Thermal runaway risk assessment method, device and storage medium

Abstract

The present application relates to a method, device, and storage medium for assessing thermal runaway risk, and relates to the field of vehicle technology. The method at least addresses the technical problem of providing a warning to a battery before thermal runaway occurs in the related art. The method comprises: obtaining time series status data of a battery, wherein the time series status data is used to indicate changes in battery parameters of the battery at different time points; inputting the time series status data into a trained thermal runaway warning model to obtain a risk assessment result. The thermal runaway warning model is used to detect whether the battery has a thermal runaway risk. The thermal runaway warning model comprises: a long short-term memory network module and a soft decision tree module. The long short-term memory network module is used to extract feature data related to thermal runaway risk, and the soft decision tree module is used to determine whether the battery has experienced thermal runaway.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, in particular to the field of battery technology, and specifically to a method, device, and storage medium for assessing thermal runaway risk. Background Art

[0002] With the rapid development of vehicle technology, people have increasingly stringent requirements for vehicle safety. Batteries, as a vehicle's power source, can experience thermal runaway if the internal temperature of a single battery cell rises abnormally. This can spread to adjacent cells through radiation or conductive materials, causing the battery separator to melt and the electrolyte to react violently, releasing large amounts of flammable gases that can lead to fire or even explosion.

[0003] Therefore, it is necessary to monitor the battery and issue an early warning before thermal runaway occurs to ensure vehicle safety. However, how to issue an early warning before thermal runaway occurs has become a technical problem that needs to be solved urgently. Summary of the Invention

[0004] This application provides a method, device, and storage medium for assessing thermal runaway risk to at least address the technical problem of providing early warning of a battery before thermal runaway occurs. The technical solution of this application is as follows:

[0005] According to a first aspect of the present application, a method for assessing thermal runaway risk is provided, comprising: obtaining time series status data of a battery, the time series status data being used to indicate changes in battery parameters at different time points; inputting the time series status data into a trained thermal runaway warning model to obtain a risk assessment result, the thermal runaway warning model being used to detect whether the battery is at risk of thermal runaway, the thermal runaway warning model comprising a long short-term memory network module and a soft decision tree module, the long short-term memory network module being used to extract feature data related to thermal runaway risk, and the soft decision tree module being used to determine whether the battery is experiencing thermal runaway.

[0006] In one possible implementation, inputting time series state data into a trained thermal runaway warning model to obtain a risk assessment result includes: inputting the time series state data into a long short-term memory network module to obtain a risk feature vector for the battery, where the risk feature vector indicates characteristic data related to thermal runaway risk; and inputting the risk feature vector into a soft decision tree module to obtain a risk assessment result.

[0007] In one possible implementation, the soft decision tree module includes multiple preset assessment paths. A risk feature vector is input into the soft decision tree module to obtain a risk assessment result, including: for each preset assessment path, obtaining initial probability assessment information corresponding to the preset assessment path based on the risk feature vector and the preset assessment path, thereby obtaining initial probability assessment information corresponding to the multiple preset assessment paths. The initial probability assessment information indicates the probability of thermal runaway risk in the battery, with one initial probability assessment information corresponding to each preset assessment path. A risk assessment result is obtained based on the multiple initial probability assessment information.

[0008] In one possible implementation, a preset assessment path includes multiple assessment factors and multiple probability weights. The assessment factors are used to indicate conditions under which a battery may experience thermal runaway risk, and the probability weights are used to indicate the degree of influence of the assessment factors on the battery's thermal runaway risk. One assessment factor corresponds to multiple probability weights. Obtaining initial probability assessment information corresponding to the preset assessment path based on the risk characteristic vector and the preset assessment path includes: obtaining initial probability assessment information corresponding to the preset assessment path based on the risk characteristic vector, the multiple assessment factors, and the multiple probability weights.

[0009] In one possible implementation, the long short-term memory network module includes a fully connected layer, which is used to input the risk feature vector into the soft decision tree module. The dimension of the fully connected layer of the long short-term memory network module is the same as the dimension of the input layer of the soft decision tree module.

[0010] In one possible implementation, a trained thermal runaway warning model is obtained by obtaining a training set, a first model loss, a second model loss, and an initial thermal runaway warning model. The first model loss is used to adjust the accuracy of the soft decision tree module, and the second model loss is used to adjust the accuracy of the long short-term memory network module. The initial thermal runaway warning model is trained based on the training set, the first model loss, and the second model loss to obtain a trained thermal runaway warning model.

[0011] In one possible embodiment, the initial thermal runaway warning model is a pruned thermal runaway warning model. The initial thermal runaway warning model is trained based on a training set, a first model loss, and a second model loss to obtain a trained thermal runaway warning model, including: training the initial thermal runaway warning model based on the training set, the first model loss, and the second model loss to obtain first model parameters. The training set and the first model parameters are sent to a server, which is deployed with an unpruned thermal runaway warning model. The server is configured to tune the first model parameters based on the training set and the unpruned thermal runaway warning model. The second model parameters are received. The second model parameters are the tuned first model parameters. Based on the second model parameters, a trained thermal runaway warning model is obtained.

[0012] In one possible embodiment, the method further includes: obtaining an initial soft decision forest model, the initial soft decision forest model including multiple initial evaluation paths, the initial evaluation paths including branch structures; determining similarities of the multiple branch structures based on the multiple branch structures; determining multiple groups of similar structure groups based on the similarities of the multiple branch structures, each similar structure group including at least two branch structures having similarities less than a preset similarity threshold; and fusing branch structure parameters in each similar structure group to obtain the soft decision forest model.

[0013] In one possible embodiment, the branch structure includes at least one of the following: a branch node and a decision edge, the branch node is used to indicate an evaluation factor, the decision edge is used to indicate a probability weight, the evaluation factor is used to indicate the conditions under which the battery has a thermal runaway risk, and the probability weight is used to indicate the degree of influence of the evaluation factor on the thermal runaway risk of the battery, and one evaluation factor corresponds to multiple probability weights.

[0014] According to a second aspect of the present application, a device for assessing thermal runaway risk is provided, the device including an acquisition module and a processing module.

[0015] The acquisition module is used to obtain the battery's time series status data, which indicates changes in the battery's parameters at different time points. The processing module is used to input the time series status data into the trained thermal runaway warning model to obtain a risk assessment result. The thermal runaway warning model is used to detect whether the battery has a thermal runaway risk. The thermal runaway warning model includes a long-short-term memory network module and a soft decision tree module. The long-short-term memory network module is used to extract feature data related to thermal runaway risk, and the soft decision tree module is used to determine whether the battery has experienced thermal runaway.

[0016] In one possible implementation, the processing module is configured to input the time series state data into a long short-term memory network module to obtain a battery risk feature vector, where the risk feature vector indicates characteristic data related to thermal runaway risk. The processing module is further configured to input the risk feature vector into a soft decision tree module to obtain a risk assessment result.

[0017] In one possible implementation, the soft decision tree module includes: multiple preset evaluation paths. A processing module is configured to, for each preset evaluation path, obtain initial probability assessment information corresponding to the preset evaluation path based on the risk feature vector and the preset evaluation path, thereby obtaining initial probability assessment information corresponding to the multiple preset evaluation paths. The initial probability assessment information indicates the probability of thermal runaway risk in the battery, with each preset evaluation path corresponding to one piece of initial probability assessment information. The processing module is further configured to obtain a risk assessment result based on the multiple pieces of initial probability assessment information.

[0018] In one possible implementation, a preset assessment path includes multiple assessment factors and multiple probability weights. The assessment factors are used to indicate the conditions under which a battery may experience thermal runaway risk, and the probability weights are used to indicate the degree of influence of the assessment factors on the battery's thermal runaway risk. One assessment factor corresponds to multiple probability weights. A processing module is configured to obtain initial probability assessment information corresponding to the preset assessment path based on the risk characteristic vector, the multiple assessment factors, and the multiple probability weights.

