CN113013514B - A thermal runaway gas-sensing alarm device for vehicle-mounted lithium-ion power battery and its detection method - Google Patents

Abstract

The invention discloses a thermal runaway gas-sensitive alarm device of a vehicle-mounted lithium ion power battery, which comprises: a box body; the battery pack modules are arranged in the box body at intervals, the battery monomers are uniformly arranged in the battery pack modules, and cooling flow field channels are formed among the battery pack modules; the gas sensor is arranged at the outlet position of the cooling flow field channel; the thermocouple temperature sensors are uniformly arranged inside the battery pack modules; the mass sensors are correspondingly arranged at the lower parts of the battery monomers one by one; the management device is connected with the gas sensor, the thermocouple temperature sensors and the mass sensors. The invention also discloses a detection method of the thermal runaway gas-sensitive alarm device of the vehicle-mounted lithium ion power battery, which judges the probability of thermal runaway of the battery by establishing a Markov chain prediction model, and adopts cooling measures for the battery and reminds drivers and passengers.

Description

Thermal runaway gas-sensing alarm device for vehicle-mounted lithium-ion power battery and detection method thereof

technical field

The invention relates to the technical field of lithium-ion battery safety, and more particularly, the invention relates to a thermal runaway gas-sensing alarm device of a vehicle-mounted lithium-ion power battery and a detection method thereof.

Background technique

Lithium-ion batteries have become the mainstream choice for automotive power batteries due to their high capacity, high output voltage, high charging rate, high energy density, low self-discharge and excellent cycle characteristics. However, the high activity of its electrode materials and the flammability of electrolyte materials determine that the risk of thermal runaway in lithium-ion batteries always exists. In recent years, with the growth of the electric vehicle market and the improvement of its power performance, vicious safety accidents caused by the thermal runaway of on-board lithium-ion power batteries have occurred frequently, which has seriously damaged consumers' confidence in electric vehicles.

For cells composed of stacked wound materials, even if the lithium-ion battery is used normally, with each charge and discharge, the cells will be structurally damaged due to the residual stress generated by carriers at the interface. Structural breakage at the interface can cause uneven deposition of charge carriers, leading to the generation of dendrites. The SEI passivation film covering the electrode surface is an important measure to prevent dendrite-induced destructive side reactions. Studies have shown that the SEI passivation film decomposes under thermal abuse, resulting in side reactions that continue to move in the positive direction. The resulting gas and thermal effects will aggravate the thermal damage to the battery structure, make the scale of side reactions exceed the threshold, and lead the bistable system of the battery to irreversible thermal runaway. Studies have shown that during the normal operation of the battery, there is no material exchange between the battery and the outside world, and no gas is produced. Therefore, the gases produced by these side reactions can be used as a measurable physical characterization of the degree of thermal damage to the cell to assess the risk of thermal runaway of the cell.

Thermal damage to batteries can be divided into two types: reversible and irreversible. Studies have shown that the occurrence of thermal runaway is highly related to the internal short circuit of the battery, and the irreversible damage to the internal structure of the battery will become the direct cause of the internal short circuit. In the actual use of electric vehicles, it is a gradual process from the occurrence of thermal damage to the battery cell to the occurrence of thermal runaway. In the early stage of thermal damage, due to the structural characteristics of the battery itself, as the main output parameters of the battery, the temperature and voltage often do not change significantly, and the length of this period of time has a great relationship with the triggering method of thermal damage, but this period of time The lag is common. This provides a time window for the intervention of engineering means, making active protection against thermal runaway possible. Expanding this time window as much as possible has become the key to the success of engineering intervention.

Thermal runaway is essentially a violent side reaction of the electrolyte. The concentration of free radicals in the battery has a decisive influence on the final direction of the side reaction. If the concentration of free radicals in the battery can be quickly reduced, that is, the electrolyte can be quickly deactivated, then It can effectively block the spread of side reactions in the battery, control the scale of the reaction, prevent the system from leading to a thermal runaway state due to the breakthrough of the threshold, and realize active safety intervention for thermal runaway. At the same time, since thermal effects and voltage fluctuations are the most significant features of battery side reactions, voltage fluctuations or thermal signals are the main triggering means for existing thermal runaway alarm systems. However, this thermal runaway early warning method based on the characterization of a single external parameter has a high probability of false positives and false negatives, and the early warning time lags, which makes the intervention window of subsequent active safety measures too small and the effect is weak.

And as the working substance of the forced fire extinguishing system, the thermal runaway blocker represented by chlorofluoroalkanes quickly exhausts the free radicals of the thermal runaway monomer, and also makes other intact battery cells in the same module due to chemical activity. lost and scrapped.

