Energy storage safety automatic fire extinguishing method and system

Abstract

The application discloses a kind of energy storage safety automatic fire extinguishing method and system, belong to thermal management technical field.Method includes: the inner wall side temperature of battery pack, outer wall side temperature and fire characteristic parameter in battery pack are detected in real time;The inner wall side temperature, outer wall side temperature and fire characteristic parameter are dynamically weighted fusion calculation, and fusion result is generated;Based on the fusion result, fire-fighting action is executed.The application can break the limitation of single monitoring mode, greatly improve detection accuracy, significantly reduce false alarm rate.

Description

Technical Field

[0001] This invention relates to the field of thermal management technology, specifically to an automatic fire extinguishing method and system for energy storage safety. Background Technology

[0002] With the rapid development of electrochemical energy storage technology, lithium-ion batteries are widely used in energy storage power stations due to their high energy density and long cycle life. However, under abnormal conditions such as overcharging, short circuits, or thermal runaway, batteries are prone to fires or even explosions, seriously threatening the safe operation of energy storage systems. Therefore, early fire warning for energy storage systems has become a focus of industry attention.

[0003] Currently, common fire monitoring methods for energy storage systems are mainly divided into two categories: temperature-sensor-based monitoring and battery pack internal environmental parameter monitoring. However, these two methods are often used independently or simply in parallel, which has obvious limitations: temperature-sensor-based monitoring, while able to monitor the overall temperature change of the battery cluster, cannot accurately identify early anomalies inside the battery pack, leading to delayed warnings and difficulty in responding in time before thermal runaway; monitoring only the internal environmental parameters of the battery pack, while capable of detecting parameters such as CO concentration, is susceptible to environmental interference, resulting in false alarms and weak risk identification capabilities; simply connecting the two monitoring methods in parallel can improve the monitoring coverage to some extent, but it still fails to achieve collaborative analysis of multi-source information and deep integration of alarm strategies, easily leading to information silos and the inability to cross-validate based on multi-dimensional information, thus posing a risk of false alarms or missed alarms. Summary of the Invention

[0004] To address the technical problems existing in the background art, this invention proposes an automatic fire extinguishing method for energy storage safety, comprising:

[0005] S1. Real-time detection of the inner wall temperature, outer wall temperature, and fire characteristic parameters within the battery pack;

[0006] S2. Perform dynamic weighted fusion calculation on the inner wall temperature, outer wall temperature and fire characteristic parameters to generate fusion results;

[0007] S3. Perform fire-fighting actions based on the fusion results;

[0008] The fire characteristic parameters include CO concentration, smoke particle concentration, and the condition of the heat-sensing cable. S2 specifically refers to:

[0009] S21 assigns dynamic confidence weights to the inner wall temperature, outer wall temperature, CO concentration, and smoke particle concentration, respectively.

[0010] S22. Based on the assigned dynamic confidence weights, the inner wall temperature and outer wall temperature are preprocessed to obtain the processed inner wall temperature and outer wall temperature. The processed inner wall temperature and outer wall temperature are then verified both temporally and spatially to generate a temperature gradient anomaly determination result.

[0011] S23. Perform multi-level decision fusion on the temperature gradient anomaly judgment result, the temperature sensing cable status, and the CO concentration and smoke particle concentration with assigned dynamic confidence weights to generate the fusion result.

[0012] This invention overcomes the limitations of single temperature or smoke detection by fusing multiple parameters, namely the temperature of the inner wall side, the temperature of the outer wall side, and fire characteristic parameters, significantly improving the comprehensiveness of fire early warning and early identification capabilities; dynamic weighted fusion calculation can adjust the importance of different parameters according to the actual situation, enhancing the accuracy and adaptability of the judgment; and finally, actions are executed based on the fusion results, realizing intelligent closed-loop control from perception to execution.

[0013] This invention introduces dynamic confidence weights, which can flexibly respond to changes in the reliability of various parameters under different operating conditions; and through dual verification of time and space, temperature data is cross-checked from two dimensions of time and physical location, which can effectively distinguish between real thermal runaway inside the battery and external environmental interference or instantaneous fluctuations, thereby reducing the false alarm rate of temperature anomalies.

[0014] Preferably, S22 specifically includes:

[0015] S221. Based on the assigned dynamic confidence weights, adaptive filtering is performed on the inner wall temperature and the outer wall temperature to obtain the filtered inner wall temperature and outer wall temperature.

[0016] S222. Based on the filtered inner wall temperature and outer wall temperature, calculate the time-series temperature gradient of each battery pack, where the time-series temperature gradient represents the rate of change of the inner wall temperature and outer wall temperature over time; and calculate the spatial temperature gradient of each battery pack, where the spatial temperature gradient represents the rate of change of the temperature difference obtained by subtracting the outer wall temperature from the inner wall temperature of the same battery pack over time.

[0017] S223. The temporal and spatial temperature gradients of the same battery pack are analyzed using a convolutional neural network feature extraction algorithm, and temperature anomaly scores for each battery pack are generated as the output of temperature gradient anomaly judgment results.

