Battery thermal runaway prediction method, device and system based on multi-mode neural network
By using a multimodal neural network to extract and fuse features from battery thermal images, voltage data, and temperature data, the problem of low accuracy in predicting battery thermal runaway in existing technologies is solved, achieving higher prediction accuracy.
Patent Information
- Application Number
- CN202410314589.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-19
- Publication Date
- 2024-11-22
AI Technical Summary
Existing methods for predicting battery thermal runaway have low accuracy when data is insufficient.
By acquiring thermal images, voltage data, and temperature data of the battery, feature extraction and fusion processing are performed using a multimodal neural network to generate fused features, and the battery is determined to be in a thermal runaway state when the temperature prediction value does not meet the preset conditions.
It improves the accuracy of predicting battery thermal runaway states and enhances the reliability of predictions by integrating information from multiple data sources.
Smart Images

Figure CN119001452A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of battery energy storage technology, and in particular to a battery thermal runaway prediction method, device and system based on a multimodal neural network. Background Art
[0002] In the related technology, lithium-ion batteries are currently widely used in battery energy storage power stations due to their higher energy and power density, longer cycle life and lower cost. Since the batteries in energy storage power stations are large in scale, the safety of the battery system is of vital importance. However, the existing prediction of battery thermal runaway only uses the voltage, current and temperature data of the energy storage system. In the case of insufficient data, the accuracy of the prediction results is low. Summary of the invention
[0003] In order to overcome the problems existing in the related art, the present disclosure provides a battery thermal runaway prediction method, device and system based on a multimodal neural network.
[0004] According to a first aspect of an embodiment of the present disclosure, a battery thermal runaway prediction method using a multimodal neural network is provided, comprising:
[0005] Obtain thermal images, voltage data, and temperature data of the battery within a preset historical time period;
[0006] Extracting features from the thermal image, voltage data and temperature data respectively to obtain temperature features, voltage features and image features;
[0007] fusing the temperature feature, the voltage feature and the image feature to obtain a fused feature;
[0008] Predicting a temperature prediction value of the battery at a target time node based on the fusion feature;
[0009] When the temperature prediction value does not meet a preset condition, it is determined that the battery is in a thermal runaway state at the target time node.
[0010] In some embodiments of the present disclosure, the fusing process of the temperature feature, the voltage feature and the image feature to obtain the fused feature includes:
[0011] Normalizing the temperature feature, the voltage feature, and the image feature respectively to obtain the normalized temperature feature, the voltage feature, and the image feature;
[0012] Performing weighted processing on the image feature using the voltage feature to obtain a first attention weighted feature;
[0013] Using the temperature feature to perform weighted processing on the image feature to obtain a second attention weighted feature;
[0014] generating a composite feature based on the first attention weighted feature and the first attention weighted feature;
[0015] The composite feature is added to the aggregate value of the temperature feature before normalization and the aggregate value of the voltage feature before normalization to obtain the fused feature.
[0016] In some embodiments of the present disclosure, the step of performing weighted processing on the image feature using the voltage feature to obtain a first attention weighted feature includes:
[0017] The voltage feature and the image feature are fused by the following formula to obtain the first temporal attention map:
[0018]
[0019] Among them, I o is the normalized image feature, V o is the normalized voltage characteristic, σ x is the Softmax function, T is the transposed symbol;
[0020] Using the first temporal attention map The image features are weighted by the following formula to obtain the first attention weighted feature I V :
[0021]
[0022] Among them, W1 is the weight matrix and b1 is the mean deviation.
[0023] In some embodiments of the present disclosure, the step of performing weighted processing on the image feature using the temperature feature to obtain a second attention weighted feature includes:
[0024] The temperature feature and the image feature are fused by the following formula to obtain the second temporal attention map:
[0025]
[0026] Among them, I o is the normalized image feature, T o is the normalized temperature characteristic, σ x is the Softmax function, T is the transposed symbol;
[0027] Using the second temporal attention map The image feature is weighted by the following formula to obtain the second attention weighted feature I T :
[0028]
[0029] Among them, W2 is the weight matrix and b2 is the average deviation value.
