Intelligent fire-fighting monitoring and early warning system for battery energy storage cabin
By constructing a thermal runaway prediction model and combining real-time and historical data, the CNN-LSTM model is used to conduct risk assessment and trend analysis of the battery energy storage compartment. This solves the shortcomings of thermal runaway monitoring and early warning in the battery energy storage compartment, realizes accurate identification and early warning of thermal runaway risks, and improves the reliability and predictability of safety protection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-10
- Publication Date
- 2026-04-03
AI Technical Summary
Existing technologies for monitoring and early warning of thermal runaway in battery energy storage compartments suffer from incomplete risk assessments, a lack of forward-looking early warning mechanisms, and difficulty in accurately identifying risk levels and trends, resulting in insufficient initiative and foresight in safety protection.
A thermal runaway prediction model is constructed. Experimental operation parameters throughout the entire life cycle are obtained through thermal runaway experiments. Combined with real-time and historical fault data, a CNN-LSTM fusion model is used for risk assessment and trend analysis to construct a thermal runaway probability prediction model and achieve early warning.
It enables accurate identification and early warning of thermal runaway risks in battery energy storage compartments, improves the reliability and predictability of safety protection, avoids misjudgment of trends in low-risk states, and ensures sufficient fire preparation time.
Smart Images
Figure CN121789366A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of energy storage fire safety technology, specifically an intelligent fire monitoring and early warning system for battery energy storage compartments. Background Technology
[0002] With the rapid development of the new energy industry, battery energy storage containers are widely used due to their advantages such as high energy density and flexible deployment. However, the batteries inside are prone to thermal runaway due to factors such as overcharging, internal short circuits, and external heating. Once runaway occurs, it will trigger a chain reaction, posing a serious threat to equipment safety and personnel life. Therefore, it is crucial to build a reliable thermal runaway monitoring and early warning system.
[0003] Existing technologies for monitoring and early warning of thermal runaway in battery energy storage compartments have significant defects and shortcomings: they rely solely on real-time operating parameters to assess risk, without integrating historical fault data to calculate the failure rate, resulting in incomplete risk assessment and difficulty in accurately identifying the true risk level; there is a gap in trend monitoring in low-risk scenarios, as existing technologies lack monitoring of the potential upward trend of probability when the risk of thermal runaway is low, relying only on a single parameter or local fluctuations, which easily leads to misjudgment of the overall trend and failure to detect early signs of risk; the early warning mechanism lacks foresight, only passively initiating warnings after the risk exceeds the limit, without building a probability prediction model to calculate the warning time, failing to reserve sufficient preparation time for fire response, resulting in insufficient initiative and foresight in safety protection.
[0004] Therefore, the present invention provides an intelligent fire monitoring and early warning system for battery energy storage compartments. Summary of the Invention
[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.
[0006] The technical solution adopted by this invention to solve its technical problem is:
[0007] Thermal runaway test module: Conduct thermal runaway experiments on the battery storage compartment, obtain experimental operation parameters of the entire life cycle of thermal runaway triggering of the battery storage compartment, and build a thermal runaway prediction model;
[0008] Risk assessment module: Real-time acquisition of actual operating parameters of the battery energy storage compartment and input into the thermal runaway prediction model, outputting the thermal runaway probability, and comprehensively assessing the thermal runaway risk of the battery energy storage compartment by combining the historical fault analysis library of the battery energy storage compartment.
[0009] Trend Analysis Module: If the risk of thermal runaway is low, a preset monitoring period is set. By analyzing the trend of thermal runaway probability changes during the monitoring period, it can be determined whether the probability of thermal runaway is on the rise during the monitoring period.
[0010] Thermal runaway early warning module: If the probability of thermal runaway is on the rise, a thermal runaway probability prediction model is constructed to determine the thermal runaway early warning time and feed it back to the fire early warning platform to provide early warning.
[0011] As a further aspect of the present invention: the process of constructing the thermal runaway prediction model is as follows:
[0012] Based on the CNN-LSTM fusion model, the preprocessed data is integrated into a dataset in chronological order. The dataset is then divided into training, validation, and test sets in a 7:2:1 ratio. Cross-entropy is used as the loss function and Adam is used as the optimizer to construct a thermal runaway prediction model.