[0018] The adaptive filtering of this invention effectively smooths out data noise, providing a high-quality data foundation for subsequent analysis; calculating temporal and spatial gradients separately can capture different types of thermal anomaly patterns, such as slow temperature rise and instantaneous surge, internal overheating and external overheating; and using convolutional neural networks for feature extraction and scoring can automatically learn complex, nonlinear temperature anomaly patterns, which is more intelligent and accurate than simply setting a fixed threshold, and the anomaly scoring results are more reliable.

[0019] Preferably, S23 specifically includes:

[0020] S231. Construct a multi-dimensional feature vector based on temperature anomaly score, temperature sensing cable status, and CO concentration and smoke particle concentration with assigned dynamic confidence weights.

[0021] S232. Input the multidimensional feature vector into a predefined decision maker based on multi-level threshold rules;

[0022] S233. The decision-maker performs logical decision-making on the multi-dimensional feature vector according to the preset multi-level threshold rules, and generates and outputs the fire risk probability value as the fusion result.

[0023] This invention achieves deep fusion and decision-making of multi-source heterogeneous information: it constructs a multi-dimensional feature vector by combining temperature anomaly scores with other fire characteristic parameters, providing a more comprehensive description of the fire state; it performs logical decision-making based on a multi-level threshold rule, which is intuitive, stable, easy to implement and adjust, and finally outputs a quantified fire risk probability value, providing a clear and accurate basis for subsequent graded response.

[0024] Preferably, the decision maker based on multi-level threshold rules is configured to execute the following decision logic:

[0025] When the temperature anomaly score of a certain battery pack is greater than the preset temperature anomaly score threshold, a low fire risk probability value is output, ranging from [0, 0.4].

[0026] When the CO concentration in a battery pack with assigned dynamic confidence weights is greater than a preset CO concentration threshold, or the smoke particle concentration with assigned dynamic confidence weights is greater than a preset smoke particle concentration threshold, the fire risk probability value is output, ranging from (0.4, 0.7].

[0027] When the temperature anomaly score of a certain battery pack is greater than the preset temperature anomaly score threshold, and the CO concentration and smoke particle concentration in the battery pack, which are assigned dynamic confidence weights, are both greater than their corresponding preset thresholds, or when the temperature sensing cable alarms, a high fire risk probability value is output, with a range of (0.7, 1).

[0028] This invention establishes a cross-validated fire prevention mechanism through the linkage of multiple parameters, namely the temperature of the inner and outer walls of the battery pack, CO concentration, smoke particle concentration, and the alarm signal from the temperature-sensing cable. Multi-level threshold rules set trigger conditions for different risk levels: low risk is triggered only by temperature anomalies, achieving early warning; medium risk is triggered by gas or smoke, indicating the presence of combustion products; high risk is triggered by simultaneous anomalies in multiple parameters or by the temperature-sensing cable alarm, indicating open flame or high temperature. This design achieves graded judgment of early fire warning, preliminary confirmation, and emergency response, effectively balancing the timeliness and accuracy of warnings and avoiding false alarms caused by fluctuations in a single parameter.

[0029] Preferably, S3 specifically comprises:

[0030] S31. Determine the alarm level based on the fire risk probability value;

[0031] S32. Locate the battery cluster or battery pack that triggered the alarm based on the alarm level;

[0032] S33. Perform the corresponding fire-fighting actions based on the alarm level and location results.

[0033] This invention determines alarm levels based on quantified risk probability values, ensuring that response measures match the severity of fire risks and avoiding either overreaction or underreaction due to a "one-size-fits-all" approach. Accurately locating faulty battery clusters or packs is a prerequisite for precise fire suppression, minimizing the amount of extinguishing agent used and its impact on normal battery cells.

[0034] Preferably, the alarm levels include alarm levels one to three, and S31 specifically includes:

[0035] When the fire risk probability value falls within the low fire risk probability value range, a level one alarm is triggered.

[0036] When the fire risk probability value falls within the range of medium fire risk probability value, a level two alarm is triggered;

[0037] When the fire risk probability value falls within the high fire risk probability value range, a level three alarm is triggered.

[0038] This invention maps continuous risk probability values ​​to discrete, explicit alarm levels, making response strategies clearer and more standardized.

[0039] Preferably, S32 specifically includes:

[0040] If a Level 1 alarm is triggered, the battery cluster containing the battery pack whose temperature anomaly score is greater than the preset temperature anomaly score threshold is identified as a risky battery cluster.

[0041] If a level 2 alarm is triggered, the battery packs with CO concentrations assigned dynamic confidence weights that are greater than a preset CO concentration threshold or smoke particle concentrations assigned dynamic confidence weights that are greater than a preset smoke particle concentration threshold are identified as risky battery packs.

[0042] If a Level 3 alarm is triggered, the target battery pack is the one whose temperature anomaly score is greater than the preset temperature anomaly score threshold, whose CO concentration and smoke particle concentration, both assigned with dynamic confidence weights, are greater than their corresponding preset thresholds, or whose temperature-sensing cable alarm is triggered.

[0043] This invention achieves hierarchical precise positioning: Level 1 alarms locate clusters, covering a larger area, for enhanced monitoring; Level 2 alarms locate packets, narrowing the scope, for preparing for intervention; Level 3 alarms precisely locate target packets, enabling accurate strikes. This hierarchical focusing positioning strategy has a clear logic and perfectly matches the severity of risk and the required response actions at each level.

