Gas sampling device and online monitoring and identification equipment for battery thermal runaway gas

Abstract

The present invention belongs to the field of gas detection, and specifically relates to a gas sampling device and an online monitoring and identification device for battery thermal runaway gases. The gas sampling device includes: a detection container, a first valve body, a gas storage container, a circulating air pump, and a sampling manager. The detection container is a container open at both ends; the multi-channel gas detector is located inside the detection container. The first valve body is used to adjust the on-off of the air inlet of the detection container. The circulating air pump includes a first pump body and a second pump body, the first pump body is used to pump the gas in the detection container into the gas storage container, and the second pump body is used to pump the gas in the gas storage container into the detection container. The online monitoring device for battery thermal runaway gases identifies the type of characteristic gas based on the gas-sensitive characteristic parameters in the response recovery curve collected in real time. The present invention solves the problems that the existing technology is difficult to accurately classify lithium battery thermal runaway gases, the gas composition analysis is low in efficiency, high in cost, and has a long detection cycle.

Description

Technical Field

[0001] The present invention belongs to the field of gas detection, and in particular relates to a gas sampling device and online monitoring and identification equipment for battery thermal runaway gas using the gas sampling device. Background Art

[0002] With the continuous development and application of new energy power generation and electric vehicle technologies, the technology and demand for power batteries are also rapidly evolving. Among them, power lithium batteries have become the most widely used battery type due to their advantages such as high energy density, strong high-rate charge and discharge performance, and long cycle life. While power lithium batteries offer superior performance, they also have inherent drawbacks, such as high safety risks. As is well known, lithium batteries contain a large amount of materials that are prone to violent chemical reactions. This can cause lithium batteries to spontaneously combust or even explode when short-circuited or subjected to external impact. Therefore, the development of more advanced lithium battery safety monitoring technologies is becoming one of the most pressing needs in this field.

[0003] When a lithium battery experiences thermal runaway due to factors such as short circuit, impact or puncture, there is a risk of combustion and explosion. However, before a lithium battery burns or explodes, some volatile gases are usually produced due to chemical reactions. These characteristic gases can usually be used as early characteristics to determine whether there are safety risks in the lithium battery, so as to achieve early monitoring of battery failures. For example, for lithium batteries using electrolyte components such as DEC, DMC, and EMC, the components of the early flue gas generally include: electrolyte vapor (80%), H2 (5%), CO (4%), hydrocarbon gas (1%), CO2 (10%), etc. Therefore, by developing a gas sensor that is highly sensitive to the mixed gas of the above components, it is possible to achieve early monitoring of thermal runaway failures of lithium batteries. At present, some specific sensor products have been developed and promoted.

[0004] However, in the field of early detection of lithium battery thermal runaway failures, while existing technologies can achieve high-precision measurement of certain specific battery thermal runaway gas compositions, they are unable to classify thermal runaway gases of different compositions. Existing solutions typically require collecting gas samples and then using instruments such as FTIR to qualitatively or quantitatively analyze the composition and content of the gas samples to ultimately determine the type of thermal runaway gas. This detection method is inefficient, costly, and requires long detection cycles, limiting its application in the field of battery thermal runaway monitoring. Summary of the Invention

[0005] In order to solve the problems that the existing technology is difficult to accurately classify thermal runaway gases in lithium batteries, and the existing gas composition analysis solutions are inefficient, costly, and have long detection cycles, the present invention provides a gas sampling device and an online monitoring and identification device for battery thermal runaway gases using the gas sampling device.

[0006] The present invention is achieved by adopting the following technical solutions:

[0007] A gas sampling device is used in conjunction with a multi-channel gas detector to assist the multi-channel gas detector in obtaining a complete "response-recovery curve" of the target gas when the multi-channel gas detector detects that the concentration of the target gas is higher than a safety threshold. The gas sampling device includes: a detection container, a first valve body, a gas storage container, a circulating air pump, and a sampling manager.

[0008] The detection container is an open-ended container with an air inlet at one end and an air outlet at the other. The center of the container contains a cavity for accommodating the multi-channel gas detector. A first valve, located at the air inlet of the detection container, regulates the flow of air through the container.

[0009] The gas storage container is used to store air in an environment that does not contain thermal runaway characteristic gases, and use it as background gas.

[0010] The circulating air pump includes a first pump body and a second pump body. The first pump body and the second pump body are connected to the gas storage container and the detection container through independent air paths via air pipes. The first pump body is used to pump the gas in the detection container into the gas storage container, and the second pump body is used to pump the gas in the gas storage container into the detection container.

[0011] The sampling manager is electrically connected to the multi-channel gas detector, the first valve body and the circulating air pump; the sampling manager is used to: (i) obtain the monitoring results of the target gas concentration in the multi-channel gas detector. (ii) when the gas concentration of the target gas is lower than the safety threshold, drive the first pump body in the circulating air pump to fill the background gas from the detection container into the gas storage container. (iii) after a preset delay period when the gas concentration of the target gas is higher than the safety threshold, close the first valve body and drive the second pump body in the circulating air pump to continuously inject the background gas in the gas storage container into the detection container, thereby "displacing" the air containing the target gas therein.

[0012] As a further improvement of the present invention, the diameter of the air outlet of the detection container is smaller than that of the air inlet, and a gas one-way valve that only allows gas to be discharged is also provided at the air outlet.

[0013] As a further improvement of the present invention, the detection container is further provided with a T-shaped vent interface communicating with the internal cavity; one end of the T-shaped vent interface is connected to the first pump body, and the other end is connected to the second pump body.