[0013] As a further aspect of the present invention: the process of comprehensively assessing the thermal runaway risk of the battery energy storage compartment is as follows:
[0014] The historical fault analysis database of the battery energy storage compartment is analyzed to determine the failure rate of the battery energy storage compartment.
[0015] The failure rate of the battery storage compartment is summed with the probability of thermal runaway to obtain the thermal runaway risk value. If the thermal runaway risk value is less than the thermal runaway risk threshold, it means that the thermal runaway risk of the battery storage compartment is low.
[0016] As a further aspect of the present invention: the failure rate determination process of the battery energy storage compartment is as follows:
[0017] Extract actual operating parameter data from historical faults of the battery energy storage compartment and construct a historical fault analysis library;
[0018] Based on the historical fault analysis database, the number of historical faults of the battery energy storage compartment is counted and compared with the number of historical operations of the battery energy storage compartment to obtain the failure rate of the battery energy storage compartment.
[0019] As a further aspect of the present invention: the process for determining whether the probability of thermal runaway shows an upward trend during the monitoring period is as follows:
[0020] Analyze the trend of thermal runaway probability during the monitoring period to determine the probability trend value;
[0021] If the probability trend value is greater than the probability trend threshold, it indicates that the probability of thermal runaway is on the rise during the monitoring period.
[0022] As a further aspect of the present invention: the process for determining the probability trend value is as follows:
[0023] The actual operating parameters of the battery energy storage compartment are continuously monitored during the monitoring period. The monitoring period is divided into several analysis periods according to equal time intervals. The actual operating parameters within the analysis period are input into the thermal runaway prediction model, and the thermal runaway probability of the analysis period is output. The thermal runaway probabilities of all analysis periods are integrated into a thermal runaway probability sequence in chronological order.
[0024] Analyze the thermal runaway probability sequence to determine the thermal runaway probability change ratio, the rise window number ratio, and the probability change degree ratio;
[0025] The weights of the thermal runaway probability change ratio, the rise window number ratio, and the probability change degree ratio are determined by using the analytic hierarchy process (AHP). The probability trend value is obtained by weighted summation of these three ratios.
[0026] As a further aspect of the present invention: the process for determining the thermal runaway probability change ratio is as follows:
[0027] The thermal runaway probability change ratio is obtained by subtracting the reversed thermal runaway probability from the first thermal runaway probability in the thermal runaway probability sequence and then comparing it with the first thermal runaway probability in the thermal runaway sequence.
[0028] As a further aspect of the present invention: the process for determining the ratio of rising windows is as follows:
[0029] The thermal runaway probabilities of two adjacent analysis periods within the thermal runaway probability sequence are used as an analysis window. The difference between the thermal runaway probability of the next analysis period and the thermal runaway probability of the previous period within the analysis window is calculated to obtain the adjacent probability change value.
[0030] If the adjacent probability change value is greater than 0, the corresponding analysis window is recorded as the probability increase window;
[0031] Calculate the percentage of windows with increasing probability across all analysis windows, denoted as the percentage of windows with increasing probability.
[0032] As a further aspect of the present invention: the process for determining the probability change ratio is as follows:
[0033] The mean probability change is obtained by summing the adjacent probability change values of all analysis windows and taking the average. The ratio of the mean probability change to the standard value of the adjacent probability change is then calculated to obtain the probability change ratio.
[0034] As a further aspect of the present invention: the process for determining the thermal runaway early warning time is as follows:
[0035] Based on a machine learning model, using thermal runaway probability sequences as the dataset, the dataset is divided into training, validation and test sets in a 7:2:1 ratio. Mean squared error is used as the loss function and Adam is used as the optimizer to build a thermal runaway probability prediction model.
[0036] A preset thermal runaway probability threshold is set, and the thermal runaway probability threshold is input into the trained thermal runaway probability prediction model. The time point when the thermal runaway probability reaches the thermal runaway probability threshold is output and recorded as the runaway time point. The difference between the runaway time point and the current time point is calculated to obtain the thermal runaway warning time.
