Energy storage fire-fighting mixed solution high-pressure precise injection control system and method

Abstract

The application discloses a high-pressure precise injection control system and method for energy storage fire-fighting mixed solution, and relates to the technical field of energy storage battery protection; the method comprises the following steps: obtaining a monitoring data set at the current time; inputting the monitoring data set into a pre-established evaluation model, calculating the residual reaction activity in the battery cell based on the temperature factor, the gas factor, the smoke factor, the pressure deformation factor and the sound factor obtained through the monitoring data set, and generating a thermal runaway reaction activity index; comparing the thermal runaway reaction activity index with a preset safety threshold, combining the monitoring data set, and outputting a spray control state; when the spray control state is ready to stop, performing a short-time spray stop test and outputting a reignition risk detection result; outputting a spray execution instruction according to the reignition risk detection result and executing; the application can accurately evaluate the residual reaction activity in the battery cell, reduce the thermal runaway reignition probability, and avoid damage to the energy storage equipment caused by excessive spraying.

Description

Technical Field

[0001] This invention relates to the field of energy storage battery protection technology, and more specifically, to a high-pressure precision injection control system and method for energy storage fire-fighting mixture. Background Technology

[0002] In the field of safety protection for electrochemical energy storage systems, lithium batteries are widely used in various energy storage power stations due to their advantages such as high energy density and long cycle life. However, the risk of thermal runaway remains a core bottleneck affecting the safe and stable operation of the system. To address the fire hazards caused by lithium battery thermal runaway, the industry generally adopts fire extinguishing technology that uses high-pressure precision injection of fire extinguishing mixture. This technology targets and delivers the extinguishing medium to the thermal runaway area to achieve temperature control and reaction inhibition, thereby preventing the spread and expansion of the fire. This type of fire extinguishing system is typically used in enclosed or semi-enclosed scenarios such as containerized energy storage tanks and large energy storage power stations. It needs to balance fire extinguishing efficiency and equipment protection, rapidly suppressing the thermal runaway process while avoiding excessive discharge that could cause secondary damage to the energy storage equipment.

[0003] The core working principle of existing high-pressure precision injection fire extinguishing technology is based on environmental parameters collected by sensors such as temperature and smoke to trigger a discharge command. A high-pressure power unit then injects the fire extinguishing mixture into the target area at a preset pressure and flow rate. This alters the external heat dissipation conditions and thermal runaway reaction rate of the lithium battery through physical cooling and chemical inhibition, thereby extinguishing the open flame and reducing the temperature. In the discharge cessation control phase, existing technologies mostly employ criteria based on surface temperature monitoring or fixed discharge duration. Specifically, when the ambient temperature inside the energy storage chamber or the battery surface temperature drops to a preset threshold, or the cumulative discharge duration reaches a set value, the system automatically terminates the injection of the fire extinguishing mixture, thus balancing the fire extinguishing effect with media consumption.

[0004] However, lithium-ion battery thermal runaway is essentially an exothermic process dominated by internal chemical reactions within the cell. Existing fire suppression technologies can only alter heat dissipation conditions and reaction rates through external intervention, failing to directly observe the reactive state within the cell. The residual level of internal reactive activity directly determines the risk of reignition after thermal runaway, but current criterion designs fail to effectively consider this crucial factor. Because the internal reactive activity of the cell is difficult to detect directly online, engineering practice often forces the use of simplified criteria, relying solely on surface temperature drop or a fixed duration as the sole basis for stopping the runaway, ignoring the potential residual internal exothermic activity behind the temperature drop. Furthermore, different battery types, states of charge, and heat propagation paths lead to significant differences in the decay patterns of internal reactive activity within the cell. A single simplified criterion cannot adapt to various operating conditions, resulting in errors in determining the timing of stopping the runaway.

[0005] The aforementioned flaws in the criteria can lead to a series of consequences, the most direct being thermal runaway reignition and secondary heat release. When the surface temperature drops to the threshold but the internal reactivity has not yet decayed to a safe level, residual heat inside the cell will continue to accumulate after discharge stops, pushing the temperature back up to the thermal runaway critical range, causing the fire to reignite. This not only causes further damage to the energy storage equipment but may also breach the initial protective barriers, triggering a chain reaction and expanding the scope of the accident. Furthermore, to mitigate the risk of reignition, some designs deliberately extend the discharge time, leading to excessive consumption of the fire extinguishing mixture. Excessive medium soaking into critical components of the energy storage equipment may damage its insulation performance, causing secondary faults such as electrical short circuits, increasing equipment maintenance costs and downtime. More critically, the simplified criteria result in uncontrollable reignition probabilities under different operating conditions, reducing the stability and reliability of the fire extinguishing system's protective effect.

[0006] In view of this, the present invention proposes a high-pressure precision injection control system and method for energy storage fire-fighting mixture to solve the above problems. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: a high-pressure precision injection control system and method for energy storage fire-fighting mixture; wherein, the high-pressure precision injection control method for energy storage fire-fighting mixture includes:

[0008] The system acquires battery surface temperature, cabin ambient temperature, smoke concentration, characteristic gas concentration, cabin air pressure, battery module deformation, and sound signals, performs time synchronization and format standardization processing, and outputs the monitoring dataset for the current moment.

[0009] The monitoring dataset is input into a pre-established evaluation model. Based on the temperature, gas, smoke, pressure deformation, and sound factors obtained from the monitoring dataset, the residual reactivity inside the battery cell is calculated, and a thermal runaway reactivity index is generated.

[0010] The thermal runaway reaction activity index is compared with a preset safety threshold, and combined with the monitoring dataset, the release control status is output; the release control status includes continue release and prepare to stop.

[0011] When the discharge control status is ready to stop, perform a short-term discharge stop test and output the reignition risk detection result;

[0012] Based on the reignition risk detection results, output and execute the discharge execution command.

[0013] Furthermore, temperature factors include the extent to which the battery surface temperature exceeds the safety threshold and the rate of change of the battery surface temperature; gas factors include the deviation of the characteristic gas concentration from the warning value; smoke factors include the amount of change in smoke concentration; pressure deformation factors include the amount of change in cabin air pressure and battery module deformation; and sound factors include the detection results of abnormal noises in the sound signal.