[0044] Preferably, S33 specifically includes:

[0045] When the alarm level is Level 1, increase the frequency of collecting the inner and outer wall temperatures of each battery pack within the risky battery cluster.

[0046] When the alarm level is level two, the charging and discharging of the at-risk battery pack will be cut off.

[0047] When the alarm level is level three, fire suppression measures are taken against the target battery pack.

[0048] This invention forms a complete closed loop from perception to decision-making to execution: response measures are escalated step by step and matched with risk levels: Level 1 only strengthens monitoring without causing damage; Level 2 implements electrical isolation to cut off the energy source and prevent deterioration; Level 3 initiates physical fire suppression to directly extinguish the fire, reflecting the safety concept of "prevention first and graded handling". It can both suppress the development of faults in the early stage and effectively extinguish fires in emergency situations, maximizing the safety of the energy storage system.

[0049] The present invention also provides an energy storage safety automatic fire extinguishing system, comprising:

[0050] The detection module is used to detect the inner wall temperature, outer wall temperature, and fire characteristic parameters inside the battery pack in real time.

[0051] The main control module is used to perform dynamic weighted fusion calculations on the inner wall temperature, outer wall temperature, and fire characteristic parameters to generate fusion results;

[0052] The execution module is used to perform fire-fighting actions based on the fusion results;

[0053] The fire characteristic parameters include CO concentration, smoke particle concentration, and temperature sensing cable status. The main control module specifically assigns dynamic confidence weights to the inner wall temperature, outer wall temperature, CO concentration, and smoke particle concentration. Based on the assigned dynamic confidence weights, the inner wall temperature and outer wall temperature are preprocessed to obtain the processed inner wall temperature and outer wall temperature. The processed inner wall temperature and outer wall temperature are then verified both temporally and spatially to generate a temperature gradient anomaly judgment result.

[0054] The results of temperature gradient anomaly determination, temperature sensing cable status, and CO concentration and smoke particle concentration with assigned dynamic confidence weights are fused into a multi-level decision to generate a fusion result.

[0055] Compared with existing technologies, the advantages of this invention are that it constructs an integrated proactive safety defense system that combines multi-parameter fusion, intelligent decision-making, hierarchical early warning, and precise handling: by collecting the temperature of the inner / outer wall of the battery pack and various fire characteristic parameters in real time, and using dynamic weighted fusion calculation instead of relying on a single threshold judgment, it greatly improves the perception and anti-interference capabilities of complex faults and early fires, and can significantly reduce the risk of false alarms and missed alarms. Attached Figure Description

[0056] Figure 1 This is a flowchart of Embodiment 1 of the present invention;

[0057] Figure 2 This is a system block diagram of the system in Embodiment 2 of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1

[0060] like Figure 1 As shown, this embodiment provides an automatic fire extinguishing method for energy storage safety, including the following steps:

[0061] S1. Real-time monitoring of the inner wall temperature, outer wall temperature, and fire characteristic parameters within the battery pack. Specifically:

[0062] Temperature is collected on the inner and outer walls of each battery pack using temperature-sensing optical fibers arranged within the battery clusters. The fibers pass through all battery packs within the cluster in the order they are arranged, employing a through-pack design and routing within the packs. In this embodiment, there are 12 battery clusters, each containing 8 battery packs, and the positioning accuracy of the temperature-sensing optical fibers is 0.5m. Twelve temperature-sensing optical fibers are deployed within the 12 battery clusters to collect temperature data on the inner and outer walls of each battery pack, particularly at the critical connection points between the busbars and tabs.

[0063] The CO concentration within the battery pack is collected by a CO sensor located within each battery pack, and the smoke particle concentration is collected by a smoke sensor located within each battery pack. In this embodiment, the CO sensor has a measurable range of 0~3000ppm, and the smoke sensor has a measurable range of 0~300mg / m³. 3 Each battery pack is also equipped with a temperature-sensing cable, which sends an alarm signal carrying the battery pack's address identifier when the temperature inside the battery pack reaches a preset internal temperature threshold. The preset internal temperature threshold range is... In this embodiment, The collected CO concentration, smoke particle concentration, and temperature sensing cable status within the battery pack are the fire characteristic parameters of the battery pack.

[0064] The internal and external wall temperatures of the battery pack, as well as fire characteristic parameters within the battery pack, are synchronously collected once per second. The internal wall temperature of the i-th battery pack at time t is denoted as... The temperature on the outer wall side is CO concentration is The concentration of smoke particles is The temperature sensing cable is in condition. , It is 0 or 1. A value of 0 indicates that the temperature sensing cable of the i-th battery is normal at time t, while a value of 1 indicates that the temperature sensing cable of the i-th battery alarms at time t, meaning that its internal temperature sensing cable emits an alarm signal.

