Automatic energy storage cabinet fire safety system

Abstract

The invention provides an automatic energy storage cabinet fire safety system, and relates to the field of energy storage cabinet fire safety, and the system comprises an acquisition module, a processing module, an early warning module and an execution module. The acquisition module acquires target data of a plurality of monitoring points in the energy storage cabinet in real time; the processing module analyzes the target data set based on a density clustering algorithm, and determines a first target area and extracts target features in combination with a three-dimensional space model, physical structure information and electrical connection relation information of the energy storage cabinet; the early warning module generates multi-level early warning information according to the target features; and the execution module triggers the cooling system, the local fire extinguishing system or the whole cabinet fire extinguishing system according to the early warning level, and sends an emergency alarm. According to the method, clustering parameters are dynamically adjusted, multi-dimensional data analysis and internal environment constraints of the energy storage cabinet are combined, abnormal areas are accurately recognized, early warning information is generated, fire fighting treatment measures are optimized, and the operation safety, reliability and resource utilization efficiency of the energy storage cabinet are remarkably improved.

Description

Technical Field

[0001] The present disclosure relates to the field of fire safety of energy storage cabinets, and in particular, to an automated fire safety system for energy storage cabinets. Background Art

[0002] With the rapid development of energy storage technology, energy storage devices are widely used in power systems to adjust peak and valley loads and improve energy efficiency. However, energy storage devices are often accompanied by high temperatures, smoke and other potential risks during operation, especially the thermal runaway problem of lithium batteries, which is an important challenge in the current safety management of energy storage devices.

[0003] For example, the Chinese patent application with publication number CN118230486A discloses an electrochemical energy storage smart fire protection system and method. The system can realize real-time monitoring, alarm and remote control of each component in the electrochemical energy storage device through a cloud monitoring platform combined with a communication protocol, and supports threshold setting and alarm data reporting.

[0004] The above methods have the problems raised by this background technology. The existing technology lacks flexibility and accuracy in identifying abnormal areas in energy storage equipment, and cannot dynamically adjust the identification range to adapt to the complex physical and electrical environment inside the energy storage system; secondly, the abnormal detection and early warning mechanism is single, usually based on simple fixed threshold monitoring, and fails to fully utilize the multi-dimensional data in the energy storage system (such as temperature, smoke concentration, distribution density, etc.) for comprehensive analysis; in addition, the fire treatment measures lack in-depth integration with the actual physical and electrical layout of the energy storage equipment, which easily leads to waste of resources or insufficient treatment effect. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present application discloses an automated energy storage cabinet fire safety system, which includes: The acquisition module is used to collect the target data of the energy storage cabinet at the preset monitoring points in real time and generate a target data set; a processing module, configured to perform cluster analysis on the target data set based on a density clustering method after performing data screening in the target data set based on a temperature threshold and / or a smoke threshold and a time threshold, determine a first target area of ​​the energy storage cabinet, and extract target features from the first target area; The density clustering method includes: expanding the initial clustering range based on the spatial coordinates, physical regions and components of each target data point in the initial clustering range, and spatial constraints; and dynamically adjusting the minimum point count threshold and neighborhood distance threshold in the density clustering method based on the abnormality index of the target data point; An early warning module matches corresponding early warning information for the target features based on a preset early warning mechanism; An execution module is used to execute fire fighting processing corresponding to the warning information in the first target area based on the warning information.

[0006] As an optional implementation, the acquisition module further includes: A sensor network, used for collecting the target data of each preset monitoring point in the energy storage cabinet in real time; wherein the sensor network includes at least one sensor; A preprocessing unit is used to preprocess the target data of each preset monitoring point in the energy storage cabinet after filtering the target data set based on the temperature threshold and / or the smoke threshold and the time threshold to generate the target data set.

[0007] As an optional implementation, the density clustering method includes: Set a minimum point count threshold and a neighborhood distance threshold, wherein the minimum point count threshold is used to limit the number of target data points required to form a cluster; the neighborhood distance threshold is used to limit the maximum distance between target data points to determine the neighborhood range; the neighborhood range is an area centered on each target data point and with a radius equal to the neighborhood distance threshold; Based on the neighborhood range, determine the core data in the target data set, where the core data is the target data whose number of target data in the neighborhood range is greater than or equal to a minimum point count threshold; Based on the neighborhood range of the core data, determine an initial cluster, and classify all target data within the neighborhood range of the core data into the initial cluster; Wherein, the target data set includes multiple target data points.

[0008] As an optional implementation, the density clustering method further includes: Under the preset spatial constraints, the boundary data within the neighborhood of the core data are included in the cluster; For the newly included boundary data, if it meets the core data conditions and conforms to the spatial constraints, continue to expand the target data within its neighborhood; Iteratively expand the clustering until no new target data is included in the clustering range.

[0009] As an optional implementation, performing data screening in the target data set based on a temperature threshold and / or a smoke threshold and a time threshold includes: Set time thresholds; Screening the target data set based on the time threshold includes: Determining whether the target data point is greater than a temperature threshold and / or a smoke threshold and continues to exceed the time threshold; In response to the target data point being greater than the temperature threshold and / or the smoke threshold but not exceeding the time threshold, it is excluded from the target data set.

[0010] As an optional implementation, the setting of the minimum point number threshold and the neighborhood distance threshold includes: Based on the abnormality index of the target data point, dynamically adjust the minimum point count threshold and the neighborhood distance threshold; Wherein, the determination of the abnormality degree index includes: quantifying the difference between the temperature value of the target data point and the temperature threshold to obtain a temperature excess metric; Monitoring the smoke concentration change rate of the target data point and determining the smoke concentration change gradient of the target data point; Inputting the temperature over-limit measure and the smoke concentration change gradient into a preset abnormality degree assessment model, and outputting the corresponding abnormality degree index; The dynamically adjusting the minimum point number threshold and the neighborhood distance threshold comprises: When the abnormality index is greater than or equal to a preset abnormality index threshold, the neighborhood distance threshold is increased and the minimum point threshold is decreased to expand the clustering range; When the abnormality index is less than a preset abnormality index threshold, the neighborhood distance threshold is reduced and the minimum point threshold is increased to narrow the clustering range.

[0011] As an optional implementation, determining the first target area of ​​the energy storage cabinet based on the clustering result includes: determining the spatial range of the first target area based on the core data and boundary data in the target data set; Extracting target features from the first target area includes: Extracting the highest temperature value of the target data; Extracting the smoke concentration change rate of the target data; Extracting the spatial distribution density of the target data; The number and proportion of abnormal points in the target data are extracted.

[0012] As an optional implementation, the density clustering method further includes: Based on the three-dimensional spatial model of the energy storage cabinet, determining the spatial coordinates of each of the target data points; Based on the spatial coordinates of the target data points, determining the physical area and component to which each target data point belongs; Determining space constraints based on the physical structure information and electrical connection relationship information of the energy storage cabinet; Based on the spatial coordinates, the physical area and the component to which each target data point belongs in the initial clustering range, and the spatial constraint condition, the initial clustering range is expanded, including: For the target data points to be expanded that are not within the initial clustering range, determine whether the following conditions are met at the same time: The physical region to which the target data point to be expanded belongs is directly adjacent to the physical region to which at least one target data point within the current clustering range belongs in structure, and there is no physical separation between the two; The component to which the target data point to be expanded belongs is directly and electrically connected to the component to which at least one target data point within the current clustering range belongs; If all the above conditions are met, the target data point to be expanded will be included in the clustering range.

[0013] As an optional implementation manner, the matching of the target feature with corresponding warning information based on the preset warning mechanism includes: Based on the target characteristics, the first target area is classified into different warning levels; wherein the warning levels include: level one warning, level two warning, and level three warning; Based on the warning level, warning information corresponding to the warning level is generated.