[0014] Furthermore, the evaluation model adopts a gated recurrent neural network model, which models the temporal correlation of the input monitoring dataset through a gated recurrent unit layer, and then outputs the thermal runaway response activity index through a feature fusion fully connected layer.

[0015] Furthermore, when training and evaluating the model, whether there is a temperature rebound and a rise in the concentration of characteristic gases within the observation window, which is limited by the observation window length parameter after the discharge stops, is used as the labeling rule for reignition judgment, thereby generating training labels for thermal runaway reaction activity indicators, so that the training target is consistent with the control target.

[0016] The data acquisition environment during the training of the evaluation model is limited to the semi-enclosed state of the energy storage cabin.

[0017] Furthermore, the methods for performing short-term discharge cessation tests include:

[0018] The discharge of the fire-fighting mixture is temporarily interrupted, and the monitoring dataset is continuously collected and updated during the short-term discharge stop test. If any sign of reignition occurs during the short-term discharge stop test, the reignition risk detection result is output as "there is a risk of reignition"; if no sign of reignition occurs during the short-term discharge stop test, the reignition risk detection result is output as "there is no risk of reignition".

[0019] Furthermore, signs of reignition include a renewed rise in battery surface temperature, a renewed rise in cabin ambient temperature, a renewed increase in smoke concentration, a renewed increase in the concentration of characteristic gases, a continued rise in cabin pressure, or the reappearance of abnormal sounds.

[0020] Furthermore, when the thermal runaway reaction activity index is higher than the preset safety threshold, the release control state is to continue releasing; when the thermal runaway reaction activity index is not higher than the preset safety threshold and the monitoring dataset meets the conditions for stopping the release, the release control state is to prepare to stop.

[0021] Furthermore, when the reignition risk detection result indicates a reignition risk, the output discharge execution command is to resume discharge; when the reignition risk detection result indicates no reignition risk, the output discharge execution command is to stop discharge.

[0022] Furthermore, after outputting the discharge execution command to resume discharge, the system enters the minimum continuous discharge duration constraint. Within the minimum continuous discharge duration, the controller maintains the discharge control state as continue discharging and does not allow the generation of preparation to stop.

[0023] The high-pressure precision injection control system for energy storage fire-fighting mixture includes:

[0024] The data acquisition module is used to acquire battery surface temperature, cabin ambient temperature, smoke concentration, characteristic gas concentration, cabin air pressure, battery module deformation, sound signals, and fire-fighting liquid discharge parameters, and to perform time synchronization and format standardization processing to output the monitoring dataset at the current moment.

[0025] The activity calculation module is used to input the monitoring dataset into the pre-established evaluation model, and calculate the residual reactivity inside the cell based on the temperature, gas, smoke, pressure deformation and sound factors obtained from the monitoring dataset, and generate thermal runaway reactivity index.

[0026] The status generation module is used to compare the thermal runaway reaction activity index with the preset safety threshold, and combine it with one or more data from the monitoring dataset to output the release control status; the release control status includes continue release and prepare to stop.

[0027] The reignition assessment module is used to perform a short-term release stop test when the release control state is ready to stop, and output the reignition risk detection results.

[0028] The instruction execution module is used to output and execute the discharge execution instructions based on the reignition risk detection results.

[0029] Compared with the prior art, the technical effects and advantages of the energy storage fire-fighting mixture high-pressure precision injection control system and method of the present invention are as follows:

[0030] The present invention relates to a high-pressure precision injection control system and method for energy storage fire-fighting mixture. A data acquisition module acquires multi-dimensional monitoring data, such as battery surface temperature and cabin ambient temperature, which is then processed through time synchronization and format standardization to form a monitoring dataset. An activity calculation module inputs the monitoring dataset into a pre-trained gated recurrent neural network model, integrating factors such as temperature, gas, and smoke to calculate and generate a thermal runaway reaction activity index. A state generation module compares this index with a preset safety threshold, and, based on the monitoring data, outputs a control state indicating continued discharge or preparation for cessation, while adaptively adjusting the discharge parameters. When preparing for cessation, a reignition assessment module performs a short-term discharge cessation test to detect reignition risk. An instruction execution module outputs an instruction to resume or stop discharge based on the detection results, while simultaneously setting a minimum continuous discharge duration and a monitoring hold duration after discharge cessation to ensure control stability.

[0031] This invention effectively solves the problems of existing technologies that rely solely on surface temperature or fixed duration as the criterion for stopping the discharge, ignoring the residual reactive activity inside the battery cell. This leads to deviations in the timing of stopping the discharge under different operating conditions, a high risk of reignition, and the potential for secondary failures due to excessive discharge. Its advantages lie in its ability to accurately assess the residual reactive activity inside the battery cell, maintaining consistency and reliability in stopping the discharge decision under different battery types, states of charge, and heat propagation paths. This reduces the probability of thermal runaway reignition and avoids damage to energy storage devices caused by excessive discharge, thus balancing fire suppression effectiveness and equipment protection. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the high-pressure precision injection control system for energy storage fire-fighting mixture according to an embodiment of the present invention;

[0033] Figure 2 This is a flowchart of the high-pressure precision injection control method for energy storage fire-fighting mixture according to an embodiment of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments below are only used to better illustrate and explain the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0035] Example 1:

[0036] Please see Figure 1 As shown, this embodiment discloses a high-pressure precision injection control system for energy storage fire-fighting mixture, including a data acquisition module, an activity calculation module, a status generation module, a reignition assessment module, and an instruction execution module. Each module is connected via wired and / or wireless means to achieve data transmission.

[0037] The data acquisition module is used to acquire battery surface temperature, cabin ambient temperature, smoke concentration, characteristic gas concentration, cabin air pressure, battery module deformation, sound signals, and fire-fighting liquid discharge parameters, and performs time synchronization and format standardization processing to output the monitoring dataset at the current moment.

