Energy storage three-level fire-fighting response system and method based on temperature and smoke linkage

Abstract

The invention relates to the technical field of energy storage fire fighting, and discloses an energy storage three-level fire fighting response system and method based on temperature and smoke linkage, and the energy storage three-level fire fighting response method based on temperature and smoke linkage comprises the steps: collecting the operation data of an energy storage system in real time, respectively establishing and updating corresponding gas concentration base lines, temperature base lines and smoke base lines for different operation conditions; calculating a threatened gas index, temperature trend energy and a smoke abnormity index, and calculating a gas information modulation factor; calculating a comprehensive risk amount; performing three-level response judgment and executing an action; the risk attenuation rate is calculated, and self-correction is carried out; according to the method, continuous coverage of thermal runaway of the energy storage system from an early symptom to emergency disposal is realized through linkage fusion of temperature, smoke and threatened gas, risks can be identified in advance in a window period in which traditional temperature and smoke information is not obvious, a unified comprehensive risk quantity is taken as a judgment basis, missing report and false report of a single sensing quantity threshold value are reduced, and the safety of thermal runaway of the energy storage system is improved. And the timeliness and consistency of early warning are improved.

Description

Technical Field

[0001] This invention relates to the field of energy storage fire protection technology, and more specifically, to an energy storage three-level fire response system and method based on temperature and smoke linkage. Background Technology

[0002] With the development of new energy technologies, lithium-ion battery energy storage systems, with their advantages of high energy density and fast response speed, are widely used in various scenarios such as grid peak shaving and frequency regulation, and industrial and commercial energy storage. They often adopt containerized or cabinet-type structures, and are equipped with battery management systems, thermal management systems, and fire safety systems. However, energy storage systems face risks such as electrochemical side reactions and short circuits during charging and discharging, which can easily lead to thermal runaway, resulting in safety accidents such as fires and explosions.

[0003] Chinese Patent CN116758712B discloses a fire response control method and system for a lithium battery energy storage box. The method involves: collecting battery compartment layout information of the lithium battery energy storage box; reading data from the lithium battery energy storage box to obtain real-time control data; deploying a temperature acquisition device based on the battery compartment layout information; obtaining temperature data acquisition results; collecting historical detection data, which includes time-series identifiers; constructing a lithium battery characteristic dataset for the lithium battery energy storage box; constructing an anomaly monitoring model; outputting anomaly warning level information; generating multi-level fire response control data; and managing the lithium battery energy storage box.

[0004] However, in practical applications, the thermal runaway evolution of energy storage systems often exhibits phased characteristics. Especially in the early stages of thermal runaway, significant side reactions may have already occurred inside the battery cell, producing flammable and toxic gases. However, due to the widespread use of thermal management devices in current energy storage systems, and the adoption of efficient heat dissipation methods such as liquid cooling in some systems, the surface temperature of the battery cell or the ambient temperature of the battery compartment are continuously suppressed within a relatively low range. This results in a delayed manifestation of temperature anomalies, and visible smoke may not appear immediately. Existing technologies lack sufficient lead time in the early, smokeless, low-temperature but gaseous stage of thermal runaway, making it difficult to achieve early identification of risks in energy storage systems and timely activation of graded responses. Summary of the Invention

[0005] The purpose of this invention is to provide a three-level fire response system and method based on temperature and smoke linkage for energy storage in order to solve the above-mentioned problems.

[0006] This invention provides a three-level fire response method for energy storage based on temperature and smoke linkage, comprising the following steps:

[0007] Real-time acquisition of energy storage system operation data; identification of current energy storage system operation conditions based on the operation data; establishment and updating of corresponding gas concentration baselines, temperature baselines and smoke baselines for different operation conditions.

[0008] Based on the gas concentration baseline, temperature baseline and smoke baseline corresponding to the current operating conditions, calculate the precursor gas index, temperature trend energy and smoke anomaly index respectively, identify whether the energy storage system is in a state of strong intervention thermal management and calculate the gas information modulation factor.

[0009] The overall risk level is calculated by combining the precursor gas index, temperature trend energy, smoke anomaly index, and gas information modulation factor.

[0010] Based on the comparison result between the comprehensive risk amount and the preset risk threshold, a three-level response judgment is made and actions are executed. When the first-level pre-control response is triggered, power limiting and cooling capacity enhancement actions are executed. When the second-level suppression preparation response is triggered, electrical isolation and fire protection system standby actions are executed. When the third-level emergency backup response is triggered, fire discharge and overall safety actions are executed.

[0011] After the action is executed, an evaluation time window is entered to calculate the risk decay rate. When the risk decay rate is less than the preset minimum effective decay rate within a preset number of consecutive evaluation time windows for correction judgments, the response level is increased. When the risk decay rate is greater than or equal to the preset rapid recovery decay rate and the total risk is less than the preset first risk threshold, the response level is reduced until normal monitoring is restored.

[0012] Furthermore, establishing and updating corresponding gas concentration baselines, temperature baselines, and smoke baselines for different operating conditions includes:

[0013] The operating data includes charging and discharging power, temperature at each monitoring point and its rate of temperature change, state of charge, ambient temperature, and operating parameters of the cooling system;

[0014] Operating conditions include standby mode, low power charging mode, low power discharging mode, high power charging mode, high power discharging mode, forced cooling mode, and equalization maintenance mode.

[0015] Identify steady-state time segments that meet low fluctuation conditions, and establish or update the sensing baseline for the current operating condition within the steady-state time segment. The sensing baseline includes gas concentration baseline, temperature baseline and smoke baseline. The gas concentration baseline includes the gas concentration baseline mean and gas concentration baseline standard deviation, the temperature baseline includes the temperature baseline mean and temperature baseline standard deviation, and the smoke baseline includes the smoke baseline mean and smoke baseline standard deviation.

[0016] When the operating conditions of the energy storage system change, the current operating condition is identified and the corresponding sensing baseline is called.

[0017] The sensor drift is monitored in real time. When the long-term mean deviation of the sensing quantity is detected to exceed the preset drift threshold, the sensing quantity baseline is slowly updated using an exponentially weighted moving average method.

[0018] Furthermore, the mean gas concentration baseline is the arithmetic mean of the gas concentration values ​​at all sampling points within the steady-state time segment, and the standard deviation of the gas concentration baseline is a measure of the dispersion of the gas concentration values ​​at each sampling point relative to the mean gas concentration baseline within the steady-state time segment.

[0019] The mean temperature baseline is the arithmetic mean of the temperature values ​​at all sampling times of the monitoring point within the steady-state time segment, and the standard deviation of the temperature baseline is a measure of the dispersion of the temperature values ​​at each sampling time of the monitoring point relative to the mean temperature baseline within the steady-state time segment.

[0020] The smoke baseline mean is the arithmetic mean of the smoke concentration values ​​at all sampling points within the steady-state time segment, and the smoke baseline standard deviation is a measure of the dispersion of the smoke concentration values ​​at each sampling point relative to the smoke baseline mean within the steady-state time segment.

[0021] The low fluctuation condition is as follows: within a time segment, the fluctuation amplitude of the charging and discharging power is less than a preset power fluctuation threshold, and the maximum absolute value of the temperature change rate at all monitoring points is less than a preset temperature change rate threshold, and the change in the state of charge of the energy storage system is less than a preset state of charge change threshold, and the change in ambient temperature is less than a preset ambient temperature change threshold; the duration of the steady-state time segment is not less than a preset minimum steady-state duration.

[0022] Furthermore, the precursor gas index, temperature trend energy, and smoke anomaly index are calculated separately to identify whether the energy storage system is under strong intervention thermal management and to calculate the gas information modulation factor, including:

[0023] Based on the gas concentration baseline corresponding to the current operating conditions, calculate the gas anomaly score and the standardized rise rate of gas concentration, and construct the precursor gas index. The precursor gas index is the sum of the limited anomaly score term and the limited gas concentration term. The limited anomaly score term is the gas anomaly score after the limited treatment multiplied by the first preset weight coefficient, and the limited gas concentration term is the standardized rise rate of gas concentration after the limited treatment multiplied by the second preset weight coefficient.

[0024] Based on the temperature baseline corresponding to the current operating conditions, calculate the temperature deviation score and temperature change rate score of each monitoring point. Perform exponential moving average processing on the temperature deviation score and temperature change rate score of each monitoring point respectively. Take the maximum value of the temperature deviation score and the maximum value of the temperature change rate score of all monitoring points after exponential moving average processing, and then add them together after limiting processing to obtain the temperature trend energy.

[0025] If the energy storage system is equipped with a smoke detector, the current smoke concentration is subtracted from the mean smoke baseline, and then divided by the effective smoke baseline standard deviation. The resulting quotient is then subjected to amplitude limiting to obtain the smoke anomaly index. The effective smoke baseline standard deviation is the larger of the smoke baseline standard deviation and the preset lower limit of the smoke standard deviation.

[0026] Determine whether the energy storage system is under strong intervention thermal management and whether there is smoke anomaly. Calculate the gas information modulation factor based on the determination results. When the system is under strong intervention thermal management and there is no smoke anomaly, the gas information modulation factor is set to 1. When the system is not under strong intervention thermal management or there is smoke anomaly, the gas information modulation factor is set to the gas weight weakening coefficient under the preset atypical scenario.

[0027] Furthermore, the calculation method for the gas anomaly score is as follows: subtract the mean gas concentration baseline from the current gas concentration, and then divide by the effective gas concentration baseline standard deviation, wherein the effective gas concentration baseline standard deviation is the larger value between the gas concentration baseline standard deviation and the preset lower limit of the gas standard deviation.

[0028] The method for calculating the standardized rate of rise of gas concentration is as follows: First, calculate the gas concentration multi-point regression slope between the current time and the previous several sampling times to obtain the gas concentration change rate. Then, take the larger value between the gas concentration change rate and zero to obtain the positive change rate of gas concentration. Finally, divide the positive change rate of gas concentration by the standard deviation of the effective gas concentration baseline to obtain the standardized rate of rise of gas concentration.

[0029] The temperature deviation score is calculated as follows: subtract the mean temperature baseline of the monitoring point from the temperature of the current monitoring point, then divide by the effective temperature baseline standard deviation to obtain the effective quotient, and finally take the larger of the effective quotient and zero as the temperature deviation score. The effective temperature baseline standard deviation is the larger of the temperature baseline standard deviation of the monitoring point and the preset lower limit of the temperature standard deviation.

[0030] The method for calculating the temperature change rate fraction is as follows: First, calculate the temperature change rate at the monitoring point, then take the larger value between the temperature change rate and zero to obtain the positive temperature change rate, and finally divide the positive temperature change rate by the standard deviation of the effective temperature baseline to obtain the temperature change rate fraction.

[0031] Furthermore, the criteria for identifying a strong intervention thermal management state include at least one of the following: the ratio of the actual operating power of the cooling system to the rated power is greater than or equal to the preset strong intervention judgment ratio of cooling power; the ratio of the actual flow rate of the coolant to the rated flow rate is greater than or equal to the preset strong intervention judgment ratio of cooling flow rate; the temperature difference between the inlet and outlet of the cooling system is greater than or equal to the preset strong intervention temperature difference judgment threshold; or the system has executed an emergency power reduction command or the forced ventilation system has been started.

