A battery cell thermal runaway detection method and device based on a convolutional neural network

Abstract

The application discloses a kind of based on convolutional neural network battery thermal runaway detection method and device, including according to the time interval of pre-set, obtain the side view image containing at least two battery, and each side view image is input to first convolutional neural network, obtain the key point feature of each side view image;Based on the key point feature of side view image judges whether the explosion vent of battery is offset, and when detecting the explosion vent of battery is offset, first pulse signal is input to the loop where low melting point resistance is located;When the second pulse signal output by loop is inconsistent with first pulse signal, the thermal runaway battery is determined based on the voltage signal at the two ends of low melting point resistance.Through visual algorithm, it is accurately and quickly judged whether the battery explosion vent that sends offset exists in image, and the position of battery that exists thermal runaway is accurately and effectively obtained by combining low melting point resistance with lower cost, not only overall effective control input cost, data error caused by temperature control can be avoided.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of new energy vehicles, and particularly relates to a battery cell thermal runaway detection method and device based on a convolutional neural network. BACKGROUND

[0002] The development of new energy electric vehicles has shown an increasing trend year by year, and the number of new energy electric vehicles on the market has also been growing. As one of the core power battery assemblies of the vehicle powertrain, the battery cell has also been broken through and developed in various technologies with the promotion of the country's electrification. Among them, high-capacity and high-specific-energy battery cells are the goal pursued by major battery manufacturers and automobile manufacturers, and the resulting challenge and rigorous test for battery safety.

[0003] Among the current market problems of new energy vehicles, the safety hazard of thermal runaway is the first. Due to various factors such as external environment and battery cell quality, the degradation of individual battery cells may be higher than that of other battery cells, leading to poor battery cell consistency and causing various problems including thermal runaway. Currently, various manufacturers mainly determine whether thermal runaway occurs based on changes in single battery cell voltage, module temperature, and daisy chain continuity. The strategy seems to be relatively perfect, but there are the following problems:

[0004] First, the conventional temperature sensor arrangement is scattered, which cannot accurately detect the temperature of the single battery cell. When the thermal runaway battery cell is far away from the temperature sensor, it is easy to lead to inaccurate thermal runaway determination, and thus there is a risk of not reporting thermal runaway. Second, the current vehicle generally uses single-chain daisy chain communication. When the daisy chain is short-circuited or virtually connected, all single battery cell voltages and module temperature data will be unreliable, i.e., there is no basis for thermal runaway judgment, and the consistency and timeliness of the communication data cannot be guaranteed. SUMMARY

[0005] To solve the technical defects of the conventional temperature sensor arrangement being scattered, the inability to accurately detect the temperature of the single battery cell, and the risk of not reporting thermal runaway when the thermal runaway battery cell is far away from the temperature sensor, and the fact that the current vehicle generally uses single-chain daisy chain communication, which will lead to all single battery cell voltages and module temperature data being unreliable when the daisy chain is short-circuited or virtually connected, i.e., there is no basis for thermal runaway judgment, and the consistency and timeliness of the communication data cannot be guaranteed, the application provides a battery cell thermal runaway detection method and device based on a convolutional neural network, and the technical solution is as follows:

[0006] In a first aspect, the application provides a battery cell thermal runaway detection method based on a convolutional neural network, comprising:

[0007] acquire side view images containing at least two battery cells at preset time intervals, and input each side view image into the trained first convolutional neural network to obtain key point features of each side view image; wherein the first convolutional neural network is trained by a plurality of sample images with known key point features and the second convolutional neural network;

[0008] determine whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, and when detecting that the explosion-proof valve of any one battery cell is offset, input a first pulse signal that periodically changes to the loop in which the low-melting-point resistor arranged on each battery cell is located;

[0009] when detecting that the second pulse signal output by the loop is inconsistent with the first pulse signal, determine the thermal runaway battery cell based on the voltage signals across all low-melting-point resistors.

[0010] In an optional implementation of the first aspect, the key point features of the side view images include confidence features of the key points;

[0011] After inputting each side view image into the trained first convolutional neural network to obtain the key point features of each side view image, before determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, the method further includes:

[0012] deleting the key points corresponding to the confidence features that are lower than a preset confidence threshold from all key points of each side view image;

[0013] determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images includes:

[0014] determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images after the deletion processing.

[0015] In another optional implementation of the first aspect, the key point features of the side view images further include offset features of the key points;

[0016] determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images includes:

[0017] determining a vertex key point in each side view image based on the offset features of all key points in the side view image; wherein the vertex key point and the two adjacent key points are not on the same straight line;

[0018] regarding any two side view images in all side view images that are in adjacent time intervals as a group, and determining whether the offset features of the vertex key points corresponding to the two side view images in each group are consistent;

[0019] When detecting that there is an inconsistent offset feature of the vertex key point in any one group, it is determined that the explosion valve of any one battery cell is offset;

[0020] When detecting that there is a completely consistent offset feature of the vertex key point in any one group, it is determined that the explosion valve of any one battery cell is not offset.

[0021] In another optional implementation of the first aspect, when detecting that there is an inconsistent offset feature of the vertex key point in any one group, it is determined that the explosion valve of any one battery cell is offset, including:

[0022] When detecting that there is an inconsistent offset feature of the vertex key point in any one group, the slope between the inconsistent vertex key point and the adjacent two vertex key points in each side view image is calculated;

[0023] When the number of different slopes contained in the two side view images is consistent, it is determined that the explosion valve of any one battery cell is offset.

[0024] In another optional implementation of the first aspect, after detecting that the explosion valve of any one battery cell is offset, before inputting the first pulse signal with periodic change to the loop in which the low-melting-point resistor arranged on each battery cell is located, the method further includes:

[0025] Based on the temperature sensor arranged above at least two battery cells, a temperature signal corresponding to a preset time interval is acquired;

[0026] The first pulse signal with periodic change is input to the loop in which the low-melting-point resistor arranged on each battery cell is located, including:

[0027] According to the temperature signal and the preset time interval, a temperature change amount is calculated, and when detecting that the temperature change amount exceeds a preset temperature threshold, the first pulse signal with periodic change is input to the loop in which the low-melting-point resistor arranged on each battery cell is located.

