CN-121995245-A - Power battery thermal runaway monitoring method based on machine smell

Abstract

The development of electric power is rapid, and the power battery is increasingly applied to electric automobiles, vertical take-off and landing aircrafts and the like. The power battery is out of control and serious in damage. When the thermal runaway enters the middle and late stages, the thermal runaway is difficult to reverse, so that a novel monitoring and early warning method is required to be developed urgently. According to the power battery thermal runaway monitoring method based on machine smell, the characteristic gases and organic volatile matters released by the battery in the thermal runaway incubation period and the initial period are captured through the plurality of gas collector arrays fused with the plurality of sensing modules, and the early warning and evolution stage tracking of the thermal runaway is realized by combining the multi-mode information fusion recognition technology, so that the problems of hysteresis, insufficient reliability and easiness in interference existing in the existing power battery thermal runaway monitoring method and system are solved, the improvement of the safety of a power battery system is facilitated, and conditions are provided for reliable operation and intelligent operation and maintenance of equipment such as electric automobiles, airplanes and energy storage systems.

Inventors

• LI LIJUN
• DING RUIZHE
• Hou Yueyao

Assignees

• 北京航空航天大学

Dates

Publication Date

20260508

Application Date

20260213

Claims (1)

1. 1. The power battery thermal runaway monitoring method based on machine smell is characterized by comprising the following steps of: Step S1, gas sampling and sensing, namely arranging a plurality of gas collectors at key positions above a power battery module, above an exhaust channel and in a converging area in a battery box, monitoring gas and volatile matters generated by the power battery module, guiding the gas to be tested into the gas collectors through a micro pump and a gas pipeline, and obtaining specific gas type and concentration information generated by the power battery module through a sensing module and a bionic snail type spiral air chamber of the gas collectors; The gas collector consists of a gas inlet, a sensing module, a bionic snail type spiral gas chamber, a gas collector shell, a gas outlet and a signal preprocessing module; the sensor module is a sensor array, at least comprises a semiconductor gas sensor for monitoring electrolyte solvent vapor, an electrochemical or metal oxide semiconductor sensor for monitoring carbon monoxide, a broad-spectrum semiconductor sensor for monitoring hydrocarbon gas, a bionic snail spiral air chamber, a spiral ascending gas flow channel, a system, a cleaning air channel, a built-in high-precision temperature and humidity sensor, a software real-time compensation device and a long-term stability protection device, wherein the semiconductor gas sensor is used for monitoring electrolyte solvent vapor, the electrochemical or metal oxide semiconductor sensor is used for monitoring hydrogen gas, the broad-spectrum semiconductor sensor is used for monitoring hydrocarbon gas, each sensor in the array has cross sensitivity to battery characteristic gas and jointly forms a response spectrum for monitoring battery thermal runaway gas; S2, preprocessing gas and volatile raw data generated by a power battery module through a signal preprocessing module, wherein the preprocessing comprises filtering denoising, baseline correction and temperature and humidity compensation, and extracting feature vectors from preprocessed response signals, wherein the feature vectors comprise, but are not limited to, time sequences and associated features such as absolute concentration of each sensor, dynamic response change rate, response curve integral area, response ratio among the sensors, sequence of occurrence of specific gas and the like, so as to form a high-dimensional feature vector; Let the gas sensor array output n-dimensional eigenvectors: Wherein g i (t) is the normalized response value of the ith gas sensor; the feature extraction uses a sliding window (window size W): Wherein: Is the mean value; Is the standard deviation; is a first order difference; is the main component characteristic; Step S3, multi-mode information fusion and identification, namely extracting characteristics of voltage, current and temperature data provided by a Battery Management System (BMS), and fusing the identified characteristic data of gas and organic volatile matters with the characteristics of the voltage, current and temperature data provided by the Battery Management System (BMS); voltage, current, temperature data vector provided by Battery Management System (BMS): Wherein m is the number of temperature sensors; the extraction formula of the voltage, current and temperature data features is as follows: because the sampling frequencies of the sensors are different, cubic spline interpolation is adopted for time alignment: wherein t k is a fusion timestamp; The weighting characteristic splicing and fusion method is adopted: Wherein, the Representing Hadamard products (Hadamard products); representing vector stitching; For the self-adaptive weight matrix, calculating by an attention mechanism, updating weight parameters on line to adapt to battery aging; let the probability distribution function of the gas and volatile data generated by the power battery module and the data provided by the BMS system be m G and m B respectively, then output probability after fusion is: Wherein, the Is a collision function; S4, judging a thermal runaway state, namely transmitting a data preprocessing and feature extraction result to a microprocessor for operation analysis, adopting a Convolutional Neural Network (CNN) and a long-short-term memory network (LSTM) fusion model, wherein the CNN is used for processing spatial features extracted from sensor array response, the LSTM is used for processing time sequence features of gas release, and the model is input into a multi-dimensional time sequence signal and outputs thermal runaway risk probability, fault type probability distribution and development stage; the method comprises the steps of inputting the extracted feature vector into a pre-trained thermal runaway identification model to judge the thermal runaway state, wherein the thermal runaway identification model is trained by a machine learning algorithm, training data of the thermal runaway identification model is derived from gas response data collected by a gas collector array and a corresponding known battery state label thereof and a voltage, current and temperature state label of a BMS system in the whole process from initial abnormality to thermal runaway under various abuse conditions and different aging states of batteries of different types; Step S5, grading early warning and linkage control, wherein the grading early warning and linkage control module executes according to the state of the power battery: if the model output is in a normal state, continuing to monitor; If the model output is in an early warning state, if slight abnormality, such as micro gassing, occurs in the corresponding battery, a first-level warning is started, data are recorded, inspection is prompted, audible and visual warning is performed, and the cloud platform is notified; If the model output is in a thermal runaway early warning state, corresponding to the starting of the thermal runaway chain reaction and the abrupt change of the concentration of the characteristic gas, immediately starting a secondary early warning, sending a highest-level alarm to the BMS system through a communication interface, triggering forced outage, starting a directional fire protection system, carrying out audible and visual alarm, prompting linkage measures of personnel evacuation, and informing a cloud platform.

