High-density energy storage battery thermal management system with multi-stage safety protection

Abstract

The invention discloses a high-density energy storage battery thermal management system with multistage safety protection, and relates to the technical field of new energy storage, and the system comprises a multi-source information obtaining module which is used for collecting and obtaining the working data of a plurality of sensors, and carrying out the preprocessing of the working data; and the fusion analysis module is electrically connected with the multi-source information acquisition module, and the fusion analysis module is used for receiving the preprocessed working data and performing fusion analysis on the working data through an algorithm model to generate a thermal runaway prediction signal. According to the thermal management system for the high-density energy storage battery, the safety and efficiency of thermal management of the battery are remarkably improved by integrating the multi-source information acquisition module, the fusion analysis module, the graded early warning and response module and the self-adaptive closed-loop control module; multi-dimensional features of the battery before thermal runaway can be comprehensively captured, and more accurate thermal runaway prediction is realized; and the early warning level is flexibly matched with the protection strategy, so that the pertinence of protection is enhanced.

Description

Technical Field

[0001] This invention relates to the field of new energy storage technology, specifically to a high-density energy storage battery thermal management system with multi-level safety protection. Background Technology

[0002] With the rapid development of the new energy storage industry, high-density energy storage batteries, with their advantages of high energy density and excellent space utilization, are increasingly widely used in fields such as grid peak shaving, distributed energy storage, and new energy vehicles. As a core component ensuring the safe operation of high-density energy storage batteries, the performance of the thermal management system directly affects the battery's cycle life, energy output stability, and the effectiveness of thermal runaway risk prevention. Currently, the industry's core requirements for thermal management systems focus on real-time and accurate acquisition of battery operating data, early prediction of potential thermal runaway risks, and dynamic adjustment of protection and cooling strategies based on actual operating conditions to adapt to the complex and variable thermal load requirements of high-density batteries during charging and discharging.

[0003] Existing high-density energy storage battery thermal management systems still have room for optimization in practical applications: some solutions lack comprehensive fusion analysis of multi-source monitoring data, often relying on signals from single sensors such as temperature sensors for judgment, making it difficult to fully capture the multi-dimensional characteristics of the early stages of thermal runaway, such as characteristic gas leaks and instantaneous voltage fluctuations; the matching degree between the early warning level classification and protection strategies of some systems needs improvement, making it difficult to flexibly adjust response measures according to dynamic factors such as battery health status and ambient temperature, which may affect the targeting and economy of protection; in addition, the adaptive adjustment capability of some cooling systems is limited, and the adjustment accuracy and response speed of operating parameters such as pump speed and fan speed need to be further improved when facing changes in charge and discharge rate and fluctuations in ambient temperature. To address these issues, we propose a high-density energy storage battery thermal management system with multi-level safety protection. Summary of the Invention

[0004] To address the aforementioned technical issues, a high-density energy storage battery thermal management system with multi-level safety protection is provided. This technical solution solves the problems of existing high-density energy storage battery thermal management systems, such as insufficient fusion of multi-source monitoring data, reliance on single sensor signals for judgment, difficulty in fully capturing multi-dimensional characteristics in the early stages of thermal runaway, the need to improve the matching degree between early warning levels and protection strategies, difficulty in dynamically adjusting according to battery health status and ambient temperature, and limited adaptive adjustment capability of the cooling system, requiring improvement in parameter adjustment accuracy and response speed.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A high-density energy storage battery thermal management system with multi-level safety protection includes: A multi-source information acquisition module is used to collect and acquire working data from multiple sensors and to preprocess the working data. The fusion analysis module is electrically connected to the multi-source information acquisition module; the fusion analysis module is used to receive preprocessed working data, perform fusion analysis on the working data through an algorithm model, and generate a thermal runaway prediction signal; the algorithm model includes: a machine learning model and a threshold logic judgment model; The graded early warning and response module is electrically connected to the fusion analysis module; the graded early warning and response module is used to define multiple early warning levels based on the thermal runaway prediction signal, including: Level 1 early warning, Level 2 alarm and Level 3 emergency, each early warning level corresponds to a different protection strategy; the protection strategy includes: enhanced cooling, activation of adjacent module heat insulation, directional spraying of fire extinguishing agent and system depressurization; The adaptive closed-loop control module is electrically connected to the multi-source information acquisition module and the hierarchical early warning and response module; the adaptive closed-loop control module is used for: Based on the real-time heat load data transmitted by the multi-source information acquisition module and the warning level transmitted by the graded warning and response module, the operating parameters of the cooling system are dynamically adjusted; the real-time heat load data includes: charge / discharge rate and ambient temperature; the operating parameters include: pump speed, fan speed and cooling power.