[0037] The beneficial effects of this invention are as follows:
[0038] 1. This invention designs repeated thermal runaway experiments covering typical causes such as overcharging, internal short circuits, and external heating. Combined with synchronous time-series monitoring and thermal runaway stage labeling of multiple parameters such as temperature, voltage, current, pressure, and CO concentration of battery cells and the chamber environment, it provides high-quality basic data covering the entire failure path and life cycle for model construction. The prediction model built based on the CNN-LSTM fusion model can accurately learn the evolution law of thermal runaway parameters and achieve effective prediction of thermal runaway probability. By fusing real-time predicted thermal runaway probability with failure rate calculated based on historical fault analysis library, a more comprehensive thermal runaway risk value is formed. With the graded response mechanism for risk value threshold judgment, it can not only accurately identify thermal runaway risk, but also promptly activate fire warning when the risk exceeds the standard, providing reliable technical support for early warning, accurate prediction and safety protection of thermal runaway in battery energy storage chambers.
[0039] 2. This invention addresses scenarios with low thermal runaway risk by constructing a thermal runaway probability sequence. Focusing on three dimensions—overall probability magnitude, continuity of increase, and intensity of change—it achieves accurate and comprehensive judgment of the potential upward trend of thermal runaway probability, effectively avoiding misjudgments of trends caused by local fluctuations or slow increases in low-risk conditions. Simultaneously, after identifying an upward trend, a thermal runaway probability prediction model is constructed based on a machine learning model to accurately calculate the warning time when the thermal runaway probability reaches a threshold. This information is promptly fed back to the fire warning platform, activating the fire warning mechanism in advance. This not only fills the gap in trend monitoring under low-risk conditions but also achieves forward-looking and proactive thermal runaway warnings, significantly improving the predictability and reliability of battery energy storage compartment safety protection. Attached Figure Description
[0040] The invention will now be further described with reference to the accompanying drawings.
[0041] Figure 1 This is a system block diagram of an intelligent fire monitoring and early warning system for a battery energy storage compartment according to an embodiment of the present invention;
[0042] Figure 2This is a flowchart illustrating the acquisition steps of an intelligent fire monitoring and early warning method for a battery energy storage compartment according to an embodiment of the present invention. Detailed Implementation
[0043] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.
[0044] Example 1
[0045] Please see Figure 1 As shown in the figure, an intelligent fire monitoring and early warning system for a battery energy storage compartment according to an embodiment of the present invention includes the following steps:
[0046] Thermal runaway test module: Conduct thermal runaway experiments on the battery storage compartment, obtain experimental operation parameters of the entire life cycle of thermal runaway triggering of the battery storage compartment, and build a thermal runaway prediction model;
[0047] The experimental operating parameters include temperature, voltage, current, pressure, and CO gas concentration.
[0048] The experiment simulates typical causes of thermal runaway in a real energy storage compartment, ensuring data coverage of different failure paths. Repeated experiments are conducted for each type of cause, including:
[0049] Overcharge trigger: Select one complete module in the energy storage compartment and charge it with constant current until a single cell starts to smoke (thermal runaway trigger indicator).
[0050] Internal short circuit triggering: Use a steel needle to puncture a single cell in the module (simulate local failure), and record the diffusion process of experimental parameters of each cell after the short circuit.
[0051] External heating trigger: The energy storage chamber environment is heated at a fixed rate, and the experimental operation parameters of all individual batteries in the module are monitored simultaneously.
[0052] Temperature sensors (attached to the cell surface) and voltage sensors (leading out from the positive and negative terminals) are synchronously deployed for each individual cell in the module; pressure sensors, gas concentration sensors, and ambient temperature sensors are evenly deployed in the compartment, and the correspondence between the ID of each sensor and its monitoring location is clearly defined.
[0053] From the start of the experiment to the triggering of thermal runaway, the timing of experimental parameters was recorded throughout the entire process, with time nodes marked according to the thermal runaway stages. The thermal runaway stages include:
[0054] Preheating phase: Experimental operating parameters are stable and there are no abnormalities (probability of thermal runaway is 0% to 10%).