[0030] In some embodiments of the present disclosure, generating a composite feature based on the first attention weighted feature and the second attention weighted feature includes:
[0031] Based on the first attention weighted feature and the second attention weighted feature, the composite feature C is generated by the following formula:
[0032]
[0033] Among them, W3 and W4 are weight matrices, b3 and b4 are bias values, σ f is the rectified linear unit activation function.
[0034] In some embodiments of the present disclosure, predicting the temperature prediction value of the battery at the target time node based on the fusion feature includes:
[0035] Based on the fusion feature, the temperature prediction value of the battery at the target time node is calculated by the following formula:
[0036]
[0037] Among them, W5 and W6 are weight matrices, b5 and b6 are deviation values, F is the fusion feature, and h is the number corresponding to the target time node.
[0038] In some embodiments of the present disclosure, after predicting the temperature prediction value of the battery at the target time node based on the fusion feature, the method further includes:
[0039] Acquire a target data set; wherein the target data set includes a target thermal image, target voltage data, target temperature data, and a target prediction value;
[0040] Extracting features from the target data set to obtain target data features;
[0041] Based on the target data feature and the temperature feature, voltage feature and image feature, generating a correction weight using a cosine similarity function;
[0042] Multiplying the temperature prediction value by the correction weight to obtain a temperature prediction value after correction;
[0043] When the temperature prediction value does not meet a preset condition, determining that the battery is in a thermal runaway state at the target time node includes:
[0044] When the predicted temperature value after the correction does not meet the preset condition, it is determined that the battery is in a thermal runaway state at the target time node.
[0045] In some embodiments of the present disclosure, when the temperature prediction value does not meet a preset condition, determining that the battery is in a thermal runaway state at the target time node includes:
[0046] When the temperature prediction value is greater than or equal to a preset threshold, it is determined that the battery is in a thermal runaway state at the target time node.
[0047] According to a second aspect of an embodiment of the present disclosure, a battery thermal runaway prediction device based on a multimodal neural network is provided, comprising:
[0048] An acquisition unit, used to acquire a thermal image, voltage data, and temperature data of a battery within a preset historical time period;
[0049] A feature extraction unit, used to extract features from the thermal image, voltage data and temperature data respectively to obtain temperature features, voltage features and image features;
[0050] A fusion unit, used for fusing the temperature feature, the voltage feature and the image feature to obtain a fusion feature;
[0051] A prediction unit, configured to predict a temperature prediction value of the battery at a target time node based on the fusion feature;
[0052] A determination unit is used to determine that the battery is in a thermal runaway state at the target time node when the temperature prediction value does not meet a preset condition.
[0053] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method described in any one of the first aspects is implemented.
[0054] According to a fourth aspect of an embodiment of the present disclosure, there is provided a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method according to any one of the first aspects is implemented.
[0055] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising a computer program, wherein the computer program implements the method as described in any one of the first aspects when executed by a processor.
[0056] The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: by acquiring the thermal image, voltage data and temperature data of the battery within a preset historical time period, using a neural network model to extract features of the thermal image, voltage data and temperature data respectively, obtaining temperature features, voltage features and image features, fusing the temperature features, voltage features and image features to obtain fusion features, predicting the temperature prediction value of the battery at the target time node based on the fusion features, and determining that the battery is in a thermal runaway state at the target time node when the temperature prediction value does not meet the preset conditions. By fusing the thermal image, voltage data and temperature data of the battery, and comprehensively predicting whether the battery is in a thermal runaway state based on the fusion result, the accuracy of the prediction of the thermal runaway state of the battery is improved.
[0057] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0059] Figure 1 The present invention is a flowchart of a method for predicting thermal runaway of a battery using a multimodal neural network according to an exemplary embodiment.
[0060] Figure 2 It is a block diagram of a battery thermal runaway prediction device using a multimodal neural network according to an exemplary embodiment.
[0061] Figure 3 It is a block diagram of a device for battery thermal runaway prediction using a multimodal neural network according to an exemplary embodiment. DETAILED DESCRIPTION
[0062] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.