[0006] Furthermore, the multi-source information acquisition module includes: Data acquisition unit and data preprocessing unit; among which, The data acquisition unit is used to acquire temperature anomaly rate data, characteristic gas concentration data, and voltage drop data in real time from a distributed sensor network; wherein, the temperature anomaly rate data is calculated by monitoring the rate of temperature change inside the battery using a temperature sensor; the characteristic gas concentration data is obtained by detecting the concentration of characteristic gases escaping from the battery using a gas sensor; and the voltage drop data is obtained by monitoring the instantaneous drop in battery terminal voltage using a voltage sensor. The data preprocessing unit is used to filter, normalize, and time-align the raw sensor working data.

[0007] Furthermore, the sensor network includes: Temperature sensors, gas sensors, and voltage sensors; among them, The temperature sensor is used to monitor the temperature of the battery cell or module and is placed in key hot spot areas; The gas sensor is used to detect the characteristic gas of a battery leak; The voltage sensor is used to monitor the battery terminal voltage; The sensor data is wirelessly transmitted to the multi-source information acquisition module; All sensor calibrations are performed automatically on a regular basis.

[0008] Furthermore, the fusion analysis module is used to apply an algorithm model to perform fusion analysis on the preprocessed data; wherein, The machine learning model employs a supervised learning algorithm and is trained based on a historical thermal runaway event dataset. The threshold logic judgment model is used to perform logical operations on multi-source data combinations based on preset threshold conditions to generate a thermal runaway probability score. The fusion analysis module is used to output thermal runaway prediction signals, including prediction time points and confidence levels.

[0009] Furthermore, the method by which the fusion analysis module generates the thermal runaway prediction signal includes: Features strongly correlated with thermal runaway are extracted from the preprocessed working data to construct a feature vector; wherein, the features strongly correlated with thermal runaway include: temperature anomaly rate features, characteristic gas concentration features, and voltage drop features; The constructed feature vector is input into a pre-trained machine learning model, which outputs a continuous probability value of thermal runaway risk and a discrete risk level label; the extracted features are logically compared and combined with a preset dynamic threshold rule, which is adjusted according to the ambient temperature or battery health status. Uncertainty assessment is performed on the output results of the two models, including: calculating the confidence score of the machine learning model output and assessing the completeness and consistency of the triggering of each rule in the threshold logic model; The weighted average method is used to fuse the probability value, level, and judgment result output by the machine learning model and the threshold logic model to obtain the final comprehensive thermal runaway risk score and unified risk level.

[0010] Furthermore, the fusion analysis module is used to generate a thermal runaway prediction signal, including prediction time points and confidence levels, based on the final comprehensive thermal runaway risk score and unified risk level; wherein, The predicted time points are derived through regression models or by statistical inference based on the time intervals from the appearance of the feature to the occurrence of thermal runaway in historical data.

[0011] Furthermore, the tiered early warning and response module includes: The warning level definition unit and the policy execution unit; among them, The warning level definition unit is used to classify warning levels according to the severity of the thermal runaway prediction signal; wherein, Level 1 warning corresponds to a low-risk signal, Level 2 alarm corresponds to a medium-risk signal, and Level 3 emergency corresponds to a high-risk signal; each warning level is associated with a specific protection strategy, Level 1 warning strategy includes increasing the coolant flow rate of the cooling system or reducing the cooling temperature setpoint, Level 2 alarm strategy includes activating the thermal insulation material between adjacent battery modules, and Level 3 emergency strategy includes starting the directional spray system to spray fire extinguishing agent at the fault point and triggering the system pressure relief valve; The strategy execution unit is used to drive the executor to implement the protection strategy through the control interface; The tiered early warning and response module also includes a feedback mechanism for monitoring the effectiveness of strategy execution and adjusting the early warning level based on real-time response data.

[0012] Furthermore, the implementation of the protection strategy in the hierarchical early warning and response module includes strategy customization and priority management; wherein, Strategy customization is based on battery type and application scenario; Prioritization management ensures that the highest risk is addressed first when multiple events occur; Policy execution includes timing control; The effectiveness of the strategy is evaluated by monitoring changes in temperature, gas, and pressure using additional sensors; All strategies are recorded in the form of logs.

[0013] Furthermore, the adaptive closed-loop control module includes: The system comprises a data acquisition unit, a control algorithm unit, and an actuator unit; among which, The data acquisition unit is used to monitor heat load data in real time, including: charge / discharge rate data, ambient temperature data, and battery temperature distribution data; The control algorithm unit is used to apply a closed-loop control algorithm to calculate the adjustment values ​​of the cooling system operating parameters, and its control objective is to maintain the battery temperature within the optimal window. The actuator unit is used to adjust the cooling components according to the control signal, and its control objective is to maintain the battery temperature within the optimal window. The adaptive closed-loop control module integrates learning capabilities and optimizes control parameters based on historical operating data.