[0014] As a further improvement of the present invention, an air pressure sensor is provided inside the gas storage container, and the air pressure sensor is electrically connected to the sampling manager. The sampling manager obtains the detection result of the air pressure sensor and adjusts the pumping in or out state of the air flow according to the pressure value inside the gas storage container.

[0015] As a further improvement of the present invention, the sampling manager is also used to regularly perform a gas pumping-in and pumping-out cycle at preset time intervals when the gas concentration of the target gas is lower than a safety threshold, so as to clean the inner cavity of the detection container.

[0016] In the present invention, the preset delay period in the sampling manager is an expert experience value determined based on statistical data. The length of the preset delay period is sufficient to enable the response signal of the multi-channel gas detector to any concentration of target gas to rise and then stabilize, and reach a stable response state.

[0017] The present invention also includes an online monitoring and identification device for battery thermal runaway gas, which includes a multi-channel gas detector, a gas sampling device, a data acquisition device, and a gas classification model.

[0018] The multi-channel gas detector is composed of multiple gas sensors of the same model suitable for detecting the target gas. The gas sampling device adopts the gas sampling device mentioned above; the multi-channel gas detector is installed in the central cavity of the gas sampling device.

[0019] The data acquisition device includes a data acquisition unit and a vector generator. The data acquisition unit is electrically connected to the multi-channel gas detector and includes a sampling control unit and a conversion unit. The sampling control unit is used to synchronously acquire the response signals of each channel in the multi-channel gas detector according to a preset sampling frequency. The conversion unit is used to convert the analog quantities corresponding to each response signal into digital quantities and output them. The vector generator is communicatively connected to the data acquisition unit and is used to first fit the discrete data collected by the data acquisition unit into a continuous response recovery curve. The vector generator is then used to extract the corresponding 12 gas-sensitive characteristic parameters based on the response recovery curve within a complete cycle. Finally, all the corresponding gas-sensitive characteristic parameters in each channel are combined into a gas-sensitive characteristic vector U with a length of 12*n, where n represents the number of channels in the multi-channel gas detector.

[0020] The gas classification model uses a trained MLP neural network; the input of the gas classification model is the gas-sensitive feature vector U, and the output is the category of the corresponding characteristic gas.

[0021] As a further improvement of the present invention, the gas-sensitive characteristic parameters extracted from the vector generator include:

[0022] Response curve offset γres , recovery curve offset γ rec , the fitting constant C of the response curve res , the fitting constant C of the recovery curve rec , time constant τ of the response curve res , time constant τ of the recovery curve rec Response time T res , maximum response rate V res , response curve area A res , recovery time T rec , maximum recovery rate V rec , and the recovery curve area A rec .

[0023] As a further improvement of the present invention, the complete response recovery curve fitted in the vector generator includes a response curve segment, a response stability curve segment, a recovery curve segment, and a recovery stability curve segment. Among them, the response curve segment and the recovery curve segment are both fitted using an exponential function with a natural constant exp as the base. The response curve G res and recovery curve G rec The function expressions are:

[0024]

[0025] In the online monitoring and identification device for battery thermal runaway gas provided by the present invention, the monitoring and identification process is as follows:

[0026] The multi-channel gas detector installed in the detection container monitors the real-time concentration of characteristic gases of battery thermal runaway in the environment in real time.

[0027] When the real-time concentration of the characteristic gas is lower than the safety threshold, a sufficient amount of ambient gas is filled into the gas storage container of the gas sampling device as background gas.

[0028] When the concentration of the characteristic gas in the environment is detected to be higher than the safety threshold, a preset delay period is waited, and then the air inlet of the detection container is closed, and the background gas in the gas storage container is injected into the detection container to replace the air containing the characteristic gas.

[0029] The data acquisition device generates a complete response recovery curve according to the detection signal of the multi-channel gas detector, extracts the gas-sensitive characteristic parameters therein, and constructs a gas-sensitive characteristic vector.

[0030] The gas classification model identifies the category corresponding to the characteristic gas based on the input gas-sensitive feature vector.

[0031] The technical solution provided by the present invention has the following beneficial effects:

[0032] The gas sampling device provided by the present invention can flexibly adjust the gas type on the detector's periphery according to the detection results of the multi-channel gas detector, thereby assisting the multi-channel gas detector to quickly complete the generation of the gas response recovery curve when the characteristic gas concentration is abnormal, thereby laying the foundation for the subsequent classification of the gas components.

[0033] Utilizing the designed gas sampling device, the present invention further designs a solution for online monitoring and rapid identification of battery thermal runaway gases in the environment. This solution uses a traditional gas sensor to form a multi-channel gas detector that can synchronously acquire multiple detection signals, thereby achieving real-time monitoring and synchronous measurement of the concentration of lithium battery thermal runaway gases. The monitoring signal is then used to promptly generate a response recovery curve, and the gas-sensitive characteristic parameters in the response recovery curve are extracted. Finally, the gas-sensitive characteristic vector composed of the gas-sensitive characteristic parameters is input into a pre-trained artificial neural network to learn the differences in the response and recovery characteristics of different gas sensors to the same battery thermal runaway sample, and thereby analyze the category of the gas to be detected.

[0034] The solution provided by the present invention provides a new approach for the detection and analysis of battery thermal runaway gases. The solution has low detection cost and high detection accuracy, and is therefore suitable for promotion and application, which can generate considerable economic value and good social benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flowchart of the steps of a method for detecting and identifying thermal runaway gas provided in Example 1 of the present invention.