[0019] In one possible implementation, the long short-term memory network module includes a fully connected layer, which is used to input the risk feature vector into the soft decision tree module. The dimension of the fully connected layer of the long short-term memory network module is the same as the dimension of the input layer of the soft decision tree module.

[0020] In one possible implementation, the acquisition module is configured to acquire a training set, a first model loss, a second model loss, and an initial thermal runaway warning model. The first model loss is used to adjust the accuracy of the soft decision tree module, and the second model loss is used to adjust the accuracy of the long short-term memory network module. The processing module is configured to train the initial thermal runaway warning model based on the training set, the first model loss, and the second model loss to obtain a trained thermal runaway warning model.

[0021] In one possible embodiment, the initial thermal runaway warning model is a pruned thermal runaway warning model. The device also includes a sending module and a receiving module. The processing module is used to train the initial thermal runaway warning model based on the training set, the first model loss, and the second model loss to obtain the first model parameters. The sending module is used to send the training set and the first model parameters to the server, which is deployed with the unpruned thermal runaway warning model. The server is used to tune the first model parameters based on the training set and the unpruned thermal runaway warning model. The receiving module is used to receive the second model parameters. The second model parameters are the tuned first model parameters. The processing module is also used to obtain the trained thermal runaway warning model based on the second model parameters.

[0022] In one possible implementation, an acquisition module is configured to acquire an initial soft decision forest model, the initial soft decision forest model including multiple initial evaluation paths, each of which includes a branch structure. A processing module is configured to determine similarities among the multiple branch structures based on the multiple branch structures. The processing module is further configured to determine multiple groups of similar structure groups based on the similarities among the multiple branch structures, each similar structure group including at least two branch structures having similarities less than a preset similarity threshold. The processing module is further configured to fuse the branch structure parameters within each similar structure group to obtain the soft decision forest model.

[0023] In one possible embodiment, the branch structure includes at least one of the following: a branch node and a decision edge, the branch node is used to indicate an evaluation factor, the decision edge is used to indicate a probability weight, the evaluation factor is used to indicate the conditions under which the battery has a thermal runaway risk, and the probability weight is used to indicate the degree of influence of the evaluation factor on the thermal runaway risk of the battery, and one evaluation factor corresponds to multiple probability weights.

[0024] According to the third aspect provided by the present application, a device for assessing thermal runaway risk is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute instructions to implement a method as in the first aspect and any possible implementation method thereof.

[0025] According to the fourth aspect provided by the present application, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are executed by the processor of the thermal runaway risk assessment device, the thermal runaway risk assessment device is able to execute the method of the first aspect and any possible implementation method thereof.

[0026] Beneficial effects of the present invention:

[0027] (1) By obtaining the time series status data of the battery and inputting the time series status data into the trained thermal runaway warning model, the risk assessment results can be obtained. The thermal runaway warning model includes: a long short-term memory network module and a soft decision tree module. Since the long short-term memory network module can capture the long-term trends and periodic changes in the time series status data, it can better determine the characteristic data related to the thermal runaway risk in the time series status data. The soft decision tree module can smooth the decision boundary and reduce the sensitivity to noise, thereby more accurately determining the probability of the battery thermal runaway risk, and then more accurately determining whether the battery has thermal runaway. In addition, since the data processing process of the soft decision tree module can perform interpretable hierarchical reasoning, the interpretability of the thermal runaway warning model can be improved.

[0028] (2) By combining the long short-term memory network model with the soft decision tree, it is possible to capture higher-level abstract features while balancing the performance and interpretability of the thermal runaway warning model, making the risk assessment results more accurate and interpretable.

[0029] (3) By obtaining the initial probability assessment information for each preset assessment path, the probability results of multiple preset assessment paths of the soft decision tree model can be obtained. In other words, the probability distribution of the battery's thermal runaway risk can be obtained. In this way, the risk assessment results can be obtained more accurately based on the probability distribution of the battery's thermal runaway risk.

[0030] (4) Since the preset evaluation path of the soft decision tree model includes multiple evaluation factors and multiple probability weights, each preset evaluation path corresponds to a clear feature combination logic. In this way, the meaning of each preset evaluation path is clear and interpretable, and the cause of the battery thermal runaway risk can be determined when the risk assessment result indicates that the battery has thermal runaway.

[0031] (5) The long short-term memory network module and the soft decision tree module can be directly connected and trained as a whole, thereby simplifying the model architecture of the thermal runaway warning model and improving the computational efficiency of the thermal runaway warning model.

[0032] (6) By obtaining the training set, the first model loss, the second model loss and the initial thermal runaway warning model, and training the initial thermal runaway warning model based on the training set, the first model loss and the second model loss, the trained thermal runaway warning model is obtained. The first model loss is used to adjust the accuracy of the soft decision tree module, and the second model loss is used to adjust the accuracy of the long short-term memory network module. In this way, the training direction of the thermal runaway warning model can be effectively guided, the thermal runaway warning model can be prevented from overfitting on the training data, and the generalization ability of the thermal runaway warning model can be improved, so that the thermal runaway warning model can have better robustness when facing different types of input data, and the performance of the thermal runaway warning model will not be degraded due to excessive use of some paths.

[0033] (7) By deploying a pruned thermal runaway warning model on a thermal runaway risk assessment device, the thermal runaway risk assessment device can quickly predict whether a battery is experiencing thermal runaway. By deploying an unpruned thermal runaway warning model on a server and optimizing the first model parameters on the server, the prediction speed of the thermal runaway warning model can be increased while making the prediction results of the thermal runaway warning model more accurate.

[0034] (8) By fusing the branch structures in each group of similar structure groups to obtain multiple fused branch structures, the thermal runaway warning model can be pruned. Moreover, since a group of similar structure groups has at least two branch structures with similarities less than a preset similarity threshold, the branch structure includes branch nodes and decision edges. This allows lightweight processing of the branch nodes and decision edges while maintaining the accuracy of the thermal runaway warning model.

[0035] It should be noted that the technical effects brought about by any implementation method in the second to fourth aspects can refer to the technical effects brought about by the corresponding implementation method in the first aspect, and will not be repeated here.

[0036] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute an improper limitation on the present application.

[0038] Figure 1 is a schematic diagram of a thermal runaway risk assessment system according to an exemplary embodiment;

[0039] Figure 2 is a flow chart illustrating a method for assessing thermal runaway risk according to an exemplary embodiment;

[0040] Figure 3 is a flow chart illustrating another method for assessing thermal runaway risk according to an exemplary embodiment;

[0041] Figure 4 is a schematic diagram showing an example of a system between a vehicle terminal and a server according to an exemplary embodiment;

[0042] Figure 5 is a schematic diagram illustrating an example of a method for evaluating thermal runaway risk according to an exemplary embodiment;

[0043] Figure 6 is a schematic structural diagram of a device for assessing thermal runaway risk according to an exemplary embodiment;

[0044] Figure 7 It is a structural schematic diagram of another device for evaluating thermal runaway risk according to an exemplary embodiment. DETAILED DESCRIPTION

[0045] In order to enable ordinary people in the art to better understand the technical solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0046] It should be noted that the terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0047] Before introducing the thermal runaway risk assessment method according to the embodiment of the present application in detail, the implementation environment and application scenarios of the embodiment of the present application are first introduced.

[0048] An embodiment of the present application provides a method for assessing thermal runaway risk, comprising: obtaining time series status data of a battery and inputting the time series status data into a trained thermal runaway warning model to obtain a risk assessment result. The thermal runaway warning model comprises: a long short-term memory network module and a soft decision tree module. Since the long short-term memory network module can capture long-term trends and periodic changes in the time series status data, it can better determine the characteristic data related to the thermal runaway risk in the time series status data. The soft decision tree module can smooth the decision boundary and reduce the sensitivity to noise, thereby more accurately determining the probability of the battery having a thermal runaway risk, and further more accurately determining whether the battery has a thermal runaway. Moreover, since the data processing process of the soft decision tree module can perform interpretable hierarchical reasoning, the interpretability of the thermal runaway warning model can be improved.