[0014] Furthermore, the cooling system includes a liquid cooling subsystem, an air cooling subsystem, and a refrigeration subsystem; The liquid cooling subsystem is used to contact the battery module through a cooling plate or cold pipe, circulate coolant to absorb heat, and the pump speed is adjustable to control the flow rate. The air-cooling subsystem is used to force airflow across the battery surface for convective heat exchange using a fan. The fan speed is adjustable to control the airflow and heat dissipation rate. The refrigeration subsystem is used to provide active refrigeration based on a vapor compression cycle, and the refrigeration power is adjustable; The cooling system operating parameters are adjusted based on the output of the adaptive closed-loop control module.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The high-density energy storage battery thermal management system proposed in this invention significantly improves the safety and efficiency of battery thermal management by integrating multi-source information acquisition, fusion analysis, hierarchical early warning and response, and adaptive closed-loop control modules. It can comprehensively capture multi-dimensional characteristics before battery thermal runaway, achieving more accurate thermal runaway prediction. The flexible matching of early warning levels and protection strategies ensures dynamic adjustment of response measures based on battery status and environmental conditions, enhancing the targeted nature of protection. The adaptive adjustment capability of the cooling system is optimized, enabling rapid response to changes in charge / discharge rate and ambient temperature, accurately adjusting operating parameters, and maintaining battery temperature within the optimal range. This effectively extends battery cycle life, improves energy output stability, reduces the risk of thermal runaway, and provides a solid guarantee for the safe operation of high-density energy storage batteries. Attached Figure Description

[0016] Figure 1 A schematic diagram of a high-density energy storage battery thermal management system. Figure 2 This is a flowchart of the thermal management system for high-density energy storage batteries. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, a high-density energy storage battery thermal management system with multi-level safety protection includes: A multi-source information acquisition module is used to collect and acquire working data from multiple sensors and to preprocess the working data. The fusion analysis module is electrically connected to the multi-source information acquisition module; the fusion analysis module is used to receive preprocessed working data, perform fusion analysis on the working data through an algorithm model, and generate a thermal runaway prediction signal; the algorithm model includes: a machine learning model and a threshold logic judgment model; The graded early warning and response module is electrically connected to the fusion analysis module; the graded early warning and response module is used to define multiple early warning levels based on the thermal runaway prediction signal, including: Level 1 early warning, Level 2 alarm and Level 3 emergency, each early warning level corresponds to a different protection strategy; the protection strategy includes: enhanced cooling, activation of adjacent module heat insulation, directional spraying of fire extinguishing agent and system depressurization; The adaptive closed-loop control module is electrically connected to the multi-source information acquisition module and the hierarchical early warning and response module; the adaptive closed-loop control module is used for: Based on the real-time heat load data transmitted by the multi-source information acquisition module and the warning level transmitted by the graded warning and response module, the operating parameters of the cooling system are dynamically adjusted; the real-time heat load data includes: charge / discharge rate and ambient temperature; the operating parameters include: pump speed, fan speed and cooling power.