Description

Power battery thermal runaway monitoring method based on machine smell Technical Field The invention relates to a power battery thermal runaway monitoring method, in particular to a power battery thermal runaway monitoring method based on machine smell, and belongs to the technical field of power battery safety. Background Along with the rapid development of global energy structure transformation and electric power generation, the power battery is increasingly applied to electric automobiles, vertical take-off and landing aircrafts and the like. However, the occurrence of thermal runaway accidents of the power battery severely restricts the safety and reliability of the large-scale application of the power battery. In the use process of the power battery, thermal runaway is easy to occur under the action of reasons such as mechanical abuse, electric abuse, thermal abuse and the like. For a power battery for a vehicle, once thermal runaway occurs, great threat may be caused to personal safety of personnel in the vehicle and surrounding personnel. At present, the power battery monitoring method mainly depends on physical parameters such as temperature, voltage, current and the like, and has the problems of hysteresis, insufficient reliability, easiness in interference and the like. The obvious change of temperature and voltage often occurs in the middle and later stages of irreversible thermal runaway, the early warning time window is extremely short, enough time cannot be reserved for personnel evacuation, and hysteresis is shown. The battery is small and local in temperature rise at the early stage of short circuit, a surface temperature sensor is difficult to capture effectively, and voltage change can be covered up due to an equalization strategy of a battery management system, so that the reliability is insufficient. Electromagnetic environments of vehicle-mounted electronic and electric equipment and an energy storage system are complex, and a temperature sensor is easily influenced by environmental temperature and heat dissipation conditions, so that false alarm or missing alarm is easily generated. Studies have shown that power cells are accompanied by the release of characteristic gases before and during the occurrence of thermal runaway. For example, under abusive conditions such as overcharging, internal short-circuiting, overheating, etc., the electrolyte may decompose to produce gases such as CO, H 2、CH4、C2H4, etc., electrolyte solvent vapors, volatiles, etc. The type, concentration and release sequence of these gases are strongly correlated with the failure mode of the battery, the stage of development, and are characteristic signals earlier and more direct than temperature and voltage. The machine olfactory technique utilizes an array of gas sensors to detect gas by mimicking the bio-olfactory system. Therefore, the high-sensitivity gas sensor is used for capturing characteristic gases, volatile organic compounds and electrolyte decomposition products released in the early stage of thermal runaway of the battery in real time, and the power battery thermal runaway monitoring method based on machine smell is provided by combining an artificial intelligence technology, so that the method has important engineering significance and is beneficial to improving the safety of the power battery. Disclosure of Invention 1. The invention aims to: The invention aims to solve the technical problems of hysteresis, insufficient reliability and easy interference of the existing power battery thermal runaway monitoring method and system, provides a power battery thermal runaway monitoring method based on machine smell, according to the method, the characteristic gas and the organic volatile matters released by the battery in the early gestation period and the initial period of the thermal runaway are captured, so that early warning and evolution stage tracking of the thermal runaway are realized, and timeliness and reliability of the warning are remarkably improved. 2. The technical scheme is as follows: The power battery thermal runaway monitoring method based on machine smell is characterized by comprising the following steps of: Step S1, gas sampling and sensing, namely arranging a plurality of gas collectors at key positions above a power battery module, above an exhaust channel and in a converging area in a battery box, monitoring gas and volatile matters generated by the power battery module, guiding the gas to be tested into the gas collectors through a micro pump and a gas pipeline, and obtaining specific gas type and concentration information generated by the power battery module through a sensing module and a bionic snail type spiral air chamber of the gas collectors; The gas collector consists of a gas inlet, a sensing module, a bionic snail type spiral gas chamber, a gas collector shell, a gas outlet and a signal preprocessing module; the sensor module is a sensor array, at least comprises a semicond