[0063] The terms used in the disclosed embodiments are only for the purpose of describing specific embodiments, and are not intended to limit the disclosed embodiments. The singular forms "a", "an" and "the" used in the disclosed embodiments and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.
[0064] It should be understood that although the terms first, second, third, etc. may be used to describe various information in the disclosed embodiments, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the disclosed embodiments, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at" or "when" or "in response to determination".
[0065] In addition, various forms of processes shown in the embodiments of the present disclosure may be used to reorder, add or delete steps. For example, the steps described in the present application may be performed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present disclosure can be achieved, and this document does not limit this.
[0066] In the related technology, lithium-ion batteries are currently widely used in battery energy storage power stations due to their higher energy and power density, longer cycle life and lower cost. Since the batteries in energy storage power stations are large in scale, the safety of the battery system is of vital importance. However, the existing prediction of battery thermal runaway only uses the voltage, current and temperature data of the energy storage system. In the case of insufficient data, the accuracy of the prediction results is low.
[0067] In order to solve the above problems, the present disclosure provides a battery thermal runaway prediction method, device and system of a multimodal neural network, which obtains the thermal image, voltage data and temperature data of the battery in a preset historical time period, and uses a neural network model to extract features of the thermal image, voltage data and temperature data respectively to obtain temperature features, voltage features and image features, and fuses the temperature features, voltage features and image features to obtain fusion features, and predicts the temperature prediction value of the battery at the target time node based on the fusion features. When the temperature prediction value does not meet the preset conditions, it is determined that the battery is in a thermal runaway state at the target time node. By fusing the thermal image, voltage data and temperature data of the battery, and comprehensively predicting whether the battery is in a thermal runaway state based on the fusion result, the accuracy of the prediction of the battery thermal runaway state is improved.
[0068] Figure 1 is a flowchart of a battery thermal runaway prediction method using a multimodal neural network according to an exemplary embodiment. Figure 1 As shown, it should be noted that the battery thermal runaway prediction method of the multimodal neural network of the embodiment of the present disclosure is applied to the battery thermal runaway prediction device of the multimodal neural network. Figure 1 As shown, the method may include the following steps:
[0069] Step 101 : obtaining a thermal image, voltage data, and temperature data of a battery within a preset historical time period.
[0070] It should be noted that the above thermal image may be a high-dimensional thermal image, which may be obtained by photographing the battery with an infrared thermal imager.
[0071] Step 102 , extracting features from the thermal image, voltage data, and temperature data to obtain temperature features, voltage features, and image features.
[0072] In the embodiment of the present disclosure, a temperature feature extractor g may be used. T Feature extraction is performed on temperature data to obtain temperature features.
[0073] In the embodiment of the present disclosure, a voltage feature extractor g may be used. V Feature extraction is performed on the voltage data to obtain voltage features.
[0074] Among them, the temperature feature extractor g T and voltage feature extractor g V Both can use a recurrent neural network (RNN) based on a gated recurrent unit (GRU) to extract voltage features or temperature features using the following formula:
[0075] z t =σ g (W z X t +U z h t-1 +b z ),
[0076] r t =σ g (W r X t +U r h t-1 +b r ),
[0077]
[0078] in, Refers to the Hadamard matrix Hadamard product, X t Indicates the input data of step t (voltage data X V Or temperature data X T ), h t represents the output vector of step t (h0=0), i.e., the extracted eigenvalue, z t represents the update gate vector of step t, r tis the reset gate vector of step t, W z , W r , W h , U z , U r and U h is the weight matrix, b z 、b r 、b h is the deviation, σ g and σ h They are Sigmoid function and hyperbolic tangent function respectively.
[0079] In the embodiment of the present disclosure, a convolutional neural network (CNN) can be used as an image feature extractor. I To extract time-sensitive visual features.
[0080] Step 103, fusing the temperature feature, the voltage feature and the image feature to obtain a fused feature.
[0081] It is understandable that the fusion of temperature features, voltage features and image features can take into account multiple influencing factors when predicting battery thermal runaway. If only one or two factors are considered, the result may be inaccurate.
