Early warning and cooling system for thermal runaway of battery

Abstract

The invention relates to the technical field of batteries, and discloses a battery thermal runaway early warning and cooling system which comprises a multi-parameter monitoring module, a data processing and early warning model and a cooling system control module. Wherein the multi-parameter monitoring module is used for acquiring the internal resistance value of each battery monomer in real time in a battery charging process, calculating the internal resistance range, the internal resistance standard deviation and the variable coefficient according to the internal resistance value at a sampling time point, monitoring the working environment gas of the battery through a mass spectrometer, and acquiring the real-time mass spectrum monitoring data of the gas in real time; release characteristics, including hydrogen, hydrocarbon compounds and the like, of escaped gas are quantitatively analyzed on the basis of an equivalent characteristic spectrum method. According to the battery thermal runaway early warning and cooling system, various key parameters in the battery charging process are comprehensively monitored, early warning of the battery thermal runaway risk is achieved, timely intervention is conducted in combination with the cooling system, the thermal runaway risk is reduced, and vehicle and personal safety is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of battery technology, and in particular to a battery thermal runaway early warning and cooling system. Background Art

[0002] With the vigorous rise of the new energy vehicle industry, battery safety issues have become increasingly prominent as the key "heart" of this field and have become a top priority for the development of the industry. Battery thermal runaway, a potential safety hazard, is like a sword of Damocles hanging over new energy vehicles, threatening the safety of vehicles and passengers at all times. When the temperature inside the battery rises abnormally, if it is not effectively controlled in a timely manner, it will trigger a series of chain reactions, causing the battery to heat up and expand rapidly, and even cause fire or explosion. The consequences are disastrous. Therefore, strengthening the early warning and prevention of battery thermal runaway has become a difficult problem that the new energy vehicle industry must overcome. From material innovation, process optimization to the research and development and application of intelligent monitoring systems, improving the safety performance of batteries in all aspects is not only an urgent need to protect the safety of consumers' lives and property, but also the key to promoting the healthy and sustainable development of the new energy vehicle industry.

[0003] Among the existing battery thermal runaway warning technologies, although a variety of methods have emerged, most are still limited to the monitoring of a single parameter. This limitation significantly weakens the accuracy and timeliness of the warning system. Specifically, many warning systems only rely on a single indicator such as battery temperature or voltage for monitoring. However, battery thermal runaway is a complex process, and its occurrence is often accompanied by the coordinated changes of multiple physical parameters. The monitoring of a single parameter is difficult to fully capture the abnormal state inside the battery, and it is easy to miss or falsely report in the early stage of thermal runaway, resulting in a significant reduction in the accuracy of the warning. In addition, due to the delayed response, even if the warning system eventually issues an alarm, it may have missed the best time for intervention and cannot effectively curb the further development of thermal runaway. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] The purpose of the present invention is to realize early warning of the risk of thermal runaway of the battery by comprehensively monitoring multiple key parameters in the battery charging process, and timely intervene in combination with the cooling system to reduce the risk of thermal runaway and ensure the safety of the vehicle and people, and to propose a battery thermal runaway early warning and cooling system.

[0006] (II) Technical solution

[0007] The technical solution of the present invention to solve the above technical problems is as follows:

[0008] Battery thermal runaway early warning and cooling system, the warning and cooling system includes: multi-parameter monitoring module, data processing and warning model and cooling system control module;

[0009] Among them, the multi-parameter monitoring module obtains the internal resistance value of each battery cell in real time during the battery charging process, and calculates the internal resistance range, internal resistance standard deviation and coefficient of variation based on the internal resistance value at the sampling time point, monitors the battery working environment gas through a mass spectrometer, obtains real-time gas mass spectrometry monitoring data in real time, and quantitatively analyzes the release characteristics of the escaped gas based on the equivalent characteristic spectrum method, including hydrogen, hydrocarbon compounds, etc., and sets a temperature sensor on the surface of the battery shell to monitor the temperature change of the battery surface in real time, especially the temperature difference between the highest point and the lowest point;

[0010] The data processing and early warning model uses the data of internal resistance monitoring, gas monitoring and temperature monitoring as input into a pre-trained machine learning model. The model is built based on deep learning algorithms such as Transformer and can process multi-dimensional time series data. It evaluates the thermal runaway risk of the battery through feature extraction and classification prediction. When the model predicts that the battery has a thermal runaway risk, it generates a corresponding early warning signal according to the risk level and sends the signal to the control module.

[0011] Among them, after the cooling system control module receives the early warning signal and the cooling system is started, it continuously monitors the battery temperature until the temperature drops to a safe range to ensure battery safety.

[0012] Based on the above technical solution, the present invention can also be improved as follows.

[0013] Furthermore, the multi-parameter monitoring module specifically includes an internal resistance monitoring unit, a gas monitoring unit and a temperature monitoring unit;

[0014] A high-precision, low-drift DC internal resistance sensor is used in the internal resistance monitoring unit to ensure that the internal resistance value of each battery cell can be stably and accurately measured during the battery charging process. The internal resistance sensor is directly connected to the positive and negative electrodes of the battery cell to ensure the accuracy and real-time nature of the measured data. A high-precision data acquisition card or integrated module is used to collect the internal resistance value at each sampling time point in real time and perform analog-to-digital conversion. The internal resistance range is calculated: within the preset time window, the maximum and minimum values ​​of the internal resistance are found, and the difference between the two is the internal resistance range. The statistical method is used to calculate the mean and standard deviation of all internal resistance values ​​within the time window. The standard deviation reflects the discrete degree of the internal resistance. The standard deviation is divided by the mean to obtain the coefficient of variation, which is used to evaluate the relative fluctuation of the internal resistance. The internal resistance monitoring module adopts an independent design for easy maintenance and upgrading. The module is connected to the main control system through standard communication interfaces such as CAN bus and RS485 to achieve real-time data transmission and command reception.

[0015] The gas monitoring unit uses a high-sensitivity, high-resolution mass spectrometer that can accurately detect trace gas components in the battery working environment. The mass spectrometer needs to be equipped with a special sampling probe and preprocessing system to ensure the representativeness of the sampled gas and the accuracy of the test results. The mass spectrometer outputs real-time mass spectrometry monitoring data of the gas, including gas composition and concentration information, and compares the detected gas composition with the predefined equivalent characteristic spectrum. The release characteristics of the escaping gas are determined by quantitative analysis methods, with special attention paid to gas components such as hydrogen and hydrocarbon compounds that may indicate thermal runaway of the battery, and their concentration change trends are recorded. It also includes the design of the monitoring module. The gas monitoring module is independently designed to reduce interference with other monitoring units. The data processing unit has a built-in data processing unit to process and analyze the data output by the mass spectrometer in real time.

[0016] High-precision and high-stability temperature sensors, such as platinum resistance thermometers and thermocouples, are selected in the temperature monitoring unit. Temperature sensors are arranged at multiple key positions of the battery shell, such as the highest point, the lowest point, and the center point, to comprehensively monitor the temperature changes on the battery surface. The data of each temperature sensor is collected in real time through the data acquisition system, and the temperature difference between the highest point and the lowest point is calculated as an important indicator for evaluating the temperature uniformity of the battery. The changing trend of the battery surface temperature is monitored to detect abnormal temperature rise in time. The temperature monitoring module is independently designed to facilitate installation and maintenance. The temperature threshold is set and an alarm signal is triggered when the temperature exceeds the threshold.

[0017] Furthermore, the multi-parameter monitoring module also includes a comprehensive data processing and early warning model, which fuses the data of internal resistance monitoring, gas monitoring and temperature monitoring to form a multi-dimensional time series data set, and uses a machine learning model based on deep learning algorithms such as Transformer to extract features and classify and predict the fused data, evaluate the thermal runaway risk of the battery, generate a corresponding early warning signal based on the model prediction results, and send it to the control module to start the corresponding cooling measures.