SUMMARY OF THE INVENTION

The purpose of the present invention is to design and develop a thermal runaway gas-sensing alarm device for vehicle-mounted lithium-ion power batteries. The BMS host computer can monitor whether the power battery is damaged and the degree of damage in real time under working conditions, and improve the early warning and intervention time of thermal runaway.

Another object of the present invention is to design and develop a detection method for a thermal runaway gas-sensing alarm device of a vehicle-mounted lithium-ion power battery, which can determine the probability of thermal runaway of the battery by establishing a Markov chain prediction model, and adopt cooling measures and The driver and occupant are alerted, enabling quantitative predictability of the risk of thermal runaway.

The technical scheme provided by the present invention is:

A thermal runaway gas-sensing alarm device for a vehicle-mounted lithium-ion power battery, comprising:

box; and

A plurality of battery pack modules are arranged inside the box at intervals, and the plurality of battery pack modules are provided with uniformly arranged battery cells, and cooling flow between the plurality of battery pack modules field channel;

a gas sensor, which is arranged at the outlet position of the cooling flow field channel;

a plurality of thermocouple temperature sensors, which are evenly arranged inside the plurality of battery pack modules;

a plurality of mass sensors, which are arranged in the lower part of the battery cells in one-to-one correspondence;

A management device, which is connected with the gas sensor, a plurality of thermocouple temperature sensors and a plurality of mass sensors, is used for signal reception and command transmission.

Preferably, the management device includes:

a signal transmission component, which is connected with the gas sensor, a plurality of thermocouple temperature sensors and a plurality of mass sensors;

a single-chip microcomputer, which is connected with the signal transmission component;

The BMS host computer is connected with the single-chip microcomputer, and is used for signal reception and command transmission.

Preferably, it also includes:

a plurality of current/voltage signal collectors, which are connected to the battery cells in one-to-one correspondence, and are used to monitor the output current and output voltage of the battery cells;

a solid-state particle detector, which is arranged on one side of the gas sensor;

Wherein, the plurality of current/voltage signal collectors and solid-state particle detectors are all connected with the signal transmission component.

A method for detecting a thermal runaway gas-sensing alarm device for a vehicle-mounted lithium-ion power battery, using the thermal-runaway gas-sensing alarm device for a vehicle-mounted lithium-ion power battery, comprises the following steps:

Step 1. According to the sampling period, collect the concentration of carbon monoxide characteristic gas in the box, the temperature in multiple battery pack modules and the mass loss percentage of the battery cell, and construct the battery state characterization matrix:

Ω=(Q, T, Δm) ^T ;

In the formula, Q is the concentration vector of carbon monoxide characteristic gas in the box, T is the temperature vector of each monitoring point in the battery pack module, and Δm is the mass loss percentage of each battery cell;

Step 2: Divide the battery state data into training set and test set randomly according to the ratio of 1:1, and count the battery state representation matrix that has appeared on the training set according to the time series principle, obtain the state space of the battery, and establish the marathon. Kov chain prediction model to obtain the state transition matrix:

为电池从状态Ω_i迁移到状态Ω_j的概率，i＝1,2,…N，j＝1,2,…N，N为电池的唯一状态数量；In the formula, the element

is the probability of the battery migrating from state Ω _i to state Ω _j , i=1,2,...N, j=1,2,...N, N is the unique number of states of the battery;

Step 3: Predict the battery state [Ω _t+T ,Ω _t+2T ,Ω _t+3T ] of each battery state in the test set in the next three sampling periods based on the established effective Markov chain model, where t is the current Time label, T is the sampling period;

Step 4: Input the battery state, Markov chain validity and battery SOC value of the next three sampling periods into the BP neural network model to obtain the stage level of the battery thermal runaway;

Step 5: Determine the probability of thermal runaway of the battery according to the stage level of the thermal runaway of the battery, and adopt cooling measures for the battery and remind drivers and passengers.

Preferably, the probability of the battery migrating from state Ω _i to state Ω _j satisfies:

为实测实验下电池由状态迁移Ω_i到状态Ω_j发生的次数，

为实测实验下电池状态Ω_i出现的总次数。In the formula,

is the number of times that the battery transitions from state Ω _i to state Ω _j under the measured experiment,

is the total number of times that the battery state Ω _i appears in the measured experiment.

Preferably, the process of judging whether the Markov chain model is valid is:

According to the state transition matrix, predict the battery state of each battery state in the test set for a sampling period in the future, and compare the predicted result with the real state. If the validity of the Markov chain is greater than 90%, it proves that the Markov chain is effective. The Markov chain prediction model is valid, otherwise the Markov chain prediction model is re-established.