[0014] Based on the warning information, executing firefighting processing corresponding to the warning information in the first target area includes: In response to the warning level being the first-level warning, triggering a cooling system in the energy storage cabinet to reduce the temperature of the first target area; In response to the warning level being the second-level warning, triggering the local fire protection system of the energy storage cabinet to spray a fire extinguishing medium toward the first target area; In response to the warning level being the third-level warning, the entire cabinet fire protection system of the energy storage cabinet is triggered, and a fire extinguishing medium is sprayed toward the first target area, and an emergency alarm signal is sent to notify an external control center.

[0015] Compared with the prior art, the beneficial effects of the present invention are: by combining multidimensional data analysis, dynamic clustering algorithms, and the comprehensive application of physical and electrical constraints of energy storage cabinets, the intelligence and precision of fire safety management of energy storage cabinets are significantly improved. Specifically, the present invention introduces a dynamic adjustment mechanism based on density clustering in the abnormal area identification process, which can flexibly adjust the clustering parameters according to the abnormal degree index of the target data point, so that the system can accurately locate potential abnormal areas in a complex energy storage cabinet environment and trigger alarms in time. In addition, by combining the three-dimensional spatial model, physical structure information and electrical connection relationship information of the energy storage cabinet, the present invention effectively overcomes the recognition errors caused by simple threshold monitoring and fixed range limitations of traditional methods, ensuring that the determination of abnormal areas is more scientific and reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of an automated energy storage cabinet fire safety system provided by an embodiment of the present disclosure; Figure 2 A schematic diagram of a collection module provided in an embodiment of the present disclosure; Figure 3 A schematic diagram of the three-dimensional structure of an energy storage cabinet provided in an embodiment of the present disclosure; Figure 4 A schematic diagram of the front structure of a cabinet provided in an embodiment of the present disclosure; Figure 5 A schematic diagram of the back structure of a cabinet provided in an embodiment of the present disclosure.

[0017] Figure numerals: 100, acquisition module; 200, processing module; 300, early warning module; 400, execution module; 110, sensor network; 120, pre-processing unit; 1, water pump; 2, fire water inlet pipe; 3, refrigerant box; 4, water inlet main pipe; 5, refrigerant box inlet; 6, return water main pipe; 7, battery PACK box; 8, return water branch pipe; 9, return water branch pipe; 10, water inlet branch pipe; 11, water inlet branch pipe; 12, fire prevention valve; 13, fire spray pipe; 14, fire sprinkler. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0019] See also Figure 1As shown, it is a schematic diagram of an automated energy storage cabinet fire safety system provided by an embodiment of the present disclosure. The present invention provides an automated energy storage cabinet fire safety system, which realizes efficient management of energy storage cabinet fire safety through modular design. The system includes a collection module 100, a processing module 200, an early warning module 300 and an execution module 400. The modules cooperate with each other to monitor the operating status of the energy storage cabinet in real time, analyze potential risks, and take timely fire treatment measures.

[0020] in: The acquisition module 100 is used to collect target data of the energy storage cabinet at a preset monitoring point in real time and generate a target data set; The processing module 200 is used to perform cluster analysis on the target data set based on a density clustering method after data screening based on a temperature threshold and / or a smoke threshold and a time threshold in the target data set, determine a first target area of ​​the energy storage cabinet, and extract target features from the first target area; wherein the density clustering method includes: expanding the initial clustering range based on the spatial coordinates, the physical area and component to which each target data point belongs, and the spatial constraint condition in the initial clustering range; and dynamically adjusting the minimum point count threshold and the neighborhood distance threshold in the density clustering method based on the abnormality index of the target data point; The warning module 300 matches the target feature with corresponding warning information based on a preset warning mechanism; The execution module 400 is used to execute fire fighting processing corresponding to the warning information in the first target area based on the warning information.

[0021] In the present disclosure, an energy storage cabinet refers to a device for storing electrical energy, which includes battery cells, control circuits, cooling systems, etc. During operation, abnormal conditions such as temperature rise and smoke generation may occur, which require real-time monitoring and safety management.

[0022] The target data refers to the monitoring data related to the operating status of the energy storage cabinet acquired by the acquisition module 100, including but not limited to parameters such as temperature, smoke concentration, voltage, and current.

[0023] The target data set refers to a set of target data of multiple preset monitoring points, which is provided for analysis by the subsequent processing module 200 .

[0024] The first target area refers to the internal space range of the energy storage cabinet where anomalies may exist or need to be paid special attention, which is determined by the processing module 200 after clustering and analyzing the target data set.

[0025] The target features refer to specific data features extracted by the processing module 200 from the first target area, including but not limited to the maximum temperature value, smoke concentration change rate, spatial distribution density, etc., which are used for subsequent early warning and fire fighting.

[0026] Early warning information refers to warning information generated based on target features, which identifies the abnormal risk level and specific location, and is used to guide the execution module 400 to take corresponding measures.

[0027] The present invention deploys a variety of sensors through the acquisition module 100 to comprehensively collect the target data of each preset monitoring point of the energy storage cabinet, form a target data set, and provide a data basis for subsequent risk analysis. The processing module 200 performs intelligent clustering analysis on the target data set to determine the first target area. This analysis process can dynamically identify abnormal areas and avoid misjudgments caused by simple threshold monitoring. The early warning module 300 generates different levels of early warning information based on the extracted target features, and guides the execution module 400 to accurately perform fire protection in the corresponding first target area to avoid unnecessary resource consumption triggered by the entire cabinet.

[0028] In a specific implementation, the acquisition module 100 deploys a sensor network 110 at multiple preset monitoring points in the energy storage cabinet to collect target data of the energy storage cabinet. The target data includes but is not limited to operating parameters such as temperature, smoke concentration, voltage, and current. Each monitoring point corresponds to at least one sensor, such as a temperature sensor and a smoke sensor deployed near the battery unit, which are used to sense changes in the operating state in real time. The sensor network 110 transmits the collected data to the processing module 200 via wired or wireless means.

[0029] The function of the processing module 200 is to perform cluster analysis on the target data set provided by the acquisition module 100 and determine the first target area. Cluster analysis uses intelligent algorithms (such as density clustering algorithms) to accurately locate areas where risks may exist based on data features. The first target area can be represented as one or more areas in the three-dimensional space of the energy storage cabinet. For example, if the temperature of two groups of battery cells in a certain energy storage cabinet is found to be abnormally high, the group of battery cells can be marked as the first target area.

[0030] It should be noted that the first target area is only the initial or priority discovered abnormal area; if new data is detected that meets the clustering conditions, the second and third target areas will be generated; The early warning module 300 stores a multi-level early warning mechanism. The module matches the target features (such as the maximum temperature value, the rate of change of smoke concentration, etc.) with the preset rules to generate corresponding early warning information. For example, when the temperature in a certain area exceeds the safety threshold and the smoke concentration rises rapidly, the system will generate a second or third level early warning information to remind the execution module 400 to take more stringent fire prevention measures.

[0031] The execution module 400 takes precise fire prevention measures according to the warning information generated by the warning module 300. For example, when a level 1 warning is generated, only the cooling system is triggered to cool down the first target area; when a level 3 warning is generated, the whole cabinet fire prevention system is triggered, including starting the fire extinguishing medium injection and issuing an emergency alarm to notify the external control center.

[0032] In a specific implementation, the processing module 200 performs data screening in the target data set based on the temperature threshold and / or the smoke threshold and the time threshold, including: Set temperature threshold and smoke threshold; The target data set is filtered based on the temperature threshold and / or the smoke threshold.