[0038] The battery surface temperature is collected by temperature sensors attached to the surface of individual battery cells or battery modules. These sensors output instantaneous measurements of the battery casing temperature, characterizing the battery's heating level and serving as a direct observation of the cooling effect. The internal ambient temperature is collected by ambient temperature sensors located within the energy storage compartment. This temperature characterizes changes in the overall thermal environment of the energy storage compartment and assesses the extent of heat diffusion within it. Smoke concentration is collected by photoelectric smoke sensors located within the energy storage compartment. This concentration characterizes the presence level of combustion or thermal decomposition products and helps determine changes in the open flame state. Characteristic gas concentrations are collected by gas sensors or gas sampling and analysis devices located within the energy storage compartment. These concentrations characterize the changing trends of thermal runaway gas production and are used to extrapolate whether the side reaction activity within the battery cell is still in the release channel. Characteristic gases include carbon monoxide, hydrogen, methane, ethylene, and carbon dioxide. Internal pressure is collected by pressure sensors located within the energy storage compartment. This pressure characterizes the gas production pressure response within the enclosed or semi-enclosed space and helps determine whether gas production behavior is still increasing. Battery module deformation is collected by strain or displacement sensors mounted on the surface of the battery module. This deformation characterizes the structural response of the battery module under internal gas expansion and helps determine whether the expansion process has stabilized. Acoustic signals are collected by acoustic sensors installed inside the energy storage compartment. These signals characterize the occurrence of abnormal acoustic events such as safety valve ejection and structural rupture, serving as supplementary evidence that violent internal reactions have not subsided. Fire extinguishing mixture discharge parameters are collected by flow meters and pressure sensors arranged in the fire extinguishing system piping. These parameters include at least the discharge flow rate and discharge pressure, characterizing the delivery intensity of the fire extinguishing mixture and providing execution-side feedback for discharge control and effectiveness evaluation.

[0039] To ensure comparability of data from different sources at the same computation time, time synchronization includes scheduling sampling from each monitoring sensor according to a unified sampling period and resampling and aligning asynchronously arriving sampled values. The unified sampling period limits the update frequency of the monitoring dataset, reducing computational load while ensuring the tracking of rapid changes in thermal runaway; for example, a unified sampling period of 1 second can be used. Resampling and alignment uses timestamp matching combined with linear interpolation to generate alignment values: when the difference between the sampling time marker of a certain sensor parameter and the target alignment time does not exceed the alignment tolerance parameter, the sampled value is directly selected as the alignment value. The alignment tolerance parameter limits the effective time error of different sensor parameters; for example, if the alignment tolerance parameter is 0.2 seconds, it will not exceed 0.2 times the unified sampling period. When the difference exceeds the alignment tolerance parameter, linear interpolation is calculated between the two sampled values ​​before and after the target alignment time according to the time ratio to obtain the alignment value at the target alignment time, thus avoiding abrupt errors caused by asynchronous sampling.

[0040] Format standardization includes dimensionless normalization and outlier suppression. Dimensionless normalization maps the values ​​of different sensor parameters to a unified data representation range for subsequent fusion calculations and threshold comparisons. Dimensionless normalization uses a linear mapping based on the range, which is determined by the nominal range of the corresponding sensor and fixed at the factory. Outlier suppression eliminates the impact of sampling noise or transient interference on the monitoring dataset. Outlier suppression uses a median filter window parameter to perform median filtering on each sensor parameter sequence and outputs the filtered value. The median filter window parameter is used to strike a trade-off between suppressing spike noise and maintaining dynamic response. For example, a median filter window parameter of 3 can cover 3 consecutive sampling points to suppress single-point spikes without significantly introducing hysteresis. After time synchronization and format standardization, the controller generates and outputs the monitoring dataset for the current moment. The monitoring dataset serves as input for thermal runaway reaction activity assessment in subsequent steps and is used for determining the discharge control status and generating reignition risk detection results.

[0041] The activity calculation module is used to input the monitoring dataset into the pre-established evaluation model, and calculate the residual reactivity inside the cell based on temperature, gas, smoke, pressure deformation and sound factors obtained from the monitoring dataset, and generate thermal runaway reactivity index.

[0042] Temperature factors include the extent to which the battery surface temperature exceeds the safety threshold and the rate of change of the battery surface temperature; gas factors include the trend of characteristic gas concentration changes or the deviation of characteristic gas concentration from the warning value; smoke factors include the stable level or amount of change of smoke concentration; pressure deformation factors include the amount of change of cabin air pressure and battery module deformation; and sound factors include the detection results of abnormal noises in the sound signal.

[0043] During the thermal runaway response activity assessment phase, the monitoring dataset is input into a pre-established assessment model, which outputs the current thermal runaway response activity index. The correlation between the input and output data of the evaluation model lies in the fact that the thermal runaway reaction activity index is used to characterize the strength of the residual reaction activity inside the cell. However, the residual reaction activity inside the cell is difficult to measure directly online and needs to be extrapolated through external observable measurements. Among them, the battery surface temperature reflects the external thermal response after the internal heat release of the cell is conducted through the structure. The rate of change of the battery surface temperature reflects whether the dynamic balance between external cooling and internal heat release is still dominated by internal heat release. The characteristic gas concentration reflects the gas production intensity of the side reactions and electrolyte decomposition inside the cell. The trend of the characteristic gas concentration or the deviation of the characteristic gas concentration from the warning value is used to characterize whether the gas production channel is still increasing. The smoke concentration reflects the accumulation level of thermal decomposition and combustion products and is used to determine whether combustion-related products are still being generated. The cabin pressure reflects the pressure response of gas production and release in the closed or semi-closed space. The battery module deformation reflects the loading response of the cell expansion to the structure and is used to determine whether the internal gas production pressure is still driving structural changes. The abnormal noise detection results in the sound signal reflect whether violent events such as safety valve ejection or structural rupture are still occurring. By simultaneously incorporating temperature, gas, smoke, pressure deformation, and sound factors, the evaluation model can identify whether residual internal reactivity is still in the release channel when the battery surface temperature drops due to external cooling. This reduces the probability of misjudging the cessation of discharge based solely on temperature drop.

[0044] The input features of the evaluation model consist of statistics and changes in the monitoring dataset within the feature extraction time window parameter. This parameter balances response to short-term thermal runaway mutations with suppression of sensor noise, covering 10 uniform sampling periods to determine a stable trend; for example, a 10-second time window parameter could be used. Based on this parameter, the controller calculates the temperature exceedance and rate of change for the battery surface temperature, the gas deviation and trend for the characteristic gas concentration, the smoke stability and trend for the smoke concentration, the pressure and deformation trends for the cabin pressure and battery module deformation, and the frequency or duration of abnormal noise detection results for the sound signal, thus forming the input vector of the evaluation model. The thermal runaway reaction activity index output by the evaluation model ranges from 0 to 100; a higher value indicates stronger residual reaction activity within the cell. This index is used for subsequent steps in determining the discharge control state and triggering discharge termination verification.