[0055] Critical stage: Abnormal experimental operating parameters of a single cell (e.g., temperature change rate > 0.5℃ / min, probability of thermal runaway 10%~50%).
[0056] Triggering phase: Two or more types of experimental operating parameters exceed critical values (e.g., temperature ≥45℃ + CO concentration ≥10ppm, thermal runaway probability 50%~90%).
[0057] Propagation stage: Multiple individual cells runaway (thermal runaway probability 90%–100%).
[0058] Data preprocessing transforms experimental parameters into effective features that the model can recognize. Data preprocessing operations include filling missing values with linear interpolation, normalizing experimental parameters of different dimensions, and identifying outlier data using the 3σ criterion.
[0059] Based on the CNN-LSTM fusion model, the preprocessed data is integrated into a dataset in chronological order. The dataset is divided into training, validation and test sets in a 7:2:1 ratio. Cross-entropy is used as the loss function and Adam is used as the optimizer to build a thermal runaway prediction model.
[0060] Risk assessment module: Real-time acquisition of actual operating parameters of the battery energy storage compartment and input into the thermal runaway prediction model, outputting the thermal runaway probability, and comprehensively assessing the thermal runaway risk of the battery energy storage compartment by combining the historical fault analysis library of the battery energy storage compartment.
[0061] Extract actual operating parameter data from historical faults of the battery energy storage compartment and construct a historical fault analysis library;
[0062] Based on the historical fault analysis database, the number of historical faults of the battery energy storage compartment is counted and compared with the number of historical operations of the battery energy storage compartment to obtain the failure rate of the battery energy storage compartment.
[0063] It should be noted that the reason for conducting a thermal runaway assessment of the battery storage compartment based on its historical failures is that the individual cells inside the battery storage compartment have experienced failures such as bulging, leakage, and abnormal voltage, which means that they have potential defects and the risk of thermal runaway is higher than that of fault-free cells. Therefore, it is necessary to conduct a thermal runaway assessment of the battery storage compartment based on its historical failures.
[0064] The failure rate of the battery storage compartment and the probability of thermal runaway are summed to obtain the thermal runaway risk value.
[0065] In some embodiments, the thermal runaway risk value is compared with a thermal runaway risk value threshold, specifically:
[0066] If the thermal runaway risk value is less than the thermal runaway risk value threshold, it means that the thermal runaway risk of the battery energy storage compartment is low.
[0067] If the thermal runaway risk value is greater than or equal to the thermal runaway risk value threshold, it indicates that the thermal runaway risk of the battery energy storage compartment is high, and the fire alarm mechanism should be activated immediately.
[0068] The technical solution of this invention is as follows: A thermal runaway experiment is conducted on the battery storage chamber to obtain experimental operating parameters for the entire lifecycle of the battery storage chamber, triggering thermal runaway, and a thermal runaway prediction model is constructed. Real-time acquisition of the actual operating parameters of the battery storage chamber is performed and input into the thermal runaway prediction model, outputting the thermal runaway probability. Combined with the historical fault analysis database of the battery storage chamber, the thermal runaway risk of the battery storage chamber is comprehensively assessed. This invention, through the design of repeated thermal runaway experiments covering typical causes such as overcharging, internal short circuits, and external heating, combined with synchronous time-series monitoring and thermal runaway stage labeling of multiple parameters such as temperature, voltage, current, pressure, and CO concentration of the battery cells and the chamber environment, provides high-quality basic data covering the entire failure path and lifecycle for model construction. Based on CNN-LSTM... The prediction model constructed by the fusion model can accurately learn the evolution law of thermal runaway parameters and achieve effective prediction of thermal runaway probability. By fusing the real-time predicted thermal runaway probability with the failure rate calculated based on the historical failure analysis library, a more comprehensive thermal runaway risk value is formed. With the graded response mechanism for risk value threshold judgment, it can not only accurately identify thermal runaway risk, but also promptly activate fire warning when the risk exceeds the standard, providing reliable technical support for early warning, accurate prediction and safety protection of thermal runaway in battery energy storage compartments.