[0028] In another optional implementation of the first aspect, based on the voltage signals across all low-melting-point resistors, a thermal runaway battery cell is determined, including:

[0029] When detecting that the voltage signal across any one low-melting-point resistor is greater than a preset first voltage threshold, and the voltage signals across all other low-melting-point resistors are all less than a preset second voltage threshold, the battery cell corresponding to the low-melting-point resistor greater than the preset first voltage threshold is determined as the thermal runaway battery cell;

[0030] The number corresponding to the low-melting-point resistor greater than the preset first voltage threshold is sent to a mobile terminal.

[0031] In a further optional implementation of the first aspect, the first convolutional neural network comprises one hourglass structure, and the second convolutional neural network comprises four hourglass structures; the loss function of the first convolutional neural network comprises loss parameters obtained after the second convolutional neural network is trained, and the second convolutional neural network is trained by using a plurality of sample images of known key point features.

[0032] In a second aspect, the embodiments of the present application provide a device for detecting thermal runaway of a battery cell based on a convolutional neural network, comprising:

[0033] a feature extraction module configured to acquire side-view images of at least two battery cells according to a preset time interval, and input each side-view image into a trained first convolutional neural network to obtain key point features of each side-view image; wherein the first convolutional neural network is trained by using a plurality of sample images of known key point features and a second convolutional neural network;

[0034] a signal detection module configured to determine whether the explosion-proof valve of any battery cell is offset based on the key point features of all side-view images, and input a first pulse signal that periodically changes into a loop in which a low-melting-point resistor arranged on each battery cell is located when detecting that the explosion-proof valve of any battery cell is offset;

[0035] a fault determination module configured to determine a thermal runaway battery cell based on voltage signals across all low-melting-point resistors when detecting that a second pulse signal output by the loop is inconsistent with the first pulse signal.

[0036] In an optional implementation of the second aspect, the key point features of the side-view images comprise confidence features of the key points.

[0037] The device further comprises:

[0038] After inputting each side-view image into the trained first convolutional neural network to obtain the key point features of each side-view image, and before determining whether the explosion-proof valve of any battery cell is offset based on the key point features of all side-view images, the key points with confidence features lower than a preset confidence threshold are deleted from all key points of each side-view image.

[0039] The signal detection module is configured to:

[0040] determine whether the explosion-proof valve of any battery cell is offset based on the key point features of all side-view images after the deletion processing.

[0041] In a further optional implementation of the second aspect, the key point features of the side-view images further comprise offset features of the key points.

[0042] The signal detection module is specifically configured to:

[0043] Determine the vertex key point in each side view image based on the offset features of all key points in each side view image; wherein the vertex key point and the two adjacent key points are not on the same straight line.

[0044] Take any two side view images in adjacent time intervals in all side view images as a group, and determine whether the offset features of the vertex key points corresponding to the two side view images in each group are consistent;

[0045] When detecting that there are inconsistent offset features of vertex key points in any one group, determine that the explosion valve of any one battery cell has been offset.

[0046] When detecting that there are completely consistent offset features of vertex key points in any one group, determine that the explosion valve of any one battery cell has not been offset.

[0047] In another optional solution of the second aspect, the signal detection module is specifically further used for:

[0048] When detecting that there are inconsistent offset features of vertex key points in any one group, calculate the slope between the inconsistent vertex key point and the two adjacent vertex key points in each side view image.

[0049] When the number of different slopes contained in the two side view images is consistent, determine that the explosion valve of any one battery cell has been offset.

[0050] In another optional solution of the second aspect, the device further comprises:

[0051] After detecting that the explosion valve of any one battery cell has been offset, before inputting the first pulse signal with periodic changes to the loop in which the low-melting-point resistor arranged on each battery cell is located, acquire the temperature signal corresponding to the preset time interval based on the temperature sensor arranged above at least two battery cells.

[0052] Input the first pulse signal with periodic changes to the loop in which the low-melting-point resistor arranged on each battery cell is located, including:

[0053] Calculate the temperature change amount according to the temperature signal and the preset time interval, and when detecting that the temperature change amount exceeds the preset temperature threshold, input the first pulse signal with periodic changes to the loop in which the low-melting-point resistor arranged on each battery cell is located.

[0054] In another optional solution of the second aspect, the fault determination module is used for:

[0055] When it is detected that the voltage signal at any one of the low-melting-point resistors is greater than a preset first voltage threshold, and the voltage signals at all other low-melting-point resistors are less than a preset second voltage threshold, the battery cell corresponding to the low-melting-point resistor greater than the preset first voltage threshold is determined as the thermal runaway battery cell.

[0056] The number corresponding to the low-melting-point resistor greater than the preset first voltage threshold is sent to the mobile terminal.

[0057] In still another optional implementation of the second aspect, the first convolutional neural network comprises one hourglass structure, and the second convolutional neural network comprises four hourglass structures; the loss function of the first convolutional neural network comprises a loss parameter obtained after the second convolutional neural network is trained, and the second convolutional neural network is trained by using sample images of known key point features.

[0058] In a third aspect, the embodiments of the present application further provide a battery cell thermal runaway detection device based on a convolutional neural network, comprising a processor and a memory;

[0059] The processor is connected with the memory;

[0060] The memory is configured to store executable program codes.

[0061] The processor runs a program corresponding to the executable program codes by reading the executable program codes stored in the memory, so as to implement the battery cell thermal runaway detection method based on the convolutional neural network provided in the first aspect or any one of the implementation manners of the first aspect.