[0065] S2. Perform dynamic weighted fusion calculation on the inner wall temperature, outer wall temperature, and fire characteristic parameters to generate a fusion result. Specifically:

[0066] S21 assigns dynamic confidence weights to the inner wall temperature, outer wall temperature, CO concentration, and smoke particle concentration, respectively. Specifically:

[0067] S211. Calculate the original dynamic confidence weights for the inner wall temperature, outer wall temperature, CO concentration, and smoke particle concentration independently. The weights are based on the data stability (standard deviation) of each parameter within a preset time window. Calculate the standard deviation. The calculation formula is: Where N is the total number of samples, This represents the value of the m-th sample. This represents the average of all sample values. The smaller the standard deviation, the more stable the data, and the higher the original weight.

[0068] In this embodiment, the preset time window is 30 seconds, and the original weights of the inner wall temperature, outer wall temperature, CO concentration, and smoke particle concentration of the i-th battery pack at time t are as follows:

[0069] ;

[0070] in, , , , The values ​​represent the standard deviations of the inner wall temperature, outer wall temperature, CO concentration, and smoke particle concentration of the i-th battery pack over the past 30 seconds at time t. It is a very small value, such as 0.001, which is used to prevent division by zero errors.

[0071] S212. Normalize the original weights of the inner wall temperature and the outer wall temperature to obtain the final dynamic confidence weights, which are also the weight coefficients used for filtering: The filtering weight coefficients of the inner wall temperature and the outer wall temperature of the i-th battery pack at time t are respectively , ,in .

[0072] The original weights for CO concentration and smoke particle concentration are normalized to obtain the final dynamic confidence weights, which are also the weights used for feature weighting: the feature weighting coefficients for CO concentration and smoke particle concentration inside the i-th battery pack at time t are respectively... , ,in .

[0073] Globally normalizing the inner and outer wall temperatures, CO concentration, and smoke particle concentration would lead to inappropriate mixing and comparison of parameters with different physical meanings. This invention employs a grouping approach, rather than simply combining all four parameters for global normalization, which is both reasonable and efficient.

[0074] S22. Based on the assigned final dynamic confidence weights, the inner wall temperature and outer wall temperature are preprocessed to obtain the processed inner wall temperature and outer wall temperature. The processed inner wall temperature and outer wall temperature are then subjected to both temporal and spatial verification to generate a temperature gradient anomaly determination result. Specifically:

[0075] S221. Based on the assigned weight coefficients for filtering, adaptive filtering is performed on the inner wall temperature and outer wall temperature of the i-th battery pack at time t, resulting in the filtered inner wall temperature. ,in Let be the filtered inner wall temperature of the i-th battery pack at time (t-1); let be the filtered outer wall temperature. ,in Let be the filtered outer wall temperature of the i-th battery pack at time (t-1).

[0076] S222. Based on the filtered inner wall temperature and outer wall temperature, calculate the time-series temperature gradient of each battery pack, where the time-series temperature gradient represents the rate of change of the inner wall temperature and outer wall temperature over time. Also calculate the spatial temperature gradient of each battery pack, where the spatial temperature gradient represents the rate of change of the temperature difference between the inner wall temperature and the outer wall temperature of the same battery pack over time. Specifically:

[0077] The time-series temperature gradient of the i-th battery pack includes the rate of change of temperature on the inner wall side and the rate of change of temperature on the outer wall side, and its rate of change of temperature on the inner wall side at time t. The calculation formula is: ,in Let be the filtered inner wall temperature of the i-th battery pack at time (tn); and be the rate of change of the outer wall temperature at time t. The calculation formula is: ,in Let be the filtered outer wall temperature of the i-th battery pack at time (tn).

[0078] Spatial temperature gradient of the i-th battery pack at time t The calculation formula is: .

[0079] This invention introduces the rate of change of temperature difference between the inner and outer walls, i.e., the spatial temperature gradient, as a key parameter. By utilizing the thermal resistance characteristics of the battery pack casing itself, it can visualize external environmental interference, thermal runaway of the battery pack itself, and thermal conduction of adjacent packs, thereby significantly improving detection accuracy and reducing false alarm rate.

[0080] S223. A convolutional neural network feature extraction algorithm is used to analyze the temporal and spatial temperature gradients of the same battery pack, generating temperature anomaly scores for each battery pack as the output of the temperature gradient anomaly determination result. Specifically:

[0081] The most recent time period for the i-th battery pack is 60 seconds in this embodiment. , , Three time series data points are combined to form a 3×60 matrix, which serves as the input to a pre-trained CNN model.

[0082] The CNN model is trained using a large amount of historical normal data and thermal runaway experimental data to identify temperature rise patterns that characterize thermal runaway, such as a much higher temperature rise rate on the inner wall side than on the outer wall side, and a continuously and rapidly increasing temperature difference between the inner and outer sides. The model ultimately outputs a value between 0 and 1. This value is the temperature anomaly score of the i-th battery pack, which is output as the temperature gradient anomaly determination result.

[0083] S23. A multi-level decision fusion is performed on the temperature gradient anomaly judgment result, the temperature sensing cable status, and the CO concentration and smoke particle concentration assigned dynamic confidence weights to generate a fusion result. Specifically:

[0084] S231. Based on the temperature anomaly score, the status of the temperature sensing cable, and the CO concentration and smoke particle concentration with dynamically assigned confidence weights, construct a multi-dimensional feature vector for the i-th battery pack: ;

[0085] S232. Input the multidimensional feature vector into a predefined decision maker based on multi-level threshold rules;

[0086] S233. The decision-maker makes the following logical decision on the multi-dimensional feature vector according to a preset multi-level threshold rule:

[0087] When the temperature of the i-th battery pack is abnormal, a score is given. When the temperature exceeds the preset abnormal temperature scoring threshold, a low fire risk probability value P1 is output. The range is [0, 0.4].