Preferably, the validity of the Markov chain satisfies:

In the formula, k is the exact number of prediction results, and K is the number of all test set states.

Preferably, the SOC value of the battery is obtained through a real-time current/voltage signal query after obtaining the battery SOC-OCV curve based on an offline bench test.

Preferably, the calculation process of the BP neural network model is:

Step 1. Determine the input layer neuron vector x={x ₁ , x ₂ , x ₃ , x ₄ , x ₅ } of the three-layer BP neural network; where x ₁ is the battery state Ω _t+T , x ₂ is the battery State Ω _t+2T , x ₃ is the battery state Ω _t+3T , x ₄ is the Markov chain effectiveness f, and x ₅ is the SOC value of the battery;

Step 2. The input layer vector is mapped to the hidden layer, and the number of hidden layer neurons is h, and

In the formula, m is the number of input nodes, n is the number of output nodes, and a is the adjustment factor;

Step 3. Obtain the output layer neuron vector o={o ₁ , o ₂ }; o ₁ is the thermal runaway risk level of the battery, and o ₂ is the reliability of the prediction result of o ₁ ;

0为零级风险，表明电池正常，1为一级风险，表明电池单体可能出现热失控并需对电池持续监测，2为二级风险，表明电池单体出现热失控并需要进行冷却处理，3为三级风险，表明电池包模组发生热失控。in,

0 is a zero-level risk, indicating that the battery is normal; 1 is a first-level risk, indicating that the battery cell may have thermal runaway and the battery needs to be continuously monitored; 2 is a second-level risk, indicating that the battery cell has thermal runaway and needs to be cooled. 3 is the third-level risk, indicating that the thermal runaway of the battery pack module occurs.

Preferably, the step 5 specifically includes:

When ο ₁ =0 and ο ₂ ≥90%, the battery has no probability of thermal runaway;

When ο ₁ = 1 and ο ₂ ≥ 80%, the probability of thermal runaway of the battery is less than 50% and the single cell of the battery needs to be cooled;

When ο ₁ =2 and ο ₂ ≥70%, the probability of thermal runaway of the battery is greater than 50% and the battery pack module needs to be cooled;

When ο ₁ = 3 and ο ₂ ≥ 60%, the battery pack module is thermally out of control, and the driver and passengers need to be reminded to avoid danger in time and sound an alarm.

The beneficial effects of the present invention:

(1) A thermal runaway gas-sensing alarm device for a vehicle-mounted lithium-ion power battery designed and developed by the present invention adopts a patch-type electrical signal sensor as the gas-sensing sensor, which is small in size, low in cost, and does not depend on an external power supply. It can be directly integrated into the battery pack module during the packaging process, independent of the external signal output of the battery pack, and is less affected by external factors; different from the traditional photosensitive detection equipment, it does not rely on the upper computer detection equipment, which can realize the Real-time monitoring of gas composition in battery pack modules.

(2) The thermal runaway gas-sensing alarm device for vehicle-mounted lithium-ion power batteries designed and developed by the present invention can coordinately measure a variety of physical quantities characterizing thermal runaway precursors, which can effectively avoid false alarms and omissions of BMS. The introduction of the signal can significantly advance the warning time and improve the overall safety performance of the electric vehicle;

(3) The detection method of the thermal runaway gas-sensing alarm device of the vehicle-mounted lithium-ion power battery designed and developed by the present invention solves the technical problem of accurately evaluating the thermal runaway risk of the vehicle-mounted lithium-ion power battery under dynamic conditions, so that it can be determined by Unmeasurable, quantitative and predictable, and based on this, realize timely warning to drivers; by systematically quantifying the risk of thermal runaway, BMS can more accurately identify potential threats and assess their degree, greatly improving the effectiveness of intervention measures .

(4) The detection method of the thermal runaway gas-sensing alarm device of the vehicle-mounted lithium-ion power battery designed and developed by the present invention, in conjunction with the safety solution of the vehicle-mounted BMS system, is a safety sub-module specially developed based on the safety requirements of electric vehicles, which is different from the existing The safety measures developed for the energy storage power station emphasize the compatibility with the existing control software of electric vehicles, taking into account the internal structure layout and power requirements of electric vehicles.

Description of drawings

FIG. 1 is a schematic diagram of the assembly structure of the battery pack module according to the present invention in a system.

FIG. 2 is a schematic diagram of the safe working range and thermal runaway characterization of the lithium battery according to the present invention.

FIG. 3 is a schematic diagram of the change of monomer mass under different SOC conditions in each thermal runaway stage according to the present invention.

FIG. 4 is a schematic diagram of the relationship between the overall heat released by the battery and the number of cells under different pressure environments according to the present invention.