[0033] As an optional implementation, screening the target data set based on the temperature threshold and / or the smoke threshold further includes: Set time thresholds; Screening the target data set based on the time threshold includes: Determining whether the target data point is greater than a temperature threshold and / or a smoke threshold and continues to exceed the time threshold; In response to the target data point being greater than the temperature threshold and / or the smoke threshold but not exceeding the time threshold, it is excluded from the target data set.

[0034] In the operation status monitoring of energy storage cabinets, the collected target data set may contain a large amount of redundant or noisy data, which not only increases the computational complexity of subsequent analysis, but also may interfere with the accuracy of positioning abnormal areas. For example, due to sensor errors or short-term fluctuations, some target data points may temporarily exceed the threshold, but do not represent a true abnormality.

[0035] The present invention adds a screening process based on temperature threshold and smoke threshold before determining the core data, preliminarily filters out obviously irrelevant data points, and significantly improves the quality of the target data set. At the same time, the time threshold screening is introduced to address the abnormal persistence problem, further ensuring that the target data point is considered abnormal only when it meets the persistence condition. This dual screening mechanism effectively solves the problem of misjudgment caused by short-term fluctuations and provides more accurate data input for subsequent density clustering analysis.

[0036] In a specific implementation, before the density cluster analysis, the target data set is screened by temperature threshold, smoke threshold and time threshold, wherein the temperature threshold can be set according to the safety operation requirements inside the energy storage cabinet. For example, the temperature threshold range can be set to 50°C to 70°C according to the heat resistance of battery cells and electrical equipment in different areas; the smoke threshold is set according to the smoke concentration range that may be generated inside the energy storage cabinet. For example, the smoke threshold range can be set to 20ppm to 50ppm with reference to the baseline value of the smoke sensor in the normal operating environment of the energy storage cabinet.

[0037] In a specific implementation, for each target data point in the target data set generated by the acquisition module 100, it is determined whether it exceeds the temperature threshold and / or the smoke threshold: For example, if the temperature value of a target data point exceeds the temperature threshold, or its smoke concentration value exceeds the smoke threshold, it can be marked as a data point "to be further screened"; If the temperature value and smoke concentration value of a target data point do not exceed the corresponding threshold, it will be directly excluded from the target data set.

[0038] In addition, the time threshold is used to determine whether the abnormal state of the target data point is persistent. For example, the time threshold can be set to 5 seconds, which means that the abnormal state of the target data point must last for more than 5 seconds to be considered a valid abnormality. For the data points marked as "pending further screening", analyze the duration of their exceeding the temperature threshold or smoke threshold: If the abnormal state of the target data point (such as the temperature or smoke concentration value exceeding the threshold) lasts longer than the time threshold, the data point is retained; If the abnormal state of the target data point does not last longer than the time threshold, it is excluded from the target data set.

[0039] Through the above screening steps, the processing module 200 finally generates an optimized target data set for subsequent density clustering analysis.

[0040] In this way, the present invention effectively filters out irrelevant or short-term fluctuating target data points by adding a dual screening mechanism based on temperature threshold, smoke threshold and time threshold before density cluster analysis, significantly improving the quality of the target data set and the accuracy of subsequent cluster analysis. The setting of temperature threshold and smoke threshold is flexibly adjusted according to the actual operating environment of the energy storage cabinet, and the introduction of time threshold further ensures the reliability of abnormal data, thereby reducing the misjudgment rate and calculation complexity, and providing more accurate basic support for the fire safety management of energy storage cabinets.

[0041] It should be noted that since the target data includes but is not limited to parameters such as temperature, smoke concentration, voltage, current, etc., in addition to temperature and smoke, parameters such as voltage and current can also be used to screen abnormal data.

[0042] For example, a certain energy storage station is composed of multiple energy storage cabinets, and multiple temperature sensors and smoke sensors are arranged inside each energy storage cabinet. The acquisition module 100 collects data in real time, and the processing module 200 finds that the temperature of two areas of a certain energy storage cabinet rises abnormally (65°C and 70°C respectively), but the smoke concentration has not yet reached the alarm threshold. The early warning module 300 generates a first-level early warning message, and the system triggers the cooling system to cool down the abnormal area. Subsequently, if the smoke concentration continues to rise, the system automatically upgrades to a second-level early warning, triggering the local fire extinguishing system to spray the fire extinguishing medium into the abnormal area. Through precise monitoring and processing, risks are controlled in a timely manner to avoid fires.

[0043] In this way, through modular design, it integrates functions such as multi-point real-time collection, cluster analysis and regional positioning, hierarchical warning and precise fire treatment, which can quickly respond to abnormal changes in the energy storage cabinet, effectively cover the operating status of the entire energy storage cabinet, and realize full-process automated management from monitoring to processing. Compared with traditional single-point monitoring and full-cabinet trigger fire protection systems, the present invention significantly improves the response speed and abnormal positioning accuracy, while greatly reducing resource consumption and equipment loss, and improving the operating safety of the energy storage cabinet and the level of system intelligence.

[0044] See also Figure 2 FIG. 1 is a schematic diagram of a collection module 100 provided in an embodiment of the present disclosure. As an optional implementation, the collection module 100 further includes: The sensor network 110 is used to collect the target data of each preset monitoring point in the energy storage cabinet in real time; wherein the sensor network 110 includes at least one sensor; The preprocessing unit 120 is used to preprocess the target data of each preset monitoring point in the energy storage cabinet to generate the target data set.

[0045] The present disclosure proposes an improved design of the acquisition module 100, which adds two core functions: a sensor network 110 and a pre-processing unit 120. The sensor network 110 realizes full coverage of multiple monitoring points in the energy storage cabinet, ensuring the comprehensiveness of monitoring and the integrity of data; the pre-processing unit 120 performs preliminary cleaning and optimization on the collected target data, providing expected data input for the cluster analysis of the subsequent processing module 200.

[0046] In a specific implementation, the sensor network 110 is composed of various types of sensors deployed at different locations inside the energy storage cabinet. Specifically, the sensors include but are not limited to temperature sensors, smoke sensors, current sensors, and voltage sensors. These sensors can be connected to the preprocessing unit 120 in a wired manner (such as CAN bus) or wirelessly (such as ZigBee protocol), and transmit target data in real time according to a set sampling period. For example, temperature sensors and smoke sensors are arranged near the battery cells of the energy storage cabinet to monitor temperature changes and abnormal smoke concentrations in real time, while current sensors and voltage sensors are arranged at the cable connection to monitor current overloads and voltage fluctuations.

[0047] It is understandable that in some cases, data collection of the energy storage cabinet can also be performed through an external sensor. In this embodiment, for example, only one infrared sensor can be used to obtain the temperature data of the entire energy storage cabinet surface and / or interior. At this time, the processing module 200, the early warning module 300, and the execution module 400 can also be linked to implement the present disclosure, so the sensor network 110 can include at least one sensor.

[0048] In a specific implementation, the function of the preprocessing unit 120 is to preliminarily process the target data transmitted by the sensor network 110 to improve the data quality and availability. Specifically, the functions of the preprocessing unit 120 may include: data cleaning, data noise reduction, data formatting, and data integration.

[0049] Exemplarily, the preprocessing unit 120 can remove invalid data points in the target data, such as abnormal readings caused by sensor failure; the preprocessing unit 120 can use a filtering algorithm (such as low-pass filtering) to smooth the target data to reduce noise introduced by external interference; the preprocessing unit 120 can uniformly convert the target data into a preset format to facilitate parsing and analysis by the subsequent processing module 200; the preprocessing unit 120 can classify and integrate the target data from different monitoring points according to the spatial distribution of the energy storage cabinet to generate a target data set.