[0036] Figure 2 This is a signal flow diagram of the working process of the multi-channel gas detector in Example 1 of the present invention.

[0037] Figure 3 This is a flowchart of obtaining the measurement results of a complete measurement cycle in Example 1 of the present invention;

[0038] Figure 4 This is the response recovery curve of a typical 10-channel gas detector in Example 1 of the present invention.

[0039] Figure 5 It is the response recovery curve of any gas sensor in the multi-channel gas detector.

[0040] Figure 6 This is a network architecture diagram of the MLP neural network created for training the gas classification model in Example 1 of the present invention.

[0041] Figure 7 This is a system module diagram of a lithium battery thermal runaway gas detection system provided in Example 2 of the present invention.

[0042] Figure 8 This is a schematic structural diagram of a gas sampling device in an online monitoring and identification device for battery thermal runaway gas provided in Example 4 of the present invention.

[0043] Figure 9 This is a schematic diagram of the structure of the detection container in Example 4 of the present invention.

[0044] Figure 10 This is a workflow diagram of the online monitoring and identification device for battery thermal runaway gas in Example 4 of the present invention.

[0045] The following are marked in the figure:

[0046] 1. Detection container; 2. Gas storage container; 3. Circulating air pump; 10. Cavity; 11. Air inlet; 12. Air outlet; 13. T-shaped ventilation interface; 31. First pump body; 32. Second pump body. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0048] Example 1

[0049] This embodiment provides a method for detecting and identifying thermal runaway gases based on gas-sensitive characteristic parameters. The idea of ​​this detection and identification method is: using a multi-channel gas detector to synchronously detect the same target gas, then using an artificial neural network to learn the statistical laws existing between the differences in detection signals in different channels, and finally using the artificial neural network to identify the category of the target gas based on the gas-sensitive characteristic parameters selected from the detection signals of each channel.

[0050] The following describes the solution of this embodiment based on the category of lithium batteries. Specifically, Figure 1 As shown, the method for detecting and identifying thermal runaway gas provided in this embodiment includes the following steps:

[0051] S1: Select a gas sensor that can be used for multiple types of target gases to be identified.

[0052] Among the characteristic gases that need to be distinguished in this embodiment, thermal runaway gas is a mixed gas containing multiple components, the main components of which are electrolyte vapor, CO gas, H2 gas, hydrocarbon gas (C x H y) etc. The main components of the electrolyte vapor are dimethyl carbonate (DMC), diethyl carbonate (DEC), ethyl methyl carbonate (EMC), etc. Targeting this scenario, this embodiment specifically designs and employs a resistive semiconductor gas sensor using a ternary compound of TiO2, SnO2, and ZnO as the sensitive material. This sensor has excellent gas-sensing properties for the thermal runaway characteristic gases of lithium iron phosphate batteries, ternary lithium batteries, lithium cobalt oxide batteries, and lithium titanate batteries that need to be distinguished in this embodiment.

[0053] The ternary composite semiconductor gas sensor used in this example is fabricated by first preparing a nanowire array material using TiO2, SnO2, and ZnO in a mass ratio of 1:5:4. Al2O3 is then used as a substrate and the nanowire array material as a sensitive layer to fabricate a gas sensor for detecting gas concentrations associated with battery thermal runaway.

[0054] S2: Design a multi-channel gas detector with n channels using n similar gas sensors. A multi-channel gas detector is a sensor array containing multiple gas sensor units that can simultaneously detect gas concentrations in the target environment through different channels.

[0055] like Figure 2 As shown, the mixed gas is passed into the multi-channel gas detector. The gas-sensitive material in the sensor unit contacts the gas to produce an oxidation-reduction reaction. The gas-sensitive material produces a resistance change and enters the multi-channel ADC acquisition circuit of the microprocessor in the form of voltage measurement. The microprocessor samples the voltage and converts it into binary data, which is then cached in the microprocessor as hexadecimal data. After the collected signal is processed by low-pass filtering and average value filtering, the microprocessor sends the collected detection value to the back end.

[0056] S3: Perform a gas-sensing test on the target gas through a multi-channel gas detector, and generate a corresponding gas-sensing characteristic vector U according to the test results. In this embodiment, the method for generating the gas-sensing characteristic vector U is as follows:

[0057] S31: Place the multi-channel gas detector into the measurement space, inject the target gas into the measurement space, and obtain the measurement results. Figure 3 As shown in Figure 2, the complete process of obtaining the measurement results is:

[0058] (1) Rapidly inject the target gas meeting the preset concentration into the measurement space and start measurement.

[0059] (2) Synchronously obtain the detection value of each gas sensor in the multi-channel gas detector according to the preset sampling frequency.

[0060] (3) After the detection values ​​of all gas sensors tend to be stable, inject background gas into the measurement space to exhaust the target gas.

[0061] (4) Continue to obtain the detection value of each gas sensor until the detection values ​​of all gas sensors tend to be stable again; end the measurement.

[0062] (5) Record a series of discrete detection values ​​corresponding to each channel during the entire measurement cycle, which is the required measurement result.

[0063] In this embodiment, the preset sampling frequency is 1 Hz / s, and the complete sampling period is greater than 250 seconds. During a complete sampling period, the entire process of each gas sensor in the multi-channel gas detector, from initial response to stabilization, recovery to initial state, and then stabilization again, can be captured. During a complete sampling period, no fewer than 250 sampling data points can be obtained.

[0064] S32: Fitting a complete response recovery curve of each channel in the current test cycle according to the measurement results.