FIG. 5 is a schematic diagram of the change of the internal pressure of the battery according to the present invention as the temperature of the cell surface increases.

FIG. 6 is a schematic diagram of the arrangement structure of the various sensors according to the present invention in the battery.

FIG. 7 is a schematic diagram of the evaluation model of the battery thermal runaway early warning system according to the present invention.

FIG. 8 is a schematic diagram of the physical architecture of the mechanism layer of the battery thermal runaway early warning system according to the present invention.

Detailed ways

The present invention will be further described in detail below, so that those skilled in the art can implement it with reference to the description.

The invention aims to design a thermal runaway probability evaluation model based on a gas sensor channel (gas signal, air pressure, gas concentration), a temperature sensor (temperature signal) and a mass sensor (mass change signal) and a combination of machine learning and big data analysis The thermal runaway warning device and detection method of the on-board lithium-ion power battery solves the technical problem of accurately assessing the thermal runaway risk of the on-board lithium-ion power battery cell under dynamic conditions, making it quantitative and predictable from unmeasurable to quantitative and predictable. Based on this, the timely warning to the driver is realized.

As shown in FIG. 1 , a thermal runaway gas-sensing alarm device for a vehicle-mounted lithium-ion power battery provided by the present invention includes: a box body 100 , a first battery pack module 111 , a second battery pack module 112 , and a third battery pack The module 113 and the fourth battery pack module 114, the first battery pack module 111, the second battery pack module 112, the third battery pack module 113 and the fourth battery pack module 114 are arranged at intervals in the Inside the box body 100 , the first battery pack module 111 , the second battery pack module 112 , the third battery pack module 113 and the fourth battery pack module 114 are all provided with uniformly arranged battery cells 120. Between the first battery pack module 111, the second battery pack module 112, the third battery pack module 113 and the fourth battery pack module 114 is a cooling flow field channel; the cooling flow field channel includes The cooling air inlet 131 and the heat dissipation air outlet 132 are provided.

As shown in Figure 2, the internal structural damage of the battery must reach a certain level before thermal runaway occurs. As thermal runaway looms, its internal damage continues to expand, triggering larger-scale side reactions that ultimately make it irreversible. For a lithium-ion battery with an intact structure, the electrolyte in the cell does not participate in chemical reactions, and there is no material exchange with the outside of the system under working conditions. When a lithium-ion battery suffers a safety-threatening damage, no matter how small the damage is, the electrolyte in the cell will undergo an oxidation reaction, and the oxidation product will exchange material and energy with the outside world in the form of gas and heat. As shown in Figure 3-5, when the pressure relief valve located at the top cover of the battery connects the internal and external environment of the battery due to the increase of the internal pressure of the battery, it will lead to three direct consequences: the increase of the pressure in the battery pack and the change of the gas composition , The mass of the battery cell is reduced, the heat release of the battery cell and the increase of the surface temperature. The main components of the gas are water, carbon monoxide, carbon dioxide and a small amount of organic vapor. Among them, carbon dioxide is the main component of the gas product, and the characteristic components are carbon monoxide and carbon monoxide. Hydrofluoric acid vapor.

According to this, the disturbance of the internal environment state of the battery module is monitored by the above three physical quantities, characteristic gas abundance, battery cell mass loss ratio and battery surface temperature, and then without relying on the external parameter output of the battery. , judging whether there is the occurrence of electrolyte decomposition and the degree of occurrence, and realizing real-time monitoring of whether the power battery is damaged and the degree of damage in working state.

As shown in FIG. 1 and FIG. 6 , in the present invention, sensors are arranged in units of individually packaged battery pack modules that are cooled by air cooling, and specifically include: a gas sensor 140, a plurality of thermocouple temperature sensors 150, a plurality of A mass sensor 160, a solid-state particle detector 170, a plurality of current/voltage signal collectors and management devices (not shown in the figure), the gas sensor 140 is arranged at the outlet position of the heat dissipation air outlet 132, in the air cooling channel The air flow generated by the battery will be used as the carrier gas for the battery produced gas, and the battery produced gas will be transported to the gas sensor 140; a plurality of thermocouple temperature sensors 150 are evenly arranged on the first battery pack module 111 and the second battery pack module 112 , the interior of the third battery pack module 113 and the fourth battery pack module 114, used to detect the first battery pack module 111, the second battery pack module 112, the third battery pack module 113 and the fourth battery pack module The temperature field distribution in the battery pack module 114; a plurality of mass sensors 160 are arranged in the lower part of the battery cells 120 in one-to-one correspondence, and are used to detect the mass loss of the battery cells 120 as the signal of the gas sensor 140. The supplementary signal is used to determine the position of the specific damaged battery cell 120; a plurality of current/voltage signal collectors are connected to the battery cell 120 in one-to-one correspondence, and are used to monitor the output current and the battery cell 120. Output voltage; the solid-state particle detector 170 is arranged on one side of the gas sensor 140; the management device is connected to the gas sensor 140, a plurality of thermocouple temperature sensors 150, a plurality of mass sensors 160, and a plurality of current/voltage signals The collector is connected to the solid state particle detector 170 for signal reception and command transmission.