[0088] When the i-th battery pack is Greater than the preset CO concentration threshold Th co or Greater than the preset smoke particle concentration threshold Th smoke At that time, the fire risk probability value P2 is output. The range is (0.4, 0.7].

[0089] When the temperature of the i-th battery pack is abnormal, a score is given. The temperature exceeds the preset abnormal temperature scoring threshold, and the battery pack is inside... and When all values ​​exceed their corresponding preset thresholds, or when a temperature sensing cable alarm signal is present, i.e. When the value is 1, the output is a high fire risk probability value P3. The range is (0.7,1).

[0090] In this embodiment, the preset temperature anomaly scoring threshold is 0.6, and the preset CO concentration threshold is Th. co The preset smoke particle concentration threshold Th is 50 ppm. smoke 50mg / m 3 .

[0091] The decision-maker ultimately generates and outputs the fire risk probability value as the fusion result.

[0092] S3. Execute fire-fighting actions based on the fusion results. Specifically:

[0093] S31. Determine the alarm level based on the fire risk probability value:

[0094] When the judgment device outputs a fire risk probability value for a certain battery pack that falls within the range of P1, a level one alarm is triggered.

[0095] When the judgment device outputs a fire risk probability value for a certain battery pack that falls within the range of P2, a level 2 alarm is triggered.

[0096] When the judgment device outputs a fire risk probability value for a certain battery pack that falls within the range of P3, a level 3 alarm is triggered.

[0097] S32. Locate the battery cluster or battery pack that triggered the alarm based on the alarm level. Specifically:

[0098] If a Level 1 alarm is triggered, the battery cluster containing the temperature-sensing fiber optic cable with a temperature anomaly score greater than the preset temperature anomaly score threshold of 0.6 is identified as a risky battery cluster.

[0099] It should be noted that in this embodiment, temperature data acquisition relies on temperature-sensing optical fibers arranged in clusters. While these fibers have area-based location capabilities, in the first-level alarm stage, temperature anomaly scoring alone cannot accurately distinguish which specific battery pack within the same fiber segment is the source of the anomaly. Therefore, in this embodiment, the location granularity for the first-level alarm is the battery cluster.

[0100] For embodiments employing other temperature monitoring schemes (such as independently setting temperature sensors with unique address identifiers in each battery pack), if they have precise positioning capabilities at the battery pack level, they can directly locate the specific battery pack with an abnormal temperature score exceeding the standard and identify the battery cluster containing that battery pack as a risky battery cluster.

[0101] If a level 2 alarm is triggered, it indicates that the CO concentration within the battery pack, assigned with dynamic confidence weights, exceeds a preset CO concentration threshold of 50 ppm, or the smoke particle concentration, assigned with dynamic confidence weights, exceeds a preset smoke particle concentration threshold of 50 mg / m³. 3 The battery pack is a high-risk battery pack;

[0102] If a Level 3 alarm is triggered, the target battery pack is the one whose temperature anomaly score is greater than the preset temperature anomaly score threshold, whose CO concentration and smoke particle concentration, both assigned with dynamic confidence weights, are greater than their corresponding preset thresholds, or whose battery pack issues a temperature sensing cable alarm signal.

[0103] For the Level 3 alarm, if triggered based on the former, since this embodiment uses a temperature-sensing fiber to collect the temperature of the inner and outer walls of each battery pack, it has positioning accuracy, and there may be multiple battery packs on a section of the temperature-sensing fiber. Therefore, it cannot directly locate the battery pack with the temperature anomaly. Thus, it is necessary to first locate the battery cluster where the temperature-sensing fiber detected the temperature anomaly is based on a temperature anomaly score greater than the preset temperature anomaly score threshold of 0.6. This cluster is then used as the target battery cluster. In this embodiment, the positioning accuracy of the temperature-sensing fiber is 0.5m, which can quickly locate the 0.5m section of the temperature-sensing fiber where the temperature anomaly occurs. After determining the temperature-sensing fiber section, the system searches for battery packs located within that section where the CO concentration, assigned with dynamic confidence weights, exceeds the preset CO concentration threshold of 50ppm, and the smoke particle concentration, assigned with dynamic confidence weights, exceeds the preset smoke particle concentration threshold of 50mg / m³. 3 The battery pack is the target battery pack. If multiple battery packs meet the conditions at the same time, then all of these battery packs are identified as the target battery pack.

[0104] If triggered by the latter, the target battery pack and its associated target battery cluster can be directly located based on the battery pack address identifier carried in the temperature sensing cable alarm signal.

[0105] S33. Execute corresponding fire-fighting actions based on the alarm level and location results, specifically:

[0106] When the alarm level is Level 1, the frequency of collecting the inner and outer wall temperatures of each battery pack within the risky battery cluster is increased, for example, from once per second to twice per second, to enhance the monitoring of the risky battery cluster and to issue an alarm 15 minutes in advance compared to the actual thermal runaway moment.