[0049] It should be noted that the thermal runaway risk assessment method provided in this application can be performed by a thermal runaway risk assessment device, which can be a battery controller or a vehicle. Furthermore, the device can also be the vehicle's central processing unit (CPU), a module within the device for assessing thermal runaway risk, or a vehicle-mounted device within the vehicle, though this application does not impose any limitations on this. In the embodiments of this application, the thermal runaway risk assessment method provided in this application is illustrated using the vehicle as an example.

[0050] It should be noted that the vehicle may be, but is not limited to, a pure electric vehicle (Pure Electric Vehicle / Battery Electric Vehicle, PEV / BEV), a hybrid electric vehicle (Hybrid Electric Vehicle, HEV), a range extended electric vehicle (REEV), a plug-in hybrid electric vehicle (PHEV), etc.

[0051] The implementation environment of the embodiments of the present application is introduced below.

[0052] Figure 1 FIG. 1 is a schematic diagram of a thermal runaway risk assessment system according to an exemplary embodiment. Figure 1 As shown in the figure, the thermal runaway risk assessment system includes: a time series data acquisition and preprocessing module, a model building module, a model joint training and parameter optimization module, and a model lightweight deployment and alarm module.

[0053] The time series data acquisition and preprocessing module includes a data acquisition module and a data preprocessing module. The data acquisition module can obtain initial battery status data, and the data preprocessing module is used to preprocess the initial battery status data.

[0054] Battery status data for a preset period of time is extracted from the initial battery status data. The battery status data for the preset period of time can then be cleaned to obtain first status data. The first status data can then be normalized to obtain second status data. The second status data can then be processed using a sliding time window algorithm to obtain a training set.

[0055] Optionally, the data acquisition module includes a voltage sensor, a current sensor, a temperature sensor, and a state of charge (SOC) and state of health (SOH) estimation unit.

[0056] It should be noted that the signal directly collected by the sensor will pass through the conditioning circuit to amplify the sensor input signal, remove noise within a certain frequency range through the filter, and indirectly calculate the SOC and SOH.

[0057] It should be noted that this application does not limit the data preprocessing method. For example, the data preprocessing method may include: data cleaning, data normalization, and sliding time window algorithm.

[0058] The model building module is used to construct a long short-term memory network model and a soft decision tree model, and based on the long short-term memory network model and the soft decision tree model, an initial thermal runaway warning model is obtained.

[0059] The joint model training and parameter optimization module includes a model training module and a parameter optimization module. The model training module is used to train the initial thermal runaway warning model. The parameter optimization module is used to optimize the parameters of the initial thermal runaway warning model.

[0060] The model lightweight deployment and alarm module includes: model lightweight deployment module and alarm module.

[0061] Optionally, the model lightweight deployment module is used to prune the long-short-term memory network model and the soft decision tree model to obtain a pruned thermal runaway warning model. Furthermore, the model lightweight deployment module is used to deploy the pruned thermal runaway warning model to the vehicle terminal embedded system. Furthermore, the model lightweight deployment module is used to send vehicle-collected data and training parameters to the cloud and obtain global model parameters from the cloud.

[0062] The alarm module is used to determine the target alarm level and issue an alarm based on the target alarm level when the risk assessment results trigger the alarm condition.

[0063] Optionally, when the target alarm level is the first alarm level, an alarm message can be issued through the instrument and an alarm signal can be sent to the cloud platform monitoring center. When the target alarm level is the second alarm level, an alarm message can be issued through the instrument, the power battery high voltage can be disconnected, and an alarm signal can be sent to the cloud platform monitoring center.

[0064] For ease of understanding, the thermal runaway risk assessment method provided in this application is described in detail below with reference to the accompanying drawings.

[0065] like Figure 2 As shown, the method includes the following steps:

[0066] S201: Acquire time series status data of a battery.

[0067] The time series status data is used to indicate changes in battery parameters of the battery at different time points.

[0068] Optionally, the time series status data includes: voltage change data within the target time period, current change data within the target time period, temperature change data within the target time period, charge state change data within the target time period and health state change data within the target time period, and the target time period includes: multiple time points.

[0069] Specifically, the battery includes: multiple single cells; voltage change data includes at least one of the following: the maximum voltage among the multiple single cell voltages; the minimum voltage among the multiple single cell voltages; and the total battery voltage; current change data includes at least one of the following: the highest temperature among the multiple single cell temperatures; the lowest temperature among the multiple single cell temperatures; and the temperature gradient; state of charge change data includes at least one of the following: actual SOC and displayed SOC; and health status change data includes SOH. The total battery voltage is the sum of the voltages of all single cells in the battery when connected in series, parallel, or in a series-parallel combination; and the temperature gradient is used to indicate the temperature difference between different locations within the battery.

[0070] S202: Input the time series state data into the trained thermal runaway warning model to obtain a risk assessment result.

[0071] Among them, the thermal runaway warning model is used to detect whether the battery has a thermal runaway risk. The thermal runaway warning model includes: a long short-term memory network (LSTM) module and a soft decision tree (SDT) module. The long short-term memory network module is used to extract feature data related to the thermal runaway risk, and the soft decision tree module is used to determine whether the battery has a thermal runaway risk.

[0072] In one possible implementation, time series state data can be input into a long-short-term memory network module to generate a battery risk feature vector. This risk feature vector indicates characteristic data related to thermal runaway risk. This risk feature vector can then be input into a soft decision tree module to generate a risk assessment result.

[0073] It can be understood that by combining the long short-term memory network model with the soft decision tree, it is possible to balance the performance and interpretability of the thermal runaway warning model while capturing higher-level abstract features, making the risk assessment results more accurate and interpretable.

[0074] In this embodiment of the present application, the LSTM module includes an input layer, a hidden layer (i.e., the LSTM layer), and a fully connected layer. The fully connected layer is used to input the risk feature vector into the soft decision tree module. The fully connected layer of the LSTM module has the same dimensions as the input layer of the soft decision tree module.

[0075] It should be understood that the long short-term memory network module is obtained based on the long short-term memory network model, and the soft decision tree module is obtained based on the soft decision tree model.

[0076] In this way, the long short-term memory network module and the soft decision tree module can be directly connected and trained as a whole, thereby simplifying the model architecture of the thermal runaway warning model and improving the computational efficiency of the thermal runaway warning model.

[0077] In one possible design, the input layer transforms the time series state data to obtain target sequence state data in a format that can be processed by the long short-term memory network module. The hidden layer then processes the target sequence state data to obtain a hidden state feature vector. The fully connected layer then maps the hidden state feature vector to obtain the battery risk feature vector.

[0078] In an embodiment of the present application, a hidden state feature vector includes: multiple hidden state sub-vectors, each of which includes at least one of the following: a voltage feature vector, a current feature vector, a temperature feature vector, a SOC feature vector, and a SOH feature vector. The voltage feature vector includes at least one of the following: a voltage mean vector, a voltage variance vector, a voltage extreme difference vector, a voltage rate of change vector, a voltage maximum value vector, and a voltage minimum value vector. The current feature vector includes at least one of the following: a current mean vector, a current variance vector, a current extreme difference vector, a current rate of change vector, a current maximum value vector, and a current minimum value vector. The temperature feature vector includes at least one of the following: a temperature mean vector, a temperature variance vector, a temperature extreme difference vector, a temperature rate of change vector, a temperature maximum value vector, and a temperature minimum value vector.

[0079] It should be noted that the types of the multiple hidden state sub-vectors are different.

[0080] Optionally, the hidden layer includes: a forget gate, an input gate, and an output gate. The forget gate is used to retain or discard the hidden state subvector (i.e., the cell state) at the previous moment of the current moment. The input gate is used to determine the stored hidden state subvector. The output gate is used to determine the output hidden state subvector.