[0045] To ensure consistency between training data and online applications, the evaluation model's training environment is limited to a full-size or equivalent-scale test environment with the energy storage chamber in a closed or semi-closed state. The chamber's free volume and ventilation conditions are consistent with the deployment scenario. The data acquisition process for training data is identical to the online acquisition process, collecting battery surface temperature, chamber ambient temperature, smoke concentration, characteristic gas concentration, chamber air pressure, battery module deformation, sound signals, and fire-fighting mixture discharge parameters according to a unified sampling period. Time synchronization and format standardization processing are performed in accordance with the online process to avoid generalization bias in the evaluation model due to differences in the acquisition process. The training samples cover thermal runaway processes under different battery types, states of charge, and heat propagation paths. Each training sample includes a continuous monitoring dataset sequence from discharge initiation to the stable phase after discharge cessation, supporting the evaluation model in learning the differences in the decay patterns of internal residual reactivity over time.

[0046] The evaluation model employs a gated recurrent neural network (RNN) model, which includes an input layer, a gated recurrent unit (ROU) layer, a feature fusion fully connected layer, and an output layer. The input layer receives an input vector sequence based on the feature extraction time window parameters. The RNU layer models the temporal correlation of the input vector sequence to extract a dynamic representation of the residual reactivity evolving over time. The feature fusion fully connected layer performs a nonlinear mapping on the hidden states output by the RNU layer and generates fused features for regression. The output layer outputs the thermal runaway reactivity index. The number of hidden units in the RNU layer is limited by the number of hidden units themselves. This number represents a trade-off between the model's expressive power and the embedded controller's computing power, ensuring that the single inference latency does not exceed 0.1 seconds under a uniform 1-second sampling period. Therefore, 64 hidden units are selected. The number of fully connected layers for feature fusion is limited by the total number of fully connected layers. The number of fully connected layers represents a trade-off between avoiding overfitting and improving fitting ability. For example, when the training sample size is at least 500 groups, a 2-layer structure is used to improve non-linear fitting ability and maintain stable convergence; therefore, the number of fully connected layers can be 2. The number of nodes in each fully connected layer is also limited by the total number of nodes; for example, 32 nodes can be used. The output layer uses linear regression output and maps it to an index range of 0 to 100 via an output scaling parameter, which is set to 100.

[0047] The training process of the evaluation model includes sample labeling, data partitioning, loss function construction, parameter iteration, and model solidification. Sample labeling generates training labels for each training sample, representing thermal runaway reactivity indicators. These training labels serve as supervisory signals for the evaluation model, enabling it to learn the correspondence between the rebound risk of the monitoring dataset after discharge cessation and the residual reactivity within the battery cell. Training labels are generated by reignition determination rules, which are based on the joint determination of whether temperature rebound and gas production recovery occur within the observation window after discharge cessation. The observation window length covers the possible rebound process after discharge cessation and includes slow rebound scenarios. The observation window length is no less than 300 seconds; for example, a 300-second observation window length is acceptable. The controller calculates the rebound amplitude of battery surface temperature and the rebound amplitude of characteristic gas concentration within the observation window, starting from the moment the discharge stops. The rebound amplitude of battery surface temperature is equal to the maximum value of battery surface temperature within the observation window minus the battery surface temperature at the moment the discharge stops, and the rebound amplitude of characteristic gas concentration is equal to the maximum value of characteristic gas concentration within the observation window minus the characteristic gas concentration at the moment the discharge stops. Thus, the maximum rebound and maximum rebound amplitude represent the upper bound of the rebound risk.

[0048] To ensure consistency between the joint decision-making and continuous label mapping logic, training labels are generated using a gated linear mapping method. The temperature rebound threshold is used to filter out minor rebounds caused by noise and is no less than 5 times the measurement error of the battery surface temperature sensor; for example, a temperature rebound threshold of 5 degrees Celsius. The gas rise threshold is used to filter out minor fluctuations caused by gas sensor drift and is no less than 0.01 of the full-scale range of the characteristic gas concentration sensor; for example, a gas rise threshold of 0.01 of the full-scale range. Further, upper limits for temperature and gas normalization are set. The upper limit for temperature normalization is used to limit the maximum reference range for normalizing the rebound amplitude and to avoid stretching the scale by extreme outliers. The logic for setting the upper limit for temperature normalization is to cover the rebound amplitude of most reignition samples in the training set while retaining a safety margin. For example, the upper limit for temperature normalization can be set to 30 degrees Celsius. The upper limit for gas normalization is used to limit the maximum reference range for normalizing the rebound amplitude and to avoid stretching the scale by extreme outliers. The logic for setting the upper limit for gas normalization is to cover the rebound amplitude of most reignition samples in the training set while retaining a safety margin. The upper limit for gas normalization is set to 0.2 of the full scale. For example, 0.2 of the full scale. The controller normalizes the battery surface temperature rebound amplitude into a temperature normalized value and the characteristic gas concentration rise amplitude into a gas normalized value. The temperature normalized value is calculated by subtracting the temperature rebound threshold from the battery surface temperature rebound amplitude, then dividing by the upper limit of temperature normalization minus the temperature rebound threshold, and limiting it to 0 to 1. The gas normalized value is calculated by subtracting the gas rise threshold from the characteristic gas concentration rise amplitude, then dividing by the upper limit of gas normalization minus the gas rise threshold, and limiting it to 0 to 1. This ensures that rebounds or rises below the threshold are mapped to 0, and rebounds or rises above the normalization upper limit are mapped to 1. To reflect the necessity of joint judgment for reignition risk, the controller calculates the joint normalized value as the product of the temperature normalized value and the gas normalized value. The joint normalized value is used to characterize the intensity level when temperature rebound and gas production rise occur simultaneously, thus avoiding excessively high training labels when the temperature rebound is large but the characteristic gas concentration rise is insufficient, and also avoiding excessively high training labels when the characteristic gas concentration rise is large but the temperature is insufficient, ensuring that the training labels are consistent with the joint judgment logic.