[0019] The multi-source information acquisition module includes a data acquisition unit and a data preprocessing unit. The data acquisition unit is used to acquire temperature anomaly rate data, characteristic gas concentration data, and voltage drop data in real time from a distributed sensor network. When acquiring temperature anomaly rate data, temperature sensors are first placed at key hot spots such as the tabs of individual battery cells, the center of the module, and edge areas with weak heat dissipation. The temperature values ​​at these locations are collected in real time at a sampling frequency of 1Hz. After collecting the temperature values ​​at 5 consecutive time points, the temperature change slope is obtained through linear fitting. When the slope exceeds 1.5 times the average temperature change rate under normal charging and discharging conditions, it is marked as temperature anomaly rate data, and the corresponding timestamp and battery number are recorded simultaneously for subsequent location. When acquiring characteristic gas concentration data, gas sensors are placed in the gaps between modules and in the ventilation of the battery box. Near the inlet, the concentrations of common early-stage thermal runaway characteristic gases such as carbon monoxide, hydrogen fluoride, and ethylene emitted from the battery are detected at a sampling frequency of 0.5 Hz. When the concentration of a certain characteristic gas exceeds 20% of the preset safety threshold, the data is marked as a characteristic gas concentration data that requires special attention. At the same time, the ambient temperature and humidity during the detection are recorded to eliminate the influence of environmental factors on the gas diffusion rate. When acquiring voltage drop data, a voltage sensor is connected in parallel across each battery cell to monitor the battery terminal voltage at a sampling frequency of 2 Hz. When the voltage drop exceeds 50 mV within two consecutive sampling cycles and the voltage after the drop is lower than 95% of the rated voltage of the battery cell, it is determined to be voltage drop data. At the same time, the start time and duration of the drop, as well as the charging and discharging current during this period, are recorded to distinguish between voltage changes caused by normal current fluctuations and abnormal voltage drops. The data preprocessing unit is used to filter, normalize, and time-align the raw sensor operating data. The filtering employs a combination of Kalman filtering and moving average filtering. First, based on the sensor's noise characteristics, the process noise covariance and measurement noise covariance are preset. Kalman filtering removes random noise generated by sensor circuit interference. Then, the filtered result is processed using a moving average with a window size of 5 to eliminate instantaneous pulse interference, ensuring that data fluctuations are controlled within the sensor's accuracy tolerance. Normalization uses a min-max normalization method, mapping data of different dimensions, such as temperature anomaly rate (unit: °C / min), characteristic gas concentration (unit: ppm), and voltage drop (unit: V), to the [0,1] interval. The calculation method is as follows: Normalized data = (original data - historical minimum value of this data type) / (historical maximum value of this data type - historical minimum value of this data type); The historical minimum and maximum values ​​are derived from sample data collected before system deployment under different operating conditions, such as normal charging and discharging, low-temperature start-up, and high-temperature operation, ensuring coverage of all possible operating scenarios of the system. Time alignment uses the timestamp of the voltage sensor with the highest sampling frequency as the reference time axis. Linear interpolation is used to complete the time points of temperature data and gas concentration data with lower sampling frequencies. One completion point is evenly inserted between every two adjacent original data points for temperature data, and three completion points are evenly inserted between every two adjacent original data points for gas concentration data. The values ​​of the completion points are calculated through linear interpolation, so that all data correspond one-to-one on the same time axis and the time interval is uniformly 0.5 seconds, which meets the time consistency requirements of subsequent fusion analysis.

[0020] The sensor network includes temperature sensors, gas sensors, and voltage sensors. The temperature sensors are used to monitor the temperature of individual battery cells or modules and are placed in key hotspot areas. The gas sensors are used to detect characteristic gases leaking from the battery. The voltage sensors are used to monitor the battery terminal voltage. Sensor data is wirelessly transmitted to the multi-source information acquisition module. Sensor calibration is performed automatically and periodically, with a calibration cycle set to once a month. During calibration, the system automatically retrieves a preset standard signal to compare the sensor output values. If the deviation exceeds the sensor's accuracy range, the sensor parameters are automatically adjusted to ensure the accuracy of data acquisition.

[0021] The fusion analysis module is used to perform fusion analysis on the preprocessed data using an algorithm model. The machine learning model adopts a supervised learning algorithm and is trained on a historical thermal runaway event dataset. This dataset includes different battery types such as ternary lithium batteries and lithium iron phosphate batteries, as well as thermal runaway experimental data and normal operation monitoring data under different capacities and cycle numbers, ensuring the diversity and representativeness of the dataset. The threshold logic judgment model performs logical operations on the combination of multi-source data based on preset threshold conditions to generate a thermal runaway probability score. The fusion analysis module outputs a thermal runaway prediction signal, including the prediction time point and confidence level.

[0022] The method steps for the fusion analysis module to generate thermal runaway prediction signals are as follows: Features strongly correlated with thermal runaway are extracted from the preprocessed working data to construct a feature vector. In addition to temperature anomaly rate features, characteristic gas concentration features, and voltage drop features, the temperature anomaly rate feature also includes the average and variance of the temperature change slopes of multiple battery cells within the same module. The average reflects the overall temperature change trend of the module, and the variance reflects the uniformity of temperature distribution within the module. If the variance exceeds a preset value, it is marked as a local overheating risk feature. The characteristic gas concentration feature also includes the concentration ratios of various characteristic gases, such as the concentration ratio of carbon monoxide to ethylene; the concentration change rate, i.e., the amount of concentration increase per unit time; the concentration ratio can determine the stage of thermal runaway, and the change rate can reflect the speed of thermal runaway development. The voltage drop feature also includes the recovery status of the voltage after the drop; if it cannot be restored to the normal range, it is marked as an irreversible internal battery damage feature. These individual and combined features are arranged in a preset order to form a feature vector of corresponding dimensions to comprehensively cover the key information of thermal runaway.