[0062] In a fourth aspect, the embodiments of the present application provide a computer storage medium, which stores a computer program, and the computer program comprises program instructions. When the program instructions are executed by a processor, the battery cell thermal runaway detection method based on the convolutional neural network provided in the first aspect or any one of the implementation manners of the first aspect can be implemented.

[0063] In the embodiment of the present application, when the thermal runaway detection of the battery cell is performed, the side view images of the at least two battery cells are obtained at a preset time interval, and each side view image is input into the trained first convolutional neural network to obtain the key point features of each side view image. Whether the explosion-proof valve of any battery cell is offset is determined based on the key point features of all side view images, and when it is detected that the explosion-proof valve of any battery cell is offset, a first pulse signal that periodically changes is input to the circuit in which the low-melting-point resistor arranged on each battery cell is located. When it is detected that the second pulse signal output by the circuit is inconsistent with the first pulse signal, the thermal runaway battery cell is determined based on the voltage signals across all low-melting-point resistors. By using the visual algorithm, it is accurately and quickly determined whether the battery cell explosion-proof valve that is offset exists in the image, and the low-melting-point resistor with low cost is combined to accurately and effectively obtain the position of the battery cell that has thermal runaway. Not only the overall cost is effectively controlled, but also the data error caused by temperature control is avoided, and the consistency and timeliness of the data are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0065] Figure 1 The overall flowchart of a battery cell thermal runaway detection method based on a convolutional neural network provided in the embodiment of the present application;

[0066] Figure 2 The network training schematic diagram of a convolutional neural network provided in the embodiment of the present application;

[0067] Figure 3 The detection schematic diagram of a pulse signal provided in the embodiment of the present application;

[0068] Figure 4 The structural schematic diagram of a battery cell provided in the embodiment of the present application;

[0069] Figure 5 The structural schematic diagram of a battery cell thermal runaway detection device based on a convolutional neural network provided in the embodiment of the present application;

[0070] Figure 6 The structural schematic diagram of another battery cell thermal runaway detection device based on a convolutional neural network provided in the embodiment of the present application. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings of the embodiments of the present application.

[0072] In the following description, the terms "first", "second", etc. are used only for the purpose of description, and should not be interpreted as indicating or implying relative importance. The following description provides a plurality of embodiments of the present application, and different embodiments can be replaced or combined, so the present application can also be considered to include all possible combinations of the same and / or different embodiments described. Thus, if one embodiment includes features A, B, C, and another embodiment includes features B, D, the present application should also be considered to include embodiments containing one or more of all other possible combinations of A, B, C, and D, even if the embodiment is not explicitly described in the following.

[0073] The following description provides examples and does not limit the scope, applicability, or examples set forth in the claims. Changes can be made in the function and arrangement of elements described without departing from the scope of the application. Various examples can omit, substitute, or add various procedures or components as appropriate. For instance, the methods described can be performed in an order different than described, and various steps can be added, omitted, or combined. Also, features described with respect to some examples can be combined in other examples.

[0074] See Figure 1 , Figure 1 An overall flowchart of a battery cell thermal runaway detection method based on a convolutional neural network is shown.

[0075] As Figure 1 shown, the battery cell thermal runaway detection method based on a convolutional neural network can at least include the following steps:

[0076] Step 102, acquire side view images containing at least two battery cells at a preset time interval, and input each side view image to a trained first convolutional neural network to obtain key point features of each side view image.

[0077] In the embodiments of the present application, the battery cell thermal runaway detection method based on a convolutional neural network can be applied in a control terminal, for example, to detect thermal runaway of battery cells in a vehicle battery pack, to effectively solve the safety hazards caused by battery cell thermal runaway. When the battery cells in the vehicle battery pack have thermal runaway, a violent chemical reaction occurs inside the battery cells, which causes the temperature of the entire surface of the battery cells to rise sharply, accompanied by the generation of carbon monoxide, hydrogen, ethylene, methane, and ethane gases. The rapid rise in gas production and expansion further leads to a rapid rise in air pressure inside the battery cells. At this time, the visual algorithm proposed in the embodiments of the present application can be used to process the acquired images containing battery cells to accurately and quickly determine whether the battery cells have abnormality.

[0078] Specifically, when the thermal runaway detection of the battery cell is performed, the control terminal can control the camera or the laser radar to obtain side-view images of at least two battery cells in the battery pack at a preset time interval. The battery cells in the battery pack are arranged in an array, and each battery cell is provided with an explosion-proof valve. In the embodiment of the present application, the camera or the laser radar can be arranged in parallel with the explosion-proof valve to obtain images of all battery cells and their corresponding explosion-proof valves in each region of the battery pack, or images of all battery cells and their corresponding explosion-proof valves in the battery pack. It can be understood that, within the allowable error range of the installation of the battery cells in the battery pack, the positions of all battery cells and their corresponding explosion-proof valves in the same column in the side-view image completely coincide, or the positions of all battery cells and their corresponding explosion-proof valves in the same row completely coincide, in other words, the distance between each battery cell and the adjacent battery cell in the side-view image remains consistent.

[0079] Further, after obtaining the side-view images of at least two battery cells and their corresponding explosion-proof valves at a preset time interval, the control terminal can input each side-view image into the trained first convolutional neural network to obtain key point features in each side-view image. The first convolutional neural network can be trained by a plurality of sample images of known key point features and a second convolutional neural network. The key point features can include, but are not limited to, confidence features and offset features of key points. The first convolutional neural network includes a sandglass structure, the second convolutional neural network includes four identical sandglass structures, and the loss function of the first convolutional neural network includes loss parameters obtained after the second convolutional neural network is trained. The second convolutional neural network is trained by a plurality of sample images of known key point features.