[0050] For example, in the actual application of a certain energy storage station, the sensor network 110 of the acquisition module 100 includes the following configuration: 20 sensors (16 temperature sensors and 4 smoke sensors) are arranged in each energy storage cabinet, distributed near the battery unit, busbar connector and vent. The sensor is connected to the preprocessing unit 120 via the CAN bus. The target data is collected once a second and transmitted to the preprocessing unit 120 in real time. The preprocessing unit 120 performs noise reduction processing on the target data transmitted by the sensor network 110, for example, by using the Kalman filter algorithm to eliminate the short-term fluctuations of the temperature sensor, and then integrates the cleaned data into a target data set, including the temperature value, smoke concentration value and its corresponding time and space position information of each monitoring point.

[0051] In this way, the present invention not only realizes comprehensive monitoring of multiple points in the energy storage cabinet by introducing the sensor network 110 and the preprocessing unit 120 in the acquisition module 100, but also significantly improves the data quality through preprocessing technologies such as data cleaning and noise reduction, avoiding misjudgment or omission caused by monitoring blind spots or data noise. The target data set generated by the acquisition module 100 provides reliable data support for the subsequent processing module 200, so that the system as a whole has higher accuracy and responsiveness, thereby effectively improving the real-time and reliability of the fire safety management of the energy storage cabinet.

[0052] In a specific implementation, the density clustering method includes: Set a minimum point count threshold and a neighborhood distance threshold, wherein the minimum point count threshold is used to limit the number of target data points required to form a cluster; the neighborhood distance threshold is used to limit the maximum distance between target data points to determine the neighborhood range; the neighborhood range is an area centered on each target data point and with a radius equal to the neighborhood distance threshold; Based on the neighborhood range, determine the core data in the target data set, where the core data is the target data whose number of target data in the neighborhood range is greater than or equal to a minimum point count threshold; Based on the neighborhood range of the core data, an initial cluster is determined, and all target data within the neighborhood range of the core data are classified into the initial cluster.

[0053] The present invention introduces a density clustering method (such as DBSCAN algorithm) in the processing module 200, and accurately identifies areas where abnormalities may exist by comprehensively analyzing the target data set. The density clustering method defines the relevance of the target data by setting a minimum point count threshold and a neighborhood distance threshold, and can dynamically discover abnormal areas, avoiding misjudgments caused by simple threshold monitoring, and at the same time improving the positioning accuracy of multi-point linkage abnormalities.

[0054] The core of the density clustering method is to use the neighborhood range to determine the core data points and their associated points, and expand the clustering range through the core data points. Using this method, the present invention realizes efficient abnormal area positioning of the energy storage cabinet, providing accurate input for subsequent early warning and fire fighting.

[0055] In a specific implementation, the processing module 200 first sets the key parameters of clustering: the minimum point count threshold (MinPts) and the neighborhood distance threshold (Eps).

[0056] The minimum point threshold is used to limit the number of target data points required to form a cluster. The specific value can be set based on the following factors: Distribution characteristics of target data: for example, the average number of target data points in a certain area; The size of the anomaly area is defined as: at least a certain number of sensors or data points.

[0057] The neighborhood distance threshold is used to limit the maximum distance between target data points to determine the neighborhood range of the target data points. The specific value can be set based on the following factors: The physical layout inside the energy storage cabinet: for example, the average spacing or arrangement density between sensors; The spatial range monitored by the sensor: such as the effective sensing radius of a temperature sensor (unit: meter or millimeter).

[0058] For each data point in the target data set, the processing module 200 calculates its neighborhood range according to the neighborhood distance threshold (Eps). If the number of target data in the neighborhood range of a data point is greater than or equal to the minimum point number threshold (MinPts), the data point is determined as a core data point. The core data point is the key to clustering and represents an area with a high density of target data.

[0059] The processing module 200 classifies each core data point and all target data points in its neighborhood into an initial cluster. If there are other core data points in the neighborhood of a core data point, the cluster is expanded and merged to form a larger cluster.

[0060] For example, in the actual operation of a certain energy storage cabinet, the acquisition module 100 transmits data from 20 monitoring points, including temperature data (unit: °C) and smoke concentration data (unit: ppm). The processing module 200 applies the density clustering method to analyze these target data. The specific steps are as follows: Neighborhood distance threshold (Eps): set to 0.5 meters, based on the average spacing of sensors inside the energy storage cabinet (0.5 meters); Minimum point threshold (MinPts): Set to 4, based on the definition that an area must contain at least 4 target data points to be characterized as an abnormal area.

[0061] Calculate the neighborhood range for each data point in the target data set. For example: The neighborhood of data point A (temperature is 60°C, smoke concentration is 15ppm) includes 5 target data points, and the neighborhood distance is less than 0.5 meters; The neighborhood of data point B (temperature is 45°C, smoke concentration is 5 ppm) contains only two target data points.

[0062] Data point A is determined to be a core data point because the number of target data points (5) contained in its neighborhood is greater than or equal to the minimum point count threshold (4); Data point B is not determined as a core data point because the number of target data points in its neighborhood (2) is less than the minimum point count threshold (4).

[0063] All data points (including data point A) in the neighborhood of data point A are classified as the initial cluster. If other data points in the neighborhood (such as data point C) are also core data points, continue to expand and merge all data points in the neighborhood of C into the cluster.

[0064] Finally, the processing module 200 identifies a certain area inside the energy storage cabinet as an abnormal area, and marks it as a first target area for use by the subsequent warning module 300 and the execution module 400.

[0065] In this way, the present invention analyzes the target data set through density clustering method, and can adaptively find abnormal areas in the energy storage cabinet, overcoming the problems of misjudgment and missed judgment of traditional single-point threshold monitoring. The density clustering method has high flexibility in parameter setting. For example, the neighborhood distance threshold can be adjusted according to the layout density and physical perception range of the sensor, and the minimum point number threshold can define the scale of the abnormal area according to the distribution characteristics of the target data.

[0066] In addition, the cluster analysis method of the present invention does not require the preset number or shape of abnormal areas, has strong adaptability and high accuracy, provides reliable data support for subsequent early warning and fire fighting, and significantly improves the efficiency and accuracy of energy storage cabinet fire safety management.

[0067] As an optional implementation, the density clustering method further includes: Under the preset spatial constraints, the boundary data within the neighborhood of the core data are included in the cluster; For the newly included boundary data, if it meets the core data conditions and conforms to the spatial constraints, continue to expand the target data within its neighborhood; Iteratively expand the clustering until no new target data is included in the clustering range.

[0068] As an optional implementation, expanding the initial clustering range includes: Based on the three-dimensional spatial model of the energy storage cabinet, determining the spatial coordinates of each of the target data points; Based on the spatial coordinates of the target data points, determining the physical area and component to which each target data point belongs; Determining space constraints based on the physical structure information and electrical connection relationship information of the energy storage cabinet; Based on the spatial coordinates, the physical area and the component to which each target data point belongs in the initial clustering range, and the spatial constraint condition, the initial clustering range is expanded, including: For the target data points to be expanded that are not within the initial clustering range, determine whether the following conditions are met at the same time: The physical region to which the target data point to be expanded belongs is directly adjacent to the physical region to which at least one target data point within the current clustering range belongs in structure, and there is no physical separation between the two; The component to which the target data point to be expanded belongs is directly and electrically connected to the component to which at least one target data point within the current clustering range belongs; If all the above conditions are met, the target data point to be expanded will be included in the clustering range.

[0069] Among them, in the complex physical environment of the energy storage cabinet, the determination of abnormal areas is not only affected by the distribution density of the target data points, but also needs to consider the physical space layout and component correlation. For example, due to the physical structural constraints such as fireproof partitions and electrical isolation inside the energy storage cabinet, the neighborhood relationship between target data points may be limited, and traditional density clustering methods are difficult to adapt to such complex scenarios. In addition, the correlation between components such as battery cells and cables in the energy storage cabinet also plays an important role in determining whether the target data points belong to the same area.