[0107] When the alarm level is level two, the charging and discharging of the risky battery pack is cut off, and the liquid cooling of the energy storage compartment is activated, which can provide an alarm 5 minutes in advance compared to the actual thermal runaway moment.

[0108] When the alarm level is level three, precise fire suppression is carried out on the target battery pack.

[0109] This invention uses a temperature-sensing optical fiber to monitor temperature changes in key parts of the battery in real time (accuracy ±1℃). Combined with the composite detection of CO concentration, smoke and temperature changes, it adopts a multi-parameter fusion judgment method to capture early signs 5-10 minutes before thermal runaway occurs.

[0110] Example 2

[0111] Corresponding to Embodiment 1 above, this embodiment provides an energy storage safety automatic fire extinguishing system, such as... Figure 2 As shown, it includes: a main control module, several relay modules, several detection modules, a temperature sensing fiber optic module, a liquid storage module, a container valve, a power conversion system PCS, several zone valves, several fire sprinklers, and an execution module (not shown).

[0112] The power conversion system PCS, relay modules, and temperature-sensing fiber optic modules are all connected to the main control module. The main control module also communicates with external PCS to inform the system of its current fire protection status. Each relay module corresponds to a zone valve, and the zone valve is controlled by the relay module. Each detection module corresponds to a fire sprinkler head, and the fire sprinkler head is controlled by the detection module.

[0113] The temperature-sensing fiber optic module has multiple temperature-sensing fibers, with one fiber evenly distributed within each battery cluster. The fiber optic cables pass through all the battery packs in the cluster according to their arrangement and are designed to penetrate the packs, allowing the cables to run inside the packs and collect the temperature of the inner and outer walls of each pack.

[0114] In this embodiment, there are 12 heat-sensing optical fibers arranged within 12 battery clusters. Each battery cluster contains 8 battery packs, for a total of 96 battery packs. Each battery pack contains a detection module, which includes a CO sensor, a smoke sensor, and a heat-sensing cable. The CO sensor is used to collect the CO concentration within the battery pack; the smoke sensor is used to collect the smoke particle concentration within the battery pack; and the heat-sensing cable is used to issue an alarm signal carrying the battery pack's address identifier when the temperature inside the battery pack reaches a preset temperature threshold. The collected CO concentration, smoke particle concentration, and heat-sensing cable status within the battery pack constitute the fire characteristic parameters of the battery pack.

[0115] The detection module has an intelligent wake-up mechanism: it can be woken up by the relay module in standby mode or automatically when the CO concentration or smoke particle concentration collected by either the internal CO sensor or the smoke sensor reaches the corresponding preset wake-up threshold. In this embodiment, the preset CO concentration wake-up threshold range is 25 ppm, and the preset smoke particle concentration wake-up threshold range is 25 mg / m³. 3 .

[0116] The temperature-sensing fiber optic module transmits the collected temperatures of the inner and outer walls of each battery pack to the main control module. The detection module transmits the collected CO concentration, smoke particle concentration, and temperature-sensing cable status of each battery pack to the relay module, which then forwards the CO concentration, smoke particle concentration, and temperature-sensing cable status back to the main control module.

[0117] The main control module includes: a weight allocation unit, a temperature gradient analysis unit, and a fusion result acquisition unit, specifically:

[0118] The weighting unit is used to assign dynamic confidence weights to the inner wall temperature, outer wall temperature, CO concentration, and smoke particle concentration, respectively.

[0119] The temperature gradient analysis unit is used to preprocess the inner wall temperature and outer wall temperature based on the assigned dynamic confidence weights to obtain the processed inner wall temperature and outer wall temperature, and to perform temporal and spatial dual verification on the processed inner wall temperature and outer wall temperature to generate a temperature gradient anomaly determination result.

[0120] The fusion result acquisition unit is used to perform multi-level decision fusion on the temperature gradient anomaly judgment result, the temperature sensing cable status, and the CO concentration and smoke particle concentration assigned with dynamic confidence weights, and generate the fusion result.

[0121] The specific working principles, execution steps, and algorithm logic of the weight allocation unit, temperature gradient analysis unit, and fusion result acquisition unit within the main control module have been described in detail in steps S21 to S23 of Embodiment 1 of this invention, and will not be repeated here. During system operation, each unit collaborates to perform the dynamic weighted fusion calculation as described in Embodiment 1, ultimately generating a fire risk probability value as the fusion result.

[0122] The main control module sends the fusion results to the execution module, which includes:

[0123] The alarm level determination unit is used to determine the alarm level based on the fire risk probability value, including alarm levels one to three.

[0124] The positioning unit is used to locate the battery cluster or battery pack that triggered the alarm based on the alarm level.

[0125] The logic for determining the alarm level and locating the battery cluster or battery pack that triggered the alarm has been described in detail in steps S31 to S32 of Embodiment 1 of the present invention, and will not be repeated here.