[0049] The training label is linearly mapped to 0 to 100 using a joint normalization value. The training label equals 100 multiplied by the joint normalization value and is limited to the range of 0 to 100. Therefore, when the battery surface temperature rebound does not exceed the temperature rebound threshold or the characteristic gas concentration rise does not exceed the gas rise threshold, the joint normalization value is 0, and the training label is 0. When the battery surface temperature rebound exceeds the upper limit of temperature normalization and the characteristic gas concentration rise exceeds the upper limit of gas normalization, the joint normalization value is 1, and the training label is 100. When both are between the threshold and the upper limit of normalization, the training label monotonically increases as both increase, thus maintaining both continuity to support regression training and consistency in joint judgment to support reignition risk modeling. Through this gated linear mapping method, the generation process of the training label has a definite mathematical expression, ensuring that those skilled in the art can reproduce the training label under the same acquisition environment and the same threshold settings.

[0050] Data partitioning is used to construct training and validation sets. The training set ratio is used to limit the proportion of the training and validation sets; for example, the training set ratio can be 0.8, with the remaining samples used as the validation set. The loss function uses the mean squared error loss function with a regularization term. The regularization coefficient is used to suppress overfitting, and the logic for setting the regularization coefficient is to take the minimum stable convergence value when the validation set loss no longer decreases; for example, the regularization coefficient can be 0.0001. Parameter iteration uses an adaptive moment estimation optimization algorithm. The learning rate is used to control the convergence speed and stability. The logic for setting the learning rate is to balance convergence and oscillation suppression in the training of the gated recurrent neural network model; for example, the learning rate can be 0.001. The number of training epochs limits the number of parameter iterations; for example, the number of training epochs can be 100, and the model parameters corresponding to the minimum validation set loss are used as the final model parameters. After training, the final model parameters and model structure are embedded into the controller, forming a pre-established evaluation model used for online calculation of thermal runaway response activity indicators.

[0051] In this way, the evaluation model takes the monitoring dataset as input and outputs thermal runaway reaction activity index, so that the discharge cessation control can judge the internal residual reaction activity based on multidimensional extrapolation results. This ensures the consistency and reliability of the discharge cessation timing judgment under different battery types, different states of charge and different heat propagation paths, reduces the risk of reignition and avoids secondary failures caused by excessive discharge.

[0052] The status generation module is used to compare the thermal runaway reaction activity index with the preset safety threshold, and combine it with one or more data from the monitoring dataset to output the release control status; the release control status includes continue release and prepare to stop.

[0053] During the discharge strategy decision-making phase, a discharge control state is generated based on the comparison between the thermal runaway reactivity index and the safety threshold. Simultaneously, the discharge parameters of the fire-fighting mixture are adaptively adjusted according to the thermal runaway reactivity index, transforming the discharge control from a fixed duration or single temperature threshold into a closed-loop regulation process that adapts to changes in internal residual reactivity. The discharge control state indicates the state transition for subsequent control processes, and its values ​​include continue discharge and prepare to stop.

[0054] The safety threshold is used to define the safe upper limit for stopping the discharge corresponding to the thermal runaway reactivity index. Its setting method is based on installed calibration data combined with sensor noise margin and worst-case safety factor. The installed calibration data consists of no-reignition samples collected in a closed or semi-closed energy storage chamber test environment. The criterion for no-reignition samples is that no reignition risk is detected within an observation window (defined by the observation window length parameter, which is 300 seconds) after the discharge is stopped. For each no-reignition sample, the thermal runaway reactivity index value corresponding to the discharge cessation time is extracted to form a stopped sample set. The upper quantile of the stopped sample set is calculated as the safety threshold benchmark parameter. The upper quantile parameter is used to cover most of the no-reignition stopped samples and avoid excessively high thresholds due to extreme outliers, balancing the risks of missed and false detections. The upper quantile parameter is set to 0.95. To offset the uncertainties introduced by evaluation model errors and sensor noise, a safety margin parameter is introduced based on the safety threshold baseline parameter. This safety margin parameter is obtained by multiplying the noise margin by a safety factor; for example, the noise margin can be set to 5, and the safety factor to 1.2. Thus, the safety threshold equals the safety threshold baseline parameter plus the safety margin parameter. Taking the 0.95 quantile of the stopping sample set as an example (35), the safety threshold is set to 41. Through this setting method, the safety threshold is directly linked to the stopping boundary of samples without reignition, transforming the stopping criterion from an empirical threshold into a safe boundary that can be statistically verified by samples. Furthermore, the safety margin parameter covers the uncertainties under worst-case conditions.

[0055] The controller compares the thermal runaway reactivity index with a safety threshold to generate a discharge control state. When the thermal runaway reactivity index is greater than the safety threshold, the discharge control state is to continue discharging, and adaptive flow adjustment is performed on the fire-fighting mixture discharge parameters. When the thermal runaway reactivity index is not greater than the safety threshold, the controller further combines the quantitative conditions of the monitoring dataset for judgment. If the following conditions are met simultaneously: battery surface temperature meets the safety temperature threshold, smoke concentration meets the smoke removal threshold, characteristic gas concentration meets the gas stability threshold, chamber pressure meets the pressure stability threshold, and battery surface temperature change rate meets the temperature rebound suppression threshold, then the discharge control state is to prepare for stopping; otherwise, the discharge control state is to continue discharging. The safe temperature threshold is used to define the upper limit of the surface temperature in the stop criterion, and it is lower than the common thermal runaway self-sustaining temperature range while allowing for a sensing error margin. For example, the safe temperature threshold can be 60 degrees Celsius. The smoke removal threshold is used to define the upper limit of the residual smoke concentration, and it is lower than the stable baseline of the photoelectric smoke sensor in a smoke-free environment plus a noise margin. For example, the smoke removal threshold can be 0.02 of the sensor's full scale. The gas stability threshold is used to define the trend of change in the characteristic gas concentration, and the concentration change should be close to zero after the gas generation channel is closed, while also considering drift error. For example, the gas stability threshold can be taken as the characteristic gas concentration being within 1... The absolute value of the rate of change within 0 seconds does not exceed 0.005 of the full scale. The pressure stabilization threshold is used to limit the trend of pressure change in the chamber, and the pressure should no longer rise after continuous gas production ends, taking into account the pressure sensor resolution. For example, the pressure stabilization threshold can be set to ensure that the rate of change of the chamber pressure within 10 consecutive seconds does not exceed 0.02 kPa per second. The temperature rebound suppression threshold is used to limit the direction and amplitude of the rate of change of temperature in the stopping criterion, and the temperature should remain decreasing or approximately stable before stopping to avoid rebound precursors. For example, the temperature rebound suppression threshold can be set to ensure that the rate of change of the battery surface temperature within 10 consecutive seconds does not exceed 0.1 degrees Celsius per second. The duration of the aforementioned 10 consecutive seconds is consistent with the feature extraction time window parameter, which is set to 10 seconds to ensure that the judgment window is consistent with the input window of the evaluation model and to avoid judgment deviations caused by different time scales.