[0018] Furthermore, the data processing and early warning model specifically includes a data preprocessing module, a machine learning model construction, a feature extraction and classification prediction, and an early warning signal generation and sending module;

[0019] The data preprocessing module synchronizes the time and formats of the data from the internal resistance monitoring unit, gas monitoring unit, and temperature monitoring unit to ensure that all data are comparable in the same time frame, removes outliers, missing values, and other noise data, and processes discontinuous or excessively fluctuating data points through interpolation and smoothing to improve data quality. Based on the mechanism and prior knowledge of battery thermal runaway, it extracts features useful for assessing thermal runaway risks from the original data, such as internal resistance range, standard deviation, coefficient of variation, gas concentration trend, temperature gradient, etc.;

[0020] Models based on deep learning algorithms such as Transformer are used in the construction of machine learning models because Transformer has advantages in processing long sequence data and capturing global dependencies, and is suitable for processing multi-dimensional time series data. A large amount of historical data is used to train the model, including data in normal charging state and data before thermal runaway events. Through supervised learning, the model learns how to extract features from input data and predict the thermal runaway risk of the battery. The performance of the model is evaluated through cross-validation, confusion matrix, ROC curve and other methods to ensure that the model has high accuracy and generalization ability;

[0021] In feature extraction and classification prediction, features useful for classification prediction are automatically extracted from the input data. These features may include fluctuation patterns of internal resistance, abnormal changes in gas concentration, and sharp increases in temperature gradients. The trained model is used to classify and predict the real-time input data to determine whether the battery currently has a thermal runaway risk and give a risk level.

[0022] In the warning signal generation and sending module, warning thresholds for different risk levels are set according to business needs and safety standards. When the risk level predicted by the model exceeds the set threshold, the warning mechanism is triggered, and the corresponding warning signal is generated according to the prediction result, including risk level, warning time, recommended measures and other information. The warning signal is sent to the control module through the communication interface so that cooling measures can be taken in time.

[0023] Furthermore, after each warning and cooling operation, the system automatically collects relevant data, including monitoring data before and after the warning, the implementation of cooling measures, etc., and uses the collected data to continuously optimize and calibrate the machine learning model to improve the accuracy and response speed of the warning. This may include adjusting model parameters, updating feature sets, introducing new training data, etc., and regularly monitoring and evaluating the performance of the system to ensure that the system can operate stably.

[0024] Furthermore, the cooling system control module specifically includes: a warning signal receiving and analyzing module, a risk level assessment module, a cooling system execution and monitoring module, and a cooling completion and subsequent processing module;

[0025] Interface design is carried out in the warning signal receiving and parsing module. The cooling system control module needs to have a communication interface that matches the data processing and warning model module to ensure that the warning signal can be transmitted accurately. After receiving the warning signal, the control module needs to immediately parse the signal content, including key information such as risk level, battery location, and current temperature.

[0026] The risk level assessment has a built-in risk assessment model, which can quickly assess the cooling measures to be taken based on the analyzed risk level and battery status such as temperature and internal resistance.

[0027] The cooling system execution and monitoring is provided with actuators, including air cooling system and liquid cooling system. Both the air cooling system and the liquid cooling system are equipped with corresponding actuators such as fans, circulation pumps, etc. The control module drives these mechanisms to work by sending control signals. After the cooling system is started, the control module will continuously monitor the battery temperature, obtain real-time data through the temperature sensor, and compare it with the preset safe temperature range. According to the monitoring results, the control module can adjust the operating status of the cooling system in real time, such as fan speed, coolant flow, etc., to ensure that the battery temperature is always kept within a safe range.

[0028] In the cooling completion and subsequent processing module, when the battery temperature drops to a safe range and remains stable for a period of time, the control module will determine that the cooling process is complete. After cooling is completed, the control module will perform a system reset operation, shut down the cooling system, and restore the normal power supply state of the battery. If an emergency cut-off operation was performed before, during the entire process, the control module will record all key data such as warning time, cooling measures, temperature changes, etc., and feed this data back to the data processing and warning model module for continuous optimization and calibration of the model.

[0029] Furthermore, a cooling strategy selection is also provided in the risk level assessment, which includes three levels: low-level risk, medium-high-level risk and emergency cut-off. If the assessment result is a low-level risk, the control module will start the air cooling system, which includes adjusting the fan speed, turning on or off the fan in a specific area, etc., to achieve effective heat dissipation of the battery. If the assessment result is a medium-high-level risk, the control module will start a more efficient liquid cooling system, which may involve turning on the circulation pump, adjusting the coolant flow and temperature, etc., to quickly reduce the battery temperature. In extreme cases, if it is predicted that thermal runaway is about to occur or the temperature can no longer be effectively controlled, the control module will perform an emergency cut-off operation to cut off the battery power supply to prevent the situation from further deteriorating.

[0030] Furthermore, the early warning and cooling system also includes a system feedback and optimization module. After each early warning and cooling operation, the system automatically collects relevant data for continuous optimization and calibration of the model to improve the accuracy and response speed of the early warning. The data mainly comes from the internal resistance monitoring unit, the gas monitoring unit, the temperature monitoring unit, and the cooling system control module. These modules will generate a large amount of real-time data during the early warning and cooling operation, including but not limited to the internal resistance value of the battery cell, gas concentration, temperature change, working status of the cooling system, etc. The system needs to synchronize the time and format the data from different modules to ensure the integrity and consistency of the data. At the same time, the collected data is preliminarily processed, such as removing noise, filling missing values, etc., to improve the data quality. The performance of the early warning model is evaluated based on the collected data, including indicators such as accuracy, recall rate, and F1 score. By comparing the model prediction results with the actual results, it is analyzed under what circumstances the model performs well and under what circumstances it has deficiencies. The system should continuously monitor the operating status of the battery and the performance of the early warning model to ensure that the model can accurately predict the thermal runaway risk of the battery and generate early warning signals in a timely manner. The performance of the model should be evaluated and feedback should be given regularly to promptly discover and resolve potential problems. At the same time, the optimization direction and goals of the model should be adjusted according to changes and needs in actual application scenarios. As the system runs and data accumulates, the data set should be continuously supplemented and updated to improve the generalization ability and adaptability of the model. At the same time, the model should be continuously optimized and iterated using historical data to ensure that the model is always in the best state.

[0031] Furthermore, when evaluating the performance of the early warning model, the input features may be re-evaluated and adjusted based on the information in the feedback report. It may be necessary to introduce new features or eliminate some features that have little effect on the prediction results to improve the model's predictive ability and adjust the hyperparameters in the model, such as learning rate, batch size, number of iterations, etc., to find the optimal model configuration. This usually requires multiple experiments and comparisons. If the evaluation results show that the model performs poorly in some aspects, it may be necessary to consider improving the algorithm itself. For example, you can try to use different deep learning architectures, optimization algorithms or loss functions to improve the performance and stability of the model. After completing feature selection, parameter adjustment and algorithm improvement, use the new data set to retrain the model. Through multiple iterations and optimization, the model gradually adapts to the needs of actual application scenarios.

[0032] (III) Beneficial effects

[0033] Compared with the prior art, the technical solution of this application has the following beneficial technical effects:

[0034] The system of the present invention realizes real-time monitoring of multiple key parameters of the battery's internal and external environment through a multi-parameter monitoring module, including the internal resistance value of the battery cell, gas release characteristics, and temperature changes on the battery surface. This multi-parameter comprehensive monitoring method overcomes the limitations of single-parameter monitoring and can more comprehensively and accurately capture the abnormal state of the battery's initial thermal runaway, thereby significantly improving the accuracy of early warning. The system adopts a data processing and early warning model based on deep learning algorithms such as Transformer, which can process multi-dimensional time series data and accurately evaluate the thermal runaway risk of the battery through feature extraction and classification prediction, which is different from the traditional method based on a single Compared with the early warning system based on indicators or simple algorithms, the model of the present invention has stronger learning ability and adaptability, can more accurately identify the potential risk of battery thermal runaway, and send out early warning signals, effectively avoiding the problems of missed reports and false alarms. The system has a built-in cooling system control module, which can automatically start corresponding cooling measures according to the risk level after receiving the early warning signal. This rapid response mechanism ensures that effective measures can be taken to intervene in the early stage of battery thermal runaway, thereby effectively curbing the further development of thermal runaway. In addition, the cooling system also has a continuous monitoring function to ensure that the battery stops working only after the temperature drops to a safe range, further ensuring the safety of the battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a structural block diagram of the battery thermal runaway early warning and cooling system of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Combination Figure 1 As shown, the battery thermal runaway early warning and cooling system of the present invention comprises: a multi-parameter monitoring module, a data processing and early warning model and a cooling system control module;