[0070] The present invention introduces spatial constraints on the basis of the density clustering method, and accurately limits the expansion process of the clustering range by combining the three-dimensional spatial model, physical structure information and electrical connection relationship information of the energy storage cabinet. Specifically, when expanding the neighborhood range of the core data point, the processing module 200 will make judgments based on the spatial coordinates, physical area and component information of the target data point to ensure that the clustering range is limited to areas that are directly adjacent to each other and have no physical partitions. In addition, during the expansion process, the processing module 200 iteratively analyzes whether each newly included target data point meets the core data conditions, thereby gradually expanding the clustering range and ultimately accurately identifying abnormal areas of the energy storage cabinet.

[0071] In a specific implementation, the processing module 200 uses the three-dimensional spatial model of the energy storage cabinet to determine the spatial coordinates of each target data point. For example, the spatial coordinates take the reference point of the energy storage cabinet as the origin, and use three-dimensional coordinates (x, y, z) to represent the position of the target data point (unit: meter).

[0072] Based on the spatial coordinates of the target data point, the processing module 200 further determines the physical region and component to which the target data point belongs. For example, a target data point may be located in the battery cell A1 region and directly associated with the cooling module component.

[0073] Among them, directly adjacent means that the physical area to which the target data point belongs must be directly adjacent to the physical area to which the target data point belongs within the current clustering range in space. For example, directly adjacent is defined as the distance between the center points of the two areas is less than a preset maximum adjacent distance threshold (such as 1 meter).

[0074] No physical separation means that there is no physical barrier such as fire partitions or barriers between the target data point and the target data point within the current clustering range.

[0075] Component relevance means that the component to which the target data point belongs must be directly electrically connected to the component to which the target data point belongs within the current clustering range, for example, confirmed through an electrical connection relationship database.

[0076] In a specific implementation, for each target data point in the neighborhood of each core data point, the processing module 200 determines whether it is boundary data one by one. The basis for determining boundary data is: the target data point is within the neighborhood, but the number of target data in its neighborhood is less than the minimum point threshold (MinPts).

[0077] On the premise of satisfying the spatial constraint conditions, the processing module 200 includes the boundary data into the clustering range.

[0078] In a specific implementation, for a newly included boundary data point, the processing module 200 determines whether it meets the core data condition (i.e., the number of target data in the neighborhood is greater than or equal to the minimum point number threshold). If so, it is used as a new core data point and the target data in its neighborhood is further expanded.

[0079] The clustering range is iteratively expanded by recursion or looping until no new target data points meet the conditions.

[0080] For example, in actual operation of a certain energy storage cabinet, the target data set transmitted by the acquisition module 100 includes data of 30 monitoring points. The processing module 200 applies a density clustering method to perform cluster expansion.

[0081] The processing module 200 first generates an initial cluster, which includes a core data point A and target data points within its neighborhood.

[0082] The target data point B in the neighborhood is determined to be boundary data because the number of target data in its neighborhood is 3, which is less than MinPts (4). The spatial coordinates of data point B are (1.2, 2.5, 0.8) meters, and the distance between it and the spatial coordinates of core data point A (1.0, 2.0, 0.8) meters is 0.6 meters, which is less than the maximum adjacent distance threshold (1 meter). There is no fire partition between data point B and core data point A, and the corresponding components of the two are directly connected electrically.

[0083] Data point B meets the spatial constraint conditions and is included in the clustering range.

[0084] The processing module 200 analyzes the neighborhood range of data point B and finds that the number of target data in its neighborhood range is 5, which is greater than or equal to MinPts (4), and upgrades it to a core data point.

[0085] Take data point B as the new core data point and continue to expand its neighborhood range.

[0086] After multiple iterations and expansions, the processing module 200 finally identifies the abnormal area in the energy storage cabinet and marks it as the first target area, which contains 200 target data points and covers a group of battery cells and cooling components of the energy storage cabinet.

[0087] In this way, the present invention effectively improves the accuracy of locating abnormal areas of energy storage cabinets by adding spatial constraints and iterative expansion mechanisms to the density clustering method. The spatial constraints combine the physical layout and component associations of the energy storage cabinet to ensure that the clustering range is limited to physically related areas, avoiding the risk of mistakenly including target data points in irrelevant areas into the cluster. In addition, the design of iterative expansion clustering can dynamically identify multi-point linkage anomalies and provide accurate input data for subsequent early warning and fire fighting. This technical solution significantly improves the efficiency and reliability of the energy storage cabinet fire safety system and is suitable for energy storage cabinet fire safety management in complex physical environments.

[0088] In the application of density clustering method, neighborhood distance threshold (Eps) and minimum point threshold (MinPts) are key parameters that directly affect the clustering effect of abnormal areas. However, the traditional fixed threshold method may not be able to adapt to the dynamic changes of target data points in the complex operating environment of energy storage cabinets. For example, target data points in different areas may show significantly different degrees of abnormality due to differences in environmental conditions, equipment types or monitoring status. If a fixed and unified clustering parameter is used, some abnormal areas may be ignored or misjudged.

[0089] The present invention introduces an abnormality index and dynamically adjusts the neighborhood distance threshold and the minimum point threshold, so that the clustering algorithm can adaptively adjust the clustering range according to the specific abnormal situation of the target data point. The abnormality index combines the temperature over-limit measurement and the smoke concentration change gradient, and quantifies the abnormality of the target data point with the help of the abnormality evaluation model. By adjusting the threshold, the system expands the clustering range when the abnormality is serious to ensure that the complete abnormal area is captured; when the abnormality is small, the clustering range is reduced to avoid misjudgment or over-expansion.

[0090] As an optional implementation, the setting of the minimum point number threshold and the neighborhood distance threshold includes: Based on the abnormality index of the target data point, dynamically adjust the minimum point count threshold and the neighborhood distance threshold; Wherein, the determination of the abnormality degree index includes: quantifying the difference between the temperature value of the target data point and the temperature threshold to obtain a temperature excess metric; Monitoring the smoke concentration change rate of the target data point and determining the smoke concentration change gradient of the target data point; Inputting the temperature over-limit measure and the smoke concentration change gradient into a preset abnormality degree assessment model, and outputting the corresponding abnormality degree index; The dynamically adjusting the minimum point number threshold and the neighborhood distance threshold comprises: When the abnormality index is greater than or equal to a preset abnormality index threshold, the neighborhood distance threshold is increased and the minimum point threshold is decreased to expand the clustering range; When the abnormality index is less than a preset abnormality index threshold, the neighborhood distance threshold is reduced and the minimum point threshold is increased to narrow the clustering range.

[0091] In a specific implementation, the processing module 200 processes the temperature value of the target data point. With temperature threshold Quantify the difference:

[0092] like , then the temperature of the target data point exceeds the threshold, Quantized to a positive value; if ,but Quantized to zero.

[0093] Monitor the rate of change of smoke concentration at the target data point over time and calculate the gradient of smoke concentration change :

[0094] in, Unit time The change of smoke concentration in the room (unit: ppm / s).

[0095] Measuring temperature excess and smoke concentration gradient Input the preset abnormality evaluation model (such as a regression model or rule model based on machine learning) to obtain the abnormality index .For example:

[0096] in, and is the model weight; abnormality index It is a dimensionless indicator that reflects the abnormality of the target data point, and its value range is Typically, The higher the value, the stronger the abnormality of the target data point. For the convenience of calculation and comparison, the abnormality index can also be normalized to the interval [0, 1].

[0097] Exemplary weights of anomaly assessment model and It can be set according to the operating characteristics of different areas inside the energy storage cabinet. For example, in the battery unit area, temperature exceeding the limit is the main abnormal feature, so it is preferred to set ; In the cooling component area, smoke concentration changes are the main abnormal characteristics, preferably set In addition, the weight values ​​can be obtained through statistical analysis of historical data and model validation optimization.