[0081] Specifically, the target sequence state data can be processed using a forget gate to obtain output data of the forget gate. The target sequence state data can be processed using an input gate to obtain a candidate state vector to be updated. The output data and the candidate state vector to be updated can then be processed using the forget gate and the input gate to obtain an updated state vector. The updated state vector and the target sequence state data can then be processed using an output gate to obtain a hidden state feature vector.

[0082] Exemplarily, the calculation formula of the forget gate satisfies Formula 1.

[0083] Formula 1.

[0084] Among them, f t is the output data of the forget gate, σ represents the sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 is the hidden state subvector of the previous moment (i.e. the output of the neuron), U f is the weight matrix of the forget gate, x t is the time series state data, b f is the bias term of the forget gate.

[0085] It should be noted that f t is a constant greater than or equal to 0 and less than or equal to 1.

[0086] Exemplarily, the input gate calculation formula satisfies Formula 2 and Formula 3.

[0087] Formula 2.

[0088] Formula 3.

[0089] Among them, i t is the update parameter of the target sequence state data, W i is the weight matrix of the input gating signal, b i is the bias term of the input gate, is the candidate state vector to be updated (i.e., the candidate cell state to be updated), tanh() is the hyperbolic tangent function, W c is the weight matrix of the state candidate vector, b c is the bias term of the state candidate vector.

[0090] Optionally, the update parameter is used to control whether the target sequence state data is updated.

[0091] It should be noted that if the update parameter is 1, it is determined that the control target sequence state data is updated; if the update parameter is 0, it is determined that the control target sequence state data is not updated.

[0092] The calculation formula of the exemplary updated state candidate vector satisfies Formula 4.

[0093] Formula 4.

[0094] Among them, C t is the updated state vector (i.e., the updated candidate cell state), C t-1 is the state vector at the previous moment.

[0095] Exemplarily, the calculation formula of the output gate satisfies Formula 5 and Formula 6.

[0096] Formula 5.

[0097] Formula 6.

[0098] Among them, t is the output gate signal, W o is the weight matrix of the output gate, b o is the bias term of the output gate, h t is the hidden state sub-vector (i.e., the battery feature output value at the current moment).

[0099] It should be noted that o t is an output value greater than or equal to 0 and less than or equal to 1.

[0100] It should be noted that the hidden state feature vector is a comprehensive representation of the current input and all previous inputs, and is the feature vector output by the hidden layer of the long short-term memory network.

[0101] Exemplarily, the hidden state feature vector satisfies Formula 7.

[0102] Formula 7.

[0103] Among them, H is the hidden state feature vector, h1 is the hidden state subvector at the first moment, h2 is the hidden state subvector at the second moment, and h t is the hidden state subvector at time t.

[0104] Optionally, the long short-term memory network module further includes: an attention mechanism layer.

[0105] In another possible design, the hidden state feature vector can be processed through the attention mechanism layer to obtain an attention vector set. Subsequently, the attention vector set can be mapped through a fully connected layer to obtain the battery risk feature vector.

[0106] It should be noted that the risk feature vector of the battery is a feature vector with the same dimension as the input layer of the soft decision tree module, that is, a low-dimensional feature vector.

[0107] In an embodiment of the present application, the attention vector set includes: multiple attention sub-vectors.

[0108] Optionally, an attention weight parameter of the hidden state feature vector can be determined by an attention mechanism layer. Afterwards, for each of the multiple hidden state sub-vectors, each hidden state sub-vector can be weightedly aggregated based on the attention weight parameter to obtain an attention sub-vector, thereby obtaining an attention vector set.

[0109] Exemplarily, the attention vector satisfies Formula 8.

[0110] Formula 8.

[0111] in, is any one of the multiple attention sub-vectors, is the attention weight parameter, h i is any hidden state sub-vector among multiple hidden state sub-vectors.

[0112] It should be noted that attention vector sets can be obtained based on different types of time series data.

[0113] Exemplarily, the attention vector set satisfies Formula 9.

[0114] Formula nine.

[0115] in, is the attention vector set, is the attention sub-vector at the first moment, is the attention sub-vector at the second moment, is the attention sub-vector at the t-th moment.

[0116] Exemplarily, the attention weight parameter satisfies Formula 10 and Formula 11.

[0117] Formula 10.

[0118] Formula 11.

[0119] Among them, e i is the similarity score.

[0120] Exemplarily, the similarity score satisfies Formula 12.

[0121] Formula twelve.

[0122] Among them, W h is the weight matrix of similarity score, b h is the bias term of the similarity score.

[0123] It should be noted that a soft decision tree is a deep neural network with a logical binary tree structure. The soft decision tree module consists of multiple differentiable soft decision trees, each of which can implement probabilistic path selection using a sigmoid activation function. This application does not limit the number of soft decision trees in the soft decision tree module. For example, the number of soft decision trees can be 5, 10, 50, 100, or 101.

[0124] It should be understood that the number of soft decision trees in the soft decision tree module needs to balance the accuracy of the model and the computational cost.

[0125] When the number of soft decision trees is 100 (a moderate number), it can ensure a certain degree of diversity without being too time-consuming. The maximum depth of a single tree is set to 10, which can balance the complexity and generalization ability of the model.

[0126] In an embodiment of the present application, the soft decision tree module includes: multiple preset evaluation paths and multiple initial evaluation paths, one preset evaluation path corresponds to one initial evaluation path, the preset evaluation path is used to indicate the probability distribution of the risk of thermal runaway of the battery, and the initial evaluation path is used to indicate the static hierarchical structure of the soft decision tree module.

[0127] In another possible design, for each preset assessment path, initial probability assessment information corresponding to the preset assessment path can be obtained based on the risk characteristic vector and the preset assessment path. Initial probability assessment information corresponding to multiple preset assessment paths can be obtained. The initial probability assessment information indicates the probability of thermal runaway risk in the battery, with one initial probability assessment information corresponding to each preset assessment path. Subsequently, a risk assessment result can be obtained based on the multiple initial probability assessment information.

[0128] The preset evaluation path includes multiple evaluation factors and multiple probability weights, and the initial evaluation path includes multiple branch nodes and multiple decision edges. The evaluation factors are used to indicate the conditions under which the battery may experience thermal runaway risk, and the probability weights are used to indicate the degree of impact of the evaluation factors on the battery's thermal runaway risk. The branch nodes indicate the evaluation factors, and the decision edges indicate the probability weights. One evaluation factor corresponds to multiple probability weights, one evaluation factor corresponds to one branch node, and one probability weight corresponds to one decision edge.

[0129] Optionally, initial probability assessment information corresponding to a preset assessment path may be obtained based on the risk characteristic vector, multiple assessment factors, and multiple probability weights.

[0130] It should be noted that this application does not limit the method for obtaining the probability weights. For example, the probability weights corresponding to each evaluation factor can be pre-set. For another example, the probability weights corresponding to each evaluation factor can be obtained based on the weight generator of the branch node.

[0131] It is understandable that because the preset evaluation paths of the soft decision tree model include multiple evaluation factors and multiple probability weights, each preset evaluation path corresponds to a clear feature combination logic. This makes the meaning of each preset evaluation path clear and interpretable, and when the risk assessment results indicate that the battery has thermal runaway, the cause of the thermal runaway risk can be determined.

[0132] Optionally, the risk assessment result is used to indicate whether the battery is experiencing thermal runaway.

[0133] Optionally, for each preset assessment path, the risk feature vector can be input into a subclassifier at the branch node, and multiple assessment probabilities can be obtained using the subclassifier and assessment factors. The risk feature vector can be input into a weight generator at the branch node to obtain multiple probability weights corresponding to each assessment factor, with one assessment probability corresponding to one probability weight. Subsequently, based on the multiple assessment probabilities and multiple probability weights, initial probability assessment information corresponding to the preset assessment path can be obtained to obtain initial probability assessment information corresponding to multiple preset assessment paths. Subsequently, based on the multiple initial probability assessment information, the initial probability assessment information with the highest probability of thermal runaway risk can be determined as the risk assessment result.