[0065] The complete response recovery curve fitted in this embodiment includes a response curve segment, a response stability curve segment, a recovery curve segment, and a recovery stability curve segment. Among them, the response curve segment and the recovery curve segment are both fitted using an exponential function with the natural constant exp as the base. res and recovery curve G rec The function expressions are:

[0066]

[0067] In the above formula, γ res Indicates the offset of the response curve; γ rec Indicates the offset of the recovery curve; C res represents the fitting constant of the response curve; C rec represents the fitting constant of the recovery curve; τ res Represents the time constant of the response curve; τ rec Represents the time constant of the recovery curve.

[0068] For example, in a typical 10-channel gas detector, the gas response recovery curve finally fitted by the detection signals of different channels is roughly as follows: Figure 4 As shown. Extract Figure 4 The response recovery curve generated in any channel of Figure 5 As shown. Figure 5As can be seen in the figure, a complete response recovery curve includes an initial segment before the characteristic gas is introduced, during which the curve remains stable. This segment is followed by a descending segment from the beginning of the response to the end of the response and stabilization, known as the response curve. This segment then includes a stabilization segment, during which the curve regains stability. This segment continues with the recovery curve after the characteristic gas is expelled, which is an ascending segment. Finally, after the gas sensor fully recovers, the signal stabilizes again, reaching the recovery stabilization stage.

[0069] exist Figure 5 The period of the falling edge in a complete cycle is called the response time T res ; The period of the rising segment is called the recovery time T res The derivative of the descending section is called the response rate, and the maximum response rate V res It can reflect the differential characteristics of the response curve; the derivative of the rising section is called the recovery rate, and the maximum recovery rate V res It can reflect the differential characteristics of the recovery curve; the integral of the descending section is called the response curve area A res , which is used to reflect the integral characteristics of the response curve; the integral of the rising section is called the recovery curve area A rec , which is used to reflect the integral characteristics of the response curve.

[0070] S33: Extract 12 gas-sensing characteristic parameters corresponding to each response recovery curve and use them as characteristic values ​​in the expert knowledge base, as shown in Table 1:

[0071] Table 1: Feature parameter table in expert knowledge base

[0072]

[0073] Among them, parameters 1 to 6 can be directly determined by the functional relationship between the response curve and the recovery curve obtained by response fitting, while among parameters 7 to 12, the maximum response rate V res , Maximum recovery rate V rec Response time T res and recovery time T rec The calculation formula is as follows:

[0074]

[0075] In the above formula, G r ' es (t) represents the derivative of the response curve; G r ' ec (t) represents the derivative of the recovery curve; max{} represents the maximum value function; t1 represents the starting time of the response curve; t2 represents the ending time of the response curve; t3 represents the starting time of the recovery curve; t4 represents the ending time of the recovery curve.

[0076] Response curve area A res and the recovery curve area A rec The calculation formula is:

[0077]

[0078] S34: Combining the gas-sensitive characteristic parameters extracted from the n channels into a gas-sensitive characteristic vector U with a length of 12*n.

[0079] In this embodiment, the 12 gas-sensing characteristic parameters in each channel can form a short vector with a length of 12. By connecting the short vectors obtained from all channel signals, a long vector with a length of 12*n can be obtained, which is the required gas-sensing characteristic vector U. For example, when the number of channels is 10, the length of the gas-sensing characteristic vector is 120.

[0080] S4: Obtain the gas-sensitive characteristic vector U of each type of target gas under different concentration conditions to form a sample data set.

[0081] In the sample data set obtained in this embodiment, each gas-sensitive feature vector U is generated by a multi-channel gas detector in the measurement space according to the target gas samples of different concentrations. Figure 3 For samples of the same type of gas with different concentrations, the concentration gradient test is used to complete the measurement. For example, in a closed experimental chamber with a volume of 1L, 1ml, 2ml..., 100ml of sample gas are injected in sequence, and the cycle is performed after each injection. Figure 3 The steps of "injecting sample gas - curve stabilization - exhausting sample gas - curve stabilization again" are then repeated. The collected signal is then fitted into a continuous response recovery curve, and 12 gas-sensitive characteristic parameters are ultimately extracted from the response recovery curve. Finally, the gas-sensitive characteristic parameters in all channels are connected to form the required gas-sensitive characteristic vector U.

[0082] In this embodiment, when establishing a sample dataset, the target gas samples are sourced from two methods: actual collection and manual configuration. "Actual collection" refers to manually triggering thermal runaway of a specified type of battery, such as performing a puncture test on a lithium battery, and then collecting sufficient amounts of characteristic gases for subsequent testing. "Manual configuration" refers to manually configuring a gas mixture based on the known composition and ratio of the characteristic gases of a specified type of battery during thermal runaway.

[0083] Of the two approaches described above, the first gas sample collection method offers the advantage of more accurate sample data, thus improving the generalization of the resulting trained artificial neural network. However, its disadvantage is significantly higher sample data acquisition costs. The second gas sample collection method is significantly more economical, but slightly less accurate. However, in practice, it can still meet the detection accuracy requirements and is therefore generally preferred.

[0084] S5: Divide the sample data into a training set and a test set, and add a category label representing the target gas type to each gas-sensitive feature vector U in the training set according to the data acquisition conditions.

[0085] During the test process of this embodiment, for the selected lithium iron phosphate battery (lithium battery 1), ternary lithium battery (lithium battery 2), lithium cobalt oxide battery (lithium battery 3), and lithium titanate battery (lithium battery 4), the actual metadata in the sample data set collected was 2000 groups, 500 groups for each lithium battery type. Among them, 400 groups were randomly selected from the sample data set of each type of battery as the training set, and the remaining 100 groups were used as the test set; the data volume of the training set and the test set accounted for 80.5% and 20% respectively. Among them, the metadata in the training set required an additional category label "category".

[0086] It's important to emphasize that in this example's sample dataset, the number of metadata with the same category label is identical, and the test concentration range of the target gas samples corresponding to each metadata with the same category label covers the full range of the gas sensor for the current characteristic gas. This has a significant positive impact on improving the generalization and robustness of the trained artificial neural network.

[0087] S6: Construct an MLP neural network consisting of an input layer, a hidden layer, and an output layer; the number of nodes in the input layer is 12*n, and the number of nodes in the output layer is the number of categories m of the target gas classification results; the structure of the hidden layer (the number of layers and the number of nodes in each layer) is adaptively adjusted according to the training effect.

[0088] For the task of identifying thermal runaway gases generated by four different types of lithium batteries, the structure of the MLP neural network finally established in this embodiment is as follows: Figure 6 As shown in the figure, the input layer has 120 nodes, corresponding to each data value in the gas-sensitive feature vector (length 10*12=120) generated by the 10-channel gas detector. The output layer has 4 nodes, meaning this embodiment constructs a 4-class artificial neural network. After repeated testing and refinement, to achieve the highest classification accuracy, the number of hidden layers was determined to be two, with 32 nodes in the first layer and 16 nodes in the second layer. Each node in each layer is activated using an activation function.

[0089] S7: Use the training set and the test set to train and test the MLP neural network, retain the model parameters of the MLP neural network that meet the training objectives, and obtain the required gas classification model.

[0090] S8: Use a multi-channel gas detector to monitor the concentration of characteristic gases in the test space in real time. Then, generate the corresponding gas-sensitive characteristic vector U based on the monitoring signal and input it into the gas classification model to obtain the classification result of the thermal runaway gas.

[0091] In the actual test process of this embodiment, after the training set training in step S7, the embodiment obtains Figure 6 The final classification accuracy achieved by the MLP neural network on the test set is shown in the following table:

[0092] Table 2: Classification and recognition accuracy of MLP neural network

[0093]

[0094] Combined with the data in Table 2, it can be seen that the MLP neural network designed and trained using the gas-sensitive feature vectors of this embodiment has an overall accuracy rate of 99.4% when classifying and identifying mixed gases in lithium battery thermal runaway. This is a very high recognition accuracy. Furthermore, for the characteristic gases of lithium batteries No. 2 and No. 4, the recognition accuracy rate can even reach 100%. This shows that the method for detecting and identifying thermal runaway gases based on gas-sensitive feature parameters provided in this embodiment can achieve good recognition results in the task of identifying the type of thermal runaway gases in lithium batteries, reaching a level that can be used in practical applications.

[0095] Example 2

[0096] This embodiment provides a detection system for lithium battery thermal runaway gas. The detection system adopts a thermal runaway gas detection and identification method based on gas-sensitive characteristic parameters as in Example 1 to detect the concentration of characteristic gas generated by lithium battery thermal runaway and quickly identify the type of characteristic gas based on the detection results.

[0097] like Figure 7 As shown, the detection system includes: a multi-channel gas detector, a data collector, a vector generator, and a gas classification model.

[0098] The multi-channel gas detector is composed of n gas sensors of the same model that can simultaneously detect multiple types of target gases; n represents the number of channels of the multi-channel gas detector. The multi-channel gas detector of this embodiment is the product in Example 1.

[0099] The data collector is electrically connected to the multi-channel gas detector; the data collector includes a sampling control unit and a conversion unit; the sampling control unit is used to synchronously obtain the response signal of each channel in the multi-channel gas detector according to a preset sampling frequency; the conversion unit is used to convert the analog quantity corresponding to each response signal collected into a digital quantity and output it.

[0100] In this embodiment, the control unit in the data collector can be selected from various existing acquisition card products and programmed according to actual sampling requirements. The conversion unit is composed of existing analog-to-digital converters (ADCs) and filters.

[0101] The vector generator is connected to the data acquisition device via a serial or parallel port. It first fits the discrete data collected by the data acquisition device into a continuous response recovery curve. Then, based on the waveform of a complete response recovery cycle of the response recovery curve, the corresponding 12 gas-sensing characteristic parameters are extracted. Finally, all corresponding gas-sensing characteristic parameters in each channel are combined into a gas-sensing characteristic vector U with a length of 12*n.

[0102] The vector generator in this embodiment is actually a data processing module. The data processing tasks of this module are mainly two points. One is to fit the required response recovery curve based on discrete data. The other is to extract the required 12 gas-sensitive characteristic parameters from the function of the fitted response recovery curve using a preset program, and encode all the gas-sensitive characteristic parameters into a gas-sensitive characteristic vector U with a length of 12*n according to the preset encoding rules.

[0103] The gas classification model uses an MLP neural network trained with the strategy of the method for detecting and identifying thermal runaway gases based on gas-sensitive characteristic parameters in Example 1; the input of the gas classification model is the gas-sensitive characteristic vector U, and the output is the category of the corresponding characteristic gas.

[0104] In the lithium battery thermal runaway gas detection system provided in this embodiment, after the technicians extract the thermal runaway gas, they use a multi-channel gas detector to complete a complete "response-recovery" cycle for it, thereby establishing a response recovery curve and realizing the classification of the target characteristic gas.

[0105] Example 3

[0106] This embodiment provides a device for analyzing lithium battery thermal runaway gas, comprising a memory, a processor, and a computer program stored in the memory and executable by the processor. When the processor executes the computer program, it creates the components corresponding to the vector generator and gas classification model in the lithium battery thermal runaway gas detection system described in Example 2. Furthermore, it generates a classification result for the thermal runaway gas based on detection signals obtained by a front-end multi-channel gas detector, etc.

[0107] The lithium battery thermal runaway gas detection device provided in this embodiment is essentially a computer device. The computer device can be a program-executing smart terminal, tablet computer, laptop computer, desktop computer, rack-mounted server, blade server, tower server, or cabinet-mounted server (including a standalone server or a server cluster consisting of multiple servers). The computer device of this embodiment includes at least, but is not limited to, a memory and a processor that can communicate with each other via a system bus.

[0108] In this embodiment, the memory (i.e., readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, magnetic disk, optical disk, etc. In some embodiments, the memory may be an internal storage unit of a computer device, such as the hard disk or internal memory of the computer device. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk equipped with the computer device, a smart media card (SMC), a secure digital (SD) card, a flash memory card, etc. Of course, the memory may also include both the internal storage unit of the computer device and its external storage device. In this embodiment, the memory is generally used to store the operating system and various application software installed on the computer device. In addition, the memory may also be used to temporarily store various types of data that have been output or are about to be output.

[0109] In some embodiments, the processor may be a central processing unit (CPU), a graphics processing unit (GPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is generally used to control the overall operation of the computer device. In this embodiment, the processor is used to run program code stored in the memory or process data.

[0110] Example 4

[0111] Based on the solution in Example 2, this embodiment further provides an online monitoring and identification device for battery thermal runaway gas, which includes a multi-channel gas detector, a gas sampling device, a data acquisition device, and a gas classification model.

[0112] The multi-channel gas detector is composed of multiple gas sensors of the same model suitable for detecting target gases. The gas sampling device adopts the gas sampling device mentioned above; the multi-channel gas detector is installed in the cavity 10 in the center of the gas sampling device.

[0113] The data acquisition device includes a data acquisition unit and a vector generator. The data acquisition unit is electrically connected to the multi-channel gas detector and includes a sampling control unit and a conversion unit. The sampling control unit is used to synchronously acquire the response signals of each channel in the multi-channel gas detector according to a preset sampling frequency. The conversion unit is used to convert the analog quantities corresponding to each response signal into digital quantities and output them. The vector generator is communicatively connected to the data acquisition unit and is used to first fit the discrete data collected by the data acquisition unit into a continuous response recovery curve. The vector generator is then used to extract the corresponding 12 gas-sensitive characteristic parameters based on the response recovery curve within a complete cycle. Finally, all the corresponding gas-sensitive characteristic parameters in each channel are combined into a gas-sensitive characteristic vector U with a length of 12*n, where n represents the number of channels in the multi-channel gas detector.

[0114] The gas classification model uses a trained MLP neural network; the input of the gas classification model is the gas-sensitive feature vector U, and the output is the category of the corresponding characteristic gas.

[0115] Comparing the online monitoring and identification device for battery thermal runaway gases in this embodiment with the lithium battery thermal runaway gas detection system in Example 2, the main difference between the two is that this embodiment adds a gas sampling device to the solution in Example 2. This gas sampling device is primarily designed to work in conjunction with a multi-channel gas detector. When the multi-channel gas detector detects a target gas concentration above a safety threshold, it assists the multi-channel gas detector in obtaining a complete "response-recovery curve" for the target gas.

[0116] like Figure 8 As shown, the gas sampling device provided in this embodiment includes: a detection container 1, a first valve body, a gas storage container 2, a circulating air pump 3, and a sampling manager.

[0117] Among them, Figure 9 As shown, the detection container 1 is a container open at both ends; one end of the detection container 1 is an air inlet 11, and the other end is an air outlet 12. The center of the detection container 1 contains a cavity 10 for accommodating the multi-channel gas detector. A first valve body is located at the air inlet 11 of the detection container 1 and is used to adjust the opening and closing of the air inlet 11 of the detection container 1. The air outlet 12 of the detection container 1 has a smaller diameter than the air inlet 11 and is also equipped with a gas check valve to allow only gas to be discharged.

[0118] The gas storage container 2 is used to store air in an environment that does not contain the characteristic gas of thermal runaway, and uses it as background gas. In this embodiment, a pressure sensor is provided inside the gas storage container 2, and the pressure sensor is electrically connected to the sampling manager.

[0119] The circulating air pump 3 includes a first pump body 31 and a second pump body 32. The first pump body 31 and the second pump body 32 are connected to the gas storage container 2 and the detection container 1 through independent air paths through air pipes. The first pump body 31 is used to pump the gas in the detection container 1 into the gas storage container 2, and the second pump body 32 is used to pump the gas in the gas storage container 2 into the detection container 1. Figure 8 and 9 It can be seen that the detection container 1 of this embodiment is further provided with a T-shaped vent interface 13 communicating with the internal cavity 10 ; one end of the T-shaped vent interface 13 is connected to the first pump body 31 , and the other end is connected to the second pump body 32 .

[0120] The sampling manager is electrically connected to the multi-channel gas detector, the first valve body and the circulating air pump 3; the sampling manager is used to: (i) obtain the monitoring results of the target gas concentration in the multi-channel gas detector. (ii) when the gas concentration of the target gas is lower than the safety threshold, drive the first pump body 31 in the circulating air pump 3 to fill the background gas from the detection container 1 into the gas storage container 2. (iii) after a preset delay period at the moment when the gas concentration of the target gas is higher than the safety threshold, close the first valve body and drive the second pump body 32 in the circulating air pump 3 to continuously inject the background gas in the gas storage container 2 into the detection container 1 to "replace" the air containing the target gas therein.

[0121] In the gas sampling device provided in this embodiment, the sampling manager obtains the detection results of the air pressure sensor in real time and adjusts the pumping state of the airflow based on the pressure value inside the gas storage container 2. For example, when the air pressure in the gas storage container 2 is too low, the sampling manager will quickly pump gas into the gas storage container 2 when the concentration of the target gas falls below a safety threshold, ensuring that a sufficient amount of background gas is always pre-stored in the gas storage container 2. In addition, the sampling manager also obtains the rate of change of the internal air pressure in real time while pumping gas from the gas storage container 2, thereby calculating the flow rate of the pumped-out background gas.

[0122] Because the cavity 10 within the detection container 1 of the sampling manager provided in this embodiment is open-ended, dust easily accumulates during long-term operation, thereby affecting the sensitivity of the internal multi-channel gas detector and the normal operation of the device. Therefore, in a more optimized embodiment of the present invention, the sampling manager is further configured to periodically perform a gas pumping cycle at preset intervals when the target gas concentration is below a safety threshold, thereby automatically cleaning the inner cavity of the detection container 1.

[0123] The online monitoring and identification device for thermal runaway gas provided in this embodiment has a special gas sampler added compared to the detection system for thermal runaway gas of lithium battery in Example 2. In the detection system for thermal runaway gas of lithium battery in Example 2, the probe of the multi-channel gas detector is directly exposed to the detection environment, while the probe of the multi-channel gas detector in the online monitoring system for thermal runaway gas provided in this embodiment is installed in the environment such as Figure 9 The difference in operating logic between the two is that in the normal working mode, the two ends of the detection container 1 in this embodiment are connected, and the detection probe is directly exposed to the air. At this time, the operating state of the lithium battery thermal runaway gas detection system in this embodiment and Example 3 is exactly the same, that is, in the "monitoring mode".

[0124] However, when the concentration of thermal runaway gas detected by the multi-channel gas detector exceeds a threshold, or a condition is triggered requiring classification of the thermal runaway gas, the online thermal runaway gas monitoring system provided in this embodiment automatically enters "classification mode." In classification mode, the sampling manager waits for a preset delay period to ensure that the multi-channel gas detector can respond to the "response curve" and "response stability curve" portions of the recovery curve. It then closes the first valve body at the air inlet 11 of the detection container 1 and pumps the "background gas" pre-stored in the gas storage container 2 into the detection container 1 via the second pump body 32. The background gas is the air in the current detection environment before thermal runaway gas is generated. The background gas will quickly clear the thermal runaway gas in the detection container 1 and gradually replace it with "clean" air, thereby restoring the gas sensors in the multi-channel gas detector to their initial state and generating the "recovery curve" and "recovery stability curve" portions of the response recovery curve, thereby forming a complete response recovery curve. Using the waveform of the completed response recovery curve, the corresponding gas-sensitive characteristic parameters can be extracted and the required gas-sensitive characteristic vector U can be constructed.

[0125] Specifically, if Figure 10 As shown, the working process of the online monitoring and identification device for battery thermal runaway gas provided in this embodiment is as follows:

[0126] First, the real-time concentration of characteristic gases of battery thermal runaway in the environment is monitored in real time by a multi-channel gas detector installed in the detection container 1 .

[0127] Then, when the real-time concentration of the characteristic gas is lower than the safety threshold, a sufficient amount of ambient gas is filled into the gas storage container 2 of the gas sampling device as background gas.

[0128] Next, after detecting that the concentration of the characteristic gas in the environment is above the safety threshold, a preset delay period is waited to allow the data acquisition device to plot the first half of the response recovery curve. Then, the air inlet 11 of the detection container 1 is closed, and background gas from the gas storage container 2 is injected into the detection container 1 to displace the air contained therein with the characteristic gas. This allows the data acquisition device to plot the second half of the response recovery curve.

[0129] After that, the data acquisition device generates a complete response recovery curve based on the detection signal of the multi-channel gas detector, extracts the gas-sensitive characteristic parameters therein, and constructs a gas-sensitive characteristic vector.

[0130] Finally, the gas classification model identifies the category corresponding to the characteristic gas based on the input gas-sensitive feature vector.

[0131] At this point, this embodiment, based on the real-time detection of thermal runaway gas concentration, can also automatically complete rapid sampling and real-time "component analysis" of the monitored thermal runaway gas, and output the category of the identified thermal runaway characteristic gas to facilitate subsequent risk prevention and control.

[0132] In practical applications, the online monitoring and identification device for battery thermal runaway gas provided in this embodiment can be deployed in places such as electric vehicles, charging stations, and energy storage power stations that use power lithium batteries. It can be linked with the safety monitoring and fire protection systems in the above-mentioned systems to detect the occurrence of battery thermal runaway accidents and promptly identify the type of faulty battery in the early stages of the accident to ensure the safety of the above-mentioned places or facilities.