[0056] When the discharge control state is set to continue discharging, the controller performs segmented linear adjustment of the discharge flow rate in the fire-fighting mixture discharge parameters based on the thermal runaway reactivity index, while maintaining the discharge pressure above the lower pressure limit to ensure high-pressure precision injection capability. The discharge flow rate adjustment is centered on a baseline discharge flow rate, determined by the system's rated discharge capacity and target cooling requirements; for example, a baseline discharge flow rate of 12 liters per minute is acceptable. The upper limit of the discharge flow rate is used to limit the risk of equipment wetting caused by excessive discharge; for example, an upper limit of 18 liters per minute is acceptable. The lower limit of the discharge flow rate is used to ensure the minimum supply intensity for continuous cooling and chemical inhibition; for example, a lower limit of 6 liters per minute is acceptable. The controller calculates the over-limit ratio parameter as the ratio of the thermal runaway reaction activity index to the safety threshold, and generates a discharge flow rate adjustment coefficient accordingly: when the over-limit ratio parameter is not greater than 1, the discharge flow rate adjustment coefficient is 0; when the over-limit ratio parameter is greater than 1 but not greater than 2, the discharge flow rate adjustment coefficient is the over-limit ratio parameter minus 1; when the over-limit ratio parameter is greater than 2, the discharge flow rate adjustment coefficient is 1. The discharge flow rate setpoint is then calculated as the baseline discharge flow rate multiplied by 1, plus the discharge flow rate gain coefficient multiplied by the discharge flow rate adjustment coefficient, where the discharge flow rate gain coefficient is 0.5. This allows for a small increase in flow rate when the thermal runaway reaction activity index is slightly higher than the safety threshold, and a maximum increase in flow rate when the thermal runaway reaction activity index is significantly higher than the safety threshold. The discharge flow rate setpoint is limited between the lower and upper limits of the discharge flow rate to prevent the discharge flow rate from increasing uncontrollably or becoming too low. The discharge pressure is constrained by the lower pressure limit, which is used to ensure the atomization and penetration capability of the discharge jet. The lower pressure limit is not lower than 0.7 times the pipeline design pressure, for example, the lower pressure limit can be 0.7 MPa. When the discharge pressure in the fire-fighting mixture discharge parameters is detected to be lower than the lower pressure limit, the controller outputs a discharge pressure compensation command to increase the output of the high-pressure power unit until the discharge pressure is not lower than the lower pressure limit.

[0057] When the discharge control state is "continue discharge," the controller outputs a discharge system execution command to continue discharge and enters the next monitoring cycle to reacquire the monitoring data set and update the thermal runaway reaction activity index. When the discharge control state is "preparing to stop," the controller outputs a discharge system execution command to prepare to stop, and uses this discharge control state as the trigger condition for subsequent discharge termination verification. Through the above-mentioned quantitative judgment and adaptive adjustment mechanism, the discharge stop criterion is not only constrained by the thermal runaway reaction activity index, but also by the joint constraints of multi-dimensional stability thresholds such as temperature, smoke, gas, and pressure. This ensures consistency in the discharge stop decision under different battery types, different states of charge, and different heat propagation paths, and reduces the risk of reignition and excessive discharge.

[0058] The reignition assessment module is used to perform a short-term release stop test when the release control state is ready to stop, and output the reignition risk detection results.

[0059] During the discharge termination verification phase, when the discharge control status is "ready to stop," the controller acquires the latest monitoring dataset as the initial data for entering the discharge termination verification phase and outputs a short-term discharge stop test command to the discharge system. This causes the discharge system to temporarily interrupt the discharge of the fire-fighting mixture within a short-term discharge stop window. The short-term discharge stop window is used to actively verify the residual reactivity before stopping the discharge, covering the early rebound range most likely to occur after the discharge stops without significantly reducing the continuity of fire suppression. For example, the short-term discharge stop window can be 30 seconds. Within the short-term discharge stop window, the controller continuously updates the monitoring dataset according to a uniform sampling period and generates a reignition risk detection result based on the multi-dimensional change criteria within the short-term discharge stop window. The reignition risk detection result is used to drive the state transition of the discharge system to execute the command.

[0060] Reignition signs are used to characterize the externally observable response when the residual internal reactivity regains dominance after the discharge is interrupted. Reignition signs include a renewed rise in battery surface temperature, a renewed rise in cabin ambient temperature, a renewed increase in smoke concentration, a renewed increase in characteristic gas concentration, a continued rise in cabin pressure, and the reappearance of abnormal sounds. To avoid misinterpreting sensor noise or short-term fluctuations as signs of reignition, the controller employs a combined constraint of a rate of change threshold and a cumulative change threshold for each reignition sign within the short-term discharge stop window: A resurgence of battery surface temperature is determined by both a surface rebound rate threshold and a surface rebound amplitude threshold. The surface rebound rate threshold is higher than the maximum false slope caused by noise from the battery surface temperature sensor and covers the minimum identifiable rebound speed; for example, the surface rebound rate threshold can be set to 0.1 degrees Celsius per second. The surface rebound amplitude threshold is higher than the measurement error of the battery surface temperature sensor and reflects the visible temperature rise; for example, the surface rebound amplitude threshold can be set to 2 degrees Celsius. A resurgence of battery surface temperature is determined when, within any consecutive 5 seconds within the short-term discharge stop window, the rate of change of battery surface temperature exceeds the surface rebound rate threshold and the cumulative increase in battery surface temperature relative to the interruption time within the short-term discharge stop window exceeds the surface rebound amplitude threshold. The re-rise of the cabin ambient temperature is determined by the combined thresholds of the ambient rebound rate and the ambient rebound amplitude. For example, the ambient rebound rate threshold can be set to 0.05 degrees Celsius per second, and the ambient rebound amplitude threshold can be set to 1 degree Celsius. The determination method is the same as that for the re-rise of the battery surface temperature.