[0032] The criteria for identifying abnormal smoke conditions is that the smoke abnormality index is greater than the preset smoke abnormality judgment threshold.

[0033] Furthermore, the comprehensive risk quantity is calculated as the sum of the gas item, temperature item, and smoke item. The gas item is the product of the gas information modulation factor, the precursor gas index, and the preset gas information fusion weight coefficient. The temperature item is the temperature trend energy multiplied by the preset temperature information fusion weight coefficient. The smoke item is the smoke anomaly index multiplied by the preset smoke information fusion weight coefficient.

[0034] Furthermore, risk thresholds are preset, including a first preset risk threshold, a second preset risk threshold, and a third preset risk threshold, such that the first preset risk threshold is less than the second preset risk threshold, the second preset risk threshold is less than the third preset risk threshold, and all risk thresholds are greater than zero.

[0035] The trigger condition for Level 1 pre-control is that the Level 1 pre-control response is triggered when the total risk level is greater than or equal to the preset first risk threshold.

[0036] The secondary suppression preparation is triggered when the overall risk level is greater than or equal to the preset second risk threshold.

[0037] The Level 3 emergency backup trigger condition is that the overall risk level is greater than or equal to the preset third risk threshold, or the temperature at any monitoring point is greater than or equal to the preset temperature direct trigger threshold, or the smoke concentration is greater than or equal to the preset smoke direct trigger threshold.

[0038] Furthermore, when the Level 1 pre-control response is triggered, the charging and discharging power of the energy storage system is limited or the high-voltage circuit is cut off, the cooling system capacity is increased to the safe limit but the fire sprinkler system is not activated, the data acquisition frequency is increased and suspected abnormal storage points are identified.

[0039] When the Level 2 suppression preparation response is triggered, electrical isolation is performed on suspected abnormal storage points, the fire protection system is controlled to enter the standby state, a higher level of audible and visual alarm is triggered, and the trend of changes in the comprehensive risk quantity, the growth rate of the precursor gas index, and the degree of abnormality of temperature and smoke are continuously monitored in the observation and decision window. If the comprehensive risk quantity in the observation and decision window continues to rise or remains at a high level, the Level 2 response is maintained and preparations are made to upgrade to the Level 3 response. If the comprehensive risk quantity in the observation and decision window drops rapidly and there is no abnormal increase in temperature and smoke, the response is automatically downgraded to the Level 1 response or normal monitoring is restored.

[0040] When a Level 3 emergency backup response is triggered, fire sprinkler or extinguishing medium injection is initiated according to the preset strategy, and pump pressure is controlled. Simultaneously, global safety actions are executed, including power outage isolation, ventilation linkage, and pressure relief valve opening. Complete event data is recorded for post-event review and analysis.

[0041] This invention provides a three-level fire response system based on temperature and smoke linkage, used to store computer-readable instructions, and to execute the aforementioned three-level fire response method based on temperature and smoke linkage when the computer-readable instructions are read; the system includes:

[0042] The baseline management module collects real-time operating data of the energy storage system, identifies the current operating conditions of the energy storage system based on the operating data, and establishes and updates corresponding gas concentration baselines, temperature baselines, and smoke baselines for different operating conditions.

[0043] The risk assessment module calculates the precursor gas index, temperature trend energy, and smoke anomaly index based on the gas concentration baseline, temperature baseline, and smoke baseline corresponding to the current operating conditions, respectively, to identify whether the energy storage system is in a state of strong intervention thermal management and to calculate the gas information modulation factor.

[0044] The risk calculation module integrates the precursor gas index, temperature trend energy, smoke anomaly index, and gas information modulation factor to calculate the overall risk level.

[0045] The graded response module determines the three-level response based on the comparison result between the comprehensive risk amount and the preset risk threshold and executes the action. When the first-level pre-control response is triggered, the power limiting and cooling capacity enhancement actions are executed. When the second-level suppression preparation response is triggered, the electrical isolation and fire protection system standby actions are executed. When the third-level emergency backup response is triggered, the fire spraying and global safety actions are executed.

[0046] The closed-loop verification module enters an evaluation time window after the action is executed, calculates the risk decay rate, and increases the response level when the risk decay rate is less than the preset minimum effective decay rate within a number of consecutive preset correction judgments within the evaluation time window; and decreases the response level until normal monitoring is restored when the risk decay rate is greater than or equal to the preset fast recovery decay rate and the total risk is less than the preset first risk threshold.

[0047] The beneficial effects of this invention are as follows: By integrating temperature, smoke, and precursor gases, this invention achieves continuous coverage of thermal runaway in energy storage systems from early signs to emergency response. It can identify risks in advance during the window period when traditional temperature and smoke information is not obvious, and uses a unified comprehensive risk quantity as the judgment basis, reducing false alarms and missed alarms caused by single sensor thresholds, thus improving the timeliness and consistency of early warning. Addressing the issues of frequent changes in energy storage system operating conditions and easy drift of sensor baselines, this invention introduces a hierarchical baseline library for operating conditions, establishing and updating gas concentration baselines, temperature baselines, and smoke baselines for corresponding operating conditions. Combined with the lower limit of standard deviation and smoothing processing, sensor data of different dimensions and noise levels are comparable on the same scale, avoiding misjudgments caused by excessively small baselines amplifying anomaly scores, thereby improving the stability and calibrability of anomaly identification. By using a strong intervention thermal management state recognition and gas information modulation factor mechanism, the weight of precursor gas information is strengthened in the case of strong intervention thermal management state and no smoke anomaly, which solves the problem of temperature anomaly lag caused by strong cooling and other interventions. In the case of non-strong intervention thermal management state or smoke anomaly, the weight of precursor gas information is weakened, and a smooth transition is adopted to suppress switching jitter, so as to achieve adaptive fusion of information reliability under different scenarios and reduce the risk of false alarm caused by environmental interference from the source.

[0048] This invention achieves a three-level fire response linkage by integrating risk quantity and risk threshold. It can sequentially trigger pre-control, suppression preparation, and emergency backup actions according to the risk development stage, forming an orderly collaboration between the battery management system, thermal management system, and fire control system. At the same time, it sets an evaluation time window to calculate the risk decay rate and perform self-correction, enabling the response strategy to have closed-loop optimization capabilities. While ensuring rapid handling of real risks, it can also mark and suppress suspected non-fire gas events, reduce alarm fatigue, and improve the overall safety and engineering practicality of the system. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the three-level fire response method for energy storage based on temperature and smoke linkage of the present invention.

[0050] Figure 2 This is an example diagram illustrating the calculation of the gas information modulation factor in the three-level fire response method for energy storage based on temperature and smoke linkage of the present invention.

[0051] Figure 3 This is an example diagram illustrating the three-level response determination of the energy storage three-level fire response method based on temperature and smoke linkage of the present invention;

[0052] Figure 4 This is a module example diagram of the three-level fire response system based on temperature and smoke linkage of the present invention. Detailed Implementation

[0053] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0054] A three-level fire response system and method based on temperature and smoke linkage for energy storage includes the following embodiments:

[0055] Example 1:

[0056] The three-level fire response method for energy storage based on temperature and smoke linkage is applicable to containerized or cabinet-type energy storage systems. The battery compartments or battery modules of these systems are equipped with temperature sensors, smoke detectors, and gas sensing units, and have a data interface for the battery management system. The gas sensing units detect at least one gas selected from volatile organic compounds (VOCs), carbon monoxide, hydrogen, or hydrogen fluoride. Preferably, VOC and carbon monoxide sensors are configured as the primary precursor gas detection methods, with a hydrogen sensor as an auxiliary detection method. A hydrogen fluoride sensor is optional. When the energy storage system uses batteries with fluorinated electrolyte and has extremely high requirements for early warning timeliness, a hydrogen fluoride sensor can be selected. However, it should be noted that the hydrogen fluoride sensor has high requirements for humidity control and corrosion resistance in the installation environment; appropriate sensor protection measures should be taken, and regular calibration and maintenance should be performed.

[0057] The installation location of the gas sensing unit should comprehensively consider the airflow organization characteristics, leakage paths, and the influence of the ventilation and exhaust system of the battery compartment. It is preferably installed above the battery modules or at the inlet of the exhaust channel to ensure that abnormal gas can reach the sensor detection area in a timely manner. Effective airflow communication should be maintained between the gas sensing unit and the furthest battery module to avoid gas failing to reach the sensor due to airflow dead zones or interruptions. During the design and installation phases, the rationality of the sensor placement should be verified through airflow simulation or tracer gas testing to ensure that the gas can be detected by the sensor within an acceptable time window when abnormal exhaust occurs from the battery module. The design target for this time window is that the gas concentration should reach more than twice the sensor's lower detection limit within 3 minutes after the abnormal exhaust occurs. This design target can be adjusted according to the ventilation conditions, airflow paths, and sensor placement of the battery compartment. For scenarios with strong ventilation or complex airflow paths, it can be appropriately extended to 5 minutes; for scenarios with good sealing and reasonable sensor placement, it can be shortened to 2 minutes. The temperature sensor should be placed on the surface of the battery module or at the connection point of the battery cluster, or other areas with concentrated heat, to ensure timely detection of temperature anomalies. The smoke detector should be placed on top of the battery compartment or in the exhaust duct to effectively detect smoke by utilizing its rising characteristics.

[0058] The energy storage system is equipped with a battery management system, a thermal management system, and a fire control system. These systems exchange data and perform coordinated control via communication interfaces. The communication interfaces utilize CAN bus, Ethernet, or other industrial communication protocols to ensure real-time and reliable data transmission. The coordinated control includes the battery management system sending power limiting commands to the thermal management system, the thermal management system sending abnormal status signals to the fire control system, and the fire control system sending isolation commands to the battery management system. The control priorities among the systems are determined according to the principle of safety priority, with emergency commands from the fire control system having the highest priority.

[0059] A three-level fire response method based on temperature and smoke linkage for energy storage, such as... Figure 1 As shown, it includes the following steps:

[0060] Step 100: Collect real-time operating data of the energy storage system, identify the current operating conditions of the energy storage system based on the operating data, and establish and update the corresponding gas concentration baseline, temperature baseline and smoke baseline for different operating conditions.

[0061] During the operation of the energy storage system, real-time operational data is collected. This data includes charging and discharging power, temperature and its rate of change at each monitoring point, state of charge (SBC), ambient temperature, and cooling system operating parameters. The monitoring point index distinguishes different temperature acquisition locations, and the sampling time indicates the current data acquisition time. These operational data are chosen as input because charging and discharging power directly reflects the battery's workload and heat generation intensity; temperature and its rate of change are the most direct indicators of thermal runaway; SBC affects the battery's thermal stability and chemical reactivity; ambient temperature determines the heat dissipation boundary conditions; and cooling system operating parameters reflect the degree of thermal management intervention. These parameters collectively constitute a complete description of the energy storage system's thermal safety status, providing a comprehensive data foundation for subsequent baseline establishment and anomaly detection.