[0038] Among them, the multi-parameter monitoring module obtains the internal resistance value of each battery cell in real time during the battery charging process, and calculates the internal resistance range, internal resistance standard deviation and coefficient of variation based on the internal resistance value at the sampling time point, monitors the battery working environment gas through a mass spectrometer, obtains real-time gas mass spectrometry monitoring data in real time, and quantitatively analyzes the release characteristics of the escaped gas based on the equivalent characteristic spectrum method, including hydrogen, hydrocarbon compounds, etc., and sets a temperature sensor on the surface of the battery shell to monitor the temperature change of the battery surface in real time, especially the temperature difference between the highest point and the lowest point;

[0039] The data processing and early warning model uses the data of internal resistance monitoring, gas monitoring and temperature monitoring as input into a pre-trained machine learning model. The model is built based on deep learning algorithms such as Transformer and can process multi-dimensional time series data. It evaluates the thermal runaway risk of the battery through feature extraction and classification prediction. When the model predicts that the battery has a thermal runaway risk, it generates a corresponding early warning signal according to the risk level and sends the signal to the control module.

[0040] Among them, after the cooling system control module receives the early warning signal and the cooling system is started, it continuously monitors the battery temperature until the temperature drops to a safe range to ensure battery safety.

[0041] In a preferred embodiment, the present invention can be further configured as follows: the multi-parameter monitoring module specifically includes an internal resistance monitoring unit, a gas monitoring unit and a temperature monitoring unit;

[0042] A high-precision, low-drift DC internal resistance sensor is used in the internal resistance monitoring unit to ensure that the internal resistance value of each battery cell can be stably and accurately measured during the battery charging process. The internal resistance sensor is directly connected to the positive and negative electrodes of the battery cell to ensure the accuracy and real-time nature of the measured data. A high-precision data acquisition card or integrated module is used to collect the internal resistance value at each sampling time point in real time and perform analog-to-digital conversion. The internal resistance range is calculated: within the preset time window, the maximum and minimum values ​​of the internal resistance are found, and the difference between the two is the internal resistance range. The statistical method is used to calculate the mean and standard deviation of all internal resistance values ​​within the time window. The standard deviation reflects the discrete degree of the internal resistance. The standard deviation is divided by the mean to obtain the coefficient of variation, which is used to evaluate the relative fluctuation of the internal resistance. The internal resistance monitoring module adopts an independent design for easy maintenance and upgrading. The module is connected to the main control system through standard communication interfaces such as CAN bus and RS485 to achieve real-time data transmission and command reception.

[0043] The gas monitoring unit uses a high-sensitivity, high-resolution mass spectrometer that can accurately detect trace gas components in the battery working environment. The mass spectrometer needs to be equipped with a special sampling probe and preprocessing system to ensure the representativeness of the sampled gas and the accuracy of the test results. The mass spectrometer outputs real-time mass spectrometry monitoring data of the gas, including gas composition and concentration information, and compares the detected gas composition with the predefined equivalent characteristic spectrum. The release characteristics of the escaping gas are determined by quantitative analysis methods, with special attention paid to gas components such as hydrogen and hydrocarbon compounds that may indicate thermal runaway of the battery, and their concentration change trends are recorded. It also includes the design of the monitoring module. The gas monitoring module is independently designed to reduce interference with other monitoring units. The data processing unit has a built-in data processing unit to process and analyze the data output by the mass spectrometer in real time.

[0044] The temperature monitoring unit uses high-precision and high-stability temperature sensors, such as platinum resistance and thermocouples, and arranges temperature sensors at multiple key positions of the battery shell, such as the highest point, the lowest point, and the center point, to comprehensively monitor the temperature changes on the battery surface. The data of each temperature sensor is collected in real time through the data acquisition system, and the temperature difference between the highest point and the lowest point is calculated as an important indicator for evaluating the uniformity of battery temperature. The temperature trend of the battery surface is monitored to detect abnormal temperature rise in time. The temperature monitoring module is independently designed for easy installation and maintenance. The temperature threshold is set and triggered when the temperature exceeds the threshold. Alarm signal, wherein a DC internal resistance sensor with high precision (error rate ≤ 0.1%) and low drift (≤ 0.01% / ℃) is selected to ensure measurement accuracy and stability. The sensor is directly welded or connected to the positive and negative poles of the battery cell using a high-conductivity connector to reduce contact resistance and signal interference. A 16-bit or higher precision data acquisition card is used to ensure that the internal resistance value at each sampling time point can be accurately captured and converted to analog-to-digital. Within a preset time window (such as 1 minute), the internal resistance range, average value, standard deviation and coefficient of variation are automatically calculated through a programming algorithm to comprehensively evaluate the dynamic changes of the internal resistance. The internal resistance monitoring unit is independently packaged, dustproof and waterproof, and seamlessly connected to the main control system through standard communication interfaces such as CAN bus or RS485. A mass spectrometer with a resolution of ≥10,000 is selected, equipped with a fast-response sampling probe and an efficient pretreatment system (such as drying, filtration, etc.) to ensure the purity and representativeness of the gas sample. The real-time output data of the mass spectrometer is preprocessed and quantitatively analyzed by the built-in data processing unit, and compared with the predefined equivalent characteristic spectrum with high precision to identify the release characteristics of key gases such as hydrogen and hydrocarbon compounds. The gas monitoring module adopts a physical isolation design , reduce electromagnetic interference and cross contamination, ensure the accuracy of the test results, arrange multiple high-precision (≤±0.1℃) temperature sensors at key positions such as the highest point, lowest point, center point and edge of the battery shell to form a three-dimensional temperature monitoring network, collect data from each sensor in real time through a high-speed data acquisition system, calculate the average temperature using the weighted average method, and focus on monitoring the temperature difference between the highest point and the lowest point to evaluate the temperature uniformity of the battery, preset temperature thresholds, and immediately trigger an alarm signal when an abnormal temperature increase is detected or the temperature difference exceeds the preset range, and transmit it to the main control system through the communication interface.

[0045] The application of high-precision, low-drift DC internal resistance sensors and data processing algorithms has significantly improved the accuracy and real-time performance of internal resistance monitoring, and can capture abnormal changes in the internal resistance of the battery earlier, providing more reliable input data for the early warning model. Secondly, the high-sensitivity and high-resolution mass spectrometer, combined with the efficient data processing unit, enables the gas monitoring unit to accurately identify the characteristics of trace gas release during battery thermal runaway, providing the early warning model with another important dimension of information in addition to temperature and internal resistance. Finally, the temperature monitoring unit comprehensively evaluates the temperature uniformity of the battery surface through multi-point layout and real-time monitoring, effectively avoiding the risk of thermal runaway caused by local overheating.