The management device includes: a signal transmission component, a single-chip microcomputer, and a BMS host computer. The signal transmission component is connected with the gas sensor 140, a plurality of thermocouple temperature sensors 150, a plurality of mass sensors 160, and a plurality of current/voltage signal acquisition. The device is connected with the solid-state particle detector 170; the single-chip microcomputer is connected with the signal transmission component; the BMS host computer is connected with the single-chip computer for signal reception and command transmission.

Different from the existing photosensitive gas detection method based on Fourier infrared spectrum analysis, the present invention adopts the patch type gas sensor based on electrical signal to detect the abundance change of characteristic gas in the module in real time. Based on the resistance change caused by the characteristic gas molecules being captured by the coating, it does not rely on the secondary detection of the purified sample by the optical detection equipment, and does not need to install additional accessories in the battery pack module or box. For the flow field distribution, by arranging the gas sensor array inside the module, the host computer can sense the characteristic gas concentration changes at different positions in the module in real time. By coupling the real-time temperature field distribution returned by the thermocouple, the position of the problem unit in the module can be preliminarily determined. With the high-precision mass sensor arranged at the bottom of the unit, the host computer can accurately lock the damaged unit in the module. Compared with the technical solution of arranging the pressure sensor at the pressure relief valve, the mass sensor method can avoid the pressure sensor. The disadvantage is that the signal is noisy and is obviously affected by environmental factors.

In this embodiment, 16 cells are arranged in the form of 4*4 in the casings of the first battery pack module 111 , the second battery pack module 112 , the third battery pack module 113 and the fourth battery pack module 114 . A thermocouple temperature sensor 150.

Compared with the current method of using a single external parameter of voltage or temperature for thermal runaway early warning, the introduction of combustible gas sensors and mass sensors can not only reduce the probability of false positives and false negatives, but also more directly reflect the internal state of the battery. The early warning time of thermal runaway can be significantly advanced, and the intervention time window can be extended to the minute level, creating conditions for subsequent active safety measures to play a role.

A thermal runaway gas-sensing alarm device for a vehicle-mounted lithium-ion power battery designed and developed by the present invention monitors the disturbance of the environmental state in the battery pack module through the characteristic gas abundance, the mass loss ratio of the battery cell and the battery surface temperature. Then, without relying on the output of external parameters of the battery, it is judged whether there is the occurrence of electrolyte decomposition and the degree of occurrence, so as to realize real-time monitoring of whether the power battery is damaged and the degree of damage in the working state. Compared with the current method of using a single external parameter of voltage or temperature for thermal runaway early warning, the introduction of combustible gas sensors and mass sensors can not only reduce the probability of false positives and false negatives, but also more directly reflect the internal state of the battery. The early warning time of thermal runaway can be significantly advanced, and the intervention time window can be extended to the minute level, creating conditions for subsequent active safety measures to play a role.

Although thermal runaway occurs, it can be considered that the system has entered an irreversible state, but due to differences in the specific conditions that trigger thermal runaway, thermal damage to the battery may not necessarily lead to thermal runaway. Therefore, the battery state data obtained by the sensor must be The risk of thermal runaway is assessed by a probabilistic analysis model.

According to the needs of electric vehicles, the model should take the rate of change of physical quantities as input, and output the probability of thermal runaway under this working condition, so as to achieve quantitative assessment of the risk of thermal runaway. Therefore, the thermal runaway risk assessment model located on the host computer should be composed of a mechanism simulation model describing the battery state and a probability model for evaluating the risk of thermal runaway. When the physical quantity fluctuation signal of the sensor is obtained, the mechanism simulation model first determines the scale of the side reaction of the monomer according to the input signal peak value, and then the probability analysis model based on the machine learning algorithm evaluates the thermal damage of the monomer according to the scale of the side reaction. Stage and size of the available safe intervention window. Based on the evaluation results of this model, the BMS system will automatically select the corresponding safety strategy while warning the driver of the risk of thermal runaway, so as to achieve the purpose of improving the safety of electric vehicles.