[0126] The fire-fighting action implementation unit is used to execute fire-fighting actions based on the alarm level and location results, specifically:

[0127] When the alarm level determination unit determines it to be a Level 1 alarm and the location unit locates the risk battery cluster where the alarm-triggered temperature sensing fiber is located, the fire-fighting action implementation unit will activate, increasing the frequency of collecting the inner and outer wall temperatures of each battery pack in the risk battery cluster, activating the detection modules in all battery packs in the risk battery cluster, and recording and reporting the data recorded by the detection modules.

[0128] When the alarm level determination unit determines it to be a level two alarm and the location unit locates the risky battery pack that triggered the alarm, the fire-fighting action implementation unit activates, cutting off the charging and discharging of the risky battery pack and reporting to the energy storage compartment to activate liquid cooling. In this embodiment, the fire-fighting action implementation unit cuts off the charging and discharging of the risky battery pack by controlling the PCS.

[0129] When the alarm level determination unit determines it to be a level three alarm and the location unit locates the target battery pack that triggered the alarm and its associated target battery cluster, fire suppression is initiated on the target battery pack. Specifically: after identifying the target battery cluster, the fire suppression action implementation unit controls the relay module corresponding to the target battery cluster, causing it to open the corresponding zone valve; after identifying the target battery pack, it controls the detection module of the target battery pack, causing it to open the corresponding fire sprinkler head. Thus, the extinguishing liquid in the storage module, through the container valve, zone valve, and fire sprinkler head, can achieve precise fire suppression of the target battery pack.

[0130] This system significantly improves the thermal runaway detection lead time from less than 1 minute in traditional solutions to 5-15 minutes, and reduces the response time for critical operations to less than 300ms through a three-level response mechanism. It employs temperature-sensing fiber optics in conjunction with multiple multi-parameter detectors, greatly enhancing the fault location accuracy at the individual PACK level.

[0131] This system utilizes multi-source data fusion for judgment, controlling the calculation error of composite detection confidence level within ±0.05. The temperature monitoring system simultaneously possesses a wide measurement range of -40~150℃ and a temperature rise rate identification sensitivity of 0.5℃ / min. The fire protection system employs water-based extinguishing agent injection technology, achieving an extinguishing agent coverage delay of less than 100ms and a local cooling rate of 30℃ / min, representing a 5-fold performance improvement over traditional solutions.

[0132] Furthermore, the system's low-power design keeps overall energy consumption below 50W, saving 67% on electricity costs compared to similar solutions. Precision fire suppression technology reduces single-event handling costs by 80%, while remote reset and automatic diagnostic functions shorten troubleshooting time from 4 hours to 15 minutes. The system's comprehensive event tracing function also improves insurance claim evidence collection efficiency by 40%.

[0133] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method of energy storage safety automatic fire extinguishing, characterized in that, The method comprises the following steps: S1, detecting the inner wall side temperature, the outer wall side temperature and the fire characteristic parameters in the battery pack in real time; S2, performing dynamic weighted fusion calculation on the inner wall side temperature, the outer wall side temperature and the fire characteristic parameters to generate a fusion result; S3, performing a fire-fighting action based on the fusion result; The fire characteristic parameters comprise CO concentration, smoke particle concentration and temperature-sensing cable state, and S2 specifically comprises the following steps: S21, assigning a dynamic confidence weight to the inner wall side temperature, the outer wall side temperature, the CO concentration and the smoke particle concentration respectively; S22, pre-processing the inner wall side temperature and the outer wall side temperature based on the assigned dynamic confidence weight to obtain processed inner wall side temperature and outer wall side temperature, and performing time sequence and space double verification on the processed inner wall side temperature and outer wall side temperature to generate a temperature gradient abnormality determination result; S23, performing multi-level decision fusion on the temperature gradient abnormality determination result, the temperature-sensing cable state and the CO concentration and the smoke particle concentration to which the dynamic confidence weight is assigned to generate the fusion result; S22 specifically comprises the following steps: S221, performing adaptive filtering on the inner wall side temperature and the outer wall side temperature based on the assigned dynamic confidence weight to obtain filtered inner wall side temperature and outer wall side temperature; S222, calculating the time sequence temperature gradient of each battery pack based on the filtered inner wall side temperature and the outer wall side temperature, wherein the time sequence temperature gradient represents the rate of change of the inner wall side temperature and the outer wall side temperature with time; and calculating the space temperature gradient of each battery pack, wherein the space temperature gradient represents the rate of change of the temperature difference between the inner wall side temperature and the outer wall side temperature in the same battery pack with time; S223, analyzing the time sequence temperature gradient and the space temperature gradient of the same battery pack by using a convolutional neural network feature extraction algorithm to generate a temperature abnormality score of each battery pack as the temperature gradient abnormality determination result output.

2. The method of claim 1, wherein the energy storage safety automatic fire extinguishing method is characterized by, S23 specifically comprises the following steps: S231, constructing a multi-dimensional feature vector based on the temperature abnormality score, the temperature-sensing cable state and the CO concentration and the smoke particle concentration to which the dynamic confidence weight is assigned; S232, inputting the multi-dimensional feature vector into a predefined decision maker based on a multi-level threshold rule; S233, the decision maker performing logical decision on the multi-dimensional feature vector according to a preset multi-level threshold rule to generate and output a fire risk probability value as the fusion result.