[0046] In a preferred embodiment, the present invention can be further configured as follows: the multi-parameter monitoring module also includes a comprehensive data processing and early warning model, which integrates the data of internal resistance monitoring, gas monitoring and temperature monitoring to form a multi-dimensional time series data set, and adopts a machine learning model based on deep learning algorithms such as Transformer to extract features and classify and predict the integrated data, evaluate the thermal runaway risk of the battery, generate corresponding early warning signals based on the model prediction results, and send them to the control module to start corresponding cooling measures. First, the data from the internal resistance monitoring unit, the gas monitoring unit and the temperature monitoring unit are strictly synchronized in time to ensure that all monitoring parameters are integrated under the same time reference. Subsequently, data normalization or standardization technology is used to eliminate the influence of different dimensions and data ranges on model training. Then, an efficient data structure (such as a tensor) is used to organize the synchronized and processed data into a multi-dimensional time series data set, and each data point contains information of multiple dimensions such as internal resistance value, gas component concentration, temperature and its change rate. The Tr based The deep learning algorithm of ansformer is the core. The algorithm is good at processing long sequence data and capturing complex dependencies between data. The model architecture includes an encoder part, which is used to extract high-level features from multi-dimensional time series data; the decoder part performs classification prediction based on the extracted features. In the model training stage, a large amount of historical data and annotated thermal runaway events are used to adjust the model parameters through the back propagation algorithm to optimize the prediction performance. During the model operation, the comprehensive data processing and early warning model first extracts features from the input multi-dimensional time series data, including the fluctuation pattern of internal resistance, the change trend of gas composition, the spatial distribution of temperature field, etc. Then, these features are used to classify and predict the thermal runaway risk, distinguishing different levels such as low risk, medium risk and high risk. According to the model prediction results, the corresponding early warning signal is generated. The early warning signal not only contains risk level information, but also may include detailed information such as the specific location of the risk source and recommended cooling measures. Through reliable communication protocols (such as TCP / IP, MQTT, etc.), the early warning signal is sent to the control module in real time for timely response.

[0047] Through data fusion processing, the organic combination of multi-source monitoring parameters is realized, which improves the comprehensiveness and accuracy of the early warning system. The construction of multi-dimensional time series data sets provides richer and more detailed input information for machine learning models, which helps to capture subtle changes in the process of battery thermal runaway. Secondly, the application of machine learning models based on deep learning algorithms such as Transformer has significantly improved the intelligence level of the early warning system. The model can automatically extract high-level features from massive data and accurately perform classification and prediction, effectively overcoming the limitations of traditional methods in dealing with complex and nonlinear problems. By continuously optimizing model parameters and training data, the prediction accuracy and robustness of the early warning system can be further improved. Finally, the real-time generation and transmission mechanism of the early warning signal ensures the timely start of cooling measures. When the model predicts that the battery has a risk of thermal runaway, it can immediately generate a warning signal and send it to the control module, thereby triggering corresponding cooling measures (such as air cooling, liquid cooling or cutting off battery power supply). This rapid response mechanism helps to curb the further development of thermal runaway in a timely manner and ensure the safe and stable operation of the battery system.

[0048] In a preferred embodiment, the present invention can be further configured as follows: the data processing and early warning model specifically includes a data preprocessing module, a machine learning model construction, a feature extraction and classification prediction, and an early warning signal generation and sending module;

[0049] The data preprocessing module synchronizes the time and formats of the data from the internal resistance monitoring unit, gas monitoring unit, and temperature monitoring unit to ensure that all data are comparable in the same time frame, removes outliers, missing values, and other noise data, and processes discontinuous or excessively fluctuating data points through interpolation and smoothing to improve data quality. Based on the mechanism and prior knowledge of battery thermal runaway, it extracts features useful for assessing thermal runaway risks from the original data, such as internal resistance range, standard deviation, coefficient of variation, gas concentration trend, temperature gradient, etc.;

[0050] Models based on deep learning algorithms such as Transformer are used in the construction of machine learning models because Transformer has advantages in processing long sequence data and capturing global dependencies, and is suitable for processing multi-dimensional time series data. A large amount of historical data is used to train the model, including data in normal charging state and data before thermal runaway events. Through supervised learning, the model learns how to extract features from input data and predict the thermal runaway risk of the battery. The performance of the model is evaluated through cross-validation, confusion matrix, ROC curve and other methods to ensure that the model has high accuracy and generalization ability;

[0051] In feature extraction and classification prediction, features useful for classification prediction are automatically extracted from the input data. These features may include fluctuation patterns of internal resistance, abnormal changes in gas concentration, and sharp increases in temperature gradients. The trained model is used to classify and predict the real-time input data to determine whether the battery currently has a thermal runaway risk and give a risk level.

[0052] In the warning signal generation and sending module, warning thresholds for different risk levels are set according to business needs and safety standards. When the risk level predicted by the model exceeds the set threshold, the warning mechanism is triggered, and the corresponding warning signal is generated according to the prediction result, including risk level, warning time, recommended measures and other information. The warning signal is sent to the control module through the communication interface so that cooling measures can be taken in time. High-precision timestamp synchronization technology is used to ensure that data from different monitoring units are accurately aligned in time. At the same time, by defining a unified data format standard, such as JSON or XML, seamless data integration is achieved. Statistical methods (such as box plots, IQR) are used to automatically identify and remove outliers. For missing values, appropriate interpolation methods are selected according to data characteristics. In addition, the sliding average or low-pass filtering technology is used to smooth the data points with excessive fluctuations to reduce noise interference. Based on the physical mechanism of battery thermal runaway and expert knowledge, a series of feature extraction algorithms are designed. For example, for internal resistance data, the range, standard deviation and coefficient of variation are calculated to evaluate the stability of the internal resistance; for gas data, the concentration change trend and abnormal peak are analyzed; for temperature data, the temperature gradient, temperature difference, etc. are calculated to fully reflect the thermal state of the battery. In view of the advantages of Transformer in processing long sequences and multi-dimensional data, this algorithm is selected as the core of the model. At the same time, ensemble learning or hybrid model strategies are considered to further improve the prediction performance. The data is divided into training set, validation set and test set. The training set is used for model training, the validation set is used for model parameter adjustment, and the test set is used to finally evaluate the model performance. The model is trained using optimization algorithms such as gradient descent, and regularization, Dropout and other technologies are used to prevent overfitting. The best model parameters are selected through cross-validation, and the classification performance of the model is evaluated using indicators such as confusion matrix and ROC curve to ensure its accuracy and robustness in various situations. The automatic feature learning ability of the deep learning model is used to automatically extract high-level features that are useful for classification prediction from the original data. These features may include complex internal resistance fluctuation patterns, time series characteristics of gas concentration, spatial distribution characteristics of temperature field, etc. The real-time collected data is input into the The model will automatically determine the current thermal runaway risk level of the battery based on the characteristics of the input data, and set warning thresholds for different risk levels based on business needs and safety standards, combined with historical data and expert experience. These thresholds should be able to accurately reflect the critical state of battery thermal runaway. When the risk level predicted by the model exceeds the set threshold, the warning mechanism is triggered. The system will automatically generate a warning signal containing information such as risk level, warning time, and recommended measures based on the prediction results. The warning signal will be sent to the control module in real time through a reliable communication interface (such as CAN bus, MQTT protocol, etc.). At the same time, the system should have multiple communication backup methods to ensure that the warning information can be transmitted in time in the event of a communication failure.

[0053] In-depth optimization of the data preprocessing module effectively improves data quality and provides more accurate and reliable input for subsequent machine learning models. Secondly, the construction of machine learning models based on deep learning algorithms such as Transformer enables the system to automatically extract complex features from multi-dimensional time series data and perform accurate classification predictions, thereby overcoming the limitations of traditional methods in processing complex data. In addition, the real-time nature of the feature extraction and classification prediction modules ensures that the system can promptly detect the risk of battery thermal runaway and give the corresponding risk level. Finally, the rapid response mechanism of the early warning signal generation and sending module ensures the timely transmission of early warning information and the timely initiation of control measures, effectively reducing the incidence of battery thermal runaway accidents and improving the safety and stability of the battery system.

