset gate: r t =σ(W r ·x t +U r ·h t-1 +b r ) Where W r For input x t The reset gate weight matrix, U r is the previous hidden state h t-1 The reset gate weight matrix, b z is the reset gate bias term; Update gate: z t =σ(W z ·x t +U z ·h t-1 +b z ) Where σ is the sigmoid activation function, W z For input x t The reset gate weight matrix, U z is the previous hidden state h t-1 The reset gate weight matrix, b z To update the gate bias term; Candidate hidden states: Where tanh is the hyperbolic tangent activation function, W h For input x t Candidate hidden state weight matrix, U h is the previous hidden state h t-1 With reset gate r t The candidate hidden state weight matrix of the element-wise product, b h is the candidate hidden state bias term, ⊙ represents the element product; Final hidden state: In the formula, h t is the final hidden state at time step t, which is the previous hidden state h determined by the update gate t-1 and the current candidate hidden state The weighted sum of .
7. The electric vehicle power battery status early warning method according to claim 1, characterized in that: The threshold is set according to the difference in battery types.
8. An electric vehicle power battery status warning system, characterized in that: It includes construction module, prediction module and early warning module, among which: Construction module: used to construct a pressure difference prediction model with total voltage, maximum cell voltage, minimum cell voltage, battery SOC as input and pressure difference as output, and to construct a minimum cell voltage prediction model with total voltage, maximum cell voltage, pressure difference, battery SOC as input and minimum cell voltage as output; Prediction module: used to output prediction results using the trained voltage difference prediction model and the minimum cell voltage prediction model; The early warning module is used to compare the prediction results with the threshold and issue corresponding early warnings when the threshold is exceeded; The trained voltage difference prediction model and the minimum cell voltage prediction model both adopt a GRU-based neural network model.
9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
Citation Information
Cited By
Vehicle lithium battery fault early warning method based on multivariate prediction and anomaly detection
CN119749253A
A fault warning method for automotive lithium batteries based on multivariate prediction and anomaly detection
CN119749253B