[0134] Alternatively, if the probability of the existence of thermal runaway risk in multiple initial probability assessment information is greater than the initial probability assessment information of a preset probability threshold, the risk assessment result can be determined as the battery has thermal runaway; if the probability of the absence of thermal runaway risk in multiple initial probability assessment information is greater than the initial probability assessment information of a preset probability threshold, the risk assessment result can be determined as the battery has not experienced thermal runaway.

[0135] By obtaining the initial probability assessment information for each preset assessment path, we can obtain the probability results for multiple preset assessment paths in the soft decision tree model. In other words, we can obtain the probability distribution of the battery's thermal runaway risk. This allows for more accurate risk assessment results based on the probability distribution of the battery's thermal runaway risk.

[0136] It's understandable that the LSTM model captures the temporal dependencies of battery parameters, the temporal attention mechanism focuses on key time steps (such as temperature or voltage surges), and the fully connected layer adjusts feature dimensions to adapt to the input of the soft decision tree model. The soft decision tree model adjusts the contribution of each tree using weight parameters, resulting in a weighted integrated prediction output. This generates an interpretable hierarchical decision result based on the power battery feature vector output by the LSTM network.

[0137] Based on the above technical solution, a risk assessment result can be obtained by obtaining the time series status data of the battery and inputting the time series status data into the trained thermal runaway warning model. The thermal runaway warning model includes: a long short-term memory network module and a soft decision tree module. Since the long short-term memory network module can capture the long-term trends and periodic changes in the time series status data, it can better determine the characteristic data related to the thermal runaway risk in the time series status data. The soft decision tree module can smooth the decision boundary and reduce the sensitivity to noise, thereby more accurately determining the probability of the battery's thermal runaway risk, and then more accurately determining whether the battery has thermal runaway. In addition, since the data processing process of the soft decision tree module can perform interpretable hierarchical reasoning, the interpretability of the thermal runaway warning model can be improved.

[0138] like Figure 3 As shown, combined Figure 2 , the method comprises the following steps:

[0139] S301: Obtain a training set, a first model loss, a second model loss, and an initial thermal runaway warning model.

[0140] Among them, the first model loss is used to adjust the accuracy of the soft decision tree module, and the second model loss is used to adjust the accuracy of the long short-term memory network module.

[0141] In one possible implementation, initial battery status data may be obtained. Battery status data for a preset period of time may then be extracted from the initial battery status data. The battery status data for the preset period of time may then be cleansed to obtain first status data. The first status data may then be normalized to obtain second status data. The second status data may then be processed using a sliding time window algorithm to obtain a training set.

[0142] In an embodiment of the present application, the thermal runaway risk assessment device includes: a computing unit and a time series database, the computing unit is used to estimate the SOC value and the SOH value of the battery, and the time series database is used to store data.

[0143] Alternatively, data can be collected using sensors such as voltage sensors, current sensors, and temperature sensors to obtain raw battery status data. The raw status data can then be transmitted to a computing unit to obtain initial battery status data. The initial battery status data can then be uploaded to a time series database for storage.

[0144] It should be noted that this application does not limit the specific processing methods of data cleaning and data normalization. For example, data cleaning includes: processing missing values ​​and processing outliers. Data normalization includes: maximum and minimum normalization algorithms. Among them, the method of processing missing values ​​includes: linear interpolation algorithm, and the method of processing outliers includes: Laida criterion (i.e. 3 in principle).

[0145] Optionally, the battery status data of a preset period may be processed based on a linear interpolation algorithm to obtain interpolated battery status data. Thereafter, the interpolated initial battery status data may be processed based on the Laida criterion to obtain the first status data.

[0146] It should be noted that the missing value can be estimated by assuming that there is a linear relationship between adjacent data points, and assuming that in the battery status data of the preset time period, the missing value is located between the first valid value and the second valid value.

[0147] The interpolated battery status data includes: multiple battery parameters.

[0148] Specifically, missing points can be determined based on the battery status data for a preset time period. A first valid value and a second valid value can then be determined based on the missing points and the battery status data for the preset time period. The missing values ​​corresponding to the missing points can then be obtained based on the first and second valid values ​​to obtain interpolated battery status data. The mean (i.e., μ) and the standard deviation (i.e., σ) of multiple battery parameters can then be obtained based on the interpolated battery status data. An upper bound (i.e., μ + 3σ) and a lower bound (i.e., μ - 3σ) of the multiple battery parameters can then be determined based on the mean and standard deviation. A target outlier can then be determined based on the multiple battery parameters, the upper bounds of the multiple battery parameters, and the lower bounds of the multiple battery parameters. The target outlier is defined as a battery parameter among the multiple battery parameters that is greater than the upper bound or less than the lower bound. The target outlier can then be removed to obtain the first status data.

[0149] For example, among the multiple battery parameters, battery parameter 1 is 25, battery parameter 2 is 28, battery parameter 3 is 22, battery parameter 4 is 26, battery parameter 5 is 27, battery parameter 6 is 23, battery parameter 7 is 24, battery parameter 8 is 29, battery parameter 9 is 30, and battery parameter 10 is 50. Subsequently, the multiple battery parameters can be calculated to obtain a mean of 27.4 and a standard deviation of 5.3. Therefore, the upper limit of the multiple battery parameters is 43.3, and the lower limit of the multiple battery parameters is 11.5. Since battery parameter 10 is 50, which is greater than the upper limit of 43.3, battery parameter 10 is the target outlier.

[0150] Exemplarily, the missing value satisfies Formula 13.

[0151] Formula 13.

[0152] Among them, x i is a missing value, x i-1 is the first valid value, x i+1 is the second effective value, t i is the timestamp of the missing value, t i-1 The timestamp of the first valid value, t i+1 The timestamp of the second valid value.

[0153] Exemplarily, the mean satisfies Formula 14.

[0154] Formula 14.

[0155] Among them, x j is any one of the multiple battery parameters in the interpolated battery state data, and N is the number of the multiple battery parameters in the interpolated battery state data.

[0156] Exemplarily, the standard deviation satisfies Formula 15.

[0157] Formula 15.

[0158] It should be noted that the preset time period is M consecutive unit time periods. This application does not limit the specific value of M. For example, M can be a positive integer greater than or equal to 3. This application does not limit the unit time period. For example, the unit time period can be 10 seconds, 30 seconds, 1 minute, 5 minutes, 30 minutes, 1 hour, 4 hours, or 12 hours.

[0159] Optionally, the first state data may be linearly transformed based on a maximum-minimum normalization algorithm to obtain the second state data.

[0160] Specifically, the first state data may be mapped to the interval [0, 1] for standardization to obtain the second state data.

[0161] In this way, the adverse effects caused by singular sample data can be eliminated.

[0162] Exemplarily, the second state data satisfies Formula 16.

[0163] Formula 16.

[0164] in, is the second state data, x is any one of the multiple battery parameters of the first state data, x min is the minimum battery parameter among the multiple battery parameters of the first state data, x max The maximum battery parameter among the multiple battery parameters of the first state data.

[0165] Optionally, a window size and a time step of the time window can be determined. Then, input samples can be obtained based on the second state data, the time window, and the time step. Then, the input samples can be divided according to a division ratio to obtain a training set, a validation set, and a test set.

[0166] The input sample is a matrix (N, d), where N is the window size and d is the number of features.

[0167] Specifically, the second state data within the first preset time can be intercepted through the time window to obtain the first intercepted data. After the second preset time interval, the battery time series data within the first preset time can be intercepted again through the sliding window to obtain the second intercepted data. This iteration is performed until the sliding window is slid from the first time step of the second state data to the last time step to obtain the target intercepted data. Afterwards, the intercepted data of X unit time periods that are sequentially continuous in time sequence can be extracted from the target intercepted data as model input items. Afterwards, the intercepted data of Y unit time periods that are located after the X unit time periods and sequentially continuous in time sequence can be extracted from the target intercepted data as model output items. Afterwards, the model input items and model output items can be used as input samples.