As shown in FIG. 7 , the evaluation model in the present invention adopts a probability evaluation model based on a physical model, wherein the reaction kinetic model, the thermal diffusion model, the thermal strain model and the dynamic strain model constitute a physical model, which is converted from the physical model. It is a mathematical model, and the mathematical model and the measured data are jointly input into the machine learning algorithm. The above is the mechanism layer; the empirical model/virtual measured object is obtained through the machine learning algorithm, and the empirical model/virtual measured object is combined with the dynamic bench. The control intervention model is obtained by combining the data, and the above is the inference layer; finally, the output intervention results and the data of the electrical model are jointly transmitted to the BMS host computer to form the operation layer.

The task of the mechanism layer is mainly to form a virtual prototype as the core of the model, the task of the reasoning layer is to obtain the probability distribution model of thermal runaway risk through data analysis on the database formed by the virtual prototype and the actual test through the machine learning method, and the task of the operation layer is According to the risk assessment results of the probability layer, the subsequent operation is decided. Among them, the mechanism layer, as the basis of the entire evaluation model, provides theoretical support for it, while the reasoning layer and operation layer of the subsequent superstructure require big data technology. Based on the support and restriction of hardware implementation methods, the operation layer makes rational suggestions based on the corresponding results judgments given by the reasoning layer, combined with the actual situation, such as taking corresponding cooling measures for the battery pack, or reminding drivers and passengers to escape safely. Wait.

The core of the mechanism layer lies in the accurate reproduction of the physical picture of thermal runaway by its physical model. The basis for this goal is the physical decoupling of key characterizations of thermal runaway and how the associated physics are coupled. The physical structure adopted by the physical model in the present invention and the coupling mode and judgment conditions of each main module are shown in FIG. 8 . As a concentration battery, the diffusion behavior is considered by this model as the first driving force of battery function, and the pyrogenic trigger is regarded as the intrinsic condition of the battery's thermal runaway trigger.

The reasoning layer is the control method of the thermal runaway gas-sensing alarm device of the vehicle-mounted lithium ion power battery, using the thermal runaway gas-sensing alarm device of the vehicle-mounted lithium-ion power battery according to any one of claims 1-3, A variety of sensors collect the gas characteristic distribution, temperature distribution, and mass loss of the battery, and use the Markov chain-Monte Carlo algorithm to make a confident prediction of the battery state in the future. Combined with the established battery, etc. The battery characteristic parameters represented by the efficient circuit model are used as the input signal of the neural network model to make a corresponding assessment of the thermal runaway risk of the battery. Specifically include the following steps:

Step 1: According to the sampling period, the concentration of carbon monoxide characteristic gas in the collection box, the temperature in multiple battery pack modules, and the mass loss percentage of the battery cells are measured and collected by sensors, and the state characterization matrix of a group of batteries is constructed:

Ω=(Q, T, Δm) ^T ;

In the formula, Q is the concentration vector of carbon monoxide characteristic gas in the box, T is the temperature vector of each monitoring point in the battery pack module, and Δm is the mass loss percentage of each battery cell;

Step 2: According to the database of the measured experimental itinerary, randomly divide the battery state data into a training set and a test set according to the ratio of 1:1, and perform corresponding statistics on the training set according to the time series principle for the battery state representation matrix that has appeared, and obtain: Out of the battery's state space:

[S ₁ , S ₂ ,...S _n ];

In the formula, n is the number of sampling time periods;

Establish the Markov chain prediction model and obtain the state transition matrix:

为电池从状态Ω_i迁移到状态Ω_j的概率，i＝1,2,…N，j＝1,2,…N，N为电池的唯一状态数量；In the formula, the element

is the probability of the battery migrating from state Ω _i to state Ω _j , i=1,2,...N, j=1,2,...N, N is the unique number of states of the battery;

Wherein, the probability of the battery migrating from state Ω _i to state Ω _j satisfies:

为实测实验下电池由状态迁移Ω_i到状态Ω_j发生的次数，

为实测实验下电池状态Ω_i出现的总次数；In the formula,

is the number of times that the battery transitions from state Ω _i to state Ω _j under the measured experiment,

is the total number of times that the battery state Ω _i appears in the measured experiment;

Step 3: Use the established Markov chain state transition matrix to predict the battery state of each battery state in the test set for a sampling period in the future. The prediction process is as follows:

Assuming that the current battery state is Ω _i , i=1,2,...N, take the maximum probability:

P _max =P _ij =max(P _i1 ,P _i2 ,...P _iN ),

In the formula, j is the maximum probability state number, Ω _j is the next predicted state of the battery;

Compare the prediction results of the model with the real state, test the accuracy of the Markov chain, and calculate the validity of the Markov chain:

In the formula, k is the exact number of prediction results, and K is the number of all test set states;

If f>90%, it proves that the Markov chain prediction model is valid, otherwise the Markov chain prediction model is re-established;