[0080] In the training process of the first convolutional neural network, the second convolutional neural network can be trained based on the plurality of sample images of known key point features, and the loss parameters in the trained second convolutional neural network can be added when the first convolutional neural network is trained based on the plurality of sample images of known key point features, to obtain the trained first convolutional neural network. For details, please refer to Figure 2 The network training diagram of the convolutional neural network provided by the embodiment of the present application is shown. The upper half of the diagram can represent the training process of the second convolutional neural network, and the second convolutional neural network includes four identical sandglass structures. Each sandglass structure can include four encoding layers, four distillation layers, and four decoding layers. The lower half of the diagram can represent the training process of the first convolutional neural network, and the first convolutional neural network includes one sandglass structure identical to the second convolutional neural network.

[0081] It can be understood that the four distillation layers of the last hourglass structure of the second convolutional neural network in the embodiment of the present application will make a loss function with the four distillation layers of the first three hourglass structures. After the training of the second convolutional neural network is completed, the first convolutional neural network can be trained based on the sample images of the known key point features mentioned above, and in the process of calculating the loss function of the first convolutional neural network, the four distillation layers of the hourglass structure of the first convolutional neural network are first made a loss with the four distillation layers of the last hourglass structure of the second convolutional neural network and added to the loss function of the first convolutional neural network, and then the confidence features output by the first convolutional neural network are made a loss with the confidence features of the second convolutional neural network and added to the loss function of the first convolutional neural network.

[0082] Here, the loss function of the distillation layer of the second convolutional neural network can be but is not limited to obtained by the following formula:

[0083]

[0084] In the above formula, S can be represented as a spatial softmax function, can be represented as the i-th channel output by the distillation layer of the m-th hourglass structure of the second convolutional neural network, can be represented as the m-th hourglass structure of the second convolutional neural network, and D can be represented as the sum of squares.

[0085] The loss function of the distillation layer of the first convolutional neural network can be but is not limited to obtained by the following formula:

[0086]

[0087] In the above formula, can be represented as the hourglass structure of the first convolutional neural network.

[0088] Here, the function of making a loss between the confidence features output by the first convolutional neural network and the confidence features of the second convolutional neural network can be but is not limited to obtained by the following formula:

[0089]

[0090] In the above formula, can be represented as the value of the confidence coordinate (i,j) output by the second convolutional neural network, can be represented as the value of the confidence coordinate (i,j) output by the first convolutional neural network.

[0091] As an option of the embodiment of the present application, after each side view image is input into the trained first convolutional neural network to obtain the key point features of each side view image, before determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, the method further includes:

[0092] deleting the key points whose corresponding confidence features in all key points of each side view image are lower than a preset confidence threshold;

[0093] determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, including:

[0094] determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images after the deletion processing.

[0095] Specifically, in order to guarantee the accuracy and reliability of all key points of each side view image, the control terminal can filter out the key points whose confidence features are higher than or equal to the preset confidence feature according to the confidence features of all key points in each side view image, and can delete the key points whose confidence features are lower than the preset confidence feature in each side view image. Then, the control terminal can determine whether the explosion-proof valve of any one battery cell is offset according to all key point features after the deletion processing in each side view image.

[0096] Step 104, determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, and when detecting that the explosion-proof valve of any one battery cell is offset, inputting a first pulse signal that periodically changes to the loop in which the low-melting-point resistor arranged on each battery cell is located.

[0097] Specifically, after obtaining the key point features of each side view image, the control terminal can determine the vertex feature points in each side view image according to the offset features of all key points in each side view image. The vertex feature points can include, but are not limited to, the vertices of the side view of the battery cell and the vertices of the side view of the explosion-proof valve. The offset feature of the key point can be understood as the coordinate of the key point in the planar rectangular coordinate system corresponding to the side view image. For example, but not limited to, when the slope calculated by comparing the coordinates of each key point with the coordinates of the adjacent two key points is inconsistent, it indicates that the key point and the adjacent two key points are not on the same straight line, and then the key point can be taken as a vertex key point. When the slope is consistent, it indicates that the key point and the adjacent two key points are on the same straight line, and then the key point cannot be taken as a vertex key point.

[0098] It can be understood that in the embodiments of the present application, a certain error interval can be applied to each calculated slope. When there is an overlapping interval between the two slopes corresponding to any one key point, the key point can be determined as a vertex key point. However, the present application is not limited to this.

[0099] Further, after determining the vertex key points in each side view image, the control terminal can take any two side view images in adjacent time intervals as a group according to the time interval corresponding to each side view image, and determine whether all the vertex key points corresponding to the two side view images in each group correspond completely, that is, whether all the vertex key points corresponding to the first side view image in each group correspond one by one to all the vertex key points corresponding to the second side view image in each group. The correspondence can be understood as complete coincidence of coordinates or a difference value between the coordinates within a preset difference interval. It can be understood that when the battery cell is in thermal runaway, the internal pressure rapidly increases, which easily causes the explosion-proof valve to be pushed by the pressure and deviate. The direct impact can be reflected in the deviation of the vertex coordinates of the explosion-proof valve. The visual algorithm used in the embodiments of the present application can accurately determine whether the explosion-proof valve deviates, so as to improve the accuracy of the detection effect.

[0100] Further, when the coordinates of the inconsistent vertex key points in any one group are detected, it can be determined that the positions of the explosion-proof valves in the two side view images in the group are different, and it can be further determined that there is a battery cell in thermal runaway, that is, the explosion-proof valve of any one battery cell deviates. When the coordinates of the consistent vertex key points in any one group are detected, it can be determined that the explosion-proof valves on each battery cell in the battery pack have never deviated during the whole process, and there is no battery cell in thermal runaway, that is, the explosion-proof valve of any one battery cell does not deviate.

[0101] As another optional embodiment of the present application, when the inconsistent vertex key point deviation feature in any one group is detected, it is determined that the explosion-proof valve of any one battery cell deviates, including:

[0102] When the inconsistent vertex key point deviation feature in any one group is detected, the slope between the inconsistent vertex key point and the adjacent two vertex key points in each side view image is calculated.