[0054] In a preferred embodiment, the present invention can be further configured as follows: after each warning and cooling operation, the system automatically collects relevant data, including monitoring data before and after the warning, the execution status of cooling measures, etc., and uses the collected data to continuously optimize and calibrate the machine learning model to improve the accuracy and response speed of the warning, which may include adjusting model parameters, updating feature sets, introducing new training data, etc. The performance of the system is regularly monitored and evaluated to ensure that the system can operate stably. The system not only collects monitoring data at the time of warning triggering, but also traces back and records monitoring data for a period of time before the warning, as well as monitoring data after the warning until the cooling measures take effect and the risk is confirmed to be lifted. These data include but are not limited to changes in internal resistance values, fluctuations in gas concentrations, evolution of temperature fields, etc. The system records in detail the execution time, execution method (such as liquid cooling start-up, air cooling intensity adjustment, etc.), and effect evaluation after execution of the cooling measures, so as to subsequently analyze the effectiveness and optimization space of the cooling measures. Based on the newly collected data, the system uses optimization algorithms (such as grid search, random search, Bayesian optimization, etc.) to automatically adjust the internal parameters of the machine learning model, such as the learning rate , batch size, Dropout ratio, etc. to improve the prediction accuracy and generalization ability of the model. By analyzing the distribution and importance of features in new data, the system may discover new features that contribute to thermal runaway risk assessment, such as specific gas concentration combinations, correlation patterns between temperature gradients and internal resistance changes, etc. These new features will be added to the feature set to enhance the prediction ability of the model. After cleaning and preprocessing, the newly collected data will be added to the training set of the model as new training data, which will help the model learn more diverse data patterns and improve its adaptability in different scenarios. The system sets a regular (such as daily, weekly or monthly) performance evaluation plan to evaluate the accuracy and stability of the model by comparing the differences between the actual warning results and the model prediction results. The system monitors the abnormal conditions in the model operation process in real time, such as the sudden deviation of the prediction results from the normal range, the abnormal fluctuation of the model parameters, etc. Once an abnormality is found, an alarm will be triggered immediately and the emergency handling process will be started. The system records the key information of each warning, cooling operation, data collection, model optimization and other processes in detail, and generates detailed log files and reports for subsequent analysis and auditing.

[0055] By meticulously collecting monitoring data before and after the warning and the implementation of cooling measures, the system can obtain rich feedback information, providing a solid foundation for the continuous optimization of the model. This optimization method based on actual operation data is more targeted and effective than models that rely solely on historical data for training, and can significantly improve the accuracy and response speed of warnings. Secondly, through regular performance monitoring and evaluation, the system can promptly discover and resolve potential problems and anomalies to ensure that the system can operate stably and continue to provide reliable warning services. This self-monitoring and self-repairing capability greatly reduces the cost and difficulty of system maintenance and improves the overall reliability and availability of the system.

[0056] In a preferred embodiment, the present invention can be further configured as follows: the cooling system control module specifically includes: an early warning signal receiving and analyzing module, a risk level assessment module, a cooling system execution and monitoring module, and a cooling completion and subsequent processing module;

[0057] Interface design is carried out in the warning signal receiving and parsing module. The cooling system control module needs to have a communication interface that matches the data processing and warning model module to ensure that the warning signal can be transmitted accurately. After receiving the warning signal, the control module needs to immediately parse the signal content, including key information such as risk level, battery location, and current temperature.

[0058] The risk level assessment has a built-in risk assessment model, which can quickly assess the cooling measures to be taken based on the analyzed risk level and battery status such as temperature and internal resistance.

[0059] The cooling system execution and monitoring is provided with actuators, including air cooling system and liquid cooling system. Both the air cooling system and the liquid cooling system are equipped with corresponding actuators such as fans, circulation pumps, etc. The control module drives these mechanisms to work by sending control signals. After the cooling system is started, the control module will continuously monitor the battery temperature, obtain real-time data through the temperature sensor, and compare it with the preset safe temperature range. According to the monitoring results, the control module can adjust the operating status of the cooling system in real time, such as fan speed, coolant flow, etc., to ensure that the battery temperature is always kept within a safe range.

[0060] In the cooling completion and subsequent processing module, when the battery temperature drops to a safe range and remains stable for a period of time, the control module will determine that the cooling process has been completed. After the cooling is completed, the control module will perform a system reset operation, shut down the cooling system, and restore the normal power supply state of the battery. If an emergency cut-off operation was performed before, during the entire process, the control module will record all key data such as warning time, cooling measures, temperature changes, etc., and feed this data back to the data processing and warning model module for continuous optimization and calibration of the model, using standardized communication protocols (such as MQTT, OPC UA, etc.), design a dedicated data exchange interface to ensure seamless connection with the data processing and warning model module, the interface realizes two-way communication capability, supports the sending of warning signals and the receiving of feedback information, and immediately starts the parsing engine after receiving the warning signal to decode and parse the signal. The parsing process is not limited to basic information such as risk level and battery location, but also includes any additional detailed data, such as specific monitoring values, timestamps, etc. At the same time, an error checking mechanism is implemented to ensure that the parsed information is accurate, with a built-in refined risk assessment algorithm. Combined with the current state of the battery (such as temperature, internal resistance, gas concentration, etc.) and historical data, through a comprehensive analysis of multiple factors, the thermal runaway risk level and potential impact range are quickly assessed. According to the assessment results, a specific cooling strategy is formulated, including but not limited to priority sorting, cooling method selection (air cooling / liquid cooling) and preliminary parameter setting (such as initial fan speed, coolant flow), and a refined actuator control logic is designed to ensure that the air cooling system and liquid cooling system can accurately respond to control signals. Actuators such as fans and circulating pumps are equipped with independent control units to support remote adjustment. Control, through the high-density deployed temperature sensor network, the battery temperature distribution is monitored in real time. Based on real-time data, a closed-loop feedback control algorithm is used to dynamically adjust the operating status of the cooling system, such as automatically adjusting the fan speed or coolant flow according to temperature changes to achieve the best cooling effect. The integrated fault detection mechanism monitors the status of the actuator and sensor in real time. Once an abnormality is found (such as fan failure, temperature sensor failure, etc.), the emergency handling process is immediately started, including fault alarm, automatic switching of backup equipment, etc. According to the preset cooling completion standard (such as the battery temperature drops to a safe range and remains stable for a certain period of time), it automatically determines whether the cooling process is completed. After the cooling is completed, a series of reset operations are performed, including shutting down the cooling system and restoring the normal power supply status of the battery. If an emergency cut-off operation has been performed before, a safety check and verification are required to ensure the stable operation of the system after recovery. Record key data of the entire cooling process, such as warning time, cooling measures, temperature change curve, etc., and feed these data back to the data processing and warning model module in real time through a dedicated interface for continuous optimization and calibration of the model.

[0061] Through standardized interface design and sophisticated signal analysis mechanism, accurate transmission and rapid response of early warning signals are ensured, laying the foundation for timely cooling measures. Secondly, the built-in risk assessment model and dynamic adjustment strategy enable the cooling system to quickly formulate and execute the optimal cooling plan according to the actual state of the battery, effectively reducing the damage of thermal runaway events to the battery system. At the same time, the introduction of real-time monitoring and fault detection mechanism further improves the reliability and stability of the system. Finally, through data recording and feedback mechanism, closed-loop optimization of the cooling system and control model is realized, and the early warning and response capabilities of the system are continuously improved, providing a strong guarantee for the long-term safe operation of the battery system. These improvements not only overcome the defects of slow response speed and inaccurate control that may exist in the background technology, but also significantly improve the intelligence level and comprehensive performance of the entire battery thermal runaway management system.