Relying on the established effective Markov chain model to predict the battery state [Ω _t+T ,Ω _t+2T ,Ω _t+3T ] of each battery state in the test set in the next three sampling periods, t is the current time label, T is the sampling period;

Step 4: Input the battery state, Markov chain validity and battery SOC value of the next three sampling periods into the BP neural network model to obtain the stage level of the battery thermal runaway;

Among them, after the battery SOC-OCV curve is obtained based on the offline bench test, the physical characterization model of the battery is established. On this basis, the real-time output current/voltage signal of the battery is obtained through the CAN bus, and then the physical characterization of the battery is established. The model is used to accurately characterize the SOC of the battery, and determine its characteristic state as S _SOC ;

The calculation process of the BP neural network model is:

Step 1. Determine the input layer neuron vector x={x ₁ , x ₂ , x ₃ , x ₄ , x ₅ } of the three-layer BP neural network; where x ₁ is the battery state S ₁ , and x ₂ is the battery state S ₂ , x ₃ is the battery state S ₃ , x ₄ is the Markov chain validity, and x ₅ is the SOC value of the battery;

Step 2. The input layer vector is mapped to the hidden layer, and the number of hidden layer neurons is h, and

In the formula, m is the number of input nodes, n is the number of output nodes, a is the adjustment factor, and the value of the adjustment factor ranges from 1 to 10, so 5 hidden layer neurons are selected;

Step 3. Obtain the output layer neuron vector o={o ₁ , o ₂ }; o ₁ is the thermal runaway risk level of the battery, and o ₂ is the reliability of the prediction result of o ₁ ;

0为零级风险，表明电池正常，1为一级风险，表明电池单体可能出现热失控并需对电池持续监测，2为二级风险，表明电池单体出现热失控并需要进行冷却处理，3为三级风险，表明电池包模组发生热失控。in,

0 is a zero-level risk, indicating that the battery is normal; 1 is a first-level risk, indicating that the battery cell may have thermal runaway and the battery needs to be continuously monitored; 2 is a second-level risk, indicating that the battery cell has thermal runaway and needs to be cooled. 3 is the third-level risk, indicating that the thermal runaway of the battery pack module occurs.

The excitation functions of the hidden layer and the output layer both adopt the sigmoid function f _j (x)=1/(1+e ^-x );

Step 5. Determine the probability of thermal runaway of the battery according to the stage level of the thermal runaway of the battery, and adopt cooling measures for the battery and remind drivers and passengers, specifically including:

When ο ₁ =0 and ο ₂ ≥90%, the battery has no probability of thermal runaway;

When ο ₁ = 1 and ο ₂ ≥ 80%, the probability of thermal runaway of the battery is less than 50% and the single cell of the battery needs to be cooled;

When ο ₁ =2 and ο ₂ ≥70%, the probability of thermal runaway of the battery is greater than 50% and the battery pack module needs to be cooled;

When ο ₁ = 3 and ο ₂ ≥ 60%, the battery pack module is thermally out of control, and the driver and passengers need to be reminded to avoid danger in time and sound an alarm.

Although the embodiment of the present invention has been disclosed as above, it is not limited to the application listed in the description and the embodiment, and it can be applied to various fields suitable for the present invention. For those skilled in the art, it can be easily Therefore, the invention is not limited to the specific details and embodiments shown and described herein without departing from the general concept defined by the appended claims and the scope of equivalents.

Claims (7)

1. A detection method of a thermal runaway gas-sensitive alarm device of a vehicle-mounted lithium ion power battery is characterized by comprising the following steps of:

step one, according to a sampling period, collecting the concentration of carbon monoxide characteristic gas in a box body, the temperature in a plurality of battery pack modules and the mass loss percentage of a battery monomer, and constructing a battery state representation matrix:

Ω＝(Q,T,Δm) ^T ；

in the formula, Q is a concentration vector of carbon monoxide characteristic gas in the box body, T is a temperature vector of each monitoring point in the battery pack module, and Δ m is mass loss percentage of each battery monomer;

step two, the battery state data is processed according to the following steps of 1: 1, randomly dividing the proportion into a training set and a testing set, carrying out statistics on the appeared battery state representation matrix on the training set according to a time sequence principle to obtain a state space of the battery, establishing a Markov chain prediction model, and obtaining a state transition matrix:

in the formula (II)