[0168] It should be noted that the present application does not limit the first preset time and the second preset time. For example, the first preset time can be 5 minutes (min), 10 minutes, 15 minutes, 20 minutes, or 25 minutes, and the second preset time can be 1 minute, 2 minutes, 3 minutes, 4 minutes, or 5 minutes.

[0169] It should be noted that X and Y are both positive integers.

[0170] It should be noted that this application does not impose any restrictions on the division ratio. For example, the division ratio can be 14:3:3, and 70% of the time series data can be used as the training set, 15% of the time series data can be used as the validation set, and 15% of the time series data can be used as the test set.

[0171] In one possible implementation, an initial long short-term memory network model and an initial soft decision tree model may be obtained, and then an initial thermal runaway warning model may be obtained based on the initial long short-term memory network model and the initial soft decision tree model.

[0172] Optionally, the initial thermal runaway warning model is a pruned thermal runaway warning model.

[0173] Optionally, a pruning threshold for the initial long short-term memory network module can be obtained. Subsequently, a structured pruning process can be performed on the initial long short-term memory network model based on the pruning threshold to obtain a pruned long short-term memory network model. Subsequently, the soft decision tree model can be pruned to obtain a pruned soft decision tree model. Subsequently, a pruned thermal runaway warning model can be obtained based on the pruned long short-term memory network model and the pruned soft decision tree model.

[0174] In one possible design, the pruned LSTM model can be initially trained to obtain an initially trained LSTM model. The initially trained LSTM model includes LSTM model parameters. Subsequently, quantization-aware training can be performed to convert multiple floating-point numbers in the LSTM model parameters into low-precision integers to obtain an optimized LSTM model.

[0175] Exemplarily, the low-precision integer satisfies Formula 17.

[0176] Formula 17.

[0177] in, is a low-precision integer, is the quantization step size, is any value among the multiple floating-point numbers in the long short-term memory network model parameters, It is the minimum value of multiple floating-point numbers in the long short-term memory network model parameters.

[0178] Exemplarily, the quantization step size satisfies Formula 18.

[0179] Formula 18.

[0180] in, is the maximum value of a floating-point number, and b is the number of quantization bits.

[0181] In some embodiments, the optimized long short-term memory network model can be evaluated.

[0182] In this way, the generalization performance and performance stability of the pruned long short-term memory network model can be ensured.

[0183] It should be understood that, in conjunction with S202 , the soft decision tree model includes: a plurality of initial evaluation paths, and the initial evaluation paths include: a branch structure.

[0184] In another possible design, the similarity of multiple branch structures can be determined based on multiple branch structures (i.e. ). Subsequently, based on the similarities of the multiple branch structures, multiple similar structure groups can be determined, where each similar structure group includes at least two branch structures whose similarity is less than a preset similarity threshold. Subsequently, the branch structures in each similar structure group can be fused to obtain multiple fused branch structures. Subsequently, based on the multiple fused branch structures, a pruned thermal runaway warning model can be obtained.

[0185] It should be understood that, in conjunction with S202 , the branch structure includes at least one of the following: a branch node and a decision edge, where the branch node is used to indicate an evaluation factor and the decision edge is used to indicate a probability weight.

[0186] It should be noted that the present application does not impose any restrictions on the preset similarity threshold. For example, the preset similarity threshold may be 1.

[0187] For example, if the similarity between branch node 1 and branch node 2 is less than 0.1, branch node 1 and branch node 2 are fused to obtain a fused branch node.

[0188] In this way, by fusing the branch structures within each similar structure group to obtain multiple fused branch structures, the thermal runaway warning model can be pruned. Furthermore, since a similar structure group contains at least two branch structures with a similarity less than a preset similarity threshold, and the branch structure includes branch nodes and decision edges, the accuracy of the thermal runaway warning model can be maintained while lightweighting the branch nodes and decision edges.

[0189] It is understood that by performing structured pruning on the LSTM network model, neurons and connections with little impact on thermal runaway prediction can be removed. By fusing the branch structures within each group of similar structures to obtain multiple fused branch structures, repeated calculations can be reduced. Furthermore, by lightweighting the LSTM network model and the soft decision tree model, the network architecture can be streamlined, the number of parameters and the amount of computation can be reduced, and the trained lightweight model can be adapted to the embedded system of the vehicle terminal while maintaining the model's prediction accuracy.

[0190] S302 : Training the initial thermal runaway warning model based on the training set, the first model loss, and the second model loss to obtain a trained thermal runaway warning model.

[0191] In one possible implementation, the parameters of the initial thermal runaway warning model can be initialized, and an optimizer for the initial thermal runaway warning model can be defined. Subsequently, a target loss function can be derived based on the first model loss and the second model loss. The training set can then be input into the initial thermal runaway warning model, allowing the initial thermal runaway warning model to be continuously optimized using the training set and the target loss function, resulting in a trained thermal runaway warning model.

[0192] Optionally, the objective loss function satisfies Formula 19.

[0193] Formula 19.

[0194] in, is the first weight parameter, is the second weight parameter, and Used to balance the training of long short-term memory network module and soft decision tree module. task is the cross entropy loss (i.e. the first model loss), Ltree is the tree structure regularization (i.e., the second model loss).

[0195] Exemplarily, the first model loss satisfies Formula 20.

[0196] Formula 20.

[0197] Where n is the number of samples, y i is the real data, y i ' is the predicted value of the long short-term memory network module.

[0198] Exemplarily, the second model loss satisfies Formula 21.

[0199] Formula 21.

[0200] in, is the hyperparameter of the soft decision tree model. left is the probability of selecting the first node, p right is the probability of selecting the second node.

[0201] It should be noted that the soft decision tree module includes multiple soft decision tree levels, and the first node and the second node are nodes located in the same soft decision tree level.

[0202] Optionally, the training set can be subjected to feature extraction through a long short-term memory network module to obtain an extracted feature vector. The extracted feature vector can then be input into a soft decision tree module, and the initial output result can be obtained through the differentiable hierarchical decision path of the soft decision tree module. The model loss of the initial thermal runaway warning model can then be calculated, and the gradient of each layer of parameters in the initial thermal runaway warning model can be reversely calculated using the chain rule. The model parameters can then be updated based on the gradient and the optimizer to obtain the initial thermal runaway warning model after initial training. The initial thermal runaway warning model after initial training can then be trained, and the training can be iteratively repeated until the initial thermal runaway warning model converges, thereby obtaining a trained thermal runaway warning model.

[0203] It should be noted that this application does not limit the optimizer. For example, the optimizer may be an Adaptive Moment Estimation (Adam) optimizer.

[0204] It should be noted that this application does not limit the initial learning rate and number of iterations during the training process of the thermal runaway warning model. For example, the initial learning rate can be 0.001 and the number of iterations can be 100.

[0205] It should be noted that when the feature extraction of the training set is performed through the long short-term memory network model to obtain the extracted feature vector, the extracted feature vector should be closely related to thermal runaway and representative, such as battery temperature, voltage, current, SOC, and SOH.

[0206] It should be noted that this application does not impose any restrictions on the number of LSTM layers, the number of hidden units in the LSTM model, the dimension of the extracted feature vector, or the data sampling frequency. For example, the number of LSTM layers can be 2, 3, or 4, the dimension of the feature vector can be 9, 10, or 11, and the number of hidden units can be 32, 64, or 128. The sensor data sampling frequency can be 1 Hz, 10 Hz, or 100 Hz.

[0207] It should be noted that to ensure model performance while reducing computational workload, the number of long short-term memory network layers can be set to 2. To capture complex nonlinear relationships and avoid model overfitting, the number of hidden units can be 64.

[0208] It should be noted that the data sampling frequency needs to be set in combination with the timeliness requirements of thermal runaway warnings. It is necessary to ensure that sufficient time dependencies can be captured while avoiding the introduction of excessive noise. Taking the sensor data sampling frequency of 1 Hz as an example, selecting 300 time steps can capture the key stages of sudden temperature rise (such as temperature rise greater than 1 degree Celsius per second (°C / s)) and sudden voltage drop (such as voltage drop greater than millivolts per second (50mV / s)).