[0098] For example, the threshold of the abnormality index can be set according to the operating environment of the energy storage cabinet and the historical data statistics. For example, in the battery unit area, due to the high fire risk, the threshold of the abnormality index is preferably set to 2.0; in the cooling component area, due to the slow development of the abnormality, the threshold is preferably set to 3.0.

[0099] For example, when the abnormality index When the value is greater than or equal to the preset abnormality index threshold, the neighborhood distance threshold Eps can be expanded by 10%-20%, and the minimum point threshold MinPts can be reduced by 1-2 data points; when the abnormality index When it is less than the preset abnormality index threshold, the neighborhood distance threshold and the minimum point threshold gradually return to the initial value.

[0100] It should be noted that the adjusted parameters must be controlled within the set upper and lower limits.

[0101] In this way, the processing module 200 calculates the abnormality index in real time during the clustering process and dynamically adjusts the clustering parameters so that the clustering range can flexibly adapt to the abnormality of the target data point.

[0102] Exemplarily, the present invention uses a linear regression model as an anomaly assessment model. The input features of the model include the temperature over-limit metric and the smoke concentration change gradient of the target data point, and the output is an abnormality degree indicator. The training data set is constructed through the actual operation data of the energy storage cabinet, and the temperature values ​​and smoke concentration change values ​​of the historical target data points are annotated with the data of the known abnormal areas to generate labels for model training. The model is optimized using the mean square error as the loss function, and the model performance is evaluated using the validation data set. Experimental results show that the model can accurately reflect the degree of abnormality of the target data point.

[0103] The present invention effectively solves the problem that fixed parameters in traditional density clustering methods are difficult to adapt to dynamic changes by dynamically adjusting clustering parameters based on the abnormality degree index. The quantification of the abnormality degree index comprehensively considers the degree of temperature exceeding the limit and the rate of change of smoke concentration, and can accurately reflect the abnormal situation of the target data point. By dynamically adjusting the neighborhood distance threshold and the minimum point number threshold, the present invention expands the clustering range when the abnormality is serious to ensure the integrity of the abnormal area; and narrows the clustering range when the abnormality is relatively minor to avoid misjudgment and over-expansion. This method significantly improves the flexibility and accuracy of the energy storage cabinet fire safety management system.

[0104] As an optional implementation, determining the first target area of ​​the energy storage cabinet based on the clustering result includes: determining the spatial range of the first target area based on the core data and boundary data in the target data set; Extracting target features from the first target area includes: Extracting the highest temperature value of the target data; Extracting the smoke concentration change rate of the target data; Extracting the spatial distribution density of the target data; The number and proportion of abnormal points in the target data are extracted.

[0105] The processing module 200 uses the results of the density clustering method, combined with the core data and boundary data in the target data set, to determine the first target area of ​​the energy storage cabinet, and extract key feature information related to the abnormal state to provide support for the subsequent warning module 300 and the execution module 400.

[0106] The process of determining the first target area includes the spatial distribution of core data and boundary data, combined with the three-dimensional spatial model, physical structure information and electrical connection relationship information of the energy storage cabinet, and finally accurately delineating the energy storage cabinet area where abnormalities may exist.

[0107] In a specific implementation, the processing module 200 first extracts core data points and boundary data points from the density clustering results. Core data points refer to target data points whose number of target data in the neighborhood is greater than or equal to the minimum point count threshold MinPts, while boundary data points refer to target data points whose number of target data in the neighborhood is less than MinPts but belongs to the neighborhood of the core data points. Combining the spatial coordinates of the core data and the boundary data, the processing module 200 determines the preliminary range of the first target area according to the three-dimensional spatial model of the energy storage cabinet. The preliminary range of the first target area can be expressed as the minimum envelope of the core data points and the boundary data points, and its spatial range is defined as the minimum three-dimensional spatial range that contains all relevant data points. For example, if the spatial coordinates of the core data points and the boundary data points are and , then the spatial range of the first target area is: .

[0108] In addition, in order to further improve the accuracy of area determination, the processing module 200 uses the physical structure information and electrical connection relationship information of the energy storage cabinet to correct the preliminary range.

[0109] In a specific implementation, the processing module 200 will analyze whether there are physical barriers between the target data points, such as fireproof partitions, electrical barriers, etc. If a target data point within the preliminary range is blocked from other areas, the target data point will be removed from the first target area. At the same time, the processing module 200 will combine the electrical connection relationship information to determine whether the component to which the target data point belongs has a direct electrical connection relationship with the component in the current area. If the component to which a target data point belongs has no direct electrical connection with the component in the current area, the target data point will also be excluded from the first target area. By combining the three-dimensional space model, physical structure information and electrical connection relationship information, the precise range of the first target area is finally determined.

[0110] For example, the physical structure information and electrical connection relationship information of the energy storage cabinet can be stored in a database, and the spatial and electrical relationships between components can be managed in the form of a graph data structure (Graph). The processing module 200 can use Python's NetworkX library to implement the judgment of the electrical connection relationship, and analyze whether there is a physical barrier in combination with the three-dimensional spatial coordinates of the target data point (such as excluding the target data points separated by fireproof partitions through spatial geometric calculations); After the first target area is determined, the processing module 200 further extracts target feature information from the area to comprehensively evaluate the state of the abnormal area and provide support for subsequent modules. These target features include the maximum temperature value, smoke concentration change rate, spatial distribution density, and the proportion of abnormal points.

[0111] Exemplarily, the maximum temperature value is determined by extracting the maximum value of the temperature values ​​of all target data points in the first target area, which is used to reflect the degree of temperature abnormality in the area; the smoke concentration change rate is obtained by calculating the average value of the smoke concentration change rate of all target data points in the first target area, which is used to evaluate the smoke change trend in the area; the spatial distribution density is determined by calculating the ratio of the number of target data points in the first target area to the volume of the area, which is used to reflect the distribution of the target data points in the area; the proportion of the number of abnormal points is determined by calculating the percentage of the number of abnormal points in the first target area to the total number of target data points, which is used to reflect the relative proportion of abnormal points.

[0112] Exemplarily, it is assumed that the three-dimensional spatial model of the energy storage cabinet includes battery cells, cooling components and other areas, and the core data points and boundary data points in the target data set are distributed in the battery cell A1 and cooling component B1 areas respectively. The processing module 200 first determines the preliminary range based on the density clustering results, and then excludes the target data points isolated by the fireproof partitions in combination with the physical structure information, and excludes the target data points that have no direct electrical connection with the current area in combination with the electrical connection relationship information. Finally, the first target area determined by the processing module 200 covers the battery cell A1 and the cooling component B1. Subsequently, the processing module 200 extracts the target features from the area: the maximum temperature value is 80°C, the smoke concentration change rate is 0.35 ppm / s, the spatial distribution density is 15 points / m³, and the number of abnormal points accounts for 60%. These features can fully reflect the state of the abnormal area and provide a basis for the early warning module 300 to generate early warning information.

[0113] In this way, the present invention can not only accurately locate the abnormal area, but also comprehensively evaluate the regional status by determining the first target area from the clustering results and extracting the target feature information, providing efficient and reliable technical support for the fire safety management of energy storage cabinets. This design is particularly suitable for scenarios with complex physical layouts and multi-component associations of energy storage cabinets, significantly improving the system's abnormality detection accuracy and response efficiency.

[0114] As an optional implementation manner, the matching of the target feature with corresponding warning information based on the preset warning mechanism includes: Based on the target characteristics, the first target area is classified into different warning levels; wherein the warning levels include: level one warning, level two warning, and level three warning; Based on the warning level, warning information corresponding to the warning level is generated.