For the battery from state omega _i Transition to state Ω _j I 1,2, … N, j 1,2, … N, N being the unique number of states of the battery;

step three, predicting the battery state [ omega ] of each battery state in the test set in the next three sampling periods by means of the established effective Markov chain model _t+T ,Ω _t+2T ,Ω _t+3T ]T is the current time tag, and T is the sampling period;

inputting the battery states, the Markov chain validity and the SOC value of the battery in the next three sampling periods into a BP neural network model to obtain the stage grade of the thermal runaway of the battery;

judging the probability of the thermal runaway of the battery according to the stage grade of the thermal runaway of the battery, and adopting a cooling measure for the battery and reminding drivers and passengers;

wherein, the thermal runaway gas-sensitive alarm device of on-vehicle lithium ion power battery includes:

a box body; and

the battery pack modules are arranged inside the box body at intervals, the battery cells are uniformly arranged inside the battery pack modules, and cooling flow field channels are formed among the battery pack modules;

the gas sensor is arranged at the outlet position of the cooling flow field channel;

a plurality of thermocouple temperature sensors uniformly arranged inside the plurality of battery pack modules;

the mass sensors are arranged at the lower parts of the battery cells in a one-to-one correspondence manner;

and the management device is connected with the gas sensor, the thermocouple temperature sensors and the mass sensors and is used for receiving signals and transmitting commands.

2. The detection method of the thermal runaway gas-sensitive alarm device of the vehicle-mounted lithium ion power battery as claimed in claim 1, wherein the battery is in a state omega _i Transition to state Ω _j The probability of (c) satisfies:

in the formula,

for the state transition omega of the battery under the actual measurement experiment _i To state omega _j The number of times that this occurs is,

for the battery state omega under the actual measurement experiment _i Total number of occurrences.

3. The detection method of the thermal runaway gas-sensitive alarm device of the vehicle-mounted lithium-ion power battery as claimed in claim 2, wherein the process of judging whether the Markov chain model is valid is as follows:

and predicting the battery state of each battery state in the test set in a future sampling period according to the state transition matrix, comparing a prediction result with a real state, if the effectiveness of the Markov chain is more than 90%, proving that the Markov chain prediction model is effective, and otherwise, reestablishing the Markov chain prediction model.

4. The detection method of the thermal runaway gas-sensitive alarm device of the vehicle-mounted lithium-ion power battery as claimed in claim 3, wherein the Markov chain validity meets the following requirements:

in the formula, K is the accurate number of the prediction results, and K is the number of the states of all the test sets.

5. The detection method of the thermal runaway gas-sensitive alarm device of the vehicle-mounted lithium ion power battery as claimed in claim 4, wherein the SOC value of the battery is obtained by real-time current/voltage signal query after a battery SOC-OCV curve is obtained based on an offline bench test.

6. The detection method of the thermal runaway gas-sensitive alarm device of the vehicle-mounted lithium ion power battery as claimed in claim 5, wherein the calculation process of the BP neural network model is as follows:

step 1, determining an input layer neuron vector x ═ x of a three-layer BP neural network ₁ ,x ₂ ,x ₃ ,x ₄ ,x ₅ }; wherein x is ₁ Is in a battery state omega _t+T ，x ₂ Is in a battery state omega _t+2T ，x ₃ Is in a battery state omega _t+3T ，x ₄ For Markov chain validity f, x ₅ Is the SOC value of the battery;

step 2, the vector of the input layer is mapped to a hidden layer, the number of hidden layer neurons is h, and

in the formula, m is the number of input nodes, n is the number of output nodes, and a is an adjusting factor;

step 3, obtaining an output layer neuron vector o ═ o ₁ ,o ₂ }；o ₁ Is a battery thermal runaway risk level, o ₂ Is o ₁ The reliability of the prediction result;

wherein,

0 is zero-order risk, which indicates that the battery is normal, 1 is first-order risk, which indicates that the battery monomer can generate thermal runaway and needs to be continuously monitored, 2 is second-order risk,the thermal runaway of the battery monomer is shown and needs to be cooled, and 3 is a three-level risk, which shows that the thermal runaway of the battery pack module occurs.

7. The detection method of the thermal runaway gas-sensitive alarm device of the vehicle-mounted lithium ion power battery as claimed in claim 6, wherein the fifth step specifically comprises:

when o is ₁ 0 and o ₂ When the temperature is more than or equal to 90 percent, the battery has no probability of thermal runaway;

when o is ₁ 1 and o ₂ When the temperature of the battery is more than or equal to 80%, the probability of thermal runaway of the battery is less than 50%, and the monomer of the battery needs to be cooled;

when o is ₁ 2 and o ₂ When the temperature of the battery pack is more than or equal to 70%, the probability of thermal runaway of the battery is more than 50%, and the battery pack module needs to be cooled;

when o is ₁ 3 and o ₂ And when the temperature is more than or equal to 60 percent, the battery pack module generates thermal runaway and needs to remind drivers and passengers to avoid danger in time and send out an alarm sound.

Images