[0062] In a preferred embodiment of the present invention, a cooling strategy selection is also provided in the risk level assessment, including three levels, namely low-level risk, medium-high-level risk and emergency cut-off. If the assessment result is a low-level risk, the control module will start the air cooling system, which includes adjusting the fan speed, turning on or off the fan in a specific area, etc., to achieve effective heat dissipation of the battery. If the assessment result is a medium-high-level risk, the control module will start a more efficient liquid cooling system, which may involve turning on the circulation pump, adjusting the coolant flow and temperature, etc., to quickly reduce the battery temperature. In extreme cases, if it is predicted that thermal runaway is about to occur or the temperature cannot be effectively controlled, the control module will perform an emergency cut-off operation to cut off the battery power supply to prevent accidents. If the condition further deteriorates, when the assessment result is a low-level risk, the control module will first analyze the temperature distribution and heat dissipation requirements of the battery, and then start the air cooling system. At this time, the system will intelligently adjust the fan speed according to the specific value and distribution of the battery temperature to ensure that it can effectively dissipate heat without wasting energy. At the same time, for areas with higher local temperatures in the battery pack, the system can selectively turn on or off the fans in specific areas to achieve precise heat dissipation. If the assessment result is a medium-to-high-level risk, the system will quickly switch to a more efficient liquid cooling system. During this process, the control module will first check the status of the liquid cooling system, including the circulation pump, coolant reserves and temperature, to ensure that the system is in an operational state. Then, based on the real-time data of the battery temperature, the control module will start the cooling system. Based on the data and prediction model, the system will automatically start the circulation pump and accurately adjust the flow and temperature of the coolant to reduce the battery temperature to a safe range as quickly as possible. In addition, the system will monitor the circulation efficiency of the coolant and make dynamic adjustments when necessary to cope with possible sudden changes in battery temperature. In extreme cases, if it is predicted that thermal runaway is about to occur or the battery temperature can no longer be effectively controlled by the cooling system, the control module will immediately perform an emergency cut-off operation, which includes quickly cutting off the battery's power supply circuit and initiating safety isolation measures, such as shutting down all non-essential equipment related to the battery to prevent the situation from further deteriorating. During the emergency cut-off process, the system will also record detailed operation logs and key data to provide a basis for subsequent accident analysis and processing. Real-time risk assessment and strategy selection: To ensure that the correct cooling strategy is selected in a very short time, the system uses high-performance computing units and optimized algorithms to accelerate the risk assessment and strategy selection process. At the same time, the system supports multi-threading and parallel computing to further improve processing efficiency. During the execution of the cooling strategy, the system will continuously monitor the changes in battery temperature and other related parameters, and dynamically adjust the cooling strategy based on feedback data. For example, during the operation of the air cooling system, if it is found that the battery temperature drops too slowly or the temperature in a local area is too high, the system will automatically increase the fan speed or adjust the fan layout; during the operation of the liquid cooling system, the system will adjust its working status in real time according to the changes in the temperature and flow of the coolant.

[0063] For different levels of thermal runaway risks, the system can automatically select and execute corresponding cooling strategies, effectively reducing the battery temperature and preventing the occurrence of thermal runaway events. At low-level risks, precise air cooling system control is used to achieve effective heat dissipation of the battery and save energy. At medium and high-level risks, the battery temperature is quickly reduced to a safe range through efficient liquid cooling system control, avoiding further deterioration of the situation. Secondly, the introduction of the emergency cut-off mechanism provides a strong guarantee for safety in extreme situations, effectively preventing potential hazards to equipment and personnel caused by thermal runaway events. Finally, the implementation of the dynamic adaptation and feedback adjustment mechanism enables the system to adjust the cooling strategy in real time according to actual conditions, thereby improving the flexibility and adaptability of the system. These improvements not only overcome the defects of the single cooling strategy and slow response speed that may exist in the background technology, but also significantly improve the intelligence level and comprehensive performance of the entire battery thermal runaway management system.

[0064] In a preferred embodiment, the present invention can be further configured as follows: the early warning and cooling system also includes a system feedback and optimization module. After each early warning and cooling operation, the system automatically collects relevant data for continuous optimization and calibration of the model to improve the accuracy and response speed of the early warning. The data mainly comes from the internal resistance monitoring unit, the gas monitoring unit, the temperature monitoring unit, and the cooling system control module. These modules will generate a large amount of real-time data during the early warning and cooling operation, including but not limited to the internal resistance value of the battery cell, gas concentration, temperature change, working status of the cooling system, etc. The system needs to synchronize the time and format the data from different modules to ensure the integrity and consistency of the data. At the same time, the collected data is The data is processed initially, such as removing noise and filling missing values, to improve data quality. The performance of the early warning model is evaluated with the collected data, including indicators such as accuracy, recall, and F1 score. By comparing the model prediction results with the actual results, the model is analyzed under what circumstances the model performs well and under what circumstances it is insufficient. The system should continuously monitor the operating status of the battery and the performance of the early warning model to ensure that the model can accurately predict the thermal runaway risk of the battery and generate early warning signals in a timely manner. The performance of the model should be evaluated and feedback should be given regularly to timely discover and solve potential problems. At the same time, the optimization direction and goals of the model should be adjusted according to the changes and needs of the actual application scenarios. As the system runs and data accumulates, the data set should be continuously supplemented and updated to ensure that the system can meet the requirements of the early warning model. Improve the generalization and adaptability of the model. At the same time, use historical data to continuously optimize and iterate the model to ensure that the model always remains in the best state. The system feedback and optimization module first ensures that the data from the internal resistance monitoring unit, gas monitoring unit, temperature monitoring unit and cooling system control module are strictly synchronized in time to avoid data mismatch problems caused by time differences. At the same time, the data formats of each module are unified, such as data units, sampling frequencies, etc., to ensure data compatibility and consistency. The collected raw data may contain noise, missing values ​​and other unfavorable factors. The module will perform preliminary processing on these data, use advanced filtering algorithms to remove noise, and use interpolation or statistical methods to fill in missing values ​​to improve data quality. , laying a solid foundation for subsequent analysis. The system has established a comprehensive evaluation index system, including accuracy, recall rate, F1 score, etc., to evaluate the performance of the early warning model in multiple dimensions. By comparing the model prediction results with the actual results, it can accurately identify the performance of the model in different scenarios. Using statistical analysis and machine learning techniques, the model prediction results are deeply analyzed to identify the situations in which the model performs better (such as specific temperature ranges, battery status, etc.) and the situations in which it is insufficient (such as extreme temperatures, rapid temperature changes, etc.). This information is crucial for guiding the direction of model optimization. The system implements a continuous monitoring and evaluation strategy to regularly evaluate the performance of the early warning model. Once it is found that the model performance has declined or cannot meet the actual needs,Immediately start the optimization process. Optimization measures may include adjusting model parameters, improving algorithm structure, introducing new feature variables, etc. to ensure that the model is always in the best state. As the system runs and data accumulates, the system should continuously add new data samples to the data set. These data may come from new battery models, different working environments or special operating conditions. By adding new data, the generalization ability and adaptability of the model can be improved. The model can be continuously iterated and optimized using historical data. By repeating the training, evaluation and optimization process, the prediction accuracy and response speed of the model can be gradually improved. At the same time, pay attention to the changes and needs of actual application scenarios, and flexibly adjust the optimization direction and goals of the model.

[0065] The synchronization and integration of multi-source data ensures the integrity and consistency of the data, providing a solid foundation for model evaluation and optimization. Secondly, the comprehensive evaluation index system and in-depth performance analysis enable the system to accurately identify the advantages and disadvantages of the model, so as to optimize it in a targeted manner. This continuous optimization mechanism ensures that the model can continue to improve with the changes in application scenarios and the accumulation of data, and always remain in the best state. Finally, the management and update strategy of the data set improves the generalization ability and adaptability of the model, enabling the system to better cope with various complex situations. These improvements not only overcome the defects of unstable model performance and inconsistent data that may exist in the background technology, but also significantly improve the intelligence level and comprehensive performance of the early warning and cooling system, providing a strong guarantee for the safe and stable operation of the battery system.