[0023] Next, the constructed feature vector is input into a pre-trained machine learning model, which outputs a continuous thermal runaway risk probability value and discrete risk level labels. The model uses a gradient boosting tree algorithm to adapt to complex scenarios with multiple influencing factors. Training employs cross-validation to alternate between training and validation to avoid overfitting. For example, the dataset is randomly divided into 5 parts, with 4 parts used as the training set and 1 part as the validation set for alternating training. Simultaneously, a grid search algorithm is used to optimize hyperparameters such as the learning rate, tree depth, and number of leaf nodes, ensuring the model's prediction accuracy on the validation set reaches over 95%. The model's output thermal runaway risk probability value ranges from [0,1], and the discrete risk level labels are divided into low risk, medium risk, and high risk, corresponding to probability value intervals of [0,0.3], [0.3,0.7], and [0.7,1], respectively. This interval division is based on historical data statistics; approximately 30% of cases in the [0.3,0.7] interval experience thermal runaway, and approximately 85% of cases in the [0.7,1] interval experience thermal runaway, effectively distinguishing risk levels. Simultaneously, the extracted features are compared with a set of preset... The dynamic threshold rules are logically compared and combined. The dynamic thresholds are adjusted according to the ambient temperature or battery health status. When the ambient temperature is higher than 35℃, the temperature anomaly rate threshold is reduced by 20% and the characteristic gas concentration threshold is reduced by 15% to improve the early warning sensitivity. When the ambient temperature is lower than 0℃, the voltage drop threshold is increased by 10% to avoid false judgments. When the battery cycle count exceeds 1000 times or the capacity decay rate exceeds 20%, all characteristic thresholds are reduced by 10% to cope with the increased risk caused by battery aging. The dynamic threshold rules include single-feature threshold rules and multi-feature combination rules. Single-feature threshold rules include temperature anomaly rate exceeding the threshold for 5 seconds, characteristic gas concentration exceeding the threshold for 10 seconds, etc. Multi-feature combination rules include temperature anomaly rate and characteristic gas concentration both exceeding the threshold, temperature anomaly rate and voltage drop amplitude both exceeding the threshold, etc. Logical operations are performed on the rules. Single-feature rule satisfaction is recorded as primary trigger, and multi-feature combination rule satisfaction is recorded as advanced trigger. The risk level (low, medium, high) and trigger basis of the threshold logic model are generated according to the trigger level and number.

[0024] Then, uncertainty assessment is performed on the output results of the two models. When calculating the confidence score of the machine learning model output, based on the probability distribution predicted by the model, the confidence score = 1 - the entropy value corresponding to the probability value. The smaller the entropy value, the higher the certainty. The score range is [0,1]. If it is lower than 0.6, it indicates that the model uncertainty is high and other results need to be considered. When assessing the completeness of the triggering of each rule in the threshold logic model, it is checked whether the current triggering rule covers all rules related to data features. For example, when both temperature and gas concentration exceed the threshold, whether single-feature and multi-feature combination rules are triggered at the same time, and whether there are analysis thresholds or data problems where relevant rules are not triggered. When assessing consistency, it is checked whether the risk levels triggered by different rules are consistent. For example, single-feature rules trigger low risk while multi-feature combination rules trigger medium risk. It is necessary to verify with the original data to eliminate the influence of data noise.

[0025] Then, a weighted average method is used to fuse the probability values ​​and levels output by the machine learning model and the judgment results output by the threshold logic model. The weights are set based on the historical prediction accuracy of the two models, such as 0.6 for the machine learning model and 0.4 for the threshold logic model. First, the risk levels of the threshold logic model are converted into corresponding probability values, such as 0.2 for low risk, 0.5 for medium risk, and 0.8 for high risk. Then, the comprehensive risk probability value is calculated as: Machine Learning Model Output Probability Value × 0.6 + Threshold Logic Model Conversion Probability Value × 0.4. A unified risk level is determined based on the comprehensive probability value ([0,0.3] Low Risk, [0.3,0.7] Medium Risk, [0.7,1] High Risk). At the same time, the weights and output values ​​of the fusion process are recorded for subsequent retrospective optimization.

[0026] Finally, based on the final comprehensive thermal runaway risk score and unified risk level, a thermal runaway prediction signal containing the predicted time point and confidence level is generated. The predicted time point is determined by regression model combined with historical data statistical inference. First, a linear regression time prediction model is trained, with the feature vector as input and the actual time interval from the appearance of the feature to the occurrence of thermal runaway as output. Through training with historical data, the prediction error is controlled within ±30 seconds. Then, the average and standard deviation of the time interval under the same risk level are statistically analyzed. If the deviation between the predicted time point of the regression model and the average value exceeds the standard deviation, it is adjusted to the average value to improve accuracy. The confidence level = (machine learning model confidence score + (1 - time prediction model error rate)) / 2, which comprehensively reflects the reliability of the prediction. The generated thermal runaway prediction signal also contains the basis of each model output, providing support for the formulation of subsequent protection strategies.