[0209] It is understandable that the Adam optimizer can better perform parameter optimization, and the trained thermal runaway warning model can better learn how to map data to the correct categories, thereby improving the accuracy of the thermal runaway warning model's predictions.

[0210] In some embodiments, a test set may be obtained, and the trained thermal runaway warning model may be evaluated based on the test set.

[0211] In another possible implementation, an initial thermal runaway warning model can be trained based on a training set, a first model loss, and a second model loss to obtain first model parameters. The training set and first model parameters can then be sent to a server, which has an unpruned thermal runaway warning model deployed on it. The server is configured to tune the first model parameters based on the training set and the unpruned thermal runaway warning model. The server can then receive second model parameters, which are the tuned first model parameters. A trained thermal runaway warning model can then be obtained based on the second model parameters.

[0212] It should be noted that this application does not limit the number of vehicles. For example, the number of vehicles can be one or more.

[0213] For example, Figure 4 As shown, an unpruned thermal runaway warning model is deployed on the cloud, and the vehicle side (i.e., the thermal runaway risk assessment device) includes multiple vehicles. A lightweight model (i.e., the pruned thermal runaway warning model) is deployed in the embedded system of each vehicle.

[0214] It should be noted that the multiple vehicles in this application may be the same or different, and this application does not limit the types of the multiple vehicles.

[0215] Vehicles can collect data, obtain real-time datasets, and input these datasets into a lightweight model for model training. The cloud (i.e., server) can obtain initial model parameters. The cloud then sends these initial model parameters to all vehicles. Each vehicle can train the lightweight model based on the initial model parameters and the real-time dataset, obtaining updated model parameters (i.e., first model parameters). Each vehicle can then send the real-time dataset and updated model parameters to the cloud.

[0216] It should be noted that the content of the real-time dataset is not shared between vehicles.

[0217] The cloud then obtains global model parameters based on the updated model parameters sent by all vehicles. It also obtains a global model training set based on the real-time datasets sent by all vehicles. The cloud then trains the unpruned thermal runaway warning model based on the global model parameters and the global model training set to obtain updated global model parameters. The cloud then sends the updated global model parameters (i.e., second model parameters) to all vehicles.

[0218] Exemplarily, the global model parameters satisfy Formula 22.

[0219] Twenty-two.

[0220] in, is the global model parameter, are the updated model parameters.

[0221] For example, vehicle 1 can collect data to obtain real-time data set 1. Afterwards, the real-time data set 1 can be input into the lightweight model 1 for training to obtain updated model parameters. Afterwards, vehicle 1 can send real-time dataset 1 and updated model parameters to the cloud. Vehicle 2 can collect data and obtain real-time data set 2. Afterwards, real-time data set 2 can be input into lightweight model 2 for training to obtain updated model parameters. Afterwards, vehicle 2 can send real-time dataset 2 and updated model parameters to the cloud. Vehicle N can collect data to obtain a real-time dataset N. Afterwards, the real-time dataset N can be input into the lightweight model N for training to obtain the updated model parameters. Afterwards, vehicle N can send real-time dataset N and updated model parameters to the cloud. Afterwards, the cloud can update the model parameters based on the real-time dataset 1. , real-time dataset 2, updated model parameters , real-time dataset N and updated model parameters Train the unpruned thermal runaway warning model to obtain the updated global model parameters Afterwards, the updated global model parameters can be sent to all vehicles .

[0222] By deploying the optimized lightweight model in the vehicle's embedded system, rapid thermal runaway prediction can be achieved. By uploading real-time collected data to the cloud, where high-precision dynamic predictions are performed, vehicle-cloud data synchronization is achieved, achieving a balance between rapid response and high-precision early warning.

[0223] By deploying the pruned thermal runaway warning model in the thermal runaway risk assessment device, the device can quickly predict whether the battery is experiencing thermal runaway. By deploying the unpruned thermal runaway warning model on a server and optimizing the first model parameters on the server, the thermal runaway warning model can be used to improve the speed of thermal runaway prediction and enhance the accuracy of its predictions.

[0224] Based on the above technical solution, a training set, a first model loss, a second model loss, and an initial thermal runaway warning model are obtained, and the initial thermal runaway warning model is trained based on the training set, the first model loss, and the second model loss to obtain a trained thermal runaway warning model. The first model loss is used to adjust the accuracy of the soft decision tree module, and the second model loss is used to adjust the accuracy of the long short-term memory network module. In this way, the training direction of the thermal runaway warning model can be effectively guided, the thermal runaway warning model can be prevented from overfitting on the training data, and the generalization ability of the thermal runaway warning model can be improved. This allows the thermal runaway warning model to have square root robustness when facing different types of input data, and the performance of the thermal runaway warning model will not be degraded due to excessive use of some paths.

[0225] The following describes the thermal runaway risk assessment method provided by the embodiment of the present application with reference to specific examples. Figure 5As shown, the thermal runaway warning model includes a soft decision tree module, an input layer, a hidden layer, an attention mechanism layer, and a fully connected layer. The input layer can be used to obtain time series state data, which includes voltage change data within a target time period, current change data within a target time period, temperature change data within a target time period, state of charge change data within a target time period, and health state change data within a target time period. The input layer can then perform data conversion on the time series state data to obtain target sequence state data, in a format that can be processed by the long short-term memory network module. The hidden layer can then process the target sequence state data to obtain a hidden state feature vector. The hidden state feature vector can then be processed using the attention mechanism layer and the fully connected layer to obtain a battery risk feature vector. For each of the multiple preset evaluation paths in the soft decision tree module, initial probability assessment information corresponding to the preset evaluation path can be obtained based on the risk feature vector and the preset evaluation path. For example, the initial probability evaluation information corresponding to the multiple preset evaluation paths may be initial probability evaluation information 1 (ie, P1), initial probability evaluation information 2 (ie, P2), ..., initial probability evaluation information m-1 (ie, P m-1 ), the initial probability assessment information m (ie, P m ). Afterwards, a risk assessment result can be obtained based on multiple initial probability assessment information.

[0226] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the method. In order to achieve the above functions, the thermal runaway risk assessment device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0227] The embodiment of the present application can divide the thermal runaway risk assessment device into functional modules according to the above method. For example, the thermal runaway risk assessment device can include various functional modules corresponding to the various functional divisions, or two or more functions can be integrated into one processing module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0228] Figure 6 FIG. 1 is a schematic diagram of a thermal runaway risk assessment device according to an exemplary embodiment. Figure 6 The thermal runaway risk assessment device includes an acquisition module 601 and a processing module 602 .

[0229] Acquisition module 601 is used to acquire time series status data of the battery. The time series status data is used to indicate changes in battery parameters at different time points. Processing module 602 is used to input the time series status data into a trained thermal runaway warning model to obtain a risk assessment result. The thermal runaway warning model is used to detect whether the battery has a thermal runaway risk. The thermal runaway warning model includes a long-short-term memory network module and a soft decision tree module. The long-short-term memory network module is used to extract feature data related to thermal runaway risk, and the soft decision tree module is used to determine whether the battery has experienced thermal runaway.

[0230] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0231] Figure 7 FIG. 1 is a schematic diagram of another thermal runaway risk assessment device according to an exemplary embodiment. Figure 7 As shown, the thermal runaway risk assessment device includes but is not limited to: a processor 701 and a memory 702 .

[0232] The memory 702 is used to store executable instructions of the processor 701. It is understandable that the processor 701 is configured to execute instructions to implement the thermal runaway risk assessment method in the above embodiment.

[0233] It should be noted that those skilled in the art can understand that Figure 7 The structure of the thermal runaway risk assessment device shown in the figure does not constitute a limitation on the thermal runaway risk assessment device. The thermal runaway risk assessment device may include Figure 7 More or fewer components may be shown, or certain components may be combined, or the components may be arranged differently.