[0103] When the number of different slopes contained in the two side view images is consistent, it is determined that the explosion-proof valve of any one battery cell deviates.

[0104] Specifically, in order to further improve the accuracy of the detection result, after detecting the offset feature of the inconsistent vertex key point in any group, the terminal can be controlled to calculate the slope between the adjacent two vertex key points for the inconsistent vertex key point in the two side view images of the group. It can be understood that when the position of the explosion-proof valve is offset, the top vertex of the explosion-proof valve is generally moved in the direction along the plane Y axis, that is, the X axis of the coordinate of the top vertex of the explosion-proof valve does not change, and the Y axis changes.

[0105] Possibly, when the number of different slopes calculated in the two side view images is consistent, it indicates that the positions of the explosion-proof valves in the two side view images are different, and it can be determined that there is a thermal runaway battery, that is, the explosion-proof valve of any battery is offset. Possibly, when the number of different slopes calculated in the two side view images is inconsistent, it indicates that the top vertex coordinates of the explosion-proof valves in the two side view images change in the X axis and the Y axis, and it is necessary to confirm again whether the top vertex coordinates of the explosion-proof valves are accurate. Here, the way to confirm again whether the top vertex coordinates of the explosion-proof valves are accurate can be but not limited to inputting the two side view images of the group into the first convolutional neural network mentioned above again, which will not be described in detail here.

[0106] As another optional embodiment of the present application, after detecting that the explosion-proof valve of any battery is offset, before inputting the first pulse signal with periodic change to the loop where the low-melting-point resistor arranged on each battery is located, the method further includes:

[0107] acquiring a temperature signal corresponding to a preset time interval based on the temperature sensor arranged above at least two batteries;

[0108] inputting the first pulse signal with periodic change to the loop where the low-melting-point resistor arranged on each battery is located, including:

[0109] calculating a temperature change amount according to the temperature signal and the preset time interval, and inputting the first pulse signal with periodic change to the loop where the low-melting-point resistor arranged on each battery is located when it is detected that the temperature change amount exceeds a preset temperature threshold.

[0110] Specifically, after detecting that the explosion-proof valve of any one of the battery cells is deviated, the control terminal can also acquire temperature signals at preset time intervals based on the temperature sensor arranged above the battery pack to determine whether the temperature of the battery pack changes. It can be understood that after obtaining the temperature signals corresponding to each time interval, the temperature change amount can be calculated based on the temperature signal difference corresponding to two consecutive time intervals and the time interval, and it is determined whether the temperature change amount exceeds the preset temperature threshold. Here, when a battery cell in the battery pack has thermal runaway or has a tendency to have thermal runaway, the detected temperature signal will change, which can be amplified by dividing by the time interval, and then an effective and rapid judgment can be made.

[0111] When it is detected that the temperature change amount exceeds the preset temperature threshold, the control terminal can input a periodically changing first pulse signal to the loop in which the low-melting-point resistors arranged on each battery cell are located. The low-melting-point resistors on each battery cell can be arranged above the explosion-proof valve to more directly receive the heat generated by the battery cell. Here, the low-melting-point resistors on each battery cell are connected in series in turn. One end of the series circuit can be connected to the output end of the control terminal to obtain the periodically changing first pulse signal output by the output end of the control terminal. The other end of the series circuit can be connected to the input end of the control terminal to obtain the pulse signal corresponding to the series circuit, and when any one of the low-melting-point resistors in the series circuit does not fuse, the pulse signal sent by the output end of the control terminal and the pulse signal received by the input end of the control terminal remain consistent.

[0112] Herein Figure 3 The detection schematic diagram of the pulse signal provided by the embodiment of the application is shown. As Figure 3 shown, the battery pack in the detection schematic diagram is provided with 10 rows and 4 columns of battery cells, each battery cell is provided with a low-melting-point resistor, and each low-melting-point resistor is connected in series in turn. One end of the series circuit can be connected to the output end of the control terminal, and the other end of the series circuit can be connected to the input end of the control terminal, so that the control terminal can determine whether the pulse signals of the input end and the output end are consistent in real time.

[0113] Step 106, when it is detected that the second pulse signal output by the loop is inconsistent with the first pulse signal, determining the thermal runaway battery cell based on the voltage signals across all low-melting-point resistors.

[0114] Specifically, when the control terminal detects that the second pulse signal output by the circuit formed by the series connection of each low-melting-point resistor is inconsistent with the first pulse signal, it indicates that at least one low-melting-point resistor in the circuit has melted, resulting in an open circuit. Next, to effectively and accurately identify the cell experiencing thermal runaway, the control terminal can detect the voltage signal across each low-melting-point resistor. When the voltage signal across any one of the low-melting-point resistors is greater than a preset first voltage threshold, and the voltage signals across all other low-melting-point resistors are less than a preset second voltage threshold, the cell corresponding to the low-melting-point resistor with the voltage greater than the preset first voltage threshold is identified as the thermal runaway cell. The number corresponding to this low-melting-point resistor is then sent to a designated mobile terminal, allowing staff to quickly replace the corresponding cell. It is understood that, in the embodiments of this application, the control terminal can also form a separate path with each low melting point resistor to effectively obtain the voltage signal corresponding to the two ends of each low melting point resistor. When the low melting point resistor melts, the voltage signal corresponding to its two ends is large; when the low melting point resistor does not melt, the voltage signal corresponding to its two ends approaches 0.

[0115] As another optional embodiment of this application, please refer to Figure 4 The diagram shown is a structural schematic of a battery cell provided in an embodiment of this application. Figure 4 As shown, a pressure sensor (i.e., a BPS installed inside the cell) can be installed at the bottom of the explosion-proof valve of the cell, and a low-melting-point resistor (i.e., R installed above the cell) can be installed above the explosion-proof valve of the cell. Either the pressure sensor or the low-melting-point resistor can individually determine whether the corresponding cell has experienced thermal runaway. Specifically, when the ratio of the pressure signal difference collected by the pressure sensor to the time interval exceeds a preset pressure threshold within a preset time interval, it indicates that the corresponding cell has experienced thermal runaway; or, when a large voltage signal is detected across the low-melting-point resistor, and the voltage signals across the low-melting-point resistors of the other cells are all close to 0, it indicates that the cell corresponding to the low-melting-point resistor with the large voltage signal has experienced thermal runaway, and the low-melting-point resistors of each cell form a series circuit.