[0069] Example 2
[0070] Please see Figure 1 As shown in the embodiment of the present invention, an intelligent fire monitoring and early warning system for a battery energy storage compartment further includes the following steps:
[0071] Trend Analysis Module: If the risk of thermal runaway is low, a preset monitoring period is set. By analyzing the trend of thermal runaway probability changes during the monitoring period, it can be determined whether the probability of thermal runaway is on the rise during the monitoring period.
[0072] The actual operating parameters of the battery energy storage compartment are continuously monitored during the monitoring period. The monitoring period is divided into several analysis periods according to equal time intervals. The actual operating parameters within the analysis period are input into the thermal runaway prediction model, and the thermal runaway probability of the analysis period is output. The thermal runaway probabilities of all analysis periods are integrated into a thermal runaway probability sequence in chronological order.
[0073] The thermal runaway probabilities of two adjacent analysis periods within the thermal runaway probability sequence are used as an analysis window. The difference between the thermal runaway probability of the next analysis period and the thermal runaway probability of the previous period within the analysis window is calculated to obtain the adjacent probability change value.
[0074] If the adjacent probability change value is greater than 0, it indicates that the probability of thermal runaway within the analysis window is increasing, and the corresponding analysis window is recorded as the probability increase window.
[0075] If the change in adjacent probability values is less than or equal to 0, it means that the probability of thermal runaway within the analysis window does not show an upward trend, and the corresponding analysis window is recorded as a non-increasing probability window.
[0076] The difference between the reversed thermal runaway probability and the first sequential thermal runaway probability in the thermal runaway probability sequence is processed, and then the ratio is processed with the first sequential thermal runaway probability to obtain the thermal runaway probability change ratio.
[0077] Count the number of windows with increasing probability in all analysis windows, and calculate the proportion of windows with increasing probability in all analysis windows, denoted as the rising window ratio.
[0078] The mean probability change is obtained by summing the adjacent probability change values of all analysis windows and taking the average. The ratio of the mean probability change to the standard value of the adjacent probability change is then calculated to obtain the probability change ratio.
[0079] It should be noted that the standard value for the change in adjacent probability was set by those skilled in the art based on historical experience;
[0080] The weights of the thermal runaway probability change ratio, the rise window number ratio, and the probability change degree ratio are determined by the analytic hierarchy process (AHP). The probability trend value is obtained by weighted summation of the thermal runaway probability change ratio, the rise window number ratio, and the probability change degree ratio.
[0081] Understandably, the thermal runaway probability change ratio focuses on the relative changes in probability at the beginning and end of the monitoring period, intuitively reflecting the overall rise and fall of probability and laying the macro foundation for trend judgment; the rising window number ratio quantifies the continuity and frequency of the trend by statistically analyzing the proportion of rising windows in adjacent periods, avoiding misjudgment of the overall trend due to local fluctuations; the probability change degree ratio combines historical experience standard values to measure the average intensity of a single probability change, accurately capturing the significance of the trend (such as the difference between a small, slow rise and a large, rapid rise). By judging whether the thermal runaway probability trend is rising, how frequently it is rising, and how significant it is rising, it ensures that potential upward trends can still be accurately identified when the risk is low, providing a scientific and reliable basis for early intervention monitoring and avoiding the spread of risk, and significantly improving the accuracy and comprehensiveness of trend analysis;
[0082] In some embodiments, the probability trend value is compared with the probability trend threshold, specifically:
[0083] If the probability trend value is greater than the probability trend threshold, it indicates that the probability of thermal runaway is on the rise during the monitoring period.
[0084] If the probability trend value is less than the probability trend threshold, it means that the probability of thermal runaway does not show an upward trend during the monitoring period, and no action is taken.
[0085] Thermal runaway early warning module: If the probability of thermal runaway is increasing, a thermal runaway probability prediction model is constructed to determine the thermal runaway early warning time and feed it back to the fire early warning platform to provide early warning.
[0086] Based on a machine learning model, using thermal runaway probability sequences as the dataset, the dataset is divided into training, validation and test sets in a 7:2:1 ratio. Mean squared error is used as the loss function and Adam is used as the optimizer to build a thermal runaway probability prediction model.
