Multi-dimensional thermal runaway prediction model for battery

Abstract

The invention is suitable for the technical field of battery safety management, and provides a battery multi-dimensional thermal runaway prediction model, which comprises a sensor layer, an edge calculation layer and a cloud analysis layer, the sensor layer comprises an electrochemical impedance spectroscopy module, a temperature matrix sensor module and a gas sensor array module; the edge calculation layer comprises a data synchronization alignment module, a feature extraction engine and an anomaly detection module; the cloud analysis layer comprises a thermal runaway prediction model and an early warning decision module; preferably, the electrochemical impedance spectroscopy module scans at a high frequency of 0.1-100 kHz, and the resolution of a phase angle is 0.1 degree; the temperature matrix sensor module comprises a 16-channel thermocouple array, and the spatial resolution is + / -1mm; the gas sensor array module comprises MEMS gas sensors, and the detection precision is 1 ppm. According to the invention, the electrochemical state in the battery can be sensed in real time, and time-space correlation analysis of multi-dimensional data and thermal runaway early warning migration of cross-model batteries can be carried out.

Description

Technical Field

[0001] The present invention relates to the technical field of battery safety management, and particularly to a battery multi-dimensional thermal runaway prediction model. Background Art

[0002] Battery safety management refers to taking various measures throughout the entire process of battery research and development, production, use, and recycling to ensure the safety, stability, and reliability of the battery, and to avoid safety problems such as overheating, short circuit, fire, and explosion during battery use.

[0003] The wide application of lithium-ion batteries in modern electronic devices, electric vehicles, and energy storage systems has brought high efficiency and convenience, but their safety issues have also become the focus of attention. Therefore, safety early warning is required. The safety early warning system for lithium-ion batteries is to identify potential risks of the battery in advance, such as overheating, overcharging, short circuit, physical damage, etc., so as to avoid dangerous situations such as thermal runaway, fire, or explosion of the battery, and is one of the important means of battery safety management.

[0004] The prediction of thermal gravimetric loss of lithium-ion batteries is to study the process of thermal runaway of batteries in a high-temperature environment. Thermal runaway refers to the phenomenon that the temperature of the battery rises sharply, gas is released, the structure is damaged, and even a fire is triggered when heat accumulates or the external temperature is too high. Thermal gravimetric loss is an important sign of the gradual reduction of the battery mass during the thermal runaway process of lithium-ion batteries, and is usually related to chemical reactions and physical changes inside the battery.

[0005] The prediction of thermal gravimetric loss of lithium-ion batteries is an important link in battery safety management, which can evaluate the performance and potential risks of the battery in a high-temperature environment in advance. Through various methods, including thermogravimetric analysis, thermal runaway simulation, data-driven prediction, etc., the thermal gravimetric loss of the battery can be effectively predicted, and measures can be taken to optimize the battery design and improve the thermal management system to ensure the safety and reliability of the battery.

[0006] Traditional thermal runaway early warning relies on single temperature monitoring, and cannot fundamentally solve the problem of fire accidents caused by battery charging. The false alarm rate is high and the response is lagged. In addition, there are also problems such as the inability to perceive the real-time electrochemical state inside the battery, the inability to perform spatio-temporal correlation analysis of multi-dimensional data, and the inability to migrate the thermal runaway early warning of cross-model batteries. Therefore, the present invention proposes a battery multi-dimensional thermal runaway prediction model, aiming to comprehensively consider multiple key parameters during the thermal runaway process of the battery to improve the accuracy and reliability of the prediction. Summary of the Invention

[0007] The object of the present invention is to provide a multi-dimensional thermal runaway prediction model for batteries, so as to solve the problems in the prior art that traditional thermal runaway early warning relies on single temperature monitoring, cannot fundamentally solve the problem of fire accidents caused by battery charging, has a high false alarm rate and a lagging response, and there are also problems such as the inability to perceive the internal electrochemical state of the battery in real time, the inability to perform spatio-temporal correlation analysis of multi-dimensional data, and the inability to migrate the thermal runaway early warning of cross-model batteries.

[0008] To achieve the above object, the present invention provides the following technical solution: A multi-dimensional thermal runaway prediction model for batteries, including a sensor layer, an edge computing layer and a cloud analysis layer;

[0009] The sensor layer includes an electrochemical impedance spectroscopy module, a temperature matrix sensor module and a gas sensor array module;

[0010] The edge computing layer includes a data synchronization and alignment module, a feature extraction engine and an anomaly detection module;

[0011] The cloud analysis layer includes a thermal runaway prediction model and a warning decision-making module.

[0012] Preferably, the electrochemical impedance spectroscopy module performs high-frequency scanning at 0.1 - 100 kHz with a phase angle resolution of 0.1°; the temperature matrix sensor module includes a 16-channel thermocouple array with a spatial resolution of ±1 mm; the gas sensor array module includes MEMS gas sensors with a detection accuracy of 1 ppm.

[0013] Preferably, the data synchronization and alignment module performs timestamp alignment and data caching queue; the feature extraction engine includes three-dimensional thermal field reconstruction and temperature gradient calculation; the anomaly detection module includes gas release rate analysis and multi-source data fusion.

