Automatic trigger control system of condensed aerosol fire extinguishing device

Abstract

The invention relates to the technical field of program control, in particular to an automatic trigger control system for a condensed aerosol fire extinguishing device, which is implemented by the following steps of: acquiring multiple types of monitoring data of the condensed aerosol fire extinguishing device in a current preset time period; determining a trend deviation anomaly degree and a fire risk degree of each type of monitoring data at each moment according to the multiple types of monitoring data, and further determining an attention necessity degree of each type of monitoring data at each moment; determining a fire judgment index at each moment according to the association condition and the fire risk consistency condition between each type of monitoring data and the comparison type of monitoring data thereof in combination with the attention necessity; and performing automatic trigger control on the hot aerosol fire extinguishing device based on the fire hazard judgment index. According to the method, the fire judgment index at each moment is determined by combining multiple types of monitoring data, so that the accuracy and timeliness of hot aerosol fire extinguishing triggering can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of program control technology, specifically to an automatic triggering control system for a thermal aerosol fire extinguishing device. Background Technology

[0002] Thermal aerosols are highly efficient fire extinguishing media that utilize solid generating agents to rapidly undergo a chemical reaction upon ignition, producing a large number of micron-sized suspended particles that diffuse into space along with inert gases. Thermal aerosol fire extinguishing devices primarily achieve rapid control of electrical fires, solid surface fires, and early-stage fires in enclosed spaces through a combination of heat absorption cooling, oxygen concentration dilution, and chemical inhibition of combustion free radicals. They are widely used in small, enclosed or semi-enclosed environments such as electrical cabinets, power distribution rooms, and energy storage boxes.

[0003] In existing technologies, to improve the accuracy of automatic triggering of thermal aerosol fire extinguishing devices, a common approach is to fuse multimodal sensor data, such as temperature, smoke, electrical status, and gas concentration, for fire identification. However, this approach has significant drawbacks in practical applications: due to the inherent differences in sampling frequency, noise level, and response speed among different sensor modalities, and the complex coupling between their features, it is difficult for a general data fusion model to stably and reliably extract the true fire characteristics.

[0004] Especially in typical application scenarios such as electrical cabