New energy automobile battery online intelligent fault diagnosis method

Abstract

According to the new energy automobile battery on-line intelligent fault diagnosis method provided by the invention, the gas sensing technology is used for carrying out early detection and early warning on thermal runaway, fire hazards are eliminated before the thermal runaway spreads, and the method has great potential in the aspect of improving safety; when flammable and explosive gas is discharged, real-time monitoring and timely alarming are carried out.

Description

Technical Field

[0001] The present invention relates to the field of battery safety technology, and in particular to an online intelligent fault diagnosis method for a new energy vehicle battery. Background Art

[0002] At present, in view of the fact that the current, voltage, internal resistance, internal pressure and surface temperature and other signals will show obvious changes during the thermal runaway process of the battery, the battery status monitoring and thermal runaway warning system mainly performs real-time monitoring and warning based on the critical conditions of these signals. However, there are limitations in early warning of thermal runaway of lithium-ion batteries by monitoring these signals. Traditional temperature and voltage sensors detect the external temperature and terminal voltage of the battery, but these parameters change little in the early stage of thermal runaway, making it difficult to achieve early warning of thermal runaway behavior. Emerging methods such as implantable temperature sensors and electrochemical AC impedance spectroscopy testing can provide the temperature conditions inside the lithium-ion battery cell, but their high cost makes it difficult to achieve commercial application at present. Moreover, in large battery modules, the voltage and temperature changes of single cells are very small, making it difficult to be detected before thermal runaway spreads to other batteries, and detecting the voltage and temperature of each single cell will weaken the energy density of the battery and increase costs.

[0003] The behavior of gas release caused by chemical / electrochemical reactions inside lithium-ion batteries is an important feature of the battery operation process. Lithium-ion batteries in different environments and operating conditions (such as low temperature / normal temperature / high temperature cycles, operation with different charge and discharge cut-off voltages, high temperature storage, etc.) will affect the gas release concentration and gas production. In the early stage of battery thermal runaway, as the internal reaction of the battery gradually becomes more intense, the concentration of characteristic gases produced by the reaction will increase sharply from 0 to hundreds or even thousands of milligrams per cubic meter. Therefore, the use of gas detection technology to achieve early warning of battery thermal runaway has great potential. Therefore, the use of gas sensing technology for early detection and warning of thermal runaway, and the elimination of fire hazards before thermal runaway spreads, has great potential in improving the safety of lithium-ion batteries. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present application provides a new energy vehicle battery online intelligent fault diagnosis method, comprising the following steps:

[0005] Step 1: Measure the temporal variation data of each gas component when batteries with different positive electrode materials are kept at a high temperature at different states of charge;

[0006] The components of the released gas include; the concentrations of CO2, CO, H2 and VOCs;

[0007] Step 2: Classify the data measured in step 1 into time series of different gases released by batteries with different positive electrode materials at different states of charge, and divide the time series into an input sequence and a target sequence;

[0008] Step 3: Perform LSTM neural network training to obtain prediction models for the components of precipitated gases when batteries with the same cathode material are at different states of charge;

[0009] Step 4: Collect real-time data on the composition of the gas released from each battery pack of the new energy vehicle;

[0010] Step 5: Select the corresponding battery precipitated gas composition prediction model according to the type of battery positive electrode material and battery charge state output by the new energy vehicle battery management system;

[0011] Step 6: Using the sliding window method, extract the gas composition data corresponding to the time steps within the sliding window range from the time steps of the real-time data extracted in step 4 to form a time series to be predicted;

[0012] Step 7: Input the time series to be predicted in step 6 into the trained battery gas composition prediction model selected in step 5, and output the result;

[0013] Step 8. Compare the output result with the threshold. If it is greater than the threshold, output a warning signal.

[0014] The precipitated gas component is the concentration of one of CO2, CO, H2 and VOCs.

[0015] The step 2 of dividing the time series into an input sequence and a target sequence includes:

[0016] The sliding window method is used to extract the time series data corresponding to the time steps within the sliding window range from each time series data set;

[0017] The time series sequence within a time window is taken as the input sequence, and the time series sequence of the time step after the input sequence is taken as the target sequence.

[0018] The step 3 includes: initializing the LSTM model parameters, forward propagating the input sequence through the LSTM model to obtain a predicted sequence, calculating the loss function according to the predicted sequence and the target sequence, updating the weight parameters of the LSTM model according to the calculated loss function gradient through the back propagation algorithm, and repeating the training process until a preset number of training times is reached or until convergence.

[0019] In step 1, for non-inherent gases CO, H2 and VOCs in the atmosphere, the precipitated gas composition refers to the process of change from zero to one in the volume concentration of the gas; for the intrinsic gas CO2 in the atmosphere, the precipitated gas composition refers to the process of change after the volume concentration of the gas deviates from the average volume concentration of CO2 in the atmosphere.

[0020] The time series divided into the input series and the target series in step 2 is obtained by processing in the following steps:

[0021] Merging data according to the state of charge of the battery corresponding to each time series, if the state of charge of the battery belongs to the same state of charge distribution range, the time series of the gas components are divided into the same group;

[0022] In each time series within the same state of charge distribution range, the average value of the gas composition data of batteries in different states of charge at the same time is calculated, and the average value data of the gas composition at all times constitutes the time series of the gas composition of each state of charge distribution range.

[0023] The step 3 is used to obtain prediction models for components of precipitated gases when batteries of the same positive electrode material are in different charge state distribution ranges.

[0024] The state of charge distribution range includes SOC∈[0%, 5%], SOC∈(5%, 15%], SOC∈(15%, 25%], SOC∈(25%, 35%], SOC∈(35%, 45%], SOC∈(45%, 55%], SOC∈(65%, 75%], SOC∈(75%, 85%], SOC∈(85%, 95%], SOC∈(95%, 105%]).

[0025] The component of the precipitated gas is CO2.

[0026] The beneficial effects of the present invention are: it can quantitatively evaluate the potential thermal runaway risk of the vehicle's power battery system, and can timely detect to reduce battery thermal runaway accidents, improve the safety performance of lithium-ion battery electric vehicles, reduce safety hazards and reduce property losses. The present invention predicts the concentration of the gas in the future time step according to the time series of the concentration of the gas component detected in real time through the LSTM network, so it can predict in advance whether the gas component will exceed the threshold, solving the problem that the concentration of a certain gas component of batteries of some types of positive electrode materials is low under a certain SOC state, which is not conducive to detection, or solving the problem that the CO2 precipitated by the battery is not easy to distinguish from the CO2 in the air, and the change in the concentration of CO2 precipitated by the battery is not significant in the early stage of precipitation. The present invention also obtains a battery precipitated gas component prediction model based on the SOC state of the battery by combining the detection data, solving the problem that the SOC of new energy vehicles changes continuously during operation, and the battery thermal runaway cannot be predicted based on only one battery precipitated gas component prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of the principle of LSTM neural network. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution of the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings of the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the described embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0029] Explanation of terms

[0030] SOC: refers to the state of charge, the ratio of the remaining capacity of a battery after it has been used for a period of time or has been left unused for a long time to its capacity in a fully charged state, usually expressed as a percentage. Its value range is 0 to 100. When SOC = 0, it means that the battery is fully discharged, and when SOC = 100, it means that the battery is fully charged.

[0031] Example 1

[0032] This embodiment proposes a new energy vehicle battery online intelligent fault diagnosis method, the steps are as follows:

[0033] S1. Training LSTM neural network

[0034] Experimental preparation:

[0035] Before the experiment, batteries with different positive electrode materials were placed in a high and low temperature box to heat the batteries. A gas detection system was set up in the high and low temperature box to monitor the precipitation of CO2, CO, H2 and VOC in real time, with a measurement frequency of 1Hz. For non-intrinsic gases CO, H2 and VOC in the atmosphere, gas precipitation refers to the process of the volume concentration of the gas changing from nothing to something; for the intrinsic gas CO2 in the atmosphere, gas precipitation refers to the process of the volume concentration of the gas changing after it deviates from the mean volume concentration of CO2 in the atmosphere.

[0036] A battery testing system is set in the high and low temperature box to charge and discharge the battery while heating the battery, and to measure the voltage, current and temperature of the battery. The measurement frequency is 10Hz and the error is 0.1%.

[0037] Data Collection:

[0038] All experimental batteries were first charged to 3.65V in a constant current and constant voltage mode (1C charging and 0.01C cut-off current, C represents the rate), and then multiple batteries of the same material were placed at different SOCs by controlling the discharge time. The initial temperature was (20±2)℃. Each battery was left to stand for more than 1h after being charged and discharged to different SOCs.

[0039] Batteries with different positive electrode materials at different SOCs were placed in different high and low temperature chambers. The temperature in each high and low temperature chamber was continuously raised to 150°C at a rate of 5°C / min on the basis of the initial temperature and maintained. Each battery was retained in the high and low temperature chamber for more than 1 hour. The gas detection system monitored the precipitation of CO2, CO, H2 and VOC in real time, and the results shown in Table 1 were obtained.

[0040] Table 1

[0041]

[0042] The time series of gas components released when batteries with different positive electrode materials are at different SOCs are obtained, including the time series of concentrations of released CO2, CO, H2 and VOC.

[0043] The time series of the evolved gas components are classified according to the positive electrode material of the battery and then classified into multiple time series data sets according to SOC.

[0044] It should be noted that when collecting data, the gas detection system can monitor the concentration of only one gas among CO2, CO, H2 and VOC in real time, or it can monitor the concentration of several or all gases.

[0045] The corresponding time series of the precipitated gas components is the time series of the concentration of one gas among CO2, CO, H2 and VOC. The time series of the concentrations of several or all gases may also be monitored.

[0046] If the gas detection system monitors the concentrations of several or all gases in real time during data collection, the time series data sets can be further classified into time series data sets of different gases according to the gas components.

[0047] Each time series data set is trained using an LSTM neural network to develop a battery gas composition prediction model for batteries with different positive electrode materials at different SOC conditions.

[0048] If the time series data set can be further classified into time series data sets of different gases according to gas composition, each time series data set is trained using an LSTM neural network to train a prediction model for each gas component released by batteries with different positive electrode materials at different SOC conditions.

[0049] Divide the data set: divide each time series data set into input sequence and target sequence;

[0050] That is, the sliding window method is used to extract the time series data corresponding to the time steps within the sliding window range from each time series data set;

[0051] Take the time series sequence within a time window as the input sequence, and the time series sequence of the next time step of the input sequence as the target sequence;

[0052] LSTM neural network training: Initialize the LSTM model parameters, forward propagate the input sequence through the LSTM model to obtain the predicted sequence, calculate the loss function based on the predicted sequence and the target sequence, and use the backpropagation algorithm to update the weight parameters of the LSTM model according to the calculated loss function gradient. Repeat the training process until the preset number of training times is reached or until convergence.

[0053] S2. Collect real-time data on the composition of the gas released from each battery pack of new energy vehicles, including the concentrations of CO2, CO, H2 and VOC;

[0054] It should be noted that the composition of the precipitated gas may be the concentration of only one gas among CO2, CO, H2 and VOC, or the concentration of multiple gases.

[0055] S3. Select the corresponding battery gas composition prediction model according to the type of battery positive electrode material and battery SOC output by the new energy vehicle battery management system;

[0056] Corresponding to the above technical solution, the battery precipitated gas component prediction model is a prediction model for a certain gas component precipitated by the battery when the battery with different positive electrode materials is in different SOC conditions, or a combined prediction model of multiple gas components.

[0057] S4. Using a sliding window method, extracting the gas composition data corresponding to the time steps within the sliding window range from the real-time data extracted in step S2 to form a time series to be predicted, wherein the time steps within the sliding window range include the time steps of the real-time data extracted in step S2;

[0058] S5. Input the predicted time series of S4 into the trained battery gas composition prediction model selected in step S3, and output the result;

[0059] S6. Compare the output result with the threshold value, and if it is greater than the threshold value, output a warning signal.

[0060] Example 2

[0061] A new energy vehicle battery online intelligent fault diagnosis method, which differs from Example 1 only in that:

[0062] In step S1

[0063] During data collection, the gas detection system only monitors the evolution of CO2 in real time, and obtains the time series of CO2 evolution when batteries with different cathode materials are at different SOCs;

[0064] After training the LSTM neural network, a prediction model for CO2 evolution from batteries with different cathode materials at different SOC conditions was obtained.

[0065] In step S2, real-time data of CO2 released from each battery pack of the new energy vehicle is collected.

[0066] In step S3, a corresponding prediction model for battery CO2 evolution is selected according to the type of battery positive electrode material and the battery SOC output by the new energy vehicle battery management system.

[0067] In step S4, a sliding window method is used to extract the data of CO2 evolution corresponding to the time steps within the sliding window range from the real-time data extracted in step S2 to form a time series to be predicted, wherein the time steps within the sliding window range include the time steps of the real-time data extracted in step S2;

[0068] In step S5, the time series to be predicted in S4 is input into the prediction model of CO2 released from the battery selected and trained in step S3, and the result is output;

[0069] Step S6 compares the output result with a threshold value, and if the output result is greater than the threshold value, an early warning signal is output.

[0070] At present, when batteries made of various positive electrode materials experience thermal runaway at different SOC values, the main types of battery pyrolysis gases are CO2, CO, H2, CxHy, CxHyOz, CH3F, C2H5F, POF3, and HF. According to the chemical characteristics,

[0071] CxHy (except CH4) and CxHyOz are VOCs, and photoionization detection sensors can accurately monitor the situation of VOCs. CO and H2 use electrochemical sensors, and CO2, CH4 and C2H5F can be accurately monitored using infrared sensors. Among them, CO2 accounts for a large volume percentage. Using CO2 as a thermal runaway warning signal, the warning response time is faster than using battery temperature and battery voltage as warning signals, and the battery safety monitoring is more immediate. However, due to the high concentration of CO2 itself in the air, it takes a longer time to observe significant changes in concentration. Therefore, the existing technology has not yet used CO2 as a thermal runaway warning signal. The existing technology mainly uses CO, H2 and alkane-olefin gases as monitoring signals. When the battery is at a lower SOC, the battery precipitates VOC, CO and H2 late and the concentration is low. It is difficult to apply it to thermal runaway warnings for batteries in all states. It is necessary to implant sensors for detecting CO and H2 in the optical fiber and implant them inside the battery pack to detect the concentration of CO and H2. Since the optical fiber sensor has a cross-response to temperature changes and pressure changes, this also requires additional design to decouple these two signals.

[0072] The LSTM neural network of this embodiment is a special RNN (recurrent neural network). The LSTM network structure inputs the concentration of CO2 at time t into the network in chronological order. In the LSTM unit, it is necessary to calculate the output of the current state based on the state of the memory cell (cell) combined with the input and update the information of the memory cell to pass it to the next hidden layer unit, and finally output the concentration of CO2 at time t+1.

[0073] In the training of the prediction model of CO2 released by batteries, it is necessary to predict the long-term data of CO2 released by batteries in the future based on the data of CO2 released by batteries in the previous period. This requires the LSTM network structure to be continuously cycled, and the output at time t+1 is used as the input at time t+2. In this way, the prediction of the data of CO2 released by batteries in the long term can be completed. The data of CO2 released by batteries with various step sizes are used as network input for experiments. During the network training process, the network evaluates and optimizes the output prediction value based on the known data of CO2 released by batteries, calculates the average difference between the predicted value and the actual value as the training target, selects the mean square error between the predicted value and the actual value as the loss function, and uses the back propagation algorithm to update the weight parameters of the LSTM model according to the calculated gradient of the loss function. The training process is repeated until the preset number of training times is reached or until convergence.

[0074] Therefore, by using the prediction model of CO2 released by the battery of this embodiment, the concentration of CO2 released by the battery at a later time can be predicted in advance based on the real-time data of CO2 released by each battery pack of the new energy vehicle collected, thus overcoming the problem of slow speed of observing significant changes in CO2 concentration in the prior art. When batteries with different positive electrode materials are at different SOCs, the time of CO2 release is earlier and the concentration of CO2 release is also higher, which makes up for the deficiency of using other gases as thermal runaway warning signals.

[0075] If the precipitated gas component of this embodiment is replaced by other gas components instead of CO2, the time for generating the thermal runaway warning signal can also be shortened.

[0076] In step S6 of this embodiment, the threshold corresponds to different values ​​according to the type of positive electrode material of the battery and the SOC state value of the battery. For example, when the battery SOC is 70%, the CO2 concentration threshold is 500×10 -6 .

[0077] Example 3

[0078] This embodiment proposes a new energy vehicle battery online intelligent fault diagnosis method, the details are as follows:

[0079] S1. Training LSTM neural network

[0080] Experimental preparation:

[0081] Before the experiment, batteries with different positive electrode materials were placed in a high and low temperature box to heat the batteries. A gas detection system was set up in the high and low temperature box to monitor the precipitation of CO2, CO, H2 and VOC in real time, with a measurement frequency of 1Hz. For non-intrinsic gases CO, H2 and VOC in the atmosphere, gas precipitation refers to the process of the volume concentration of the gas changing from nothing to something; for the intrinsic gas CO2 in the atmosphere, gas precipitation refers to the process of the volume concentration of the gas changing after it deviates from the mean volume concentration of CO2 in the atmosphere.

[0082] A battery testing system is set in the high and low temperature box to charge and discharge the battery while heating the battery, and to measure the voltage, current and temperature of the battery. The measurement frequency is 10Hz and the error is 0.1%.

[0083] Data Collection:

[0084] All experimental batteries were first charged to 3.65V in a constant current and constant voltage mode (1C charging and 0.01C cut-off current, C represents the rate), and then multiple batteries of the same material were placed at different SOCs by controlling the discharge time. The initial temperature was (20±2)℃. Each battery was left to stand for more than 1h after being charged and discharged to different SOCs.

[0085] Batteries with the same positive electrode material at different SOCs were placed in different high and low temperature chambers. The temperature in each high and low temperature chamber was continuously raised to 150°C at a rate of 5°C / min based on the initial temperature and maintained there. Each battery stayed in the high and low temperature chamber for more than 1 hour, and the gas detection system monitored the precipitation of CO2, CO, H2 and VOC in real time.

[0086] The time series of gas components released when batteries with different positive electrode materials are at different SOCs are obtained, including the time series of concentrations of released CO2, CO, H2 and VOC.

[0087] The time series of the evolved gas components are classified according to the positive electrode material of the battery and then classified into multiple time series data sets according to SOC.

[0088] It should be noted that the gas detection system can only monitor the concentration of CO2 in real time during data collection. The corresponding time series of the precipitated gas components is the time series of the concentration of CO2.

[0089] Before using the time series data set for training the LSTM neural network, this embodiment also includes a data merging step:

[0090] Since the time series of the gas evolution components of batteries with the same positive electrode material change similarly when the SOC values ​​are close, the time series of the gas evolution components are divided into several groups according to the SOC of the battery being divided into multiple continuous SOC distribution ranges (state of charge distribution ranges), and the SOC corresponding to the time series of the gas evolution components in each group is within the divided SOC distribution range, and the SOC distribution range includes SOC∈[0%, 5%], SOC∈(5%, 15%], SOC∈(15%, 25%], SOC∈(25%, 35%], SOC∈(35%, 45%], SOC∈(45%, 55%], SOC∈(65%, 75%], SOC∈(75%, 85%], SOC∈(85%, 95%], SOC∈(95%, 105%].

[0091] During the charging and discharging process of new energy vehicle batteries, it often takes tens of minutes for the SOC distribution range to change from one boundary to another. During this process, the chemical composition of the battery is relatively stable, and the same model can be used to predict changes in gas composition.

[0092] After grouping, the time series of the gas evolution components in each group are summed to obtain the time series of the average gas evolution components, that is, in each time series within the same SOC distribution range, the gas evolution component data of batteries with different SOC values ​​at the same time are averaged, and the average value data of the gas evolution components at all times constitute the time series of the gas evolution components in each SOC distribution range.

[0093] The time series data set of the gas evolution components in each SOC distribution range is trained using an LSTM neural network into a battery gas evolution composition prediction model for batteries with different positive electrode materials in different SOC distribution ranges.

[0094] Divide the data set: Use the sliding window method to extract the time series data corresponding to the time steps within the sliding window range from each time series data set;

[0095] Take the time series sequence within a time window as the input sequence, and the time series sequence of the next time step of the input sequence as the target sequence;

[0096] After dividing the time series of the gas components in each SOC distribution range into an input series and a target series;

[0097] LSTM neural network training: Initialize the LSTM model parameters, forward propagate the input sequence through the LSTM model to obtain the predicted sequence, calculate the loss function based on the predicted sequence and the target sequence, and use the backpropagation algorithm to update the weight parameters of the LSTM model according to the calculated loss function gradient. Repeat the training process until the preset number of training times is reached or until convergence.

[0098] S2. Collect real-time data on the composition of the gas released from each battery pack of new energy vehicles, including the concentrations of CO2, CO, H2 and VOC;

[0099] It should be noted that the component of the precipitated gas is only the concentration of CO2.

[0100] S3. Select the corresponding battery gas composition prediction model according to the type of battery positive electrode material and battery SOC output by the new energy vehicle battery management system;

[0101] Corresponding to the above technical solution, the battery evolution gas composition prediction model is a prediction model for CO2 evolved from the battery when the battery with different positive electrode materials is in different SOC distribution ranges.

[0102] S4. Using a sliding window method, extracting the gas composition data corresponding to the time steps within the sliding window range from the real-time data extracted in step S2 to form a time series to be predicted, wherein the time steps within the sliding window range include the time steps of the real-time data extracted in step S2;

[0103] S5. Input the predicted time series of S4 into the trained battery gas composition prediction model selected in step S3, and output the result;

[0104] S6. Compare the output result with the threshold value, and if it is greater than the threshold value, output a warning signal.

[0105] In this embodiment, the training data are grouped according to the SOC distribution range, and the LSTM neural network is trained separately. The prediction model is then selected according to the positive electrode material and SOC of the battery pack. The prediction model is used to predict the real-time data of the gas evolution composition of the battery pack. The SOC of the new energy vehicle that lasts for tens of minutes during the dynamic process is all within the same SOC distribution range. The selected prediction model is used to predict the gas evolution composition of the battery pack and has good stability, and there is no need to frequently replace the prediction model.

Claims

1. A new energy vehicle battery online intelligent fault diagnosis method, characterized in that: The method comprises the following steps: Step 1: measuring the time-dependent variation data of each gas component released when batteries with different positive electrode materials are kept at a high temperature at different charge states; The components of the released gas include; the concentrations of CO2, CO, H2 and VOCs; Step 2: Classify the data measured in step 1 into time series of different gases released by batteries with different positive electrode materials at different states of charge, and divide the time series into an input sequence and a target sequence; Step 3: Perform LSTM neural network training to obtain prediction models for the components of precipitated gases when batteries with the same cathode material are at different states of charge; Step 4: Collect real-time data on the composition of the gas released from each battery pack of the new energy vehicle; Step 5: Select the corresponding battery precipitated gas composition prediction model according to the type of battery positive electrode material and battery charge state output by the new energy vehicle battery management system; Step 6: Using the sliding window method, extract the gas composition data corresponding to the time steps within the sliding window range from the time steps of the real-time data extracted in step 4 to form a time series to be predicted; Step 7: Input the time series to be predicted in step 6 into the trained battery gas composition prediction model selected in step 5, and output the result; Step 8. Compare the output result with the threshold. If it is greater than the threshold, output a warning signal.

2. The method according to claim 1, characterized in that: The precipitated gas component is the concentration of one of CO2, CO, H2 and VOCs.

3. The method according to claim 1, characterized in that The step 2 of dividing the time series into an input sequence and a target sequence includes: The sliding window method is used to extract the time series data corresponding to the time steps within the sliding window range from each time series data set; The time series sequence within a time window is taken as the input sequence, and the time series sequence of the time step after the input sequence is taken as the target sequence.

4. The method according to claim 1, characterized in that: The step 3 includes: initializing the LSTM model parameters, forward propagating the input sequence through the LSTM model to obtain a predicted sequence, calculating the loss function according to the predicted sequence and the target sequence, updating the weight parameters of the LSTM model according to the calculated loss function gradient through the back propagation algorithm, and repeating the training process until a preset number of training times is reached or until convergence.

5. The method according to claim 1, characterized in that In step 1, for non-inherent gases CO, H2 and VOCs in the atmosphere, the precipitated gas composition refers to the process of change from zero to one in the volume concentration of the gas; for the intrinsic gas CO2 in the atmosphere, the precipitated gas composition refers to the process of change after the volume concentration of the gas deviates from the average volume concentration of CO2 in the atmosphere.

6. The method according to claim 3, characterized in that The time series divided into the input series and the target series in step 2 is obtained by processing in the following steps: Merging data according to the state of charge of the battery corresponding to each time series, if the state of charge of the battery belongs to the same state of charge distribution range, the time series of the gas components are divided into the same group; In each time series within the same state of charge distribution range, the average value of the gas composition data of batteries in different states of charge at the same time is calculated, and the average value data of the gas composition at all times constitutes the time series of the gas composition of each state of charge distribution range.

7. The method according to claim 6, characterized in that The step 3 is used to obtain prediction models for components of precipitated gases when batteries of the same positive electrode material are in different charge state distribution ranges.

8. The method according to claim 7, characterized in that The battery precipitated gas composition prediction model selected in step 5 is a battery precipitated gas composition prediction model when batteries with different positive electrode materials are in different charge state distribution ranges.

9. The method according to claim 7, characterized in that: The state of charge distribution range includes SOC∈[0%, 5%], SOC∈(5%, 15%], SOC∈(15%, 25%], SOC∈(25%, 35%], SOC∈(35%, 45%], SOC∈(45%, 55%], SOC∈(65%, 75%], SOC∈(75%, 85%], SOC∈(85%, 95%], SOC∈(95%, 105%]).

10. The method according to any one of claims 1 to 9, characterized in that: The component of the precipitated gas is CO2.

Images