[0062] First, a hierarchical baseline library for operating conditions is established. Based on the operating characteristics of the energy storage system, operating conditions are divided into several typical states, including standby, low-power charging, low-power discharging, high-power charging, high-power discharging, forced cooling, and equalization maintenance. This hierarchical baseline library approach is adopted because the temperature distribution, background gas concentration, and smoke levels of the energy storage system differ significantly under different operating conditions. For example, during high-power charging and discharging, the battery temperature and gas emissions are significantly higher than in standby. Using a single baseline would lead to normal operation under high-power conditions being misjudged as abnormal, or early anomalies under low-power conditions being masked by the baseline. By establishing independent baselines for each operating condition, changes in operating condition characteristics and genuine safety anomalies can be accurately distinguished, significantly improving the accuracy and sensitivity of anomaly detection. Each operating condition corresponds to an independent set of sensing baselines, including the gas concentration baseline, temperature baseline, and smoke baseline for that condition. The identification of operating conditions is based on multi-dimensional information including the magnitude and direction of charging and discharging power, the trend of state of charge change, the operating intensity of the cooling system, and system control commands. Once a certain operating condition is entered and the system has been running stably under that condition for a period of time, the corresponding baseline data for that operating condition is established or updated.

[0063] For each operating condition, a steady-state time segment that meets the low-fluctuation condition is identified. The low-fluctuation condition is defined as follows: within the time segment, the fluctuation amplitude of the charging and discharging power is less than a preset power fluctuation threshold, the maximum absolute value of the temperature change rate at all monitoring points is less than a preset temperature change rate threshold, the change in the state of charge of the energy storage system is less than a preset state of charge change threshold, and the change in ambient temperature is less than a preset ambient temperature change threshold. The purpose of setting the low-fluctuation condition is to ensure that the data used when establishing the baseline comes from the system's stable operating state, avoiding the inclusion of transient processes, abnormal fluctuations, or external interference in the baseline calculation. This is because only sensor data collected under steady-state conditions can truly reflect the normal characteristics of the operating condition. Establishing a baseline during non-steady-state periods will result in the baseline containing too much noise and transient components, reducing the reliability of subsequent anomaly detection. The default value of the preset power fluctuation threshold is 10% of the typical power of the current operating condition. This threshold is determined according to the power fluctuation characteristics of different operating conditions. For standby and low-power operating conditions, it can be reduced to 5% of the typical power, and for high-power and regulating operating conditions, it can be relaxed to 15% of the typical power. The default value for the preset temperature change rate threshold is 0.5 degrees Celsius per minute. This threshold is determined based on the thermal management capability of the energy storage system and the characteristics of ambient temperature changes. For systems with strong heat dissipation capabilities, it can be appropriately increased to 1 degree Celsius per minute. The default value for the preset state of charge change threshold is 5%. This threshold is used to exclude unsteady-state conditions caused by rapid changes in the state of charge during charging and discharging. The default value for the preset ambient temperature change threshold is 3 degrees Celsius. This threshold is used to exclude the influence of drastic changes in ambient temperature on the sensor baseline.

[0064] The duration of the steady-state time segment should not be less than the preset minimum steady-state duration. The default value of the preset minimum steady-state duration is 20 minutes. This duration ensures that a sufficient number of sample points are collected to calculate a reliable statistical baseline. For systems with a high data sampling frequency, it can be appropriately shortened to 15 minutes, and for systems with a low data sampling frequency, it should be extended to more than 30 minutes.

[0065] Within the identified steady-state time segment, a sensing baseline is established or updated for the current operating condition. The sensing baselines include gas concentration baseline, temperature baseline, and smoke baseline. The gas concentration baseline includes the mean and standard deviation of the gas concentration baseline; the temperature baseline includes the mean and standard deviation of the temperature baseline; and the smoke baseline includes the mean and standard deviation of the smoke baseline. The mean and standard deviation are used as the statistical characteristics of the baselines because the mean represents the central trend of the sensing quantity under this operating condition, i.e., the normal level, while the standard deviation represents the inherent fluctuation range of the sensing quantity under this operating condition, i.e., the noise level. These two statistics can fully describe the characteristics of normally distributed or approximately normally distributed data, providing a reference benchmark for subsequent standardized anomaly detection, ensuring that anomaly judgment considers both the absolute degree of deviation and the significance relative to normal fluctuations. The gas concentration baseline is established by calculating the mean and standard deviation of the gas concentration within a steady-state time segment. The mean of the gas concentration baseline is the arithmetic mean of the gas concentration values ​​at all sampling points within the steady-state time segment. The standard deviation of the gas concentration baseline is a measure of the dispersion of the gas concentration values ​​at each sampling point relative to the mean of the gas concentration baseline within the steady-state time segment. Specifically, it is calculated by summing the squares of the differences between the gas concentration values ​​at each sampling point and the mean of the gas concentration baseline, dividing by the number of sampling points, and then taking the square root. The temperature baseline is established by calculating the mean and standard deviation of the temperature at each monitoring point within a steady-state time segment. The mean of the temperature baseline is the arithmetic mean of the temperature values ​​at each sampling time at that monitoring point within the steady-state time segment. The standard deviation of the temperature baseline is a measure of the dispersion of the temperature values ​​at each sampling time at that monitoring point relative to the mean of the temperature baseline within the steady-state time segment. Specifically, it is calculated by summing the squares of the differences between the temperature values ​​at each sampling time and the mean of the temperature baseline, dividing by the number of sampling points, and then taking the square root. The smoke baseline is established by calculating the mean and standard deviation of the smoke concentration within a steady-state time segment if a smoke detector is configured. The mean of the smoke baseline is the arithmetic mean of the smoke concentration values ​​at all sampling points within the steady-state time segment, and the standard deviation of the smoke baseline is a measure of the dispersion of the smoke concentration values ​​at each sampling point relative to the mean of the smoke baseline. Specifically, it is calculated by summing the squares of the differences between the smoke concentration values ​​at each sampling point and the mean of the smoke baseline, dividing by the number of sampling points, and then taking the square root. Simultaneously, the normal fluctuation range of the sensor quantity, noise level characteristics, and short-term drift trend under this operating condition are recorded as additional attributes of the baseline for this operating condition.

[0066] When the operating conditions of the energy storage system change, the system automatically identifies the current operating condition and uses the corresponding baseline data for subsequent anomaly detection. If a corresponding baseline exists in the baseline library for the current operating condition, it is used directly, and fine-tuning updates are allowed during the duration of the current operating condition. If no corresponding baseline exists in the baseline library for the current operating condition, or if the current operating condition is a frequently fluctuating transitional condition, a temporary baseline protection mode is entered. In this mode, the baseline of the most recent stable operating condition is used as a reference, but the update speed of the baseline is strictly limited, allowing only a slow change in the baseline to follow the current data, preventing the absorption of abnormal data as normal baselines.

[0067] During the continuous operation of the baseline under each operating condition, sensor drift is monitored in real time. If the long-term mean deviation of a certain sensor quantity under the current operating condition exceeds a preset drift threshold, the baseline for that operating condition is slowly updated using an exponentially weighted moving average method. The exponentially weighted moving average method is used for baseline updates because it assigns different weights to historical and current data, achieving slow baseline tracking with a smaller update coefficient. This method adapts to long-term sensor drift and slow environmental changes while effectively suppressing the contamination of the baseline by short-term abnormal events, preventing the misinterpretation of early signs of thermal runaway as normal baseline. Compared to the sliding window averaging method, the exponentially weighted moving average method has better computational efficiency and smoother update characteristics. The calculation period for the long-term mean is a preset drift monitoring period, with a default value of 24 hours. This period is determined based on the sensor's drift characteristics and the changing cycle of environmental factors. For sensors with faster drift speeds, it can be shortened to 12 hours, and for sensors with slower drift speeds, it can be extended to 48 hours. The default value of the preset drift threshold is 3 times the initial baseline standard deviation for this operating condition. This threshold is determined based on the long-term stability of the sensor and the degree of influence of environmental factors. For high-precision sensors, it can be reduced to 2 times the initial baseline standard deviation, and for sensors that are susceptible to environmental interference, it can be increased to 4 times the initial baseline standard deviation.

[0068] When a long-term mean deviation of the sensed quantity is detected to exceed a preset drift threshold, an exponentially weighted moving average method is used to update the baseline. The update process involves weighting the current long-term mean and the old baseline mean using a preset baseline update coefficient to obtain a new baseline mean, and weighting the current long-term standard deviation and the old baseline standard deviation using the same preset baseline update coefficient to obtain a new baseline standard deviation. In the new baseline mean, the weight of the current long-term mean is the preset baseline update coefficient, and the weight of the old baseline mean is 1 minus the preset baseline update coefficient. Similarly, in the new baseline standard deviation, the weight of the current long-term standard deviation is the preset baseline update coefficient, and the weight of the old baseline standard deviation is 1 minus the preset baseline update coefficient. The default value of the preset baseline update coefficient is 0.05. This coefficient is much less than 1 to ensure a slow baseline update speed and avoid misjudging short-term anomalies as normal baselines. For scenarios requiring faster adaptation to environmental changes, it can be increased to 0.1; for scenarios requiring stronger anti-interference capabilities, it can be decreased to 0.02.

[0069] To further distinguish between sensor drift and changes in the actual environment, the changing trends of other relevant sensor quantities are monitored simultaneously during baseline updates. If multiple sensor quantities show long-term shifts in the same direction, it is more likely due to environmental changes or changes in operating characteristics rather than sensor drift. In this case, the baseline update speed can be appropriately accelerated. If only a few sensor quantities show shifts, it is more likely to be sensor drift. In this case, updates should be slowed down and sensor calibration prompts should be triggered. By establishing a hierarchical baseline library for operating conditions, it is possible to adapt to the diverse operating modes of energy storage power stations, avoiding baseline establishment delays or failures caused by the inability to identify a single steady-state window, and improving the stability and availability of the baseline.

[0070] Step 200: Based on the current operating conditions' corresponding gas concentration baseline, temperature baseline, and smoke baseline, calculate the precursor gas index, temperature trend energy, and smoke anomaly index respectively to identify whether the energy storage system is in a state of strong intervention thermal management and calculate the gas information modulation factor, specifically as follows: Figure 2 As shown.

[0071] In this embodiment, the strong intervention thermal management state refers to a high-intensity thermal management intervention state implemented by the energy storage system to suppress abnormal heat generation or temperature rise. Under this state, at least one of the following in the thermal management system—cooling system operating power, coolant flow rate, inlet and outlet temperature difference, emergency power reduction control, or forced ventilation control—reaches a preset strong intervention judgment condition, which may cause the battery surface temperature characterization to lag behind the abnormal heat generation state inside the battery.