[0234] The processor 701 is the control center of the thermal runaway risk assessment device. It uses various interfaces and lines to connect the various parts of the entire thermal runaway risk assessment device. By running or executing software programs and / or modules stored in the memory 702 and calling data stored in the memory 702, it performs various functions of the thermal runaway risk assessment device and processes data, thereby monitoring the thermal runaway risk assessment device as a whole. The processor 701 may include one or more processing units. Optionally, the processor 701 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, etc., and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 701.

[0235] Memory 702 can be used to store software programs and various data. Memory 702 may primarily include a program storage area and a data storage area. The program storage area may store an operating system and application programs required by at least one functional module (such as a determination unit, a processing unit, etc.). Furthermore, memory 702 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0236] In an exemplary embodiment, a computer-readable storage medium including instructions is further provided, such as a memory 702 including instructions. The instructions can be executed by the processor 701 of the thermal runaway risk assessment apparatus to implement the method in the above embodiment.

[0237] In actual implementation, Figure 6 The functions of the acquisition module 601 and the processing module 602 can be obtained by Figure 7 The processor 701 in the embodiment calls the computer program stored in the memory 702. The specific execution process can be referred to the description of the method part in the above embodiment, which will not be repeated here.

[0238] Optionally, the computer-readable storage medium may be a non-temporary computer-readable storage medium, for example, the non-temporary computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0239] In an exemplary embodiment, the present application further provides a computer program product comprising one or more instructions, which can be executed by the processor 701 of the thermal runaway risk assessment device to implement the method in the above embodiment.

[0240] It should be noted that when the instructions in the above-mentioned computer-readable storage medium or one or more instructions in the computer program product are executed by the processor of the thermal runaway risk assessment device, the various processes of the above-mentioned method embodiment are implemented and the same technical effect as the above-mentioned method can be achieved. To avoid repetition, they will not be repeated here.

[0241] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0242] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0243] Units described as separate components may or may not be physically separate, and components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0244] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0245] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The software product is stored in a storage medium and includes a number of instructions for causing a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, ROM, RAM, disk or optical disk, etc. Various media that can store program code.

[0246] The above are only specific embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for assessing thermal runaway risk, characterized in that: The method comprises: Acquiring time series status data of a battery, wherein the time series status data is used to indicate changes in battery parameters of the battery at different time points; Inputting the time series state data into a trained thermal runaway warning model to obtain a risk assessment result, wherein the thermal runaway warning model is used to detect whether the battery has a thermal runaway risk, and the thermal runaway warning model includes: a long short-term memory network module and a soft decision tree module, wherein the long short-term memory network module is used to extract feature data related to the thermal runaway risk, and the soft decision tree module is used to determine whether the battery has a thermal runaway risk; The trained thermal runaway warning model is obtained in the following way: Obtaining a training set, a first model loss, a second model loss, and an initial thermal runaway warning model, wherein the first model loss is used to adjust the accuracy of the soft decision tree module, the second model loss is used to adjust the accuracy of the long short-term memory network module, and the initial thermal runaway warning model is a pruned thermal runaway warning model; Training the initial thermal runaway warning model based on the training set, the first model loss, and the second model loss to obtain first model parameters; Sending the training set and the first model parameters to a server, where the server is deployed with an unpruned thermal runaway warning model, and the server is configured to tune the first model parameters based on the training set and the unpruned thermal runaway warning model; receiving a second model parameter, wherein the second model parameter is the first model parameter after being tuned; Based on the second model parameters, the trained thermal runaway warning model is obtained.

2. The method according to claim 1, characterized in that Inputting the time series state data into the trained thermal runaway warning model to obtain a risk assessment result includes: Inputting the time series state data into the long short-term memory network module to obtain a risk feature vector of the battery, wherein the risk feature vector is used to indicate feature data related to thermal runaway risk; The risk feature vector is input into the soft decision tree module to obtain the risk assessment result.

3. The method according to claim 2, characterized in that The soft decision tree module includes: a plurality of preset evaluation paths; inputting the risk feature vector into the soft decision tree module to obtain the risk evaluation result includes: For each of the preset assessment paths, obtaining initial probability assessment information corresponding to the preset assessment path based on the risk feature vector and the preset assessment path, so as to obtain the initial probability assessment information corresponding to the multiple preset assessment paths, wherein the initial probability assessment information is used to indicate a probability of thermal runaway risk occurring in the battery, and each preset assessment path corresponds to one piece of initial probability assessment information; The risk assessment result is obtained based on the multiple initial probability assessment information.

4. The method according to claim 3, characterized in that The preset evaluation path includes multiple evaluation factors and multiple probability weights, wherein the evaluation factors are used to indicate the conditions under which the battery has a thermal runaway risk, and the probability weights are used to indicate the degree of influence of the evaluation factors on the thermal runaway risk of the battery, and one evaluation factor corresponds to multiple probability weights; The obtaining, based on the risk feature vector and the preset assessment path, the initial probability assessment information corresponding to the preset assessment path includes: The initial probability assessment information corresponding to the preset assessment path is obtained based on the risk feature vector, the multiple assessment factors, and the multiple probability weights.

5. The method according to any one of claims 2 to 4, characterized in that The long short-term memory network module includes: a fully connected layer, which is used to input the risk feature vector into the soft decision tree module; wherein the dimension of the fully connected layer is the same as the dimension of the input layer of the soft decision tree module.

6. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Acquire an initial soft decision forest model, wherein the initial soft decision forest model includes a plurality of initial evaluation paths, and the initial evaluation paths include: a branch structure; Based on the plurality of branch structures, determining similarities of the plurality of branch structures; Determining a plurality of similar structure groups based on the similarities of the plurality of branch structures, wherein one of the similar structure groups includes at least two branch structures whose similarities are less than a preset similarity threshold; The branch structure parameters in each group of similar structure groups are fused to obtain the soft decision forest model.

7. The method according to claim 6, characterized in that The branch structure includes at least one of the following: a branch node and a decision edge, the branch node is used to indicate an evaluation factor, the decision edge is used to indicate a probability weight, the evaluation factor is used to indicate the conditions under which the battery has a thermal runaway risk, and the probability weight is used to indicate the degree of influence of the evaluation factor on the thermal runaway risk of the battery, and one evaluation factor corresponds to multiple probability weights.

8. A device for assessing thermal runaway risk, characterized in that: The device comprises: An acquisition module, configured to acquire time series status data of a battery, wherein the time series status data is used to indicate changes in battery parameters of the battery at different time points; a processing module, configured to input the time series state data into a trained thermal runaway warning model to obtain a risk assessment result, wherein the thermal runaway warning model is configured to detect whether the battery has a thermal runaway risk, and the thermal runaway warning model includes a long short-term memory network module and a soft decision tree module, wherein the long short-term memory network module is configured to extract feature data related to the thermal runaway risk, and the soft decision tree module is configured to determine whether the battery has a thermal runaway risk; The trained thermal runaway warning model is obtained in the following way: The acquisition module is further used to acquire a training set, a first model loss, a second model loss, and an initial thermal runaway warning model, wherein the first model loss is used to adjust the accuracy of the soft decision tree module, the second model loss is used to adjust the accuracy of the long short-term memory network module, and the initial thermal runaway warning model is a pruned thermal runaway warning model; The processing module is further configured to train the initial thermal runaway warning model based on the training set, the first model loss, and the second model loss to obtain first model parameters; a sending module, configured to send the training set and the first model parameters to a server, where the server is deployed with an unpruned thermal runaway warning model, and the server is configured to tune the first model parameters based on the training set and the unpruned thermal runaway warning model; A receiving module, configured to receive a second model parameter; the second model parameter being the first model parameter after optimization; The processing module is further configured to obtain the trained thermal runaway warning model based on the second model parameters.

9. A device for assessing thermal runaway risk, characterized in that: include: processor; A memory for storing the processor-executable instructions; wherein the processor is configured to execute the instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that When the computer-executable instructions stored in the computer-readable storage medium are executed by a processor of the thermal runaway risk assessment device, the thermal runaway risk assessment device can perform the method according to any one of claims 1 to 7.

Images