[0082] In some embodiments of the present disclosure, step 103 may specifically include the following steps:
[0083] Step a1, normalizing the temperature feature, voltage feature and image feature respectively to obtain normalized temperature feature, voltage feature and image feature.
[0084] In the embodiment of the present disclosure, the temperature feature T, the voltage feature V and the image feature I are normalized respectively to obtain the normalized temperature feature T o 、Voltage characteristics V o and image feature I o .
[0085] Step a2: perform weighted processing on the image features using the voltage features to obtain the first attention weighted features.
[0086] In some embodiments of the present disclosure, step a2 may specifically include the following steps:
[0087] Step a21, the voltage feature and the image feature are fused by the following formula to obtain the first temporal attention map
[0088]
[0089] Among them, I o is the normalized image feature, V o is the normalized voltage characteristic, σ x is the normalized exponential function Softmax, σ x To focus data The mapping is normalized to the range [0,1].
[0090] Step a22, using the first temporal attention map The image features are weighted by the following formula to obtain the first attention weighted feature I V :
[0091]
[0092] Among them, W1 is the weight matrix and b1 is the mean deviation.
[0093] Step a3, using the temperature feature to perform weighted processing on the image feature to obtain the second attention weighted feature.
[0094] In some embodiments of the present disclosure, step a3 may specifically include the following steps:
[0095] Step a31, the temperature feature and the image feature are fused by the following formula to obtain the second temporal attention map
[0096]
[0097] Among them, I o is the normalized image feature, T o is the normalized temperature characteristic, σ x is the Softmax function, σ x To focus data The mapping is normalized to the range [0,1].
[0098] Step a32, using the second temporal attention map The image features are weighted by the following formula to obtain the second attention weighted feature I T :
[0099]
[0100] Among them, W2 is the weight matrix and b2 is the average deviation value.
[0101] Step a4, generating a composite feature based on the first attention weighted feature and the second attention weighted feature.
[0102] In some embodiments of the present disclosure, step a4 may specifically include the following steps:
[0103] Based on the first attention weighted feature and the second attention weighted feature, the composite feature C is generated by the following formula:
[0104] C=σ f (W3I V +b3)°σ f (W4I T +b4)
[0105] Among them, W3 and W4 are weight matrices, b3 and b4 are bias values, σ f is the rectified linear unit (ReLU) activation function.
[0106] Step a5, adding the composite feature to the aggregate value of the temperature feature before normalization and the aggregate value of the voltage feature before normalization to obtain a fused feature.
[0107] In the embodiment of the present disclosure, the aggregate value of the above temperature characteristics can be obtained by Calculated, where N p is the number of temperature features in the preset historical time period, T t is the tth temperature characteristic value, and the aggregate value of the above voltage characteristics can be obtained by Calculated, where N p is the number of voltage features in the preset historical time period, V t is the tth voltage eigenvalue.
[0108] In one embodiment, the fusion feature
[0109] Step 104 : predicting the temperature prediction value of the battery at the target time node based on the fusion feature.
[0110] The temperature prediction value is any one or more of a thermal image prediction value, a voltage prediction value and a temperature prediction value.
[0111] In some embodiments of the present disclosure, step 104 may specifically include the following steps:
[0112] Based on the fusion features, the predicted temperature value of the battery at the target time node is calculated using the following formula
[0113] Among them, W5 and W6 are weight matrices, b5 and b6 are bias values, F is the fusion feature, and h is the number corresponding to the above target time node.
[0114] Step 105 : when the temperature prediction value does not meet the preset condition, determine that the battery is in a thermal runaway state at the target time node.
[0115] In some embodiments of the present disclosure, step 105 may specifically include the following steps:
[0116] When the temperature prediction value is greater than or equal to the preset threshold, it is determined that the battery is in a thermal runaway state at the target time node.
[0117] In some embodiments of the present disclosure, after step 104, the method may further include the following steps:
[0118] Step b1, obtaining the target data set.
[0119] Among them, the target data set includes target thermal image, target voltage data, target temperature data and target prediction value.