[0014] Preferably, the thermal runaway prediction model includes LSTM spatio-temporal feature extraction and a transfer learning adapter; the warning decision-making module includes risk level assessment and pre-start of the fire extinguishing system.

[0015] Preferably, the LSTM spatio-temporal feature extraction adopts an LSTM-GNN hybrid model, including:

[0016] LSTM extracts the time series features of a single battery with a time window of 60 s;

[0017] The graph neural network models the multi-battery correlation effect and analyzes the propagation path.

[0018] Preferably, the transfer learning adapter is based on the feature transfer of abnormal data of power batteries.

[0019] Preferably, the risk level assessment is divided into two levels:

[0020] Level 1: EIS phase angle shift > 5% + CO concentration > 50 ppm → Early warning 15 - 25 minutes in advance;

[0021] Level 2: Temperature gradient > 8 °C / cm + H2 concentration > 100 ppm → Trigger pre - startup of the fire extinguishing system.

[0022] Preferably, the operations performed by the feature extraction engine include:

[0023] Extract feature points of the EIS Nyquist plot, including real - part / imaginary - part curvature changes;

[0024] Calculate the temperature field gradient vector, and the calculation formula is

[0025] Gas release rate fitting, fitting with the d[CO] / dt exponential growth model.

[0026] The present invention has at least the following beneficial effects:

[0027] A multi - dimensional thermal runaway prediction model for batteries provided by the present invention can perform real - time perception of the internal electrochemical state of the battery, conduct spatio - temporal correlation analysis of multi - dimensional data, and thermal runaway warning migration of cross - model batteries, and is applicable to the safety warning scenarios of lithium - ion batteries for electric bicycles, new energy vehicles, and energy storage devices. Brief Description of the Drawings

[0028] Figure 1 It is a schematic diagram of the system architecture of the present invention;

[0029] Figure 2 It is the hardware structure of the sensor layer of the present invention;

[0030] Figure 3 It is the algorithm flow chart of the model of the present invention. Detailed Description of the Embodiments

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Apparently, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0032] Embodiment 1

[0033] As Figure 1 shown, this embodiment provides a multi - dimensional thermal runaway prediction model for batteries, including a sensor layer, an edge computing layer, and a cloud analysis layer;

[0034] The sensor layer includes an Electrochemical Impedance Spectroscopy (EIS) module, a temperature matrix sensor module, and a gas sensor array module. The hardware structure of the sensor layer is as Figure 2 shown;

[0035] The edge computing layer includes a data synchronization and alignment module, a feature extraction engine, and an anomaly detection module;

[0036] The cloud analysis layer includes a thermal runaway prediction model and a warning decision module.

[0037] Among them, the Electrochemical Impedance Spectroscopy module performs high-frequency scanning at 0.1 - 100 kHz with a phase angle resolution of 0.1°; the temperature matrix sensor module includes a 16-channel thermocouple array with a spatial resolution of ±1 mm; the gas sensor array module includes MEMS gas sensors with a detection accuracy of 1 ppm.

[0038] Among them, the data synchronization and alignment module performs timestamp alignment and data caching queue; the feature extraction engine includes three-dimensional thermal field reconstruction and temperature gradient calculation; the anomaly detection module includes gas release rate analysis and multi-source data fusion.

[0039] Among them, the thermal runaway prediction model includes LSTM spatio-temporal feature extraction and a transfer learning adapter; the warning decision module includes risk level assessment and pre-start of the fire extinguishing system.

[0040] 1. The core of the prediction model algorithm adopted in the present invention includes:

[0041] 1) Multi-dimensional sensing fusion:

[0042] ① Electrochemical layer: High-frequency EIS scanning (0.1 - 100 kHz) monitors the impedance change of the SEI film in real time (sensitivity increased by 5 times);

[0043] ② Physical layer: A 16-channel temperature matrix constructs a three-dimensional thermal field map (spatial error < 2°C);

[0044] ③ Chemical layer: The MEMS gas sensor array detects the CO / H2 release rate (response time < 3 s).

[0045] 2) Spatio-temporal correlation algorithm:

[0046] ① LSTM-GNN hybrid model:

[0047] LSTM extracts the single-cell time series features (time window 60 s);

[0048] The Graph Neural Network (GNN) models the multi-cell correlation effect (propagation path analysis).

[0049] ② Transfer learning adaptation:

[0050] Feature migration based on abnormal data of power batteries (adaptability > 85%).

[0051] 3) Hierarchical early warning mechanism

[0052] ①Level 1 (Early warning): EIS phase angle deviation > 5% + CO concentration > 50 ppm → Early warning 15 - 25 minutes in advance;

[0053] ②Level 2 (Emergency response): Temperature gradient > 8℃ / cm + H2 concentration > 100 ppm → Trigger pre - start of the fire extinguishing system.

[0054] 2. Work carried out by the feature extraction engine in the present invention:

[0055] Extract feature points of the EIS Nyquist diagram (real part / imaginary part curvature change);

[0056] Calculate the temperature field gradient vector

[0057] Fit the gas release rate (d[CO] / dt exponential growth model).