[0087] A preset thermal runaway probability threshold is set, and the thermal runaway probability threshold is input into the trained thermal runaway probability prediction model. The time point when the thermal runaway probability reaches the thermal runaway probability threshold is output and recorded as the runaway time point. The difference between the runaway time point and the current time point is calculated to obtain the thermal runaway warning time.
[0088] The thermal runaway warning time is fed back to the fire warning platform to provide early warning, and the fire warning mechanism is activated before the thermal runaway warning time arrives.
[0089] The technical solution of this embodiment is as follows: If the risk of thermal runaway is low, a monitoring period is preset. By analyzing the trend of thermal runaway probability changes during the monitoring period, it is determined whether the thermal runaway probability is increasing during the monitoring period. If the thermal runaway probability is increasing, a thermal runaway probability prediction model is constructed to determine the thermal runaway warning time and feed it back to the fire warning platform for warning. This invention targets scenarios with low thermal runaway risk. By constructing a thermal runaway probability sequence, focusing on three dimensions—overall probability magnitude, continuity of increase, and intensity of change—it achieves accurate and comprehensive judgment of the potential upward trend of thermal runaway probability, effectively avoiding trend misjudgment caused by local fluctuations or slow increases in low-risk conditions. At the same time, after identifying the upward trend, a thermal runaway probability prediction model is constructed based on a machine learning model to accurately calculate the warning time when the thermal runaway probability reaches the threshold. This is promptly fed back to the fire warning platform and the fire warning mechanism is activated in advance. This not only fills the gap in trend monitoring in low-risk conditions but also realizes the foresight and initiative of thermal runaway warning, significantly improving the predictability and reliability of battery energy storage compartment safety protection.
[0090] Example 3
[0091] Based on the same inventive concept as the intelligent fire monitoring and early warning system for battery energy storage compartments in the foregoing embodiments, such as Figure 2 As shown, this application provides an intelligent fire monitoring and early warning method for battery energy storage compartments, which specifically includes the following steps:
[0092] Step 1: Conduct thermal runaway experiments on the battery storage compartment to obtain experimental operating parameters for the entire life cycle of thermal runaway triggering of the battery storage compartment, and construct a thermal runaway prediction model;
[0093] Step 2: Acquire the actual operating parameters of the battery energy storage compartment in real time and input them into the thermal runaway prediction model. Output the thermal runaway probability and combine it with the historical fault analysis database of the battery energy storage compartment to comprehensively assess the thermal runaway risk of the battery energy storage compartment.
[0094] Step 3: If the risk of thermal runaway is low, preset a monitoring period and analyze the trend of thermal runaway probability changes during the monitoring period to determine whether the probability of thermal runaway is increasing during the monitoring period.
[0095] Step 4: If the probability of thermal runaway is increasing, construct a thermal runaway probability prediction model, determine the thermal runaway early warning time, and feed it back to the fire early warning platform to issue an early warning.
[0096] 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 illustrative of the 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 present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.
Claims
1. A smart fire monitoring and early warning system for battery energy storage compartments, characterized in that: include: Thermal runaway test module: Conduct thermal runaway experiments on the battery storage compartment, obtain experimental operation parameters of the entire life cycle of thermal runaway triggering of the battery storage compartment, and build a thermal runaway prediction model; Risk assessment module: Real-time acquisition of actual operating parameters of the battery energy storage compartment and input into the thermal runaway prediction model, outputting the thermal runaway probability, and comprehensively assessing the thermal runaway risk of the battery energy storage compartment by combining the historical fault analysis library of the battery energy storage compartment. Trend Analysis Module: If the risk of thermal runaway is low, a preset monitoring period is set. By analyzing the trend of thermal runaway probability changes during the monitoring period, it can be determined whether the probability of thermal runaway is on the rise during the monitoring period. Thermal runaway early warning module: If the probability of thermal runaway is on the rise, a thermal runaway probability prediction model is constructed to determine the thermal runaway early warning time and feed it back to the fire early warning platform to provide early warning.
2. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 1, characterized in that: The process of constructing the thermal runaway prediction model is as follows: Based on the CNN-LSTM fusion model, the preprocessed data is integrated into a dataset in chronological order. The dataset is then divided into training, validation, and test sets in a 7:2:1 ratio. Cross-entropy is used as the loss function and Adam is used as the optimizer to construct a thermal runaway prediction model.
3. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 1, characterized in that: The process of comprehensively assessing the thermal runaway risk of the battery energy storage compartment is as follows: The historical fault analysis database of the battery energy storage compartment is analyzed to determine the failure rate of the battery energy storage compartment. The failure rate of the battery storage compartment and the probability of thermal runaway are summed to obtain the thermal runaway risk value. If the thermal runaway risk value is less than the thermal runaway risk threshold, it means that the thermal runaway risk of the battery energy storage compartment is low.
4. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 3, characterized in that: The process for determining the failure rate of the battery energy storage compartment is as follows: Extract actual operating parameter data from historical faults of the battery energy storage compartment and construct a historical fault analysis library; Based on the historical fault analysis database, the number of historical faults of the battery energy storage compartment is counted and compared with the number of historical operations of the battery energy storage compartment to obtain the failure rate of the battery energy storage compartment.
5. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 1, characterized in that: The process for determining whether the probability of thermal runaway shows an upward trend during the monitoring period is as follows: Analyze the trend of thermal runaway probability during the monitoring period to determine the probability trend value; If the probability trend value is greater than the probability trend threshold, it indicates that the probability of thermal runaway is on the rise during the monitoring period.
6. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 5, characterized in that: The process for determining the probability trend value is as follows: The actual operating parameters of the battery energy storage compartment are continuously monitored during the monitoring period. The monitoring period is divided into several analysis periods according to equal time intervals. The actual operating parameters within the analysis period are input into the thermal runaway prediction model, and the thermal runaway probability of the analysis period is output. The thermal runaway probabilities of all analysis periods are integrated into a thermal runaway probability sequence in chronological order. Analyze the thermal runaway probability sequence to determine the thermal runaway probability change ratio, the rise window number ratio, and the probability change degree ratio; The weights of the thermal runaway probability change ratio, the rise window number ratio, and the probability change degree ratio are determined by using the analytic hierarchy process (AHP). The probability trend value is obtained by weighted summation of these three ratios.
7. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 6, characterized in that: The process for determining the thermal runaway probability change ratio is as follows: The thermal runaway probability change ratio is obtained by subtracting the reversed thermal runaway probability from the first thermal runaway probability in the thermal runaway probability sequence and then comparing it with the first thermal runaway probability in the thermal runaway sequence.
8. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 6, characterized in that: The process for determining the ratio of rising windows is as follows: The thermal runaway probabilities of two adjacent analysis periods within the thermal runaway probability sequence are used as an analysis window. The difference between the thermal runaway probability of the next analysis period and the thermal runaway probability of the previous period within the analysis window is calculated to obtain the adjacent probability change value. If the adjacent probability change value is greater than 0, the corresponding analysis window is recorded as the probability increase window; Calculate the percentage of windows with increasing probability across all analysis windows, denoted as the percentage of windows with increasing probability.
9. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 8, characterized in that: The process for determining the probability change ratio is as follows: The mean probability change is obtained by summing the adjacent probability change values of all analysis windows and taking the average. The ratio of the mean probability change to the standard value of the adjacent probability change is then calculated to obtain the probability change ratio.
10. The intelligent fire monitoring and early warning system for a battery energy storage compartment according to claim 6, characterized in that: The process for determining the thermal runaway early warning time is as follows: Based on a machine learning model, using thermal runaway probability sequences as the dataset, the dataset is divided into training, validation and test sets in a 7:2:1 ratio. Mean squared error is used as the loss function and Adam is used as the optimizer to build a thermal runaway probability prediction model. A preset thermal runaway probability threshold is set, and the thermal runaway probability threshold is input into the trained thermal runaway probability prediction model. The time point when the thermal runaway probability reaches the thermal runaway probability threshold is output and recorded as the runaway time point. The difference between the runaway time point and the current time point is calculated to obtain the thermal runaway warning time.