[0072] Based on the gas concentration baseline established in step 100 corresponding to the current operating condition, the gas anomaly score at the current moment is calculated. The gas anomaly score is calculated as follows: subtract the mean of the gas concentration baseline from the current gas concentration, and then divide by the standard deviation of the effective gas concentration baseline. The quotient is the gas anomaly score. A standardized method is used to calculate the gas anomaly score because different types of gas sensors have vastly different output units and ranges. For example, the output of a volatile organic compound sensor may be at the level of hundreds of ppm, while the output of a carbon monoxide sensor may be at the level of tens of ppm. Directly using the absolute concentration value cannot provide a unified threshold for judgment. By dividing by the baseline standard deviation for standardization, the outputs of different sensors are mapped to a unified dimensionless space, allowing the anomaly score to directly reflect the multiple of deviation from the normal fluctuation level. This facilitates setting a unified judgment threshold and performing multi-sensor information fusion. The effective gas concentration baseline standard deviation is the larger of the gas concentration baseline standard deviation and the preset lower limit of the gas standard deviation. The preset lower limit is determined based on the actual noise level and measurement accuracy of the gas sensor. It is obtained by statistically analyzing the sensor's long-term output under stable conditions to obtain the standard deviation of the sensor's inherent noise, which serves as the basis for the preset lower limit. The default value of this lower limit is twice the sensor's noise standard deviation under typical operating conditions. For high-precision sensors, this can be reduced to 1.5 times the noise standard deviation, and for low-precision sensors, it can be increased to 3 times the noise standard deviation. By setting the effective gas concentration baseline standard deviation, the unreasonable amplification of gas anomaly scores is avoided when the baseline standard deviation is too small. This ensures that the gas anomaly scores accurately reflect the degree of anomaly relative to the sensor's inherent fluctuation level. This processing is particularly important under extremely stable operating conditions or when the sensor's accuracy is extremely high, causing the baseline standard deviation to approach zero, preventing minor concentration fluctuations from being misjudged as serious anomalies. This gas anomaly score reflects the degree of deviation of the current gas concentration from the statistical baseline during normal operation. Standardization processing makes gas concentration data of different dimensions and magnitudes comparable.

[0073] The standardized rate of increase in gas concentration is calculated. First, the multi-point regression slope of the gas concentration between the current time and several previous sampling times is calculated. Then, the least squares method is used to linearly fit the gas concentration values ​​of the most recent sampling points within a preset rate calculation window, yielding the slope of the gas concentration change over time. This slope is the rate of change of gas concentration. The least squares method, using multi-point linear fitting instead of simple two-point difference, is used to calculate the rate of change because sensor output contains measurement noise and random fluctuations. The two-point difference method is extremely sensitive to noise and prone to severe jitter, while the least squares method, through linear regression on multiple sampling points, effectively smooths out the noise effect, obtaining a more stable and reliable estimate of the changing trend. This is crucial for accurately capturing the rising trend of gas concentration rather than instantaneous fluctuations. The default value for the preset rate calculation window length is 5 sampling points. This window length is determined based on the sampling frequency and the desired smoothness of the rate. For systems with higher sampling frequencies, this can be increased to 7 sampling points to enhance noise immunity; for systems with lower sampling frequencies, it can be reduced to 3 sampling points to maintain response speed. Then, the larger of the gas concentration change rate and zero is taken to obtain the positive gas concentration change rate. Finally, the positive gas concentration change rate is divided by the standard deviation of the effective gas concentration baseline to obtain the standardized rise rate of gas concentration. This standardization process unifies the outputs of gas sensors with different dimensions and ranges into a dimensionless standardized space, making subsequent calculations transferable and comparable. Regardless of whether the gas sensor output unit is ppm, mg / m³, %LEL, or other units, the standardized rise rate has the same physical meaning, that is, the change factor relative to the normal fluctuation level of the sensor. This process ensures that only the case of gas concentration increase is considered, ignoring the case of gas concentration decrease.

[0074] A precursor gas index is constructed. The precursor gas index consists of a limited anomaly score term plus a limited gas concentration term. The limited anomaly score term is the gas anomaly score after limiting processing multiplied by a first preset weighting coefficient, and the limited gas concentration term is the gas concentration standardized rise rate after limiting processing multiplied by a second preset weighting coefficient. The precursor gas index is constructed by combining the gas anomaly score and the gas concentration standardized rise rate because these two quantities reflect gas anomaly characteristics from different dimensions. The gas anomaly score reflects the static deviation of the current concentration from the normal level, enabling the detection of absolute concentration anomalies. The gas concentration standardized rise rate reflects the dynamic trend of concentration changes over time, enabling the capture of rapid concentration increases. Early thermal runaway may manifest as a slow rise in concentration without significant deviation from the baseline, or it may manifest as a significantly elevated concentration with a slowing rate of increase. Using either indicator alone may miss certain anomaly patterns. The weighted combination of the two can simultaneously consider both static anomalies and dynamic trends, improving the comprehensiveness and robustness of early warning. The amplitude limiting process restricts the input value to between zero and a preset upper limit. If the input value is less than zero, it is set to zero; if the input value is greater than the preset upper limit, it is set to the preset upper limit; if the input value is between zero and the preset upper limit, it remains unchanged. The default value for the preset upper limit of the gas anomaly fraction is 5. This upper limit is determined based on the sensitivity of the gas sensor and the maximum fluctuation amplitude under normal operating conditions. For high-sensitivity sensors, it can be reduced to 3, and for low-sensitivity sensors, it can be increased to 7. The default value for the preset upper limit of the gas concentration normalized rise rate is 5. This upper limit indicates that when the gas concentration rise rate reaches 5 times the baseline standard deviation, it is considered to have entered the high-risk zone. This value is based on experimental observations of the thermal runaway process of lithium iron phosphate batteries and ternary lithium batteries. The gas concentration normalized rise rate is usually 3 to 7 times the baseline standard deviation during the exhaust phase. Taking 5 as the default upper limit is suitable for most battery types. For scenarios with rapid gas release, it can be increased to 7, and for scenarios with slow gas release, it can be reduced to 3.

[0075] The default value of the first preset weighting coefficient is 0.6, and the default value of the second preset weighting coefficient is 0.4. These two preset weighting coefficients are used to balance the contributions of the gas anomaly score and the standardized rate of increase of gas concentration to the precursor gas index. Their values ​​ensure that the precursor gas index maintains a high detection rate while keeping the false alarm rate within an acceptable range. Specifically, the gas anomaly score reflects the degree of concentration deviation, and the standardized rate of increase of gas concentration reflects the trend of change. The combination of the two can improve the accuracy of early warning. The ratio of weighting coefficients 0.6 and 0.4 makes the contributions of the two parts to the precursor gas index in a typical thermal runaway event basically equal.

[0076] Based on the temperature baseline and smoke baseline established in step 100 corresponding to the current operating condition, the temperature deviation score and temperature change rate score of each monitoring point are first calculated. The temperature deviation score is calculated as follows: subtract the mean temperature baseline of the monitoring point from the temperature of the current monitoring point, then divide by the effective temperature baseline standard deviation, and finally take the larger value between the quotient and zero. For temperatures below the temperature baseline, the value is taken as zero. The effective temperature baseline standard deviation is the larger value between the standard deviation of the temperature baseline of the monitoring point and the preset lower limit of the temperature standard deviation. The preset lower limit of the temperature standard deviation is determined based on the actual noise level and measurement accuracy of the temperature sensor. The standard deviation of the sensor's inherent noise is obtained through statistical analysis of the long-term output of the sensor under stable conditions, and is used as the basis for determining the value of the preset lower limit. The default value of this lower limit is 0.2 degrees Celsius. For high-precision temperature sensors, it can be reduced to 0.1 degrees Celsius, and for low-precision temperature sensors, it can be increased to 0.5 degrees Celsius.

[0077] The calculation method for the temperature change rate score is as follows: First, calculate the temperature change rate at the monitoring point. Then, use the least squares method to linearly fit the temperature values ​​of the most recent preset rate calculation window lengths to obtain the slope of temperature change over time. Next, take the larger value between this temperature change rate and zero to obtain the positive temperature change rate. Finally, divide the positive temperature change rate by the effective temperature baseline standard deviation to obtain the temperature change rate score. Then, perform exponential moving average processing on the temperature deviation score and temperature change rate score for each monitoring point. The exponential moving average processing method is to multiply the current input value by a preset smoothing coefficient, add the exponential moving average result from the previous time step multiplied by 1 minus the preset smoothing coefficient, and sum the results. Exponential moving average processing is used on the temperature deviation score and temperature change rate score because both the temperature sensor output and the temperature change rate calculation results contain certain noise and short-term fluctuations. Directly using instantaneous values ​​can easily lead to drastic fluctuations in temperature trend energy. Exponential moving average processing can smooth these short-term fluctuations, extract the persistent trend characteristics of temperature anomalies, avoid false alarms triggered by instantaneous temperature spikes or measurement noise, and maintain the ability to respond to the true upward trend of temperature. The default value of the preset smoothing coefficient is 0.3. This coefficient is a value greater than zero and less than 1. Its value is determined according to the time scale of temperature change and the desired smoothness. The smaller the smoothing coefficient, the longer the memory time of historical data and the stronger the smoothing effect. The larger the smoothing coefficient, the faster the response speed to current data. Choosing 0.3 as the default smoothing coefficient can ensure timely response to the temperature rise trend while maintaining an appropriate smoothing effect. For scenarios with rapid temperature changes, it can be increased to 0.5 to enhance the response speed. For scenarios with slow temperature changes, it can be reduced to 0.1 to enhance the smoothing effect.

[0078] Finally, the maximum value of the temperature deviation score and the maximum value of the temperature change rate score after exponential moving average processing of all monitoring points are taken, and both are summed after limiting processing to obtain the temperature trend energy. The maximum value among all monitoring points is chosen because energy storage systems typically have multiple temperature monitoring points distributed across different battery modules or locations. Thermal runaway often occurs first at a localized location; taking the maximum value captures temperature anomalies at the most dangerous locations, avoiding the masking of local hotspots by averaging multiple points. This maximum value strategy ensures sensitivity to any single-point anomaly. The upper limit of the limiting value of the temperature deviation score is a preset upper limit of temperature deviation, and the upper limit of the limiting value of the temperature change rate score is a preset upper limit of temperature change. The default values ​​of the preset upper limits of temperature deviation and temperature change are both 5. These two upper limits ensure that the value range of the temperature trend energy is on the same numerical scale as the precursor gas index and smoke anomaly index, making the subsequent calculation of the comprehensive risk quantity comparable in terms of each component and calibrable in terms of threshold. For battery types experiencing drastic temperature changes, the upper limit of the threshold can be appropriately increased to 7, while for those with gradual temperature changes, it can be decreased to 3. By simultaneously considering both temperature deviation and the rate of temperature change, the temperature trend energy can capture both abnormal increases in absolute temperature values ​​and the upward trend in temperature, avoiding the potential for missing slow heat accumulation or high-temperature plateaus that might occur when relying solely on the rate of temperature change. By limiting the temperature trend energy, it is ensured that it exists within the same scale space as the precursor gas index and the smoke anomaly index, providing a consistent numerical basis for subsequent fusion calculations of comprehensive risk quantities and threshold setting.