[0116] Please see Figure 5 , Figure 5 This illustration shows a schematic diagram of a cell thermal runaway detection device based on a convolutional neural network, provided in an embodiment of this application.

[0117] like Figure 5 As shown, the cell thermal runaway detection device based on convolutional neural networks may include at least a feature extraction module 501, a signal detection module 502, and a fault determination module 503, wherein:

[0118] The feature extraction module 501 is configured to acquire side view images of the at least two battery cells at preset time intervals, and input each side view image into the trained first convolutional neural network to obtain key point features of each side view image; wherein the first convolutional neural network is trained by a plurality of sample images of known key point features and the second convolutional neural network.

[0119] The signal detection module 502 is configured to determine whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, and input a first pulse signal that periodically changes to a loop in which a low-melting-point resistor arranged on each battery cell is located when detecting that the explosion-proof valve of any one battery cell is offset.

[0120] The fault determination module 503 is configured to determine a thermal runaway battery cell based on voltage signals across all low-melting-point resistors when detecting that a second pulse signal output by the loop is inconsistent with the first pulse signal.

[0121] In some possible embodiments, the key point features of the side view images include confidence features of the key points.

[0122] The device further includes:

[0123] After inputting each side view image into the trained first convolutional neural network to obtain the key point features of each side view image, and before determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, the key points with confidence features lower than a preset confidence threshold are deleted from all key points of each side view image.

[0124] The signal detection module is configured to:

[0125] Determine whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images after the deletion processing.

[0126] In some possible embodiments, the key point features of the side view images further include offset features of the key points.

[0127] The signal detection module is specifically configured to:

[0128] Determine a vertex key point in each side view image based on the offset features of all key points in the side view image; wherein the vertex key point and two adjacent key points are not on the same straight line.

[0129] Take any two side view images in all side view images as a group, and determine whether the offset features of the vertex key points corresponding to the two side view images in each group are consistent.

[0130] When detecting that there is an inconsistent offset feature of the vertex key point in any one group, it is determined that the explosion valve of any one battery cell is offset;

[0131] When detecting that there is a completely consistent offset feature of the vertex key point in any one group, it is determined that the explosion valve of any one battery cell is not offset.

[0132] In some possible embodiments, the signal detection module is specifically further configured to:

[0133] When detecting that there is an inconsistent offset feature of the vertex key point in any one group, the slope between the inconsistent vertex key point and the adjacent two vertex key points in each side view image is calculated;

[0134] When the number of different slopes contained in the two side view images is consistent, it is determined that the explosion valve of any one battery cell is offset.

[0135] In some possible embodiments, the device further comprises:

[0136] After detecting that the explosion valve of any one battery cell is offset, before inputting the first pulse signal with periodic change to the loop in which the low-melting-point resistor arranged on each battery cell is located, the temperature signal corresponding to the preset time interval is acquired based on the temperature sensor arranged above at least two battery cells;

[0137] The first pulse signal with periodic change is input to the loop in which the low-melting-point resistor arranged on each battery cell is located, including:

[0138] According to the temperature signal and the preset time interval, the temperature change amount is calculated, and when it is detected that the temperature change amount exceeds the preset temperature threshold, the first pulse signal with periodic change is input to the loop in which the low-melting-point resistor arranged on each battery cell is located.

[0139] In some possible embodiments, the fault determination module is configured to:

[0140] When it is detected that the voltage signal across any one low-melting-point resistor is greater than a preset first voltage threshold, and the voltage signals across all other low-melting-point resistors are all less than a preset second voltage threshold, the battery cell corresponding to the low-melting-point resistor greater than the preset first voltage threshold is taken as a thermal runaway battery cell;

[0141] The number corresponding to the low-melting-point resistor greater than the preset first voltage threshold is sent to the mobile terminal.

[0142] In some possible embodiments, the first convolutional neural network comprises one hourglass structure, and the second convolutional neural network comprises four hourglass structures; the loss function of the first convolutional neural network comprises loss parameters obtained after the second convolutional neural network is trained, and the second convolutional neural network is trained by using sample images of the plurality of known key point features.

[0143] Those skilled in the art can clearly understand that the technical solutions of the embodiments of the present application can be implemented by means of software and / or hardware. The "unit" and "module" in the specification refer to software and / or hardware that can independently complete or cooperate with other components to complete a specific function, and the hardware may, for example, be a field programmable gate array (FPGA), an integrated circuit (IC), and the like.

[0144] Please refer to Figure 6 , Figure 6 A structure schematic diagram of still another kind of battery cell thermal runaway detection device based on a convolutional neural network provided by an embodiment of the present application is shown.

[0145] As Figure 6 shown, the battery cell thermal runaway detection device 600 based on a convolutional neural network can comprise at least one processor 601, at least one network interface 604, a user interface 603, a memory 605, and at least one communication bus 602.

[0146] The communication bus 602 can be used to realize the connection and communication of the above-mentioned various components.

[0147] The user interface 603 can comprise a key, and the optional user interface can further comprise a standard wired interface, a wireless interface.

[0148] The network interface 604 can comprise, but is not limited to, a Bluetooth module, an NFC module, a Wi-Fi module, and the like.