[0066] In a preferred embodiment, the present invention can be further configured as follows: when evaluating the performance of the early warning model, the input features can be re-evaluated and adjusted according to the information in the feedback report. It may be necessary to introduce new features or eliminate some features that have little effect on the prediction results to improve the prediction ability of the model, and adjust the hyperparameters in the model, such as learning rate, batch size, number of iterations, etc., to find the optimal model configuration. This usually requires multiple experiments and comparisons. If the evaluation results show that the model performs poorly in some aspects, it may be necessary to consider improving the algorithm itself. For example, different deep learning architectures, optimization algorithms or loss functions can be tried to improve the performance and stability of the model. After completing feature selection, parameter adjustment, and algorithm improvement, the model is retrained using a new data set. Through multiple iterations and optimizations, the model is gradually adapted to the needs of actual application scenarios. According to the detailed data in the feedback report, a comprehensive evaluation is conducted on the input features used by the current model, and the actual contribution of each feature to the prediction results is analyzed. Features with low redundancy or correlation are identified. Based on the feature evaluation results, the feature set is dynamically adjusted. For new features that have a significant impact on the prediction results, they are introduced into the model in a timely manner. At the same time, features that do not contribute much to the prediction results or introduce noise are eliminated to reduce computational complexity and improve model efficiency. Key hyperparameters in the model, such as learning rate, batch size, and number of iterations, are adjusted. etc., and make fine adjustments. These adjustments usually rely on multiple experiments and comparisons. Through grid search, random search, or Bayesian optimization, the optimal hyperparameter combination is found. Taking into account the changes in actual application scenarios, the system should support dynamic adjustment strategies for hyperparameters. According to the real-time performance of the model and data distribution, the hyperparameter settings are automatically adjusted to ensure that the model is always in the best working state. If the evaluation results show that the model performs poorly in some aspects, such as accuracy, recall rate, or F1 score, the system will consider introducing new deep learning architectures, optimization algorithms, or loss functions. These improvements are designed to fundamentally improve the predictive ability and stability of the model, and encourage innovative attempts at algorithms. For example, combining the advantages of multiple algorithms to build a hybrid model, introducing an attention mechanism to enhance the model's ability to capture key information, etc. These innovative attempts will be based on the latest research results and technological trends to ensure that the model is always at the forefront of the industry. After completing feature selection, parameter adjustment, and algorithm improvement, the model is retrained using an updated dataset. The new dataset should contain more diverse samples and richer information to further improve the generalization ability of the model. Model training should be viewed as a continuous iterative process. Through multiple iterations and optimizations, the model structure and parameter settings are gradually adjusted to gradually adapt to the needs of actual application scenarios. At the same time, keep an eye on new technologies and methods so that new improvements can be introduced when needed.

[0067] Feature selection and optimization ensure the effectiveness and efficiency of model input, reduce the waste of computing resources and improve prediction accuracy. Secondly, hyperparameter tuning and algorithm improvement enable the model to better adapt to complex and changeable data distribution and actual application scenarios, improve the stability and reliability of the model. Finally, the model retraining and iteration mechanism ensures that the model can continue to learn and improve, and constantly adapt to new needs and challenges. These improvements not only overcome the defects of poor model generalization ability and low prediction accuracy that may exist in the background technology, but also significantly improve the intelligence level and comprehensive performance of the early warning model, providing a more solid technical guarantee for the safe and stable operation of the battery system.

[0068] First, the multi-parameter monitoring module, as the front-end sensing unit of the system, works continuously during the battery charging process. The module obtains the key parameters of each battery cell in real time through high-precision sensors, including the internal resistance value, the working environment gas composition and the temperature distribution on the battery surface. For the internal resistance value, the module records the data at the sampling time point and calculates the internal resistance range, standard deviation and coefficient of variation to reflect the fluctuation of the battery internal resistance. At the same time, the mass spectrometer is used to perform real-time mass spectrometry monitoring of the gas around the battery, and quantitatively analyze the release characteristics of escaped gases such as hydrogen and hydrocarbon compounds. These gases are often signs of abnormal reactions inside the battery. In addition, by setting a temperature sensor on the surface of the battery shell, the module can accurately capture the temperature changes on the battery surface, especially focusing on the temperature difference between the highest point and the lowest point, to evaluate the thermal distribution of the battery. Next, the data processing and early warning model receives the data from the multi-parameter monitoring module as input. The model is based on advanced technologies such as Transformer. The deep learning algorithm is constructed with the ability to process multi-dimensional time series data. The model comprehensively evaluates the thermal runaway risk of the battery by extracting features and classifying and predicting the input data. Once the model predicts that the battery is at risk of thermal runaway, it will automatically generate corresponding warning signals according to the risk level. These signals are then sent to the cooling system control module. Finally, after receiving the warning signal, the cooling system control module immediately starts the cooling system. During the operation of the cooling system, the module will continuously monitor the temperature changes of the battery to ensure that the cooling effect meets the expectations. By continuously lowering the battery temperature until it drops to a safe range, the cooling system effectively controls the further development of the battery thermal runaway risk, thereby ensuring the safety and stability of the battery. In summary, the battery thermal runaway early warning and cooling system achieves comprehensive monitoring and effective control of the battery thermal runaway risk through the synergy of multi-parameter monitoring, intelligent warning and timely cooling, providing reliable technical guarantee for the safe and stable operation of the battery system.

[0069] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0070] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Battery thermal runaway early warning and cooling system, characterized in that: The early warning and cooling system includes: a multi-parameter monitoring module, a data processing and early warning model, and a cooling system control module; Among them, the multi-parameter monitoring module obtains the internal resistance value of each battery cell in real time during the battery charging process, and calculates the internal resistance range, internal resistance standard deviation and coefficient of variation based on the internal resistance value at the sampling time point, monitors the battery working environment gas through a mass spectrometer, obtains real-time gas mass spectrometry monitoring data in real time, and quantitatively analyzes the release characteristics of the escaped gas based on the equivalent characteristic spectrum method, including hydrogen, hydrocarbon compounds, etc., and sets a temperature sensor on the surface of the battery shell to monitor the temperature change of the battery surface in real time, especially the temperature difference between the highest point and the lowest point; The data processing and early warning model uses the data of internal resistance monitoring, gas monitoring and temperature monitoring as input into a pre-trained machine learning model. The model is built based on deep learning algorithms such as Transformer and can process multi-dimensional time series data. It evaluates the thermal runaway risk of the battery through feature extraction and classification prediction. When the model predicts that the battery has a thermal runaway risk, it generates a corresponding early warning signal according to the risk level and sends the signal to the control module. Among them, after the cooling system control module receives the warning signal and the cooling system is started, it continuously monitors the battery temperature until the temperature drops to a safe range to ensure battery safety.

2. The battery thermal runaway early warning and cooling system according to claim 1, characterized in that: The multi-parameter monitoring module specifically includes an internal resistance monitoring unit, a gas monitoring unit and a temperature monitoring unit; A high-precision, low-drift DC internal resistance sensor is used in the internal resistance monitoring unit to ensure that the internal resistance value of each battery cell can be stably and accurately measured during the battery charging process. The internal resistance sensor is directly connected to the positive and negative electrodes of the battery cell to ensure the accuracy and real-time nature of the measured data. A high-precision data acquisition card or integrated module is used to collect the internal resistance value at each sampling time point in real time and perform analog-to-digital conversion. The internal resistance range is calculated: within the preset time window, the maximum and minimum values ​​of the internal resistance are found, and the difference between the two is the internal resistance range. The statistical method is used to calculate the mean and standard deviation of all internal resistance values ​​within the time window. The standard deviation reflects the discrete degree of the internal resistance. The standard deviation is divided by the mean to obtain the coefficient of variation, which is used to evaluate the relative fluctuation of the internal resistance. The internal resistance monitoring module adopts an independent design for easy maintenance and upgrading. The module is connected to the main control system through standard communication interfaces such as CAN bus and RS485 to achieve real-time data transmission and command reception. The gas monitoring unit uses a high-sensitivity, high-resolution mass spectrometer that can accurately detect trace gas components in the battery working environment. The mass spectrometer needs to be equipped with a special sampling probe and preprocessing system to ensure the representativeness of the sampled gas and the accuracy of the test results. The mass spectrometer outputs real-time mass spectrometry monitoring data of the gas, including gas composition and concentration information, and compares the detected gas composition with the predefined equivalent characteristic spectrum. The release characteristics of the escaping gas are determined by quantitative analysis methods, with special attention paid to gas components such as hydrogen and hydrocarbon compounds that may indicate thermal runaway of the battery, and their concentration change trends are recorded. It also includes the design of the monitoring module. The gas monitoring module is independently designed to reduce interference with other monitoring units. The data processing unit has a built-in data processing unit to process and analyze the data output by the mass spectrometer in real time. High-precision and high-stability temperature sensors, such as platinum resistance thermometers and thermocouples, are selected in the temperature monitoring unit. Temperature sensors are arranged at multiple key positions of the battery shell, such as the highest point, the lowest point, and the center point, to comprehensively monitor the temperature changes on the battery surface. The data of each temperature sensor is collected in real time through the data acquisition system, and the temperature difference between the highest point and the lowest point is calculated as an important indicator for evaluating the temperature uniformity of the battery. The changing trend of the battery surface temperature is monitored to detect abnormal temperature rise in time. The temperature monitoring module is independently designed to facilitate installation and maintenance. The temperature threshold is set and an alarm signal is triggered when the temperature exceeds the threshold.