[0115] Based on the warning information, executing firefighting processing corresponding to the warning information in the first target area includes: In response to the warning level being the first-level warning, triggering a cooling system in the energy storage cabinet to reduce the temperature of the first target area; In response to the warning level being the second-level warning, triggering the local fire protection system of the energy storage cabinet to spray a fire extinguishing medium toward the first target area; In response to the warning level being the third-level warning, the entire cabinet fire protection system of the energy storage cabinet is triggered, and a fire extinguishing medium is sprayed toward the first target area, and an emergency alarm signal is sent to notify an external control center.

[0116] In a specific implementation, the abnormal characteristics of the first target area are graded based on a preset early warning mechanism and corresponding early warning information is generated to guide the firefighting operation of the energy storage cabinet. Specifically, the first target area is divided into different early warning levels according to the target characteristics, and corresponding early warning information is generated by the early warning module 300. The execution module 400 takes corresponding firefighting measures in the first target area according to the early warning information to ensure that the abnormal situation is responded to quickly and accurately.

[0117] Among them, the early warning module 300 stores a multi-level early warning mechanism, which is based on target characteristics, including key parameters such as the maximum temperature value, the rate of change of smoke concentration, the spatial distribution density, and the proportion of abnormal points. The early warning module 300 determines the severity of the abnormality of the first target area based on these target characteristics, and classifies it into a first-level early warning, a second-level early warning, and a third-level early warning. The first-level early warning is used to indicate a minor abnormality, the second-level early warning is used to indicate a moderate abnormality, and the third-level early warning is used to indicate a serious abnormality or an emergency state.

[0118] In specific implementations, the classification of warning levels is based on preset threshold rules. For example: 1. Level 1 warning: If the highest temperature value in the first target area is arrive Between (such as ), and the smoke concentration change rate is low (such as less than 0.3 ppm / s), and the spatial distribution density is normal, it is determined to be a first-level warning; 2. Second level warning: If the maximum temperature exceeds But lower than (like ), or the smoke concentration change rate is high (such as within 0.3 to 0.6 ppm / s), it is judged as a second-level warning; 3. Level 3 warning: If the maximum temperature exceeds , or the smoke concentration change rate exceeds 0.6 ppm / s, or the number of abnormal points accounts for more than 50%, it will be judged as a level 3 warning.

[0119] The warning module 300 generates the above classification results as corresponding warning information and sends it to the execution module 400. The warning information includes: location information of the abnormal area, warning level, key target characteristic values ​​and recommended fire fighting measures. For example, if the first target area of ​​a certain energy storage cabinet is determined to be a level 3 warning, the warning information will indicate the spatial range of the area, the maximum temperature value (such as 95°C), the smoke concentration change rate (such as 0.7 ppm / s) and the recommended emergency fire fighting measures.

[0120] The execution module 400 responds to the abnormal situation in the first target area according to the warning information and takes fire prevention measures corresponding to the warning level: 1. Level 1 warning: When the warning level is level 1, the execution module 400 triggers the cooling system in the energy storage cabinet, such as starting a cooling fan or starting a liquid cooling cycle, to reduce the temperature of the first target area to prevent the temperature from rising further; 2. Level 2 warning: When the warning level is level 2, the execution module 400 triggers the local fire protection system of the energy storage cabinet, for example, spraying a fire extinguishing medium (such as carbon dioxide or inert gas) to the first target area to quickly suppress possible fire sources; 3. Level 3 warning: When the warning level is level 3, the execution module 400 triggers the whole cabinet fire protection system, sprays the fire extinguishing medium to the first target area, and sends an emergency alarm signal to notify the external control center. For example, high priority alarm information can be sent to the monitoring center through the remote communication module, and relevant data can be recorded for subsequent analysis.

[0121] For example, in a certain energy storage station, the early warning module 300 detects the target characteristic values ​​in the first target area: the maximum temperature is 85°C, the smoke concentration change rate is 0.5 ppm / s, the spatial distribution density is 12 points / m³, and the number of abnormal points accounts for 40%. According to the preset rules, the area is judged as a secondary warning. After the early warning information is generated, the execution module 400 triggers the local fire fighting system, sprays carbon dioxide fire extinguishing medium into the area, and starts the cooling system to assist in cooling, avoiding the risk of further expansion of the abnormality.

[0122] See also Figure 3~Figure 5 , Figure 3 A schematic diagram of the three-dimensional structure of an energy storage cabinet provided in an embodiment of the present disclosure; Figure 4 A schematic diagram of the front structure of a cabinet provided in an embodiment of the present disclosure; Figure 5A schematic diagram of the back structure of a cabinet provided in an embodiment of the present disclosure. For example, in a certain energy storage station, the energy storage cabinet is equipped with a cooling circulation system and a fire protection piping system for temperature control management and fire protection of the battery PACK box 7 in the energy storage cabinet and the surrounding environment. During system operation, the processing module 200 obtains the target data in the energy storage cabinet in real time through the acquisition module 100, including the temperature distribution of the battery PACK box 7, changes in smoke concentration, etc. The processing module 200 uses a clustering algorithm combined with physical layout analysis to identify abnormal areas (first target areas) in the energy storage cabinet, and evaluates the warning level of the area through the warning module 300 to generate corresponding warning information.

[0123] When the warning module 300 detects that the highest temperature value in the first target area is 75°C, the smoke concentration change rate is 0.2 ppm / s, and the spatial distribution density is within the normal range (10 points / m³), it is determined to be a first-level warning. At this time, the execution module 400 starts the cooling circulation system of the energy storage cabinet and activates the refrigeration pipeline for local cooling. The specific processing process is as follows: the low-temperature coolant in the refrigerant box 3 is pumped into the water inlet main pipe 4 by the water pump 1, diverted to the water inlet branch pipe 11, and then flows into the battery PACK box 7 through the water inlet branch pipe 10. After the coolant exchanges heat with the heat generated by the battery heating, the high-temperature coolant flows out of the battery PACK box 7, enters the return water branch pipe 8, flows into the return water branch pipe 9 from the return water branch pipe 8, and then flows into the return water main pipe 6 from the return water branch pipe 9, and flows back to the refrigerant box 3 through the refrigerant box inlet 5 to complete the cooling cycle. The system monitors the inlet and return water temperatures of the coolant in real time to ensure the cooling effect. As the temperature gradually drops, the temperature of the target data point returns to the normal range (below 60°C), and the system cancels the first-level warning.

[0124] When the warning module 300 detects that the maximum temperature of the first target area rises to 85°C, the smoke concentration change rate reaches 0.5 ppm / s, and the number of abnormal points accounts for 40%, it is determined to be a secondary warning. The execution module 400 triggers the local fire protection system of the energy storage cabinet, and quickly cools down and extinguishes the abnormal area through the fire protection pipeline. The specific processing process is as follows: After the external fire protection measures are activated, the high-pressure fire protection water enters the fire protection pipeline from the fire protection water inlet pipe 2, flows into the return water main pipe 6 through the fire protection anti-reverse valve 12, and then enters the battery PACK box 7 through the return water branch pipe 9 and the return water branch pipe 8, and directly contacts the high-temperature area in the box for cooling and extinguishing. Part of the fire protection water flows into the refrigerant box 3 through the return water main pipe 6, and the coolant further reduces the temperature and provides support for the internal cooling system of the energy storage cabinet. The system monitors the temperature and smoke concentration changes of the battery PACK box 7 in real time to confirm the effectiveness of the cooling and fire extinguishing measures.