[0079] If the energy storage system is equipped with a smoke detector, a smoke anomaly index is calculated. The smoke anomaly index is calculated by subtracting the mean smoke baseline from the current smoke concentration, then dividing by the effective smoke baseline standard deviation, and finally limiting the quotient. The lower bound is zero, and the upper bound is a preset limit for the smoke anomaly index. The effective smoke baseline standard deviation is the larger of the smoke baseline standard deviation and the preset lower limit of the smoke standard deviation. The preset lower limit is determined based on the actual noise level and measurement accuracy of the smoke detector. It is obtained by statistically analyzing the long-term output of the detector under stable conditions to obtain the standard deviation of the detector's inherent noise, which serves as the basis for the preset lower limit. The default value for the preset upper limit of the smoke anomaly index is 5. This upper limit is determined based on the sensitivity of the smoke detector and the maximum smoke fluctuation amplitude under normal operating conditions. For high-sensitivity smoke detectors, it can be reduced to 3, and for low-sensitivity smoke detectors, it can be increased to 7. If no smoke detector is configured, the smoke anomaly index is set to the default value of zero.

[0080] To identify scenarios where temperature rise lag occurs due to strong intervention in thermal management, an intervention identification and advance compensation mechanism is established. The reason for establishing this mechanism is that when the thermal management system of an energy storage system intervenes with strong cooling, abnormal heat generation or early thermal runaway reactions may have already begun inside the battery. However, due to the forced cooling, the battery surface temperature is effectively suppressed, causing temperature sensors to fail to reflect internal anomalies in a timely manner. Relying solely on temperature information would miss early warning windows. Gas sensors, on the other hand, can detect precursor gas emissions generated by internal chemical reactions, unaffected by direct external cooling. Therefore, under strong intervention conditions, the weight of gas information should be increased to fully utilize gas precursors for early warning and compensate for the lag in temperature information. The first step is to identify the intervention state to determine whether the energy storage system is under strong intervention in thermal management. The criteria for identifying a strong intervention thermal management status include at least one of the following: the ratio of the actual operating power of the cooling system to its rated power is greater than or equal to a preset strong intervention judgment ratio for cooling power; the ratio of the actual flow rate of the coolant to its rated flow rate is greater than or equal to a preset strong intervention strong intervention judgment ratio for cooling flow rate; the temperature difference between the inlet and outlet of the cooling system is greater than or equal to a preset strong intervention temperature difference judgment threshold; the system has executed an emergency power reduction command; or the forced ventilation system has been activated. The default values ​​for the preset strong intervention judgment ratio for cooling power and the preset strong intervention judgment ratio for cooling flow rate are 70% and 5 degrees Celsius, respectively. These judgment conditions are used to identify whether the system is undergoing strong thermal management intervention.

[0081] Under strong intervention thermal management conditions, temperature anomalies may be suppressed, leading to a lag in temperature information. In this case, a pre-control response can be triggered solely based on the persistence and growth of precursor gases in the absence of obvious temperature and smoke anomalies, but spatial consistency and sensor health requirements must be met simultaneously to suppress false alarms. The spatial consistency requirement is that multiple gas sensing units simultaneously detect abnormal gas concentrations, or the duration of an anomaly detected by a single gas sensing unit exceeds a preset gas anomaly duration threshold. The default value of the preset gas anomaly duration threshold is 3 minutes, which is used to exclude false alarms caused by instantaneous gas fluctuations. The sensor health requirement is that the gas sensor is not in a faulty state, the sensor output is not saturated or ineffective, and the sensor's most recent calibration time is within the validity period.

[0082] Simultaneously, smoke anomaly detection is performed to determine whether an abnormal smoke condition exists. The detection criteria for an abnormal smoke condition are that the smoke anomaly index exceeds a preset smoke anomaly judgment threshold. The default value of this preset smoke anomaly judgment threshold is 0.5, which is determined based on the noise level and baseline fluctuation characteristics of the smoke detector. When the smoke anomaly index exceeds this threshold, it indicates the presence of a suspicious smoke signal.

[0083] Based on the results of intervention state recognition and smoke anomaly state recognition, a gas information modulation factor is calculated. When under strong intervention thermal management conditions and without smoke anomalies, the gas information modulation factor is set to 1, indicating full utilization of gas precursor information for early warning. When under non-strong intervention thermal management conditions or with smoke anomalies, the gas information modulation factor is set to a preset gas weight weakening coefficient for atypical scenarios, indicating a reduction in gas information weight to balance the contributions of gas information and temperature / smoke information. The method of dynamically adjusting gas information weight using the gas information modulation factor is adopted because the reliability of gas sensors varies in different scenarios. Under strong intervention thermal management conditions, gas information is the most reliable early warning signal and should be given sufficient weight. Under non-strong intervention conditions, temperature information is more reliable, and gas sensors may be affected by environmental factors. In this case, the contributions of gas and temperature information should be balanced to avoid over-reliance on gas information leading to false alarms. When smoke anomalies are present, it indicates that the situation has developed to a more serious stage, at which point temperature / smoke information becomes more critical. Dynamic modulation achieves an adaptive information fusion strategy for different scenarios. The default value of the gas weight weakening coefficient under the preset atypical scenario is 0.3. This coefficient is a value greater than zero and less than 1. Its value is determined based on the reliability of gas information in atypical scenarios. In non-strong intervention conditions or in the presence of smoke anomalies, gas sensors may be interfered with by other factors. Therefore, the weight of gas information should be appropriately reduced to avoid false alarms. For systems with high gas sensor reliability, it can be increased to 0.5, and for systems with gas sensors that are susceptible to interference, it can be reduced to 0.2.

[0084] To avoid abrupt changes in the gas information modulation factor when switching between strong and non-strong intervention states, a smooth transition process is adopted. When the intervention state or smoke anomaly changes, the gas information modulation factor does not immediately jump to the target value, but gradually transitions to the target value according to a preset modulation factor transition rate. The default value of the preset modulation factor transition rate is 0.1 per minute. This rate ensures that the gas information modulation factor completes the transition from 0.3 to 1 or from 1 to 0.3 in about one minute, avoiding abrupt changes in the overall risk level and alarm jitter caused by sudden switching. By establishing an intervention identification and advance compensation mechanism, early warning can be achieved by fully utilizing gas precursor information under strong intervention thermal management conditions. At the same time, the gas information weight is appropriately reduced in non-strong intervention states or when smoke anomalies are present to balance the contributions of various information sources and avoid false alarms.

[0085] Step 300: Calculate the overall risk level by combining the precursor gas index, temperature trend energy, smoke anomaly index, and gas information modulation factor.

[0086] The comprehensive risk quantity is calculated by combining the precursor gas index, temperature trend energy, smoke anomaly index, and gas information modulation factor constructed in step 200. The comprehensive risk quantity is calculated as the sum of the gas, temperature, and smoke terms. A weighted summation method is used to fuse multi-source information to calculate the comprehensive risk quantity because gas, temperature, and smoke information reflect the development process of thermal runaway from different physical dimensions. Gas information is sensitive in the early stages but may be affected by environmental interference; temperature information is reliable but suffers from thermal inertia lag; and smoke information is significant in the mid-to-late stages. A single information source cannot cover the entire thermal runaway process and is prone to false alarms or missed alarms. By merging the three types of information into a unified risk quantity index through weighted summation, the advantages of each information source can be comprehensively utilized to achieve continuous monitoring and accurate assessment of the entire thermal runaway process. At the same time, by reasonably setting weight coefficients to balance the contribution of each information source, the robustness and reliability of the system are improved. The gas component is the product of the gas information modulation factor, the precursor gas index, and the preset gas information fusion weighting coefficient. The temperature component is the temperature trend energy multiplied by the preset temperature information fusion weighting coefficient. The smoke component is the smoke anomaly index multiplied by the preset smoke information fusion weighting coefficient. The default values ​​for the preset gas information fusion weighting coefficient, the preset temperature information fusion weighting coefficient, and the preset smoke information fusion weighting coefficient are 0.4, 0.4, and 0.2, respectively. These three preset fusion weighting coefficients ensure that the contribution of gas information, temperature information, and smoke information to the overall risk level matches their early warning value in actual thermal runaway events. Specifically, gas information has high early warning value in the early stages of thermal runaway but may be affected by environmental interference; temperature information is relatively reliable but has a certain lag; and smoke information usually appears in a later stage. Therefore, the weights for gas and temperature are both set to 0.4, and the weight for smoke is set to 0.2.

[0087] The preset fusion weight coefficients are fixed values ​​and do not change with operating conditions to maintain the stability and consistency of the comprehensive risk calculation, avoiding risk fluctuations caused by frequent weight adjustments. Simultaneously, the smooth adjustment of the gas information modulation factor has achieved adaptive response to different operating conditions. Since the precursor gas index, temperature trend energy, and smoke anomaly index are all amplitude-limited to a range of 0 to 5 or 0 to 10, and the standard deviation lower limit based on the sensor's inherent noise level is adopted, the numerical scale of the comprehensive risk quantity is consistent and comparable across different projects and operating conditions. This provides a calibrable numerical basis for setting the subsequent three-level response threshold and cross-project application.

[0088] To further improve the calibrability of comprehensive risk parameters, during the initial operation of an energy storage system or when performing threshold calibration, the comprehensive risk parameters for each stage of a typical thermal runaway event can be recorded through thermal runaway simulation experiments or historical event data playback. A mapping relationship between the comprehensive risk parameters and the stages of thermal runaway development can be established. Based on this mapping relationship, the three-level response thresholds can be optimized and adjusted to make the threshold settings more consistent with the actual system response characteristics. This calibration process can be performed offline without affecting the normal operation of the system. The calibration results serve as a reference for threshold setting and are combined with the default thresholds to determine the final response thresholds.

[0089] Step 400: Based on the comparison result between the comprehensive risk amount and the preset risk threshold, a three-level response determination is made. When a Level 1 pre-control response is triggered, power limiting and cooling capacity enhancement actions are executed. When a Level 2 suppression preparation response is triggered, electrical isolation and fire suppression system standby actions are executed. When a Level 3 emergency fallback response is triggered, fire suppression and overall safety actions are executed, specifically as follows: Figure 3 As shown.