[0120] It should be noted that the data in the above target data set may be manually selected data that meets actual requirements.
[0121] In one embodiment, the thermal runaway prediction neural network is embedded into an adaptive learning framework. The learning framework uses a labeled dataset (i.e., a target dataset) To provide supervision signals, the superscript s indicates that it has been labeled. is the voltage data with labels, is the marked temperature data, is the labeled thermal image data, that is, for Based on the supervision signal, the thermal runaway prediction neural network can reasonably predict the thermal runaway state under limited training samples, and at the same time improve the unlabeled data set D = {X V ,X T ,X I} classification performance.
[0122] Step b2: extract features from the target data set to obtain target data features.
[0123] In one embodiment, the target data features can be used To represent, m is the number of labeled data sets, is the nth target data feature, which includes the marked temperature feature, voltage feature and image feature.
[0124] Step b3, based on the target data features and temperature features, voltage features and image features, a cosine similarity function is used to generate a correction weight.
[0125] In one embodiment, the temperature characteristics, voltage characteristics and image characteristics can be represented by {K v} η=1,…,n To indicate that K ηis the nth target data feature, which includes temperature features, voltage features and image features corresponding to a preset historical time period.
[0126] In one embodiment, the correction weight γ is calculated by the following formula:
[0127]
[0128] Step b4, multiplying the temperature prediction value by the correction weight to obtain the temperature prediction value after correction.
[0129] In one embodiment, the predicted temperature value after correction is Calculated by the following formula:
[0130]
[0131] in, is the temperature forecast value at the target time node.
[0132] It is understandable that by correcting the temperature prediction value, it is possible to ensure that the temperature prediction value is within a reasonable range and avoid large deviations.
[0133] In the above step 105, when the temperature prediction value does not meet the preset condition, determining that the battery is in a thermal runaway state at the target time node may specifically include:
[0134] When the predicted temperature value after correction does not meet the preset conditions, it is determined that the battery is in a thermal runaway state at the target time node.
[0135] In some embodiments of the present disclosure, different convolutional neural networks may be used to extract features of thermal images, temperature data, and voltage data to obtain temperature features, voltage features, and image features.
[0136] In some embodiments of the present disclosure, before step 101, historical data may be used to train the above-mentioned different convolutional neural networks to obtain a trained convolutional neural network.
[0137] In one example, first use N P The length window slices the collected historical data samples. Then, the historical voltage data X is generated. V 、Historical temperature dataX T and historical thermal imagesX I .
[0138] The historical voltage data X V 、Historical temperature dataX T and historical thermal imagesX IThe features are respectively input into the corresponding neural network models to be trained, and the feature values extracted by the neural network model are obtained. The feature values are compared with the true values. If the comparison results do not meet the preset requirements, the neural network model is adjusted according to the comparison results. If the comparison results meet the preset requirements, the trained neural network model is obtained.
[0139] In one example of a possible implementation, Compared with the temperature threshold β, if T H Less than β, y h It is marked as a normal state, otherwise it is a thermal runaway state. Since 313.15K is the starting temperature of the SEI film decomposition reaction, the battery temperature will rise rapidly after reaching this temperature and will soon cause thermal runaway, so β can be set to 313.15K.
[0140] According to the battery thermal runaway prediction method of the multimodal neural network proposed in the embodiment of the present disclosure, by acquiring the thermal image, voltage data and temperature data of the battery in a preset historical time period, the neural network model is used to extract features of the thermal image, voltage data and temperature data respectively to obtain temperature features, voltage features and image features, and the temperature features, voltage features and image features are fused to obtain fusion features. The temperature prediction value of the battery at the target time node is predicted based on the fusion features. When the temperature prediction value does not meet the preset conditions, it is determined that the battery is in a thermal runaway state at the target time node. By fusing the thermal image, voltage data and temperature data of the battery, whether the battery is in a thermal runaway state is comprehensively predicted based on the fusion result, thereby improving the accuracy of the prediction of the battery thermal runaway state.