[0061] A resurgence of smoke concentration is determined by a smoke rise amplitude threshold. This threshold is higher than the baseline fluctuation of smoke concentration in a smoke-free state and reflects a new trend in smoke generation. For example, the smoke rise amplitude threshold can be set to 0.02 of the full scale of the smoke sensor. A resurgence of smoke concentration is determined when the cumulative increase in smoke concentration relative to the interruption time within the short-term discharge stop window exceeds the smoke rise amplitude threshold. A resurgence of characteristic gas concentration is determined by a gas rise amplitude threshold. This threshold is higher than the drift of the characteristic gas concentration sensor and reflects a renewed enhancement of the gas generation channel. For example, the gas rise amplitude threshold can be set to 0.01 of the full scale of the characteristic gas concentration sensor. A resurgence of characteristic gas concentration is determined when the cumulative increase in characteristic gas concentration relative to the interruption time within the short-term discharge stop window exceeds the gas rise amplitude threshold. The continued rise in cabin pressure is determined by a pressure rise rate threshold. This threshold is higher than the spurious changes caused by the pressure sensor resolution and reflects the net pressurization caused by continuous gas production. For example, the pressure rise rate threshold can be set to 0.01 kPa per second. If the cabin pressure change rate exceeds the pressure rise rate threshold for any consecutive 5 seconds within the short-term discharge stop window, it is determined that the cabin pressure continues to rise. The recurrence of abnormal sounds is determined by an abnormal sound count threshold. This threshold excludes background noise triggers and captures repeated occurrences of safety valve ejection or structural rupture. For example, the abnormal sound count threshold can be set to 2. If the number of triggers of the abnormal sound detection result within the short-term discharge stop window is not less than the abnormal sound count threshold, it is determined that an abnormal sound has recurred.

[0062] The controller generates a reignition risk detection result based on the judgment results within the short-term discharge stop window. When any reignition sign is determined to be valid within the short-term discharge stop window, the reignition risk detection result is "reignition risk present"; when none of the reignition signs are determined to be valid within the short-term discharge stop window, the reignition risk detection result is "no reignition risk." The reignition risk detection result is then output as input for subsequent discharge stop or restart decisions, making discharge termination verification a verifiable control action before stopping discharge, thereby reducing the probability of false judgments regarding discharge stop caused by relying solely on the drop in battery surface temperature or a fixed discharge duration.

[0063] The instruction execution module is used to output and execute the discharge execution instructions based on the reignition risk detection results.

[0064] During the decision-making phase for stopping or restarting the fire suppression system, the controller receives the reignition risk detection result and generates a fire suppression execution command within the corresponding decision window. This command directly controls the start / stop of the fire suppression system and the state transition of the fire suppression control status, ensuring that the determination result from the fire suppression termination verification phase is applied to the fire suppression mixture fire suppression process through a deterministic control loop. The decision window is designed to prevent frequent start / stop of the fire suppression system due to short-term fluctuations in the reignition risk detection result. The decision window is set to cover the shortest stable confirmation time after the reignition risk detection result is output without significantly delaying the reignition suppression response; for example, a 2-second decision window is acceptable. Within the decision window, the controller performs stability verification on the reignition risk detection result. When the reignition risk detection result within the decision window remains consistent, the corresponding fire suppression execution command is output.

[0065] When the reignition risk detection result indicates a reignition risk, the controller outputs a resumed discharge execution command and sets the discharge control state to continue discharging, causing the discharge system to restart the fire-fighting mixture discharge and restore high-pressure injection. Simultaneously, the controller uses the current monitoring dataset as the starting data for subsequent evaluations and continues to acquire monitoring datasets, update thermal runaway reaction activity indicators, and execute discharge strategy decisions and discharge termination verification in the next sampling cycle. This forms a recurring cycle of monitoring dataset updates and discharge control state determination until the reignition risk detection result indicates no reignition risk, at which point the final discharge stop is allowed. To avoid repeated stopping and starting of the discharge system in a short period of time due to frequent reignition risks, the controller enters a minimum continuous discharge duration constraint after outputting the discharge execution command to resume discharge. The minimum continuous discharge duration is used to ensure sufficient continuous cooling and chemical inhibition time after the discharge is resumed to suppress residual internal reactivity. The setting logic of the minimum continuous discharge duration is to cover the typical duration range of thermal runaway rebound and take into account the consumption of fire-fighting mixture. For example, the minimum continuous discharge duration can be set to 60 seconds. Within the minimum continuous discharge duration, the controller maintains the discharge control state as continuous discharge and does not allow the generation of preparation to stop, so as to reduce the accumulation of reignition risk caused by premature re-stopping.

[0066] When the reignition risk detection result indicates no reignition risk, the controller outputs a stop-discharge execution command, causing the sprinkler system to shut down the fire-fighting mixture discharge and stop the high-pressure power unit output. Simultaneously, the discharge control state is set to stop to lock the control process in its terminated state. To avoid missed detections due to hysteresis after the discharge stops, the controller enters a post-stop monitoring hold period after outputting the stop-discharge execution command. This post-stop monitoring hold period is used to: continuously acquire monitoring data and confirm the fire has stabilized and subsided under the condition that the fire-fighting mixture is no longer being sprayed. The logic for setting the post-stop monitoring hold period is to cover the possible hysteresis rebound process after the discharge stops without significantly increasing system occupancy; for example, a 300-second post-stop monitoring hold period is acceptable. During this period, the controller continuously acquires monitoring data and updates the reignition risk detection result. If the reignition risk detection result changes to a reignition risk again within the post-stop monitoring hold period, the controller outputs a resume-discharge execution command to achieve a safety backup after the discharge stops. When the monitoring period ends after the fire suppression is stopped and the risk of reignition remains at zero, the controller outputs a post-fire cleanup procedure start command to initiate ventilation, exhaust, and residual heat cooling to end the fire suppression process.

[0067] When the reignition risk detection result indicates a reignition risk and stability is confirmed, a reignition execution command to resume discharge is output. When the reignition risk detection result indicates no reignition risk and stability is confirmed, a reignition execution command to stop discharge is output. The discharge execution command serves as the final control signal of the discharge system, driving the discharge system to continue operating or to shut down safely. Together with the monitoring dataset, thermal runaway reaction activity index, discharge control status, and reignition risk detection result, it forms a closed-loop control link, thereby reducing the probability of reignition and suppressing secondary damage caused by excessive discharge.