[0027] The graded early warning and response module includes an early warning level definition unit and a strategy execution unit. The early warning level definition unit is used to classify early warning levels according to the severity of the thermal runaway prediction signal. Level 1 early warning corresponds to a low-risk signal, Level 2 alarm corresponds to a medium-risk signal, and Level 3 emergency corresponds to a high-risk signal. Each early warning level is associated with a specific protection strategy. The Level 1 early warning strategy includes increasing the coolant flow rate of the cooling system or reducing the cooling temperature setpoint. The Level 2 alarm strategy includes activating the thermal insulation material between adjacent battery modules. The Level 3 emergency strategy includes starting the directional spray system to spray fire extinguishing agent at the fault point and triggering the system pressure relief valve. The strategy execution unit is used to drive the actuator through the control interface to implement the protection strategy. The graded early warning and response module also includes a feedback mechanism to monitor the effect of strategy execution and adjust the early warning level based on real-time response data.

[0028] The implementation of the protection strategy in the graded early warning and response module includes strategy customization and priority management. Strategy customization is based on battery type and application scenario. For example, ternary lithium batteries have a high risk of thermal runaway, so their level 2 alarm strategy will increase the monitoring frequency of characteristic gas concentration in advance. In the application scenario of energy storage power stations, the level 3 emergency strategy will link the grid dispatch system to reduce the load on the faulty battery pack. Priority management ensures that the highest risk is dealt with first in the event of multiple events. For example, when a level 1 warning and a level 3 emergency event occur at the same time, the protection strategy corresponding to the level 3 emergency is executed first. Strategy execution includes timing control. For example, in the level 3 emergency strategy, the directional spray extinguishing agent is started first, and the system pressure relief valve is triggered 10 seconds later to avoid leakage caused by a sudden increase in internal pressure when the extinguishing agent is sprayed. Strategy effectiveness evaluation is carried out by monitoring temperature, gas and pressure changes through additional sensors. For example, if the temperature continues to rise at a rate higher than normal after the level 1 warning is implemented, the warning level will be upgraded to level 2. All strategy logs record information such as strategy execution time, execution parameters, and changes in monitoring data, which facilitates subsequent fault analysis and strategy optimization.

[0029] The adaptive closed-loop control module includes a data acquisition unit, a control algorithm unit, and an actuator unit. The data acquisition unit monitors heat load data in real time, including charge / discharge rate data, ambient temperature data, and battery temperature distribution data. Charge / discharge rate data is acquired via a current sensor at a sampling frequency of 1Hz; ambient temperature data is acquired via a temperature sensor located outside the battery compartment at a sampling frequency of 0.5Hz; and battery temperature distribution data is acquired via temperature sensors at different locations within the module to ensure coverage of the overall temperature distribution. The control algorithm unit applies a closed-loop control algorithm to calculate adjustment values ​​for the cooling system's operating parameters. Its control objective is to maintain the battery temperature within the optimal window. The operating temperature window is determined based on the battery type. For example, the optimal operating temperature window for lithium iron phosphate batteries is 15℃-35℃, and for ternary lithium batteries it is 20℃-30℃. The closed-loop control algorithm adopts the PID control algorithm, which adjusts the parameters of the proportional, integral, and derivative links to quickly respond to changes in heat load and reduce temperature fluctuations. The actuator unit is used to adjust the cooling components according to the control signal. The adaptive closed-loop control module integrates learning capabilities and optimizes control parameters based on historical operating data. For example, by analyzing the control effect under different charge / discharge rates and ambient temperatures, the proportional coefficient, integral time, and derivative time of the PID algorithm are adjusted to keep the overshoot of temperature control within 5% and shorten the response time to within 10 seconds.

[0030] The cooling system includes a liquid cooling subsystem, an air cooling subsystem, and a refrigeration subsystem. The liquid cooling subsystem uses a cooling plate or cooling pipe to contact the battery module, circulating coolant to absorb heat. The pump speed is adjustable to control the flow rate. The cooling plate is made of aluminum alloy, and the contact area with the battery module is no less than 80% of the module's surface area to ensure efficient heat transfer. The coolant is an ethylene glycol aqueous solution with a freezing point below -30℃ and a boiling point above 100℃, adaptable to different ambient temperatures. The air cooling subsystem uses a fan to force airflow across the battery surface for convective heat exchange. The fan speed is adjustable to control the airflow and heat dissipation rate. The fan is positioned at the air inlet and outlet of the battery box, forming a constant airflow. The airflow channel improves heat dissipation uniformity. The refrigeration subsystem provides active cooling based on a vapor compression cycle, with adjustable cooling power. The evaporator of the refrigeration subsystem is in contact with the coolant pipeline of the liquid cooling subsystem, achieving active cooling through the coolant to meet the heat dissipation requirements in high-temperature environments. The operating parameters of the cooling system are adjusted based on the output of the adaptive closed-loop control module. For example, when the adaptive closed-loop control module detects that the battery temperature exceeds the upper limit of the optimal window, it first increases the pump speed of the liquid cooling subsystem to increase the coolant flow. If the temperature still does not drop, it increases the fan speed of the air cooling subsystem. If necessary, it starts the refrigeration subsystem and adjusts the cooling power to ensure that the battery temperature returns to the optimal window.