[0090] Three risk thresholds are pre-set: a first pre-set risk threshold, a second pre-set risk threshold, and a third pre-set risk threshold. These three thresholds satisfy the following conditions: the first pre-set risk threshold is less than the second pre-set risk threshold; the second pre-set risk threshold is less than the third pre-set risk threshold; and all three are greater than zero. The purpose of setting these three risk thresholds is to implement a tiered response strategy, taking intervention measures of appropriate intensity according to different levels of risk. This avoids over-response causing frequent system shutdowns or under-response leading to accident escalation. The three thresholds divide the thermal runaway development process into an early warning stage, a development stage, and an emergency stage, corresponding to pre-control response, suppression preparation response, and emergency fallback response, respectively, achieving full coverage from prevention to emergency response. The default value of the first pre-set risk threshold is 1.5. This threshold is determined based on the safety margin of the energy storage system and the early warning requirements. When the comprehensive risk level reaches 1.5, it usually corresponds to the early warning stage of thermal runaway. At this point, pre-control measures can effectively prevent the situation from escalating. For scenarios with high safety requirements, the threshold can be lowered to 1.0 to improve warning sensitivity; for scenarios where a certain false alarm rate is acceptable, the threshold can be increased to 2.0 to reduce the frequency of false alarms. The default value of the preset second risk threshold is 3.0. This threshold is determined based on the typical comprehensive risk level of the energy storage system entering a dangerous state. When the comprehensive risk level reaches 3.0, it usually corresponds to the development stage of thermal runaway. At this time, suppression preparation measures need to be initiated to prevent fire. For battery types with rapid thermal runaway development, the threshold can be lowered to 2.5 to allow more response time. For battery types with slow thermal runaway development, the threshold can be increased to 3.5 to reduce unnecessary activation of the fire suppression system. The default value of the preset third risk threshold is 5.0. This threshold is determined based on the comprehensive risk level of the energy storage system entering an emergency state and is used to trigger the highest level of emergency response. When the comprehensive risk level reaches 5.0, it usually corresponds to the severe stage of thermal runaway or the occurrence of open flame, and fire suppression must be initiated immediately.

[0091] Based on the comprehensive risk level calculated in step 300, a three-level response determination is made according to the following rules.

[0092] The Level 1 pre-control trigger condition is that the overall risk level is greater than or equal to the preset first risk threshold, which triggers the Level 1 pre-control response.

[0093] The secondary suppression preparation is triggered when the overall risk level is greater than or equal to the preset second risk threshold.

[0094] The Level 3 emergency backup trigger condition is that the overall risk level is greater than or equal to a preset third risk threshold, or the temperature at any monitoring point is greater than or equal to a preset direct temperature trigger threshold, or the smoke concentration is greater than or equal to a preset direct smoke trigger threshold. The default value for the preset direct temperature trigger threshold is 80℃. This threshold is determined based on the critical temperature and safety margin of battery thermal runaway. For battery types with poor thermal stability, it can be reduced to 70℃, and for battery types with good thermal stability, it can be increased to 90℃. The default value for the preset direct smoke trigger threshold is 10 times the average smoke baseline plus the standard deviation of the smoke baseline. This threshold is determined based on the alarm characteristics of the smoke detector and the typical concentration level of fire smoke, and is used to immediately trigger the highest level response when the smoke concentration is significantly abnormal.

[0095] By integrating gas, temperature, and smoke information based on comprehensive risk data, and adaptively adjusting the weight of gas information under different operating conditions through a gas information modulation factor, the system can trigger a response in advance during the early window period when there is no smoke or low temperature but abnormal gas is present. This solves the problem of insufficient advance amount in traditional temperature and smoke detection, while avoiding the risk of double weighting and false alarms caused by setting up an independent gas bypass channel.

[0096] Based on the above response determination results, execute the corresponding level of response action.

[0097] When a Level 1 pre-control response is triggered, the following actions are performed: Limit the charging and discharging power of the energy storage system or disconnect the high-voltage circuit. The power limiting command is sent from the fire control system to the battery management system. Upon receiving the command, the battery management system adjusts its power control strategy or disconnects the high-voltage contactor. Increase the cooling system capacity to the safe limit without activating fire suppression. The cooling capacity increase command is sent from the fire control system to the thermal management system. Upon receiving the command, the thermal management system increases the coolant flow rate or increases the fan speed. Increase the data acquisition frequency, identify suspected abnormal storage areas, and generate key monitoring tags. Suspected abnormal storage areas refer to the battery compartment or battery box location where the sensor triggering the current response level is located. By analyzing the gas concentration, temperature, and smoke concentration data of each monitoring point when the response is triggered, the monitoring point with the most significant numerical anomaly or the greatest contribution to the overall risk is identified. The battery compartment or battery box containing this monitoring point is marked as a suspected abnormal storage area. The data acquisition frequency increase command is sent from the fire control system to the battery management system and the data acquisition system, increasing the acquisition frequency from the normal once per minute to once every 10 seconds or higher. To eliminate gas concentration fluctuations caused by factors other than fire, after the Level 1 response is triggered, a simultaneous check is performed to see if there are any situations that may introduce interfering gases, such as maintenance work, cleaning operations, or external exhaust gas backflow. If such interference sources are confirmed and the temperature and smoke are not abnormal, the data for that period is marked as a non-fire gas event and the response judgment time window is appropriately extended to avoid false alarms caused by environmental interference.

[0098] When a Level 2 suppression response is triggered, the following actions are performed: Electrical isolation is applied to suspected anomalous battery compartments or battery boxes. These suspected anomalous battery compartments refer to the locations of the battery compartments or battery boxes where the sensors triggering the current response level are located. By analyzing the gas concentration, temperature, and smoke concentration data of each monitoring point at the time of triggering the response, the monitoring point with the most significant numerical anomalies or the greatest contribution to the overall risk is identified. The battery compartment or battery box containing this monitoring point is marked as a suspected anomalous battery compartment, and the circuit of that battery cluster or battery box is disconnected. The electrical isolation command is sent from the fire control system to the battery management system. Upon receiving the command, the battery management system disconnects the corresponding battery cluster contactor or battery box connector. The fire protection system is put into a standby state, including pre-pressurization, valve self-check, and establishment of the shortest discharge path. The standby command is sent from the fire control system to the fire sprinkler system. Upon receiving the command, the fire sprinkler system starts the pre-pressurization pump, checks the status of the solenoid valves, and opens the discharge pipeline valves leading to the suspected anomalous battery compartment. A higher-level audible and visual alarm is triggered, and a linkage notification is sent to the remote monitoring system. A rollback mechanism is maintained to cancel the response upon confirmation.

[0099] To reduce false alarms caused by non-fire factors, an observation and decision-making window is entered after a Level 2 response is triggered. Within this window, the system continuously monitors the changing trend of the overall risk level, the growth rate of the precursor gas index, and the degree of temperature and smoke anomalies. The purpose of setting up this observation and decision-making window is to increase the decision-making buffer time at this critical juncture of Level 2 response. By observing the subsequent evolution trend of the risk level, it determines whether it is a genuine thermal runaway or a transient anomaly caused by non-fire factors. This avoids unnecessary economic losses and system downtime caused by immediately activating the fire suppression system due to instantaneous fluctuations or environmental interference. Simultaneously, it ensures that a genuine thermal runaway event can be promptly escalated to a Level 3 response based on the continuously rising risk level. This mechanism significantly reduces the false alarm rate while ensuring safety. If the overall risk level continues to rise or remains high within the observation and decision-making window, or the growth rate of the precursor gas index exceeds the preset gas index growth rate threshold, or temperature and smoke show a coordinated upward trend, the Level 2 response is maintained, and preparations are made to escalate to a Level 3 response. The term "continuous increase in comprehensive risk" refers to an increasing trend in the comprehensive risk value across three or more consecutive sampling periods within the observation and decision window, with each increase exceeding 5% of the comprehensive risk value of the previous sampling period. The term "maintaining a high comprehensive risk level" refers to the comprehensive risk value consistently remaining above 90% of the preset second risk threshold within the observation and decision window, with fluctuations less than 10% of the preset second risk threshold. If the comprehensive risk value rapidly decreases within the observation and decision window, and neither temperature nor smoke shows abnormal increases, and the precursor gas index growth rate is below the preset gas index growth rate threshold, the response will automatically downgrade to Level 1 or resume normal monitoring. The term "rapid decrease in comprehensive risk" refers to a sustained decrease in the comprehensive risk value from its peak value at the time of triggering the Level 2 suppression preparation response within the observation and decision window, with the comprehensive risk value decreasing to below 70% of the preset second risk threshold by the end of the observation and decision window, while there is no situation within the observation and decision window where the comprehensive risk value shows an increasing trend across two or more consecutive sampling periods. The absence of abnormal increases in temperature and smoke means that, throughout the entire observation and decision window, the temperature at all monitoring points does not exceed three times the sum of the mean temperature baseline and the standard deviation of the temperature baseline, and the smoke concentration does not exceed three times the sum of the mean smoke baseline and the standard deviation of the smoke baseline. The duration of the observation and decision window is determined based on the system response characteristics and the timescale of the gas temperature and smoke information, with a default value of 5 minutes. For battery types with rapid thermal runaway development, this can be shortened to 3 minutes, and for battery types with slow thermal runaway development, it can be extended to 8 minutes. The default value for the preset gas index growth rate threshold is 0.5 per minute. This threshold indicates that when the precursor gas index continues to increase at a rate of 0.5 per minute, it signifies that the gas anomaly is intensifying, requiring the response level to be maintained or upgraded.

[0100] To avoid instability caused by frequent upgrades and downgrades of Level 2 responses, a minimum hold time constraint is set. After a Level 2 response is triggered, it must maintain a minimum hold time before downgrading is allowed. The default value for this minimum hold time is 10 minutes, ensuring the system has a sufficient observation window to determine whether the risk has truly subsided. A multi-confirmation mechanism for rollback conditions is also implemented. Downgrading from Level 2 response to Level 1 response or normal monitoring is only allowed when the overall risk level is below a preset first risk threshold for a consecutive preset number of rollback confirmations across multiple sampling periods, and both temperature and smoke are normal, and the gas concentration has fallen back to near the baseline level. The default number of preset rollback confirmations is 3, designed to prevent misjudgment of risk elimination due to a brief drop in risk level.

[0101] When a Level 3 emergency backup response is triggered, the following actions are performed: Fire sprinkler or extinguishing medium injection is initiated according to a preset strategy, and pump pressure is controlled. The fire sprinkler command is sent from the fire control system to the fire sprinkler device. Upon receiving the command, the fire sprinkler device opens the corresponding solenoid valve at the designated compartment and starts the sprinkler pump, controlling the pump pressure according to preset discharge pressure and flow curves. Simultaneously, global safety actions are executed, including power-off isolation, exhaust linkage, and pressure relief valve opening. The power-off isolation command is sent from the fire control system to the battery management system and power distribution system; the exhaust linkage command is sent from the fire control system to the ventilation system; and the pressure relief valve opening command is sent from the fire control system to the pressure relief device. Complete event data is recorded for post-event review and analysis, and threshold self-calibration. Event data includes all sensor readings at the trigger time, the overall risk level and its components, system operating conditions, the sequence of response actions, and data records of the entire event development process.

[0102] Step 500: After the action is executed, an evaluation time window is entered to calculate the risk decay rate. When the risk decay rate is less than the preset minimum effective decay rate within the evaluation time window for a consecutive preset number of correction judgments, the response level is increased. When the risk decay rate is greater than or equal to the preset fast recovery decay rate and the total risk is less than the preset first risk threshold, the response level is decreased until normal monitoring is restored.