[0149] The processor 601 can include one or more processing cores. The processor 601 connects various parts in the battery cell thermal runaway detection apparatus 600 based on a convolutional neural network by running or executing instructions, programs, code sets or instruction sets stored in the memory 605, and calling data stored in the memory 605, to perform various functions of the battery cell thermal runaway detection apparatus 600 based on a convolutional neural network and process data. Optionally, the processor 601 can be implemented in at least one of the hardware forms of DSP, FPGA, and PLA. The processor 601 can integrate one or a combination of CPU, GPU, and modem. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content required to be displayed on the display screen; and the modem is used to process wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 601, but can be realized by a separate chip.

[0150] The memory 605 can include RAM and can also include ROM. Optionally, the memory 605 includes a non-transitory computer readable medium. The memory 605 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 605 can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area can store data involved in the above-mentioned various method embodiments, etc. The memory 605 can also be at least one storage device located away from the aforementioned processor 601. As shown, the memory 605 as a computer storage medium can include an operating system, a network communication module, a user interface module, and a battery cell thermal runaway detection application based on a convolutional neural network. Figure 6 As shown, the memory 605 as a computer storage medium can include an operating system, a network communication module, a user interface module, and a battery cell thermal runaway detection application based on a convolutional neural network.

[0151] Specifically, the processor 601 can be used to call the battery cell thermal runaway detection application based on a convolutional neural network stored in the memory 605, and specifically perform the following operations:

[0152] Obtain side view images containing at least two battery cells at a preset time interval, and input each side view image into the trained first convolutional neural network to obtain key point features of each side view image; wherein the first convolutional neural network is trained by a plurality of sample images with known key point features and the second convolutional neural network;

[0153] determine whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, and when detecting that the explosion-proof valve of any one battery cell is offset, input a first pulse signal that periodically changes to the loop in which the low-melting-point resistor arranged on each battery cell is located;

[0154] When detecting that the second pulse signal output by the loop is inconsistent with the first pulse signal, determine the thermal runaway battery cell based on the voltage signals across all low-melting-point resistors.

[0155] In some possible embodiments, the key point features of the side view images include confidence features of the key points;

[0156] After inputting each side view image into the trained first convolutional neural network to obtain the key point features of each side view image, before determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images, the method further includes:

[0157] deleting the key points corresponding to the confidence features of the key points that are lower than a preset confidence threshold from all the key points of each side view image;

[0158] Determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images includes:

[0159] Determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images includes:

[0160] In some possible embodiments, the key point features of the side view images further include offset features of the key points;

[0161] Determining whether the explosion-proof valve of any one battery cell is offset based on the key point features of all side view images includes:

[0162] Determining the vertex key point in each side view image based on the offset features of all the key points in each side view image; wherein the vertex key point and the two adjacent key points are not on the same straight line;

[0163] Taking any two side view images in all side view images that are in adjacent time intervals as a group, and determining whether the offset features of the vertex key points corresponding to the two side view images in each group are consistent;

[0164] When detecting that there are inconsistent offset features of the vertex key points in any one group, determining that the explosion-proof valve of any one battery cell is offset;

[0165] When detecting that there are completely consistent offset features of the vertex key points in any one group, determining that the explosion-proof valve of any one battery cell is not offset.

[0166] In some possible embodiments, when it is detected that there is an inconsistent offset feature of the vertex key point in any one of the groups, it is determined that the explosion-proof valve of any one of the battery cells is offset, including:

[0167] When it is detected that there is an inconsistent offset feature of the vertex key point in any one of the groups, the slope between the inconsistent vertex key point and the adjacent two vertex key points in each side view image is calculated;

[0168] When the number of different slopes contained in the two side view images is consistent, it is determined that the explosion-proof valve of any one of the battery cells is offset.

[0169] In some possible embodiments, after it is detected that the explosion-proof valve of any one of the battery cells is offset, before the first pulse signal with periodic changes is input to the loop in which the low-melting-point resistor arranged on each battery cell is located, the method further includes:

[0170] Based on the temperature sensor arranged above at least two battery cells, a temperature signal corresponding to a preset time interval is acquired;

[0171] The first pulse signal with periodic changes is input to the loop in which the low-melting-point resistor arranged on each battery cell is located, including:

[0172] According to the temperature signal and the preset time interval, a temperature change amount is calculated, and when it is detected that the temperature change amount exceeds a preset temperature threshold, the first pulse signal with periodic changes is input to the loop in which the low-melting-point resistor arranged on each battery cell is located.

[0173] In some possible embodiments, based on the voltage signals across all low-melting-point resistors, a thermal runaway battery cell is determined, including:

[0174] When it is detected that the voltage signal across any one of the low-melting-point resistors is greater than a preset first voltage threshold, and the voltage signals across all other low-melting-point resistors are all less than a preset second voltage threshold, the battery cell corresponding to the low-melting-point resistor greater than the preset first voltage threshold is determined as the thermal runaway battery cell;

[0175] The number corresponding to the low-melting-point resistor greater than the preset first voltage threshold is sent to a mobile terminal.

[0176] In some possible embodiments, the first convolutional neural network includes one hourglass structure, and the second convolutional neural network includes four hourglass structures; the loss function of the first convolutional neural network includes a loss parameter obtained after the second convolutional neural network is trained, and the second convolutional neural network is trained by a plurality of sample images of known key point features.

[0177] The application further provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the steps of the method. The computer readable storage medium can include, but is not limited to, any type of disk, including a floppy disk, an optical disk, a DVD, a CD-ROM, a micro drive, a magneto-optical disk, a ROM, a RAM, an EPROM, an EEPROM, a DRAM, a VRAM, a flash memory device, a magnetic card or an optical card, a nanosystem (including a molecular memory IC), or any type of medium or device suitable for storing instructions and / or data.

[0178] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a series of action combinations, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily necessary for the present application.

[0179] In the above embodiments, the description of each embodiment is focused on, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0180] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented by other means. For example, the device embodiments described above are only illustrative, and the division of units is only a logical function division, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some service interface, device or unit, which can be electrical or other forms.