[0031] refer to Figure 2 As shown, the workflow of the above system is as follows: The system first works through a multi-source information acquisition module. In the distributed sensor network, temperature sensors, gas sensors, and voltage sensors collect temperature anomaly rate, characteristic gas concentration, and voltage drop data at specific frequencies. After the data is wirelessly transmitted, the preprocessing unit uses combined filtering to remove noise, min-max normalization to unify dimensions, and linear interpolation for time alignment.

[0032] The preprocessed data is fed into the fusion analysis module, where multi-dimensional thermal runaway-related features are extracted to construct vectors. Then, the vectors are input into a pre-trained machine learning model to obtain risk probabilities and levels. Simultaneously, dynamic threshold rules are used for calculation, and after uncertainty assessment and weighted fusion, a thermal runaway prediction signal containing prediction time points and confidence levels is generated.

[0033] The graded early warning and response module classifies warning levels according to signals, triggers corresponding protection strategies, and monitors the effectiveness of the strategies to adjust the level accordingly. The adaptive closed-loop control module obtains heat load data in real time, combines it with the warning level, and uses a PID algorithm to dynamically adjust parameters such as cooling system pump speed and fan speed to maintain the battery temperature within the optimal window.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A high-density energy storage battery thermal management system with multi-level security protection, characterized in that, The application relates to a thermal runaway prediction system and method. The application comprises: a multi-source information acquisition module, which is used for collecting and acquiring working data of a plurality of sensors and pre-processing the working data; a fusion analysis module which is electrically connected with the multi-source information acquisition module; the fusion analysis module is used for receiving the pre-processed working data, performing fusion analysis on the working data through an algorithm model, and generating a thermal runaway prediction signal; the algorithm model comprises a machine learning model and a threshold logic judgment model; a hierarchical early warning and response module which is electrically connected with the fusion analysis module; the hierarchical early warning and response module is used for defining a plurality of early warning levels according to the thermal runaway prediction signal, wherein the early warning levels comprise a first-level early warning, a second-level early warning and a third-level emergency, and each early warning level corresponds to a different protection strategy; the protection strategy comprises strengthening cooling, starting adjacent module heat insulation, directional injection of fire extinguishing agent and system pressure relief; an adaptive closed-loop control module which is electrically connected with the multi-source information acquisition module and the hierarchical early warning and response module; the adaptive closed-loop control module is used for:

2. The high-density energy storage battery thermal management system with multi-level safety protection according to claim 1, characterized in that, based on real-time thermal load data transmitted by the multi-source information acquisition module and early warning levels transmitted by the hierarchical early warning and response module, dynamically adjusting operation parameters of a cooling system; the real-time thermal load data comprises a charging / discharging rate and an environmental temperature; the operation parameters comprise pump speed, fan rotating speed and refrigeration power. The multi-source information acquisition module comprises: a data acquisition unit and a data preprocessing unit; wherein the data acquisition unit is used for acquiring temperature abnormal rate data, characteristic gas concentration data and voltage sudden drop data from a distributed sensor network in real time; wherein the temperature abnormal rate data is calculated by monitoring the internal temperature change rate of a battery through a temperature sensor; the characteristic gas concentration data is obtained by detecting the characteristic gas concentration of battery leakage through a gas sensor; and the voltage sudden drop data is obtained by monitoring the instantaneous drop of battery terminal voltage through a voltage sensor; 3. The high-density energy storage battery thermal management system with multi-level safety protection of claim 2, wherein, the data preprocessing unit is used for filtering, normalizing and time aligning the original sensor working data. The sensor network comprises: temperature sensors, gas sensors and voltage sensors; wherein the temperature sensors are used for monitoring the temperature of a battery monomer or module and are arranged in key hot spot areas; the gas sensors are used for detecting characteristic gases leaked by the battery; the voltage sensors are used for monitoring the terminal voltage of the battery; the sensor data is transmitted to the multi-source information acquisition module through wireless transmission; 4. The high-density energy storage battery thermal management system with multi-level safety protection of claim 1, wherein, all the sensors are calibrated periodically and automatically. The fusion analysis module is used for applying an algorithm model to perform fusion analysis on the pre-processed data; wherein the machine learning model adopts a supervised learning algorithm and is trained based on a historical thermal runaway event data set; the threshold logic judgment model is used for performing logic operation on a multi-source data combination based on a preset threshold condition to generate a thermal runaway probability score; 5. The high-density energy storage battery thermal management system with multi-level safety protection of claim 1, wherein, the fusion analysis module is used for outputting a thermal runaway prediction signal, which comprises a prediction time point and a confidence degree. The method for generating a thermal runaway prediction signal by the fusion analysis module comprises: extracting features strongly related to thermal runaway from the preprocessed working data to construct a feature vector; wherein the features strongly related to thermal runaway include: temperature abnormal rate features, characteristic gas concentration features, and voltage sudden drop features; inputting the constructed feature vector into a pre-trained machine learning model to output continuous thermal runaway risk probability values and discrete risk level labels; performing logical comparison and combination operation on the extracted features and a preset dynamic threshold rule, wherein the dynamic threshold is adjusted according to the ambient temperature or the battery health state; performing uncertainty evaluation on the output results of the two models, including: calculating the confidence score of the machine learning model output and evaluating the completeness and consistency of each rule triggered in the threshold logic model; fusing the probability values, levels output by the machine learning model and the judgment results output by the threshold logic model by using the weighted average method to obtain the final comprehensive thermal runaway risk score and the unified risk level.