3. The battery thermal runaway early warning and cooling system according to claim 1, characterized in that: The multi-parameter monitoring module also includes a comprehensive data processing and early warning model, which fuses the data of internal resistance monitoring, gas monitoring and temperature monitoring to form a multi-dimensional time series data set, and uses a machine learning model based on deep learning algorithms such as Transformer to extract features and classify and predict the fused data, evaluate the thermal runaway risk of the battery, generate a corresponding early warning signal based on the model prediction results, and send it to the control module to start the corresponding cooling measures.

4. The battery thermal runaway early warning and cooling system according to claim 1, characterized in that: The data processing and early warning model specifically includes a data preprocessing module, a machine learning model construction, a feature extraction and classification prediction, and an early warning signal generation and sending module; The data preprocessing module synchronizes the time and formats of the data from the internal resistance monitoring unit, gas monitoring unit, and temperature monitoring unit to ensure that all data are comparable in the same time frame, removes outliers, missing values, and other noise data, and processes discontinuous or excessively fluctuating data points through interpolation and smoothing to improve data quality. Based on the mechanism and prior knowledge of battery thermal runaway, it extracts features useful for assessing thermal runaway risks from the original data, such as internal resistance range, standard deviation, coefficient of variation, gas concentration trend, temperature gradient, etc.; Models based on deep learning algorithms such as Transformer are used in the construction of machine learning models because Transformer has advantages in processing long sequence data and capturing global dependencies, and is suitable for processing multi-dimensional time series data. A large amount of historical data is used to train the model, including data in normal charging state and data before thermal runaway events. Through supervised learning, the model learns how to extract features from input data and predict the thermal runaway risk of the battery. The performance of the model is evaluated through cross-validation, confusion matrix, ROC curve and other methods to ensure that the model has high accuracy and generalization ability; In feature extraction and classification prediction, features useful for classification prediction are automatically extracted from the input data. These features may include fluctuation patterns of internal resistance, abnormal changes in gas concentration, and sharp increases in temperature gradients. The trained model is used to classify and predict the real-time input data to determine whether the battery currently has a thermal runaway risk and give a risk level. In the warning signal generation and sending module, warning thresholds for different risk levels are set according to business needs and safety standards. When the risk level predicted by the model exceeds the set threshold, the warning mechanism is triggered, and the corresponding warning signal is generated according to the prediction result, including risk level, warning time, recommended measures and other information. The warning signal is sent to the control module through the communication interface so that cooling measures can be taken in time.

5. The battery thermal runaway early warning and cooling system according to claim 4, characterized in that: After each warning and cooling operation, the system automatically collects relevant data, including monitoring data before and after the warning, the implementation of cooling measures, etc., and uses the collected data to continuously optimize and calibrate the machine learning model to improve the accuracy and response speed of the warning. This may include adjusting model parameters, updating feature sets, introducing new training data, etc., and regularly monitoring and evaluating the performance of the system to ensure that the system can operate stably.

6. The battery thermal runaway early warning and cooling system according to claim 1, characterized in that: The cooling system control module specifically includes: a warning signal receiving and analyzing module, a risk level assessment module, a cooling system execution and monitoring module, and a cooling completion and subsequent processing module; Interface design is carried out in the warning signal receiving and parsing module. The cooling system control module needs to have a communication interface that matches the data processing and warning model module to ensure that the warning signal can be transmitted accurately. After receiving the warning signal, the control module needs to immediately parse the signal content, including key information such as risk level, battery location, and current temperature. The risk level assessment has a built-in risk assessment model, which can quickly assess the cooling measures to be taken based on the analyzed risk level and battery status such as temperature and internal resistance. The cooling system execution and monitoring is provided with actuators, including air cooling system and liquid cooling system. Both the air cooling system and the liquid cooling system are equipped with corresponding actuators such as fans, circulation pumps, etc. The control module drives these mechanisms to work by sending control signals. After the cooling system is started, the control module will continuously monitor the battery temperature, obtain real-time data through the temperature sensor, and compare it with the preset safe temperature range. According to the monitoring results, the control module can adjust the operating status of the cooling system in real time, such as fan speed, coolant flow, etc., to ensure that the battery temperature is always kept within a safe range. In the cooling completion and subsequent processing module, when the battery temperature drops to a safe range and remains stable for a period of time, the control module will determine that the cooling process is complete. After cooling is completed, the control module will perform a system reset operation, shut down the cooling system, and restore the normal power supply state of the battery. If an emergency cut-off operation was performed before, during the entire process, the control module will record all key data such as warning time, cooling measures, temperature changes, etc., and feed this data back to the data processing and warning model module for continuous optimization and calibration of the model.

7. The battery thermal runaway early warning and cooling system according to claim 1, characterized in that: The risk level assessment also provides a cooling strategy selection, which includes three levels: low-level risk, medium-high-level risk, and emergency cut-off. If the assessment result is a low-level risk, the control module will start the air cooling system, which includes adjusting the fan speed, turning on or off the fan in a specific area, etc., to achieve effective heat dissipation of the battery. If the assessment result is a medium-high-level risk, the control module will start a more efficient liquid cooling system, which may involve turning on the circulation pump, adjusting the coolant flow and temperature, etc., to quickly reduce the battery temperature. In extreme cases, if it is predicted that thermal runaway is about to occur or the temperature can no longer be effectively controlled, the control module will perform an emergency cut-off operation to cut off the battery power supply to prevent the situation from further deteriorating.

8. The battery thermal runaway early warning and cooling system according to claim 1, characterized in that: The early warning and cooling system also includes a system feedback and optimization module. After each early warning and cooling operation, the system automatically collects relevant data for continuous optimization and calibration of the model to improve the accuracy and response speed of the early warning. The data mainly comes from the internal resistance monitoring unit, the gas monitoring unit, the temperature monitoring unit, and the cooling system control module. These modules will generate a large amount of real-time data during the early warning and cooling operation, including but not limited to the internal resistance value of the battery cell, gas concentration, temperature change, working status of the cooling system, etc. The system needs to synchronize the time and format the data from different modules to ensure the integrity and consistency of the data. At the same time, the collected data is preliminarily processed, such as removing noise, filling missing values, etc., to improve the data quality. The performance of the early warning model is evaluated by data, including indicators such as accuracy, recall rate, and F1 score. By comparing the model prediction results with the actual results, it is analyzed in which cases the model performs well and in which cases it has deficiencies. The system should continuously monitor the operating status of the battery and the performance of the early warning model to ensure that the model can accurately predict the thermal runaway risk of the battery and generate early warning signals in a timely manner. The performance of the model should be evaluated and feedback should be given regularly to promptly discover and resolve potential problems. At the same time, the optimization direction and goals of the model should be adjusted according to changes and needs in actual application scenarios. As the system runs and data accumulates, the data set should be continuously supplemented and updated to improve the generalization ability and adaptability of the model. At the same time, the model should be continuously optimized and iterated using historical data to ensure that the model always remains in the best state.

9. The battery thermal runaway early warning and cooling system according to claim 1, characterized in that: When evaluating the performance of the early warning model, the input features may be re-evaluated and adjusted based on the information in the feedback report. It may be necessary to introduce new features or eliminate some features that have little impact on the prediction results to improve the prediction ability of the model, and adjust the hyperparameters in the model, such as learning rate, batch size, number of iterations, etc., to find the optimal model configuration. This usually requires multiple experiments and comparisons. If the evaluation results show that the model performs poorly in some aspects, it may be necessary to consider improving the algorithm itself. For example, you can try to use different deep learning architectures, optimization algorithms or loss functions to improve the performance and stability of the model. After completing feature selection, parameter adjustment and algorithm improvement, use a new data set to retrain the model. Through multiple iterations and optimization, the model gradually adapts to the needs of actual application scenarios.

Images