[0181] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0182] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0183] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable memory. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned memory includes: a U disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0184] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by a program instructing relevant hardware, and the program can be stored in a computer readable memory, and the memory can include a flash disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, etc.

[0185] The above are only exemplary embodiments of the present disclosure, and cannot limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. Those skilled in the art will easily think of embodiments of the present disclosure after considering the specification and practicing the disclosure herein. The present application is intended to cover any variations, uses or adaptive changes of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and examples are only considered as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for detecting thermal runaway of a battery cell based on a convolutional neural network, characterized in that, The method comprises: acquiring side view images containing at least two battery cells at preset time intervals, and inputting each side view image into a trained first convolutional neural network to obtain key point features of each side view image; wherein the first convolutional neural network is trained by a plurality of sample images of known key point features and a second convolutional neural network; judging whether the explosion-proof valve of any one of the battery cells is offset based on the key point features of all the side view images, and inputting a first pulse signal that periodically changes into a loop in which a low-melting-point resistor arranged on each battery cell is located when detecting that the explosion-proof valve of any one of the battery cells is offset; when detecting that a second pulse signal output by the loop is inconsistent with the first pulse signal, determining a thermal runaway battery cell based on voltage signals across all the low-melting-point resistors.

2. The method of claim 1, wherein, The key point features of the side view images include confidence features of key points; after the inputting of each side view image into the trained first convolutional neural network to obtain the key point features of each side view image, before the judging of whether the explosion-proof valve of any one of the battery cells is offset based on the key point features of all the side view images, the method further comprises: deleting key points with confidence features lower than a preset confidence threshold from all the key points of each side view image; the judging of whether the explosion-proof valve of any one of the battery cells is offset based on the key point features of all the side view images comprises: judging whether the explosion-proof valve of any one of the battery cells is offset based on the key point features of all the side view images after the deletion processing.

3. The method of claim 1, wherein, The key point features of the side view images further include offset features of key points; the judging of whether the explosion-proof valve of any one of the battery cells is offset based on the key point features of all the side view images comprises: determining a vertex key point in each side view image based on the offset features of all the key points in each side view image; wherein the vertex key point and two adjacent key points are not on the same straight line; taking any two side view images in all the side view images as a group, and judging whether the offset features of the vertex key points corresponding to the two side view images in each group are consistent; when detecting that the offset features of the vertex key points in any group are inconsistent, determining that the explosion-proof valve of any one of the battery cells is offset; when detecting that the offset features of the vertex key points in any group are completely consistent, determining that the explosion-proof valve of any one of the battery cells is not offset.

4. The method of claim 3, wherein, the determining of that the explosion-proof valve of any one of the battery cells is offset when detecting that the offset features of the vertex key points in any group are inconsistent comprises: when detecting that the offset features of the vertex key points in any group are inconsistent, calculating the slope between the inconsistent vertex key point and two adjacent vertex key points in each side view image; when the number of different slopes contained in the two side view images is consistent, determining that the explosion-proof valve of any one of the battery cells is offset.

5. The method of claim 1, wherein, Before the inputting of the first pulse signal with periodic variation to the loop in which the low-melting-point resistor provided on each of the battery cells is arranged, after the detecting of the deflection of the explosion-proof valve of any one of the battery cells, the method further comprises: acquiring a temperature signal corresponding to the preset time interval based on a temperature sensor arranged above at least two of the battery cells; the inputting of the first pulse signal with periodic variation to the loop in which the low-melting-point resistor provided on each of the battery cells is arranged comprises: calculating a temperature variation amount according to the temperature signal and the preset time interval, and inputting the first pulse signal with periodic variation to the loop in which the low-melting-point resistor provided on each of the battery cells is arranged when it is detected that the temperature variation amount exceeds a preset temperature threshold.

6. The method of claim 1, wherein, the determining of the thermal runaway battery cell based on the voltage signals across all the low-melting-point resistors comprises: when it is detected that the voltage signal across any one of the low-melting-point resistors is greater than a preset first voltage threshold, and the voltage signals across all the other low-melting-point resistors are all less than a preset second voltage threshold, the battery cell corresponding to the low-melting-point resistor greater than the preset first voltage threshold is taken as the thermal runaway battery cell; sending the number corresponding to the low-melting-point resistor greater than the preset first voltage threshold to a mobile terminal.

7. The method of claim 1, wherein, The first convolutional neural network comprises one hourglass structure, and the second convolutional neural network comprises four hourglass structures; a loss function of the first convolutional neural network comprises a loss parameter obtained after the second convolutional neural network is trained, and the second convolutional neural network is trained by a plurality of sample images of known key point features. 8.A device for detecting thermal runaway of a battery cell based on a convolutional neural network, characterized in that, comprise: a feature extraction module configured to acquire side view images containing at least two battery cells according to a preset time interval, and input each of the side view images to a trained first convolutional neural network to obtain key point features of each of the side view images; wherein the first convolutional neural network is trained by a plurality of sample images of known key point features and a second convolutional neural network; a signal detection module configured to determine whether the explosion-proof valve of any one of the battery cells is deflected based on the key point features of all the side view images, and input a first pulse signal with periodic variation to a loop in which a low-melting-point resistor provided on each of the battery cells is arranged when it is detected that the explosion-proof valve of any one of the battery cells is deflected; a fault determination module configured to determine a thermal runaway battery cell based on voltage signals across all the low-melting-point resistors when it is detected that a second pulse signal output by the loop is inconsistent with the first pulse signal. 9.A device for detecting thermal runaway of a battery cell based on a convolutional neural network, characterized in that, comprise a processor and a memory; the processor is connected with the memory; the memory is configured to store executable program codes; the processor runs a program corresponding to the executable program codes by reading the executable program codes stored in the memory, so as to execute the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, the computer readable storage medium stores instructions, when the instructions are run on a computer or a processor, the computer or the processor executes the steps of the method according to any one of claims 1-7.

Images