6. The high-density energy storage battery thermal management system with multi-level safety protection of claim 5, wherein, The fusion analysis module is configured to generate a thermal runaway prediction signal containing a prediction time point and a confidence level based on the final comprehensive thermal runaway risk score and the unified risk level; wherein The prediction time point is obtained by a regression model or statistical inference based on the time interval from the appearance of the feature to the occurrence of thermal runaway in the historical data.

7. The high-density energy storage battery thermal management system with multi-level safety protection of claim 1, wherein, The hierarchical early warning and response module includes: a warning level definition unit and a strategy execution unit; wherein The warning level definition unit is configured to divide the warning levels according to the severity of the thermal runaway prediction signal; wherein the first-level warning corresponds to the low-risk signal, the second-level warning corresponds to the medium-risk signal, and the third-level emergency corresponds to the high-risk signal; each warning level is associated with a specific protection strategy, the first-level warning strategy includes increasing the flow of cooling liquid of the cooling system or reducing the refrigeration temperature set value, the second-level warning strategy includes activating the thermal insulation material between adjacent battery modules, and the third-level emergency strategy includes starting the directional injection system to spray fire extinguishing agent to the fault point and triggering the system pressure relief valve; The strategy execution unit is configured to drive the actuators to implement the protection strategy through the control interface; The hierarchical early warning and response module further includes a feedback mechanism for monitoring the strategy execution effect and adjusting the warning level based on real-time response data.

8. The high-density energy storage battery thermal management system with multi-level safety protection of claim 1, wherein, The implementation of the protection strategy in the hierarchical early warning and response module includes strategy customization and priority management; wherein The strategy customization is based on the battery type and the application scenario; The priority management ensures that the highest risk is processed first in multiple events; The strategy execution includes timing control; The strategy effect evaluation monitors the temperature, gas and pressure changes through additional sensors; All strategies are recorded in the form of logs.

9. The high-density energy storage battery thermal management system with multi-level safety protection of claim 1, wherein, The adaptive closed-loop control module includes: a data acquisition unit, a control algorithm unit, and an actuator unit; wherein The data acquisition unit is configured to monitor thermal load data in real time, including: charge-discharge rate data, ambient temperature data, and battery temperature distribution data; The control algorithm unit is configured to apply a closed-loop control algorithm to calculate the adjustment value of the cooling system operating parameters, and the control target is to maintain the battery temperature in the optimal window; The executor unit is configured to adjust the cooling assembly according to the control signal, and a control target is to maintain the battery temperature in an optimal window. The adaptive closed-loop control module integrates learning ability, and optimizes control parameters based on historical operation data.

10. The high-density energy storage battery thermal management system with multi-level safety protection of claim 1, wherein, The cooling system comprises a liquid cooling subsystem, an air cooling subsystem and a refrigeration subsystem. The liquid cooling subsystem is configured to contact the battery module through a cooling plate or a cooling pipe, and to absorb heat by circulating cooling liquid. The air cooling subsystem is configured to use a fan to force air to flow through the surface of the battery for convective heat exchange. The refrigeration subsystem is configured to provide active refrigeration based on a vapor compression cycle. The adaptive closed-loop control module integrates learning ability, and optimizes control parameters based on historical operation data.

Images