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A gas sampling device, characterized in that: It is used in conjunction with a multi-channel gas detector to assist the multi-channel gas detector in obtaining a complete "response-recovery curve" of the target gas when the multi-channel gas detector detects that the concentration of the target gas is higher than a safety threshold. The gas sampling device includes: The detection container is a container open at both ends; one end of the detection container is an air inlet, and the other end is an air outlet; the center of the detection container contains a cavity for accommodating a multi-channel gas detector; a first valve body, located at the air inlet of the detection container, for regulating the on / off state of the air inlet of the detection container; A gas storage container, which is used to store air in an environment that does not contain thermal runaway characteristic gases, and uses it as background gas; A circulating air pump comprising a first pump body and a second pump body, wherein the first pump body and the second pump body are connected to a gas storage container and a detection container respectively through air pipes in independent air paths; the first pump body is used to pump the gas in the detection container into the gas storage container, and the second pump body is used to pump the gas in the gas storage container into the detection container; A sampling manager is electrically connected to the multi-channel gas detector, the first valve body and the circulating air pump; the sampling manager is used to: (i) obtain the monitoring result of the target gas concentration in the multi-channel gas detector; (ii) when the gas concentration of the target gas is lower than a safety threshold, drive the first pump body in the circulating air pump to fill the background gas from the detection container into the gas storage container; (iii) after a preset delay period when the gas concentration of the target gas is higher than the safety threshold, close the first valve body and drive the second pump body in the circulating air pump to continuously inject the background gas in the gas storage container into the detection container to "replace" the air containing the target gas therein.

2. The gas sampling device according to claim 1, wherein: The diameter of the air outlet of the detection container is smaller than that of the air inlet, and a gas one-way valve that only allows gas to be discharged is also provided at the air outlet.

3. The gas sampling device according to claim 1, wherein: The detection container is further provided with a T-shaped ventilation interface connected to the internal cavity; one end of the T-shaped ventilation interface is connected to the first pump body, and the other end is connected to the second pump body.

4. The gas sampling device according to claim 1, wherein: An air pressure sensor is provided inside the gas storage container, and the air pressure sensor is electrically connected to the sampling manager. The sampling manager obtains the detection result of the air pressure sensor and adjusts the pumping state of the air flow according to the pressure value inside the gas storage container.

5. The gas sampling device according to claim 1, wherein: The sampling manager is also used to periodically perform a gas pumping-in and pumping-out cycle at preset time intervals when the gas concentration of the target gas is lower than a safety threshold, so as to clean the inner cavity of the detection container.

6. The gas sampling device according to claim 1, wherein: The preset delay period in the sampling manager is an expert experience value determined based on statistical data. The duration of the preset delay period is sufficient to allow the response signal of the multi-channel gas detector to any concentration of target gas to rise and then stabilize, and reach a stable response state.

7. An online monitoring and identification device for battery thermal runaway gas, characterized in that: It includes: Multi-channel gas detector, which is composed of multiple gas sensors of the same model and suitable for detecting target gases; A gas sampling device, comprising the gas sampling device according to any one of claims 1 to 6; wherein the multi-channel gas detector is installed in a cavity in the center of the gas sampling device; A data acquisition device, comprising a data acquisition unit and a vector generator; the data acquisition unit is electrically connected to the multi-channel gas detector; the data acquisition unit comprises a sampling control unit and a conversion unit; the sampling control unit is used to synchronously acquire the response signal of each channel in the multi-channel gas detector according to a preset sampling frequency; the conversion unit is used to convert the analog quantity corresponding to each response signal collected into a digital quantity and output it; the vector generator is communicatively connected to the data acquisition unit, and the vector generator is used to first fit the discrete data collected by the data acquisition unit into a continuous response recovery curve; then, according to the waveform diagram of a complete response recovery cycle of the response recovery curve, the corresponding 12 gas-sensitive characteristic parameters are extracted; finally, all the corresponding gas-sensitive characteristic parameters in each channel are combined into a gas-sensitive characteristic vector U with a length of 12*n; wherein n represents the number of channels of the multi-channel gas detector; The gas classification model adopts a trained MLP neural network; the input of the gas classification model is the gas-sensitive feature vector U, and the output is the category of the corresponding characteristic gas.

8. The online monitoring and identification device for battery thermal runaway gas according to claim 7, characterized in that: The complete response recovery curve fitted by the vector generator includes a response curve segment, a response stability curve segment, a recovery curve segment, and a recovery stability curve segment; wherein the response curve segment and the recovery curve segment are both fitted using an exponential function with a natural constant exp as the base, and the response curve G res and recovery curve G rec The function expressions are: ； The gas-sensitive characteristic parameters extracted by the vector generator include: Response curve offset , recovery curve offset , fitting constant of the response curve C res , fitting constant of the recovery curve C rec , time constant of the response curve , time constant of the recovery curve , response time T res , maximum response rate V res , the area under the response curve A res Recovery time T rec , maximum recovery rate V rec , and the area under the recovery curve A rec .

9. The online monitoring and identification device for battery thermal runaway gas according to claim 7, characterized in that: The monitoring and identification process is as follows: The multi-channel gas detector installed in the detection container monitors the real-time concentration of characteristic gases of battery thermal runaway in the environment in real time; When the real-time concentration of the characteristic gas is lower than the safety threshold, a sufficient amount of ambient gas is filled into the gas storage container of the gas sampling device as background gas; When the concentration of the characteristic gas in the environment is detected to be higher than the safety threshold, a preset delay period is waited, and then the air inlet of the detection container is closed, and the background gas in the gas storage container is injected into the detection container to replace the air containing the characteristic gas; The data acquisition device generates a complete response recovery curve based on the detection signal of the multi-channel gas detector, extracts the gas-sensitive characteristic parameters therein, and constructs a gas-sensitive characteristic vector; The gas classification model identifies the category corresponding to the characteristic gas according to the input gas-sensitive feature vector.

Images