[0068] Through the above steps, this embodiment achieves intelligent control of the lithium battery thermal runaway extinguishing process: it utilizes multi-sensor data fusion to assess internal reaction activity, dynamically decides the discharge strategy, verifies safety through brief stop tests, and ultimately ensures that discharge is only stopped when the internal exothermic reaction is completely under control. This innovative method effectively avoids misjudgments that may occur based solely on surface temperature or a fixed duration, reducing the probability of thermal runaway reignition. It also prevents secondary damage to the equipment caused by excessive spraying of extinguishing agents, ensuring the reliability of fire extinguishing while protecting the energy storage device.

[0069] Please see Figure 2 As shown, this embodiment provides a high-pressure precision injection control method for energy storage fire-fighting mixture, including:

[0070] The system acquires battery surface temperature, cabin ambient temperature, smoke concentration, characteristic gas concentration, cabin air pressure, battery module deformation, and sound signals, performs time synchronization and format standardization processing, and outputs the monitoring dataset for the current moment.

[0071] The monitoring dataset is input into a pre-established evaluation model. Based on the temperature, gas, smoke, pressure deformation, and sound factors obtained from the monitoring dataset, the residual reactivity inside the battery cell is calculated, and a thermal runaway reactivity index is generated.

[0072] The thermal runaway reaction activity index is compared with a preset safety threshold, and combined with one or more data points from the monitoring dataset, to output the release control status; the release control status includes continue release and prepare to stop.

[0073] When the discharge control status is ready to stop, perform a short-term discharge stop test and output the reignition risk detection result;

[0074] Based on the reignition risk detection results, output and execute the discharge execution command.

[0075] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0076] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for high-pressure precision injection control of energy storage fire-fighting mixture, characterized in that, include: The system acquires battery surface temperature, cabin ambient temperature, smoke concentration, characteristic gas concentration, cabin air pressure, battery module deformation, and sound signals, performs time synchronization and format standardization processing, and outputs the monitoring dataset for the current moment. The monitoring dataset is input into a pre-established evaluation model. Based on the temperature, gas, smoke, pressure deformation, and sound factors obtained from the monitoring dataset, the residual reactivity inside the battery cell is calculated, and a thermal runaway reactivity index is generated. The thermal runaway reaction activity index is compared with the preset safety threshold, and the release control status is output by combining the monitoring dataset. The discharge control status includes continue discharge and prepare to stop; When the discharge control status is ready to stop, perform a short-term discharge stop test and output the reignition risk detection result; Based on the reignition risk detection results, output and execute the discharge execution command.

2. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 1, characterized in that, Temperature factors include the extent to which the battery surface temperature exceeds the safety threshold and the rate of change of the battery surface temperature; gas factors include the deviation of the characteristic gas concentration from the warning value; smoke factors include the amount of change in smoke concentration; pressure deformation factors include the amount of change in cabin air pressure and battery module deformation; and sound factors include the detection results of abnormal noises in the sound signal.

3. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 1, characterized in that, The evaluation model adopts a gated recurrent neural network model, which models the temporal correlation of the input monitoring dataset through a gated recurrent unit layer, and then outputs the thermal runaway response activity index through a feature fusion fully connected layer.

4. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 1, characterized in that, When training and evaluating the model, whether there is a temperature rebound and a rise in the concentration of characteristic gases within the observation window after the discharge stops is used as the labeling rule for reignition judgment, thereby generating training labels for thermal runaway reaction activity indicators to make the training target consistent with the control target. The data acquisition environment during the training of the evaluation model is limited to the semi-enclosed state of the energy storage cabin.

5. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 1, characterized in that, Methods for performing short-term discharge stop tests include: The discharge of the fire-fighting mixture is temporarily interrupted, and the monitoring dataset is continuously collected and updated during the short-term discharge stop test. If any sign of reignition occurs during the short-term discharge stop test, the reignition risk detection result is output as "there is a risk of reignition"; if no sign of reignition occurs during the short-term discharge stop test, the reignition risk detection result is output as "there is no risk of reignition".

6. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 5, characterized in that, Signs of reignition include a renewed rise in battery surface temperature, a renewed rise in cabin ambient temperature, a renewed increase in smoke concentration, a renewed increase in the concentration of characteristic gases, a continued rise in cabin air pressure, or the reappearance of abnormal sounds.

7. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 1, characterized in that, When the thermal runaway reaction activity index is higher than the preset safety threshold, the release control state is to continue releasing; when the thermal runaway reaction activity index is not higher than the preset safety threshold and the monitoring data set meets the conditions for stopping the release, the release control state is to prepare to stop.

8. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 1, characterized in that, When the reignition risk detection result indicates a reignition risk, the output discharge execution command is to resume discharge; when the reignition risk detection result indicates no reignition risk, the output discharge execution command is to stop discharge.

9. The high-pressure precision injection control method for energy storage fire-fighting mixture according to claim 8, characterized in that, After the output resumes the discharge execution command, it enters the minimum continuous discharge duration constraint. Within the minimum continuous discharge duration, the controller maintains the discharge control state as continue discharging and does not allow the generation of preparation to stop.

10. A high-pressure precision injection control system for energy storage fire-fighting mixture, used to implement the high-pressure precision injection control method for energy storage fire-fighting mixture as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire battery surface temperature, cabin ambient temperature, smoke concentration, characteristic gas concentration, cabin air pressure, battery module deformation, sound signals, and fire-fighting liquid discharge parameters, and to perform time synchronization and format standardization processing to output the monitoring dataset at the current moment. The activity calculation module is used to input the monitoring dataset into the pre-established evaluation model, and calculate the residual reactivity inside the cell based on the temperature, gas, smoke, pressure deformation and sound factors obtained from the monitoring dataset, and generate thermal runaway reactivity index. The status generation module is used to compare the thermal runaway reaction activity index with the preset safety threshold, and combine one or more data from the monitoring dataset to output the release control status. The discharge control status includes continue discharge and prepare to stop; The reignition assessment module is used to perform a short-term release stop test when the release control state is ready to stop, and output the reignition risk detection results. The instruction execution module is used to output and execute the discharge execution instructions based on the reignition risk detection results.

Images