3. The automatic fire extinguishing method for energy storage safety according to claim 2, characterized in that, The decision maker based on the multi-level threshold rule is configured to perform the following decision logic: When the temperature abnormality score of a certain battery pack is greater than a preset temperature abnormality score threshold, a low fire risk probability value is output, ranging from 0 to 0.4; When the CO concentration to which the dynamic confidence weight is assigned in a certain battery pack is greater than a preset CO concentration threshold, or the smoke particle concentration to which the dynamic confidence weight is assigned is greater than a preset smoke particle concentration threshold, a medium fire risk probability value is output, ranging from 0.4 to 0.

7. When the temperature anomaly score of a certain battery pack is greater than a preset temperature anomaly score threshold, and the CO concentration assigned with a dynamic confidence weight and the smoke particle concentration assigned with a dynamic confidence weight in the battery pack are both greater than their corresponding preset thresholds, or when the temperature sensing cable alarms, a high fire risk probability value is output, ranging from (0.7, 1].

4. The method of claim 3, wherein the energy storage is a compressed gas. S3 specifically is: S31, determining an alarm level based on the fire risk probability value; S32, locating the battery cluster or battery pack that triggers the alarm according to the alarm level; S33, performing corresponding fire-fighting actions according to the alarm level and the location result.

5. The method of claim 4, wherein the energy storage safety automatic fire extinguishing method is characterized by, The alarm level includes one to three levels of alarms, and S31 specifically is: When the fire risk probability value belongs to the low fire risk probability value range, a first-level alarm is triggered; When the fire risk probability value belongs to the medium fire risk probability value range, a second-level alarm is triggered; When the fire risk probability value belongs to the high fire risk probability value range, a third-level alarm is triggered.

6. The method of claim 5, wherein the energy storage safety automatic fire extinguishing method is characterized by, S32 specifically is: If a first-level alarm is triggered, the battery cluster in which the battery pack with a temperature anomaly score greater than a preset temperature anomaly score threshold is located is a risk battery cluster; If a second-level alarm is triggered, the battery pack in which the CO concentration assigned with a dynamic confidence weight is greater than a preset CO concentration threshold or the smoke particle concentration assigned with a dynamic confidence weight is greater than a preset smoke particle concentration threshold is a risk battery pack; If a third-level alarm is triggered, the battery pack in which the temperature anomaly score is greater than a preset temperature anomaly score threshold, the CO concentration assigned with a dynamic confidence weight, and the smoke particle concentration assigned with a dynamic confidence weight are all greater than their corresponding preset thresholds, or the temperature sensing cable alarms is a target battery pack.

7. The method of claim 6, wherein the energy storage safety automatic fire extinguishing method is characterized by, S33 specifically is: When the alarm level is one, the collection frequency of the inner wall side temperature and the outer wall side temperature of each battery pack in the risk battery cluster is increased; When the alarm level is two, the charging and discharging of the risk battery pack is cut off; When the alarm level is three, fire extinguishing is performed on the target battery pack.

8. An energy storage safety automatic fire extinguishing system, characterized in that, It includes: A detection module for detecting the inner wall side temperature, the outer wall side temperature, and the fire characteristic parameters in the battery pack in real time; A main control module for dynamically weighting and fusing the inner wall side temperature, the outer wall side temperature, and the fire characteristic parameters to generate a fusion result; An execution module for performing fire-fighting actions based on the fusion result; The fire characteristic parameters include CO concentration, smoke particle concentration, and temperature sensing cable state, and the main control module specifically is: assigning dynamic confidence weights to the inner wall side temperature, the outer wall side temperature, and the CO concentration, and the smoke particle concentration; based on the assigned dynamic confidence weights, pre-processing the inner wall side temperature and the outer wall side temperature to obtain processed inner wall side temperature and outer wall side temperature, and performing time sequence and space double verification on the processed inner wall side temperature and the outer wall side temperature to generate a temperature gradient anomaly determination result; Multi-level decision fusion is performed on the temperature gradient anomaly determination result, the temperature sensing cable state, and the CO concentration and the smoke particle concentration assigned with dynamic confidence weights to generate a fusion result. The temperature gradient anomaly determination result is generated by performing time sequence and space double verification on the processed inner wall side temperature and the processed outer wall side temperature based on the assigned dynamic confidence weight, specifically: based on the assigned dynamic confidence weight, the inner wall side temperature and the outer wall side temperature are adaptively filtered to obtain filtered inner wall side temperature and filtered outer wall side temperature; based on the filtered inner wall side temperature and the filtered outer wall side temperature, the time sequence temperature gradient of each battery pack is calculated, and the time sequence temperature gradient represents the rate of change of the inner wall side temperature and the outer wall side temperature with time; and the spatial temperature gradient of each battery pack is calculated, and the spatial temperature gradient represents the rate of change of the temperature difference obtained by subtracting the outer wall side temperature from the inner wall side temperature in the same battery pack with time; the time sequence temperature gradient and the spatial temperature gradient of the same battery pack are analyzed by using a convolutional neural network feature extraction algorithm, and the temperature anomaly score of each battery pack is generated as the temperature gradient anomaly determination result output.

Images