[0141] Figure 2 is a block diagram of a battery thermal runaway prediction device using a multimodal neural network according to an exemplary embodiment. Figure 2 The device includes an acquisition unit 201, a feature extraction unit 202, a fusion unit 203, a prediction unit 204 and a determination unit 205.
[0142] The acquisition unit 201 is used to acquire the thermal image, voltage data and temperature data of the battery within a preset historical time period;
[0143] A feature extraction unit 202 is used to extract features from the thermal image, voltage data and temperature data to obtain temperature features, voltage features and image features;
[0144] A fusion unit 203 is used to fuse the temperature feature, the voltage feature and the image feature to obtain a fusion feature;
[0145] A prediction unit 204, configured to predict a temperature prediction value of the battery at a target time node based on the fusion feature;
[0146] The determination unit 205 is used to determine that the battery is in a thermal runaway state at a target time node when the temperature prediction value does not meet a preset condition.
[0147] In some embodiments of the present disclosure, the fusion unit 203 is specifically used to:
[0148] Normalizing the temperature feature, the voltage feature and the image feature respectively to obtain normalized temperature feature, voltage feature and image feature;
[0149] The image features are weighted using the voltage features to obtain the first attention weighted features;
[0150] The temperature feature is used to weight the image feature to obtain the second attention weighted feature;
[0151] Generate a composite feature based on the first attention weighted feature and the second attention weighted feature;
[0152] The composite feature is added to the aggregate value of the temperature feature before normalization and the aggregate value of the voltage feature before normalization to obtain a fused feature.
[0153] In some embodiments of the present disclosure, the fusion unit 203 is specifically used to:
[0154] The voltage features and image features are fused by the following formula to obtain the first temporal attention map:
[0155]
[0156] Among them, I o is the normalized image feature, V o is the normalized voltage characteristic, σ x is the Softmax function;
[0157] Using the first temporal attention map The image features are weighted by the following formula to obtain the first attention weighted feature I V :
[0158]
[0159] Among them, W1 is the weight matrix and b1 is the mean deviation.
[0160] In some embodiments of the present disclosure, the fusion unit 203 is specifically used to:
[0161] The temperature feature and image feature are fused by the following formula to obtain the second temporal attention map:
[0162]
[0163] Among them, I o is the normalized image feature, T o is the normalized temperature characteristic, σ x is the Softmax function;
[0164] Using the second temporal attention map The image features are weighted by the following formula to obtain the second attention weighted feature I T :
[0165]
[0166] Among them, W2 is the weight matrix and b2 is the average deviation value.
[0167] In some embodiments of the present disclosure, the prediction unit 204 is specifically used to:
[0168] Based on the first attention weighted feature and the second attention weighted feature, the composite feature C is generated by the following formula:
[0169] C=σ f (W3I V +b3)°σ f (W4I T +b4)
[0170] Among them, W3 and W4 are weight matrices, and b3 and b4 are bias values.
[0171] In some embodiments of the present disclosure, the device further includes:
[0172] The acquisition unit is further used to acquire a target data set; the target data set includes a target thermal image, target voltage data, target temperature data and a target prediction value;
[0173] The feature extraction unit is also used to extract features from the target data set to obtain target data features;
[0174] A generating unit, used for generating a correction weight by using a cosine similarity function based on target data features, temperature features, voltage features, and image features;
[0175] A correction unit, used for multiplying the temperature prediction value by the correction weight to obtain the temperature prediction value after correction;
[0176] The determination unit is also used to determine that the battery is in a thermal runaway state at a target time node when the temperature prediction value after correction does not meet a preset condition.
[0177] Regarding the device 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.
[0178] According to the battery thermal runaway prediction device of the multimodal neural network proposed in the embodiment of the present disclosure, by acquiring the thermal image, voltage data and temperature data of the battery in a preset historical time period, the neural network model is used to extract features of the thermal image, voltage data and temperature data respectively to obtain temperature features, voltage features and image features, and the temperature features, voltage features and image features are fused to obtain fusion features. The temperature prediction value of the battery at the target time node is predicted based on the fusion features. When the temperature prediction value does not meet the preset conditions, it is determined that the battery is in a thermal runaway state at the target time node. By fusing the thermal image, voltage data and temperature data of the battery, whether the battery is in a thermal runaway state is comprehensively predicted based on the fusion result, thereby improving the accuracy of the prediction of the battery thermal runaway state.