[0125] When the warning module 300 detects that the maximum temperature of the first target area rises to above 95°C, the smoke concentration change rate exceeds 0.7 ppm / s, and the number of abnormal points accounts for 60%, it is judged as a third-level warning. At this time, the execution module 400 starts the full cabinet fire protection system of the energy storage cabinet, and cools and extinguishes the inside and outside of the cabinet through the fire protection pipeline. The specific processing process is as follows: high-pressure fire water enters the fire protection pipeline from the fire protection water inlet pipe 2, and flows into the return water main pipe 6 through the fire protection anti-reverse valve 12, and is divided into three routes: the first and second routes flow into the battery PACK box 7 through the return water branch 8 and the refrigerant box 3 to cool the internal high-temperature battery unit and extinguish the fire; the third route flows into the fire sprinkler 14 on the top of the cabinet through the fire spray pipe 13 to cool and extinguish the outside of the cabinet and the surrounding area.

[0126] The system also sends three-level warning information to the external monitoring center, including the specific location of the abnormal area, the current temperature (such as 95°C), the rate of change of smoke concentration (such as 0.8 ppm / s) and the firefighting measures taken. After receiving the information, the monitoring center can further intervene remotely.

[0127] In this way, by combining the cooling pipeline and fire protection pipeline system of the energy storage cabinet, the present invention can achieve accurate response of graded warning. In the first-level warning, the local cooling treatment of the cooling circulation system can timely control minor abnormalities; in the second-level and third-level warnings, the local or full-cabinet fire extinguishing measures of the fire protection pipeline system can quickly suppress the fire source and ensure the safe operation of the energy storage cabinet. In addition, the external alarm function of the third-level warning further improves the overall safety and reliability of the system.

[0128] In this way, the present invention can effectively deal with various abnormal situations in the energy storage cabinet by combining a multi-level warning mechanism with flexible fire treatment measures. By performing detailed analysis and classification of the target characteristics of the first target area, the present invention can accurately generate warning information and select the most appropriate fire fighting means according to the abnormality level, thereby significantly improving the system response efficiency while ensuring the safety of the energy storage cabinet and reducing the cost of using fire fighting resources.

[0129] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0130] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0131] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0132] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0133] The preferred embodiments of the present invention disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details in detail, nor do they limit the present application to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can understand and use the present application well. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. An automated energy storage cabinet fire safety system, characterized in that: The system comprises: The acquisition module is used to collect the target data of the energy storage cabinet at the preset monitoring points in real time and generate a target data set; a processing module, configured to perform cluster analysis on the target data set based on a density clustering method after performing data screening in the target data set based on a temperature threshold and / or a smoke threshold and a time threshold, determine a first target area of ​​the energy storage cabinet, and extract target features from the first target area; The density clustering method includes: expanding the initial clustering range based on the spatial coordinates, physical regions and components of each target data point in the initial clustering range, and spatial constraints; and dynamically adjusting the minimum point count threshold and neighborhood distance threshold in the density clustering method based on the abnormality index of the target data point; An early warning module matches corresponding early warning information for the target features based on a preset early warning mechanism; An execution module is used to execute fire fighting processing corresponding to the warning information in the first target area based on the warning information.

2. The system according to claim 1, characterized in that The acquisition module also includes: A sensor network, used for collecting the target data of each preset monitoring point in the energy storage cabinet in real time; wherein the sensor network includes at least one sensor; The preprocessing unit is used to preprocess the target data of each preset monitoring point in the energy storage cabinet to generate the target data set.

3. The system according to claim 2, characterized in that The density clustering method includes: Set a minimum point count threshold and a neighborhood distance threshold, wherein the minimum point count threshold is used to limit the number of target data points required to form a cluster; the neighborhood distance threshold is used to limit the maximum distance between target data points to determine the neighborhood range; the neighborhood range is an area centered on each target data point and with a radius equal to the neighborhood distance threshold; Based on the neighborhood range, determine the core data in the target data set, where the core data is the target data whose number of target data in the neighborhood range is greater than or equal to a minimum point count threshold; Based on the neighborhood range of the core data, determine an initial cluster, and classify all target data within the neighborhood range of the core data into the initial cluster; Wherein, the target data set includes multiple target data points.

4. The system according to claim 3, characterized in that The density clustering method also includes: Under the preset spatial constraints, the boundary data within the neighborhood of the core data are included in the cluster; For the newly included boundary data, if it meets the core data conditions and conforms to the spatial constraints, continue to expand the target data within its neighborhood; Iteratively expand the clustering until no new target data is included in the clustering range.

5. The system according to claim 4, characterized in that Performing data screening based on the temperature threshold and / or the smoke threshold and the time threshold in the target data set includes: Set time thresholds; Screening the target data set based on the time threshold includes: Determining whether the target data point is greater than a temperature threshold and / or a smoke threshold and continues to exceed the time threshold; In response to the target data point being greater than the temperature threshold and / or the smoke threshold but not exceeding the time threshold, it is excluded from the target data set.

6. The system according to claim 5, characterized in that The setting of the minimum point number threshold and the neighborhood distance threshold includes: Based on the abnormality index of the target data point, dynamically adjust the minimum point count threshold and the neighborhood distance threshold; Wherein, the determination of the abnormality degree index includes: quantifying the difference between the temperature value of the target data point and the temperature threshold to obtain a temperature excess metric; Monitoring the smoke concentration change rate of the target data point and determining the smoke concentration change gradient of the target data point; Inputting the temperature over-limit measure and the smoke concentration change gradient into a preset abnormality degree assessment model, and outputting the corresponding abnormality degree index; The dynamically adjusting the minimum point number threshold and the neighborhood distance threshold comprises: When the abnormality index is greater than or equal to a preset abnormality index threshold, the neighborhood distance threshold is increased and the minimum point threshold is decreased to expand the clustering range; When the abnormality index is less than a preset abnormality index threshold, the neighborhood distance threshold is reduced and the minimum point threshold is increased to narrow the clustering range.

7. The system according to claim 6, characterized in that Determining the first target area of ​​the energy storage cabinet based on the clustering result includes: determining the spatial range of the first target area based on the core data and boundary data in the target data set; Extracting target features from the first target area includes: Extracting the highest temperature value of the target data; Extracting the smoke concentration change rate of the target data; Extracting the spatial distribution density of the target data; The number and proportion of abnormal points in the target data are extracted.

8. The system according to claim 7, characterized in that The density clustering method also includes: Based on the three-dimensional spatial model of the energy storage cabinet, determining the spatial coordinates of each of the target data points; Based on the spatial coordinates of the target data points, determining the physical area and component to which each target data point belongs; Determining space constraints based on the physical structure information and electrical connection relationship information of the energy storage cabinet; The expanding the initial clustering range based on the spatial coordinates, the physical area and the component to which each target data point belongs, and the spatial constraint condition in the initial clustering range includes: For the target data points to be expanded that are not within the initial clustering range, determine whether the following conditions are met at the same time: The physical region to which the target data point to be expanded belongs is directly adjacent to the physical region to which at least one target data point within the current clustering range belongs in structure, and there is no physical separation between the two; The component to which the target data point to be expanded belongs is directly and electrically connected to the component to which at least one target data point within the current clustering range belongs; If all the above conditions are met, the target data point to be expanded will be included in the clustering range.

9. The system according to claim 8, characterized in that The preset warning mechanism is based on which the target feature is matched with corresponding warning information, including: Based on the target characteristics, the first target area is classified into different warning levels; wherein the warning levels include: level one warning, level two warning, and level three warning; Based on the warning level, generating warning information corresponding to the warning level; Based on the warning information, executing firefighting processing corresponding to the warning information in the first target area includes: In response to the warning level being the first-level warning, triggering a cooling system in the energy storage cabinet to reduce the temperature of the first target area; In response to the warning level being the second-level warning, triggering the local fire protection system of the energy storage cabinet to spray a fire extinguishing medium toward the first target area; In response to the warning level being the third-level warning, the entire cabinet fire protection system of the energy storage cabinet is triggered, and a fire extinguishing medium is sprayed toward the first target area, and an emergency alarm signal is sent to notify an external control center.

Images