[0058] 3. Model training and deployment of the present invention:

[0059] Train the LSTM - GNN model with millions of battery abnormal samples (F1 - score = 0.93);

[0060] Deploy a lightweight model in the edge computing layer (optimized by TensorRT, inference latency < 200 ms).

[0061] A battery multi - dimensional thermal runaway prediction model provided by the present invention, the algorithm flow chart is as Figure 3 shown.

[0062] Apply the technical model provided in this embodiment to the case for evaluation, and the results are as follows:

[0063] a. Implementation case (electric bicycle battery pack):

[0064] Deploy this system in a 2 kWh lithium - ion battery pack, and successfully trigger a Level 1 warning 18 minutes before thermal runaway, accurately identifying an internal micro - short circuit.

[0065] b. Experimental data:

[0066] Dataset: 1.05 million groups of battery abnormal data (covering types such as NMC / LFP / lithium titanate, etc.).

[0067] Performance indicators:

[0068] Recall: 96.3%;

[0069] Precision: 94.7%;

[0070] Model inference speed: 187 ms / sample (NVIDIA Jetson Xavier edge device).

[0071] Comparative Example 1

[0072] This comparative example provides a battery multi-dimensional thermal runaway prediction model, which is the same as Example 1, except that infrared thermal imaging is used to replace contact temperature sensing.

[0073] Compared with Example 1, the cost increases by 3 times and the accuracy drops by ±3°C.

[0074] Comparative Example 2

[0075] This comparative example provides a battery multi-dimensional thermal runaway prediction model, which is the same as Example 1, except that a convolutional neural network (CNN) is used to process the temperature matrix.

[0076] Compared with Example 1, the spatial gradient information of the loss is lost and the F1-score drops by 12%.

[0077] Comparing Example 1 with the traditional solution of the present invention, the results are as follows:

[0078] Indicator Traditional solution (single temperature monitoring) This invention (multi-dimensional fusion) Early warning lead time 5 - 10 minutes 15 - 25 minutes False alarm rate 18％-25％ <3％ Cross-model adaptability Retraining required (fitness < 50%) Transfer learning (adaptability > 85%) Computing resource occupancy Cloud centralized processing (delay > 2s) Edge computing (latency < 200 ms)

[0079] The above shows and describes the basic principles, main features and advantages of the present invention. For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic features of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to embrace all changes within the meaning and scope of the equivalent elements of the claims in the present invention.

[0080] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional thermal runaway prediction model for a battery, characterized in that, It includes a sensor layer, an edge computing layer, and a cloud analysis layer; The sensor layer includes an electrochemical impedance spectroscopy module, a temperature matrix sensor module, and a gas sensor array module; The edge computing layer includes a data synchronization and alignment module, a feature extraction engine, and an anomaly detection module; The cloud analysis layer includes a thermal runaway prediction model and a warning decision module.

2. The multi-dimensional thermal runaway prediction model of a battery according to claim 1, wherein The electrochemical impedance spectroscopy module performs high-frequency scanning at 0.1 - 100 kHz with a phase angle resolution of 0.1°; the temperature matrix sensor module includes a 16-channel thermocouple array with a spatial resolution of ±1 mm; the gas sensor array module includes MEMS gas sensors with a detection accuracy of 1 ppm.

3. The multi-dimensional thermal runaway prediction model of a battery according to claim 1, wherein The data synchronization and alignment module performs timestamp alignment and data caching queue; the feature extraction engine includes three-dimensional thermal field reconstruction and temperature gradient calculation; the anomaly detection module includes gas release rate analysis and multi-source data fusion.

4. The multi-dimensional thermal runaway prediction model of a battery according to claim 1, characterized in that, The thermal runaway prediction model includes LSTM spatio-temporal feature extraction and a transfer learning adapter; the warning decision module includes risk level assessment and pre-start of the fire extinguishing system.

5. The multi-dimensional thermal runaway prediction model for a battery according to claim 4, wherein The LSTM spatio-temporal feature extraction uses an LSTM-GNN hybrid model, including: LSTM extracts the time series features of a single battery with a time window of 60 s; The graph neural network models the correlation effects of multiple batteries and analyzes the propagation path.

6. The multi-dimensional thermal runaway prediction model for a battery according to claim 4, wherein The transfer learning adapter is based on the feature transfer of abnormal data of power batteries.

7. The multi-dimensional thermal runaway prediction model of a battery according to claim 4, wherein The risk level assessment is divided into two levels: Level 1: EIS phase angle shift > 5% + CO concentration > 50 ppm → early warning 15 - 25 minutes in advance; Level 2: temperature gradient > 8℃ / cm + H2 concentration > 100 ppm → trigger pre-start of the fire extinguishing system.

8. The multi-dimensional thermal runaway prediction model for a battery according to claim 1, wherein The work performed by the feature extraction engine includes: Extract the feature points of the EIS Nyquist diagram, including the curvature changes of the real part / imaginary part; Calculate the temperature field gradient vector, and the calculation formula is Fit the gas release rate with a d[CO] / dt exponential growth model.

Images