[0179] Figure 3 3 is a block diagram of a device for battery thermal runaway prediction using a multimodal neural network according to an exemplary embodiment. For example, the device 300 may be an electronic device, such as a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.
[0180] Reference Figure 3 , the device 300 may include one or more of the following components: a processing component 302 , a memory 304 , a power component 306 , a multimedia component 308 , an audio component 310 , an input / output (I / O) interface 312 , a sensor component 314 , and a communication component 316 .
[0181] The processing component 302 generally controls the overall operation of the device 300, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 302 may include one or more processors 320 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 302 may include one or more modules to facilitate the interaction between the processing component 302 and other components. For example, the processing component 302 may include a multimedia module to facilitate the interaction between the multimedia component 308 and the processing component 302.
[0182] The memory 304 is configured to store various types of data to support operations on the device 300. Examples of such data include instructions for any application or method operating on the device 300, contact data, phone book data, messages, pictures, videos, etc. The memory 304 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.
[0183] The power component 306 provides power to the various components of the device 300. The power component 306 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 300.
[0184] The multimedia component 308 includes a screen that provides an output interface between the device 300 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 308 includes a front camera and / or a rear camera. When the device 300 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.
[0185] The audio component 310 is configured to output and / or input audio signals. For example, the audio component 310 includes a microphone (MIC), and when the device 300 is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 304 or sent via the communication component 316. In some embodiments, the audio component 310 also includes a speaker for outputting audio signals.
[0186] I / O interface 312 provides an interface between processing component 302 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.
[0187] The sensor assembly 314 includes one or more sensors for providing various aspects of status assessment for the device 300. For example, the sensor assembly 314 can detect the open / closed state of the device 300, the relative positioning of components, such as the display and keypad of the device 300, the sensor assembly 314 can also detect the position change of the device 300 or a component of the device 300, the presence or absence of user contact with the device 300, the orientation or acceleration / deceleration of the device 300 and the temperature change of the device 300. The sensor assembly 314 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 314 can also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 314 can also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor or a temperature sensor.
[0188] The communication component 316 is configured to facilitate wired or wireless communication between the device 300 and other devices. The device 300 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 316 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 316 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
[0189] In an exemplary embodiment, the apparatus 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.
[0190] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 304 including instructions, which can be executed by a processor 320 of the device 300 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.
[0191] In an exemplary embodiment, a computer program product is also provided, comprising a computer program, which implements the above method when executed by the processor 320 of the device 300 .
[0192] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.
[0193] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.
Claims
1. A battery thermal runaway prediction method based on a multimodal neural network. It is characterized in that include: Obtain thermal images, voltage data, and temperature data of the battery within a preset historical time period; Extracting features from the thermal image, voltage data and temperature data respectively to obtain temperature features, voltage features and image features; fusing the temperature feature, the voltage feature and the image feature to obtain a fused feature; Predicting a temperature prediction value of the battery at a target time node based on the fusion feature; When the temperature prediction value does not meet a preset condition, it is determined that the battery is in a thermal runaway state at the target time node.
2. The battery thermal runaway prediction method of the multimodal neural network according to claim 1, It is characterized in that The fusing the temperature feature, the voltage feature and the image feature to obtain the fused feature includes: Normalizing the temperature feature, the voltage feature, and the image feature respectively to obtain the normalized temperature feature, the voltage feature, and the image feature; Performing weighted processing on the image feature using the voltage feature to obtain a first attention weighted feature; Using the temperature feature to perform weighted processing on the image feature to obtain a second attention weighted feature; generating a composite feature based on the first attention weighted feature and the first attention weighted feature; The composite feature is added to the aggregate value of the temperature feature before normalization and the aggregate value of the voltage feature before normalization to obtain the fused feature.