[0103] After the graded action in step 400 is executed, an evaluation time window is entered, the action execution time is recorded, and the risk decay rate is calculated. The risk decay rate is calculated by subtracting the comprehensive risk amount after the evaluation time window from the comprehensive risk amount at the action execution time, and then dividing by the larger of the comprehensive risk amount at the action execution time and a preset risk amount scale reference value. The quotient is the risk decay rate. The purpose of calculating the risk decay rate is to quantify the actual effect of the evaluation response action. By comparing the change in comprehensive risk amount before and after the action execution, it is determined whether the current response level is sufficient to control the risk development, providing an objective basis for subsequent response escalation or downgrading decisions. This closed-loop feedback mechanism allows the system to dynamically adjust the response strategy based on the actual effect, avoiding insufficient response leading to situation deterioration or excessive response causing resource waste. The default value of the preset risk amount scale reference value is 1. This reference value is used to avoid abnormal risk decay rate calculation when the comprehensive risk amount at the action execution time is close to zero, while ensuring that the numerical scale of the risk decay rate is comparable under different initial risk levels. The duration of the evaluation time window is a preset evaluation time window length, the default value of which is five minutes. This time length is determined based on the effective time of the response action and the time scale of the risk change. For systems with faster response speeds, it can be shortened to three minutes, while for systems with slower response speeds, it should be extended to 10 minutes.

[0104] In this embodiment, the response level refers to the current fire response status determined based on the comparison result of the comprehensive risk amount and the preset risk threshold and the direct triggering conditions, including normal monitoring, level one pre-control response, level two suppression preparation response and level three emergency backup response.

[0105] Self-correction based on risk decay rate.

[0106] If the risk decay rate remains persistently low, indicating insufficient effectiveness of the current response level, the response level will be automatically upgraded by one level. If the current response is at Level 1 (pre-control), it will be upgraded to Level 2 (suppression preparation response); if the current response is at Level 2 (suppression preparation response), it will be upgraded to Level 3 (emergency backup response); if the current response is already at Level 3 (emergency backup response), it will remain at Level 3 or expand the isolation scope, and return to step 400 to execute a higher-level response. The condition for determining a persistently low risk decay rate is that within a preset number of consecutive correction judgments and evaluation time windows, the risk decay rate is always less than the preset minimum effective decay rate. Using a persistently low risk decay rate as the trigger for response escalation is because a single evaluation may be affected by measurement fluctuations or short-term factors. Confirming a persistently low risk decay rate through multiple consecutive evaluations reliably indicates that the current response measures are indeed unable to effectively control the risk development. In this case, the response level must be upgraded or the isolation scope expanded to prevent further deterioration. This adaptive escalation mechanism based on effect feedback ensures that the system can take sufficiently strong countermeasures when facing thermal runaway events of varying severity. The default value for the preset number of correction judgments is 3. This number is used to avoid triggering response escalation due to accidental fluctuations in a single assessment. For scenarios requiring rapid response, the number can be reduced to 2, and for scenarios requiring stability, it can be increased to 4. The default value for the preset minimum effective attenuation rate is 20%. This threshold is determined based on the expected effect of the response action. When the risk attenuation rate is lower than this threshold, it indicates that the response action has failed to effectively control the development of risk, and the response measures need to be upgraded.

[0107] If the risk decay rate quickly recovers to normal levels, and the temperature and smoke concentration do not show any abnormal increases within the assessment time window, the response level is gradually reduced. If the current emergency backup response level is Level 3, it is first reduced to Level 2 suppression preparation response level. If subsequent assessment results still meet the downgrade conditions, it is further reduced to Level 1 pre-control response level until normal monitoring is restored. The data during this period is marked as a suspected non-fire gas event to update the sensor baseline in step 100 and reduce the false alarm rate of similar events in the future. The condition for determining that the risk decay rate quickly recovers to normal levels is that the risk decay rate is greater than or equal to the preset rapid recovery decay rate, and the overall risk level has been reduced to below the preset first risk threshold. The rapid recovery of risk decay rate is used as the trigger condition for response degradation because when the risk level drops rapidly and the temperature and smoke information remain normal, it indicates that the main cause of the trigger response is likely a temporary fluctuation in gas concentration caused by non-fire factors rather than a true thermal runaway event. In this case, the response should be downgraded promptly to avoid excessive intervention. Simultaneously, the characteristics of this event are recorded to optimize baseline and threshold settings. Through this self-learning mechanism, the system's ability to identify non-fire factors is gradually improved, continuously reducing the false alarm rate while ensuring safety. The default value of the preset rapid recovery decay rate is 60%, indicating that the risk has been effectively controlled and has rapidly subsided. The criterion for ensuring that temperature and smoke concentration do not show abnormal increases is that, throughout the entire evaluation time window, the temperature at all monitoring points does not exceed three times the average temperature baseline plus the standard deviation of the temperature baseline, and the smoke concentration does not exceed three times the average smoke baseline plus the standard deviation of the smoke baseline.

[0108] When the above conditions are met, it indicates that the main cause of the triggering response is gas anomaly rather than actual thermal runaway or fire. It may be due to gas concentration fluctuations caused by non-fire factors. In this case, the data for that period is marked as a suspected non-fire gas event, and the tolerance for such gas fluctuations is appropriately increased during subsequent baseline updates, thereby reducing the false alarm rate. Specifically, for periods marked as suspected non-fire gas events, the system analyzes the gas concentration fluctuation characteristics, duration, and change patterns of that period, storing these characteristics as typical patterns of non-fire gas events. If similar gas fluctuation patterns are detected in subsequent operations, the gas information modulation factor is appropriately reduced during the comprehensive risk calculation, or the observation decision window is extended during response determination, to reduce false alarms caused by similar non-fire factors.

[0109] Through the aforementioned closed-loop verification and self-correction mechanism, false alarms are effectively suppressed and response performance is dynamically optimized, avoiding alarm fatigue caused by frequent false alarms, while ensuring timely and effective responses to real thermal runaway events. This closed-loop mechanism enables the system to have self-learning and adaptive capabilities, continuously optimizing response strategies and threshold settings based on actual operating experience, thereby improving the intelligence and practicality of the fire response system.

[0110] The three-level fire response method for energy storage based on temperature and smoke linkage provided in this embodiment achieves significant technical effects through several technological innovations. First, by establishing a hierarchical baseline library for operating conditions, it solves the problem that traditional single-baseline methods cannot adapt to the diverse operating modes of energy storage systems. This enables anomaly detection to accurately distinguish between changes in operating characteristics and genuine safety anomalies, avoiding the misjudgment of normal operation under high-power conditions as anomalies or the masking of early anomalies under low-power conditions by the baseline. Second, by employing standardized processing methods to calculate gas anomaly scores and temperature deviation scores, it unifies the outputs of sensors with different dimensions and ranges into a dimensionless space, achieving effective fusion of multi-sensor information and unified threshold determination, thus improving the system's portability and calibrability. Third, by constructing a precursor gas index that simultaneously considers the static deviation and dynamic upward trend of gas concentration, it can comprehensively capture multiple anomaly modes in the early stages of thermal runaway, significantly improving the detection rate and robustness of early warning compared to a single index. Fourth, by establishing an intervention identification and advance compensation mechanism, the system fully utilizes gas precursor information to compensate for the lag in temperature information under strong intervention thermal management conditions, solving the key problem of insufficient advance in traditional temperature and smoke detection in forced cooling scenarios, and extending the warning time window from several minutes in traditional methods to more than ten minutes. Fifth, by employing a gas information modulation factor to achieve adaptive information fusion under different scenarios, the system increases gas weight to achieve early warning under strong intervention conditions, and balances gas and temperature / smoke information to avoid false alarms under non-strong intervention conditions. Compared with fixed-weight fusion methods, it reduces the false alarm rate by more than 50% while maintaining a high detection rate. Sixth, by setting an observation decision window and a risk decay rate assessment mechanism, the system achieves response escalation and de-escalation decisions based on effect feedback, avoiding over-response triggered by instantaneous fluctuations or deterioration of the situation due to insufficient response, enabling the system to have adaptive adjustment capabilities. Seventh, by using the characteristics of suspected non-fire gas events for baseline updates and threshold optimization, the system achieves self-learning capabilities, continuously improving its ability to identify non-fire factors based on actual operating experience, and gradually reducing the false alarm rate over long-term operation. In summary, the method provided in this embodiment, compared with traditional temperature and smoke detection methods, advances the thermal runaway warning time by more than ten minutes, reduces the false alarm rate by more than 60%, and reduces the missed alarm rate to near zero, significantly improving the safety, reliability, and practicality of fire response in energy storage systems.

[0111] Example 2

[0112] See Figure 4 As shown, a three-level fire response system based on temperature and smoke linkage is provided. This system stores computer-readable instructions, which, when read, execute the aforementioned three-level fire response method based on temperature and smoke linkage. The system includes:

[0113] The baseline management module 101 collects the operating data of the energy storage system in real time, identifies the current operating conditions of the energy storage system based on the operating data, and establishes and updates the corresponding gas concentration baseline, temperature baseline and smoke baseline for different operating conditions.

[0114] Risk assessment module 102 calculates precursor gas index, temperature trend energy and smoke anomaly index based on the gas concentration baseline, temperature baseline and smoke baseline corresponding to the current operating conditions, respectively, to identify whether the energy storage system is in a state of strong intervention thermal management and to calculate gas information modulation factor.

[0115] Risk calculation module 103 calculates the comprehensive risk amount by combining the precursor gas index, temperature trend energy, smoke anomaly index and gas information modulation factor.

[0116] The graded response module 104 determines the three-level response based on the comparison result between the comprehensive risk amount and the preset risk threshold and executes the action. When the first-level pre-control response is triggered, it executes the power limiting and cooling capacity enhancement actions. When the second-level suppression preparation response is triggered, it executes the electrical isolation and fire protection system standby actions. When the third-level emergency backup response is triggered, it executes the fire spraying and global safety actions.

[0117] The closed-loop verification module 105 enters the evaluation time window after the action is executed, calculates the risk decay rate, and increases the response level when the risk decay rate is less than the preset minimum effective decay rate within the evaluation time window for a consecutive preset number of correction judgments; and decreases the response level until normal monitoring is restored when the risk decay rate is greater than or equal to the preset fast recovery decay rate and the comprehensive risk amount is less than the preset first risk threshold.

[0118] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A three-level fire response method based on temperature and smoke linkage for energy storage, characterized in that, Includes the following steps: Real-time acquisition of energy storage system operation data; identification of current energy storage system operation conditions based on the operation data; establishment and updating of corresponding gas concentration baselines, temperature baselines and smoke baselines for different operation conditions. Based on the gas concentration baseline, temperature baseline and smoke baseline corresponding to the current operating conditions, calculate the precursor gas index, temperature trend energy and smoke anomaly index respectively, identify whether the energy storage system is in a state of strong intervention thermal management and calculate the gas information modulation factor. The overall risk level is calculated by combining the precursor gas index, temperature trend energy, smoke anomaly index, and gas information modulation factor. Based on the comparison result between the comprehensive risk amount and the preset risk threshold, a three-level response judgment is made and actions are executed. When the first-level pre-control response is triggered, power limiting and cooling capacity enhancement actions are executed. When the second-level suppression preparation response is triggered, electrical isolation and fire protection system standby actions are executed. When the third-level emergency backup response is triggered, fire discharge and overall safety actions are executed. After the action is executed, an evaluation time window is entered to calculate the risk decay rate. When the risk decay rate is less than the preset minimum effective decay rate within a preset number of consecutive evaluation time windows for correction judgments, the response level is increased. When the risk decay rate is greater than or equal to the preset fast recovery decay rate and the total risk is less than the preset first risk threshold, the response level is reduced until normal monitoring is restored.

2. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 1, characterized in that, Establish and update corresponding gas concentration baselines, temperature baselines, and smoke baselines for different operating conditions, including: The operating data includes charging and discharging power, temperature at each monitoring point and its rate of temperature change, state of charge, ambient temperature, and operating parameters of the cooling system; Operating conditions include standby mode, low power charging mode, low power discharging mode, high power charging mode, high power discharging mode, forced cooling mode, and equalization maintenance mode. Identify steady-state time segments that meet low fluctuation conditions, and establish or update the sensing baseline for the current operating condition within the steady-state time segment. The sensing baseline includes gas concentration baseline, temperature baseline and smoke baseline. The gas concentration baseline includes the gas concentration baseline mean and gas concentration baseline standard deviation, the temperature baseline includes the temperature baseline mean and temperature baseline standard deviation, and the smoke baseline includes the smoke baseline mean and smoke baseline standard deviation. When the operating conditions of the energy storage system change, the current operating condition is identified and the corresponding sensing baseline is called. The sensor drift is monitored in real time. When the long-term mean deviation of the sensing quantity is detected to exceed the preset drift threshold, the sensing quantity baseline is slowly updated using an exponentially weighted moving average method.

3. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 2, characterized in that, The mean gas concentration baseline is the arithmetic mean of the gas concentration values ​​at all sampling points within the steady-state time segment, and the standard deviation of the gas concentration baseline is a measure of the dispersion of the gas concentration values ​​at each sampling point relative to the mean gas concentration baseline within the steady-state time segment. The mean temperature baseline is the arithmetic mean of the temperature values ​​at all sampling times of the monitoring point within the steady-state time segment, and the standard deviation of the temperature baseline is a measure of the dispersion of the temperature values ​​at each sampling time of the monitoring point relative to the mean temperature baseline within the steady-state time segment. The smoke baseline mean is the arithmetic mean of the smoke concentration values ​​at all sampling points within the steady-state time segment, and the smoke baseline standard deviation is a measure of the dispersion of the smoke concentration values ​​at each sampling point relative to the smoke baseline mean within the steady-state time segment. The low fluctuation condition is as follows: within a time segment, the fluctuation amplitude of the charging and discharging power is less than a preset power fluctuation threshold, and the maximum absolute value of the temperature change rate at all monitoring points is less than a preset temperature change rate threshold, and the change in the state of charge of the energy storage system is less than a preset state of charge change threshold, and the change in ambient temperature is less than a preset ambient temperature change threshold; the duration of the steady-state time segment is not less than a preset minimum steady-state duration.

4. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 1, characterized in that, The precursor gas index, temperature trend energy, and smoke anomaly index are calculated separately to identify whether the energy storage system is under strong intervention thermal management and to calculate the gas information modulation factor, including: Based on the gas concentration baseline corresponding to the current operating conditions, calculate the gas anomaly score and the standardized rise rate of gas concentration, and construct the precursor gas index. The precursor gas index is the sum of the limited anomaly score term and the limited gas concentration term. The limited anomaly score term is the gas anomaly score after the limited treatment multiplied by the first preset weight coefficient, and the limited gas concentration term is the standardized rise rate of gas concentration after the limited treatment multiplied by the second preset weight coefficient. Based on the temperature baseline corresponding to the current operating conditions, calculate the temperature deviation score and temperature change rate score of each monitoring point. Perform exponential moving average processing on the temperature deviation score and temperature change rate score of each monitoring point respectively. Take the maximum value of the temperature deviation score and the maximum value of the temperature change rate score of all monitoring points after exponential moving average processing, and then add them together after limiting processing to obtain the temperature trend energy. If the energy storage system is equipped with a smoke detector, the current smoke concentration is subtracted from the mean smoke baseline, and then divided by the effective smoke baseline standard deviation. The resulting quotient is then subjected to amplitude limiting to obtain the smoke anomaly index. The effective smoke baseline standard deviation is the larger of the smoke baseline standard deviation and the preset lower limit of the smoke standard deviation. Determine whether the energy storage system is under strong intervention thermal management and whether there is smoke anomaly. Calculate the gas information modulation factor based on the determination results. When the system is under strong intervention thermal management and there is no smoke anomaly, the gas information modulation factor is set to 1. When the system is not under strong intervention thermal management or there is smoke anomaly, the gas information modulation factor is set to the gas weight weakening coefficient under the preset atypical scenario.

5. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 4, characterized in that, The calculation method for the gas anomaly score is as follows: subtract the mean gas concentration baseline from the current gas concentration, and then divide by the effective gas concentration baseline standard deviation, where the effective gas concentration baseline standard deviation is the larger of the gas concentration baseline standard deviation and the preset lower limit of the gas standard deviation. The method for calculating the standardized rate of rise of gas concentration is as follows: First, calculate the gas concentration multi-point regression slope between the current time and the previous several sampling times to obtain the gas concentration change rate. Then, take the larger value between the gas concentration change rate and zero to obtain the positive change rate of gas concentration. Finally, divide the positive change rate of gas concentration by the standard deviation of the effective gas concentration baseline to obtain the standardized rate of rise of gas concentration. The temperature deviation score is calculated as follows: subtract the mean temperature baseline of the monitoring point from the temperature of the current monitoring point, then divide by the effective temperature baseline standard deviation to obtain the effective quotient, and finally take the larger of the effective quotient and zero as the temperature deviation score. The effective temperature baseline standard deviation is the larger of the temperature baseline standard deviation of the monitoring point and the preset lower limit of the temperature standard deviation. The method for calculating the temperature change rate fraction is as follows: First, calculate the temperature change rate at the monitoring point, then take the larger value between the temperature change rate and zero to obtain the positive temperature change rate, and finally divide the positive temperature change rate by the standard deviation of the effective temperature baseline to obtain the temperature change rate fraction.

6. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 4, characterized in that, The criteria for identifying a strong intervention thermal management state include at least one of the following: the ratio of the actual operating power of the cooling system to the rated power is greater than or equal to the preset strong intervention judgment ratio of cooling power; the ratio of the actual flow rate of the coolant to the rated flow rate is greater than or equal to the preset strong intervention judgment ratio of cooling flow rate; the temperature difference between the inlet and outlet of the cooling system is greater than or equal to the preset strong intervention temperature difference judgment threshold; or the system has executed an emergency power reduction command or the forced ventilation system has been started. The criteria for identifying abnormal smoke conditions is that the smoke abnormality index is greater than the preset smoke abnormality judgment threshold.

7. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 1, characterized in that, The comprehensive risk quantity is calculated as the sum of the gas item, temperature item, and smoke item. The gas item is the product of the gas information modulation factor, the precursor gas index, and the preset gas information fusion weight coefficient. The temperature item is the temperature trend energy multiplied by the preset temperature information fusion weight coefficient. The smoke item is the smoke anomaly index multiplied by the preset smoke information fusion weight coefficient.

8. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 7, characterized in that, Pre-set risk thresholds, including a preset first risk threshold, a preset second risk threshold, and a preset third risk threshold, satisfying that the preset first risk threshold is less than the preset second risk threshold, the preset second risk threshold is less than the preset third risk threshold, and all risk thresholds are greater than zero; The trigger condition for Level 1 pre-control is that the Level 1 pre-control response is triggered when the total risk level is greater than or equal to the preset first risk threshold. The secondary suppression preparation is triggered when the overall risk level is greater than or equal to the preset second risk threshold. The Level 3 emergency backup trigger condition is that the overall risk level is greater than or equal to the preset third risk threshold, or the temperature at any monitoring point is greater than or equal to the preset temperature direct trigger threshold, or the smoke concentration is greater than or equal to the preset smoke direct trigger threshold.

9. The three-level fire response method for energy storage based on temperature and smoke linkage according to claim 1, characterized in that, When the Level 1 pre-control response is triggered, the charging and discharging power of the energy storage system is limited or the high-voltage circuit is cut off, the cooling system capacity is increased to the safe limit but the fire sprinkler is not activated, the data acquisition frequency is increased and the suspected abnormal storage point is locked. When the Level 2 suppression preparation response is triggered, electrical isolation is performed on suspected abnormal storage points, the fire protection system is controlled to enter the standby state, a higher level of audible and visual alarm is triggered, and the trend of changes in the comprehensive risk quantity, the growth rate of the precursor gas index, and the degree of abnormality of temperature and smoke are continuously monitored in the observation and decision window. If the comprehensive risk quantity in the observation and decision window continues to rise or remains at a high level, the Level 2 response is maintained and preparations are made to upgrade to the Level 3 response. If the comprehensive risk quantity in the observation and decision window drops rapidly and there is no abnormal increase in temperature and smoke, the response is automatically downgraded to the Level 1 response or normal monitoring is restored. When a Level 3 emergency backup response is triggered, fire sprinkler or extinguishing medium injection is initiated according to the preset strategy, and pump pressure is controlled. Simultaneously, global safety actions are executed, including power outage isolation, ventilation linkage, and pressure relief valve opening. Complete event data is recorded for post-event review and analysis.

10. A three-level fire response system based on temperature and smoke linkage, characterized in that: The system is used to store computer-readable instructions, which, when read, execute the energy storage three-level fire response method based on temperature and smoke linkage as described in any one of claims 1-9; the system includes: The baseline management module collects real-time operating data of the energy storage system, identifies the current operating conditions of the energy storage system based on the operating data, and establishes and updates corresponding gas concentration baselines, temperature baselines, and smoke baselines for different operating conditions. The risk assessment module calculates the precursor gas index, temperature trend energy, and smoke anomaly index based on the gas concentration baseline, temperature baseline, and smoke baseline corresponding to the current operating conditions, respectively, to identify whether the energy storage system is in a state of strong intervention thermal management and to calculate the gas information modulation factor. The risk calculation module integrates the precursor gas index, temperature trend energy, smoke anomaly index, and gas information modulation factor to calculate the overall risk level. The graded response module determines the three-level response based on the comparison result between the comprehensive risk amount and the preset risk threshold and executes the action. When the first-level pre-control response is triggered, the power limiting and cooling capacity enhancement actions are executed. When the second-level suppression preparation response is triggered, the electrical isolation and fire protection system standby actions are executed. When the third-level emergency backup response is triggered, the fire spraying and global safety actions are executed. The closed-loop verification module enters an evaluation time window after the action is executed, calculates the risk decay rate, and increases the response level when the risk decay rate is less than the preset minimum effective decay rate within a number of consecutive preset correction judgments within the evaluation time window; and decreases the response level until normal monitoring is restored when the risk decay rate is greater than or equal to the preset fast recovery decay rate and the total risk is less than the preset first risk threshold.

Images