3. The battery thermal runaway prediction method of the multimodal neural network according to claim 2, It is characterized in that The step of performing weighted processing on the image feature by using the voltage feature to obtain a first attention weighted feature includes: The voltage feature and the image feature are fused by the following formula to obtain the first temporal attention map: Among them, I o is the normalized image feature, V o is the normalized voltage characteristic, σ x is the Softmax function, T is the transposed symbol; Using the first temporal attention map The image features are weighted by the following formula to obtain the first attention weighted feature I V : Among them, W 1 is the weight matrix, b 1 is the average deviation.
4. The battery thermal runaway prediction method of the multimodal neural network according to claim 2, It is characterized in that The step of performing weighted processing on the image feature by using the temperature feature to obtain a second attention weighted feature includes: The temperature feature and the image feature are fused by the following formula to obtain the second temporal attention map: Among them, I o is the normalized image feature, T o is the normalized temperature characteristic, σ x is the Softmax function, T is the transposed symbol; Using the second temporal attention map The image feature is weighted by the following formula to obtain the second attention weighted feature I T : Among them, W 2 is the weight matrix, b 2 is the average deviation value.
5. The battery thermal runaway prediction method of the multimodal neural network according to claim 4, It is characterized in that The generating a composite feature based on the first attention weighted feature and the second attention weighted feature comprises: Based on the first attention weighted feature and the second attention weighted feature, the composite feature C is generated by the following formula: C=σ f (W 3 I V +b 3 )°σ f (W 4 I T +b 4 ) Among them, W 3 and W 4 is the weight matrix, b 3 and b 4 is the deviation value, σ f is the rectified linear unit activation function.
6. The battery thermal runaway prediction method of the multimodal neural network according to claim 1, It is characterized in that The predicting the temperature prediction value of the battery at the target time node based on the fusion feature includes: Based on the fusion feature, the temperature prediction value of the battery at the target time node is calculated by the following formula: Among them, W 5 and W 6 is the weight matrix, b 5 and b 6 is the deviation value, F is the fusion feature, and h is the number corresponding to the target time node.
7. The battery thermal runaway prediction method of the multimodal neural network according to claim 1, It is characterized in that After predicting the temperature prediction value of the battery at the target time node based on the fusion feature, the method further includes: Acquire a target data set; wherein the target data set includes a target thermal image, target voltage data, target temperature data, and a target prediction value; Extracting features from the target data set to obtain target data features; Based on the target data feature and the temperature feature, voltage feature and image feature, generating a correction weight using a cosine similarity function; Multiplying the temperature prediction value by the correction weight to obtain a temperature prediction value after correction; When the temperature prediction value does not meet a preset condition, determining that the battery is in a thermal runaway state at the target time node includes: When the predicted temperature value after the correction does not meet the preset condition, it is determined that the battery is in a thermal runaway state at the target time node.
8. The battery thermal runaway prediction method of the multimodal neural network according to claim 1, It is characterized in that When the temperature prediction value does not meet a preset condition, determining that the battery is in a thermal runaway state at the target time node includes: When the temperature prediction value is greater than or equal to a preset threshold, it is determined that the battery is in a thermal runaway state at the target time node.
9. A battery thermal runaway prediction device based on a multimodal neural network. It is characterized in that include: An acquisition unit, used to acquire a thermal image, voltage data, and temperature data of a battery within a preset historical time period; A feature extraction unit, used to extract features from the thermal image, voltage data and temperature data respectively to obtain temperature features, voltage features and image features; A fusion unit, used for fusing the temperature feature, the voltage feature and the image feature to obtain a fusion feature; A prediction unit, configured to predict a temperature prediction value of the battery at a target time node based on the fusion feature; A determination unit is used to determine that the battery is in a thermal runaway state at the target time node when the temperature prediction value does not meet a preset condition.
10. An electronic device, It is characterized in that include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.
11. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.
12. A computer program product comprising a computer program, It is characterized in that The computer program implements the method according to any one of claims 1 to 8 when executed by a processor.
Citation Information
Cited By
Wireless power transmitting chip and method thereof
CN119276105A
Method and device for monitoring thermal runaway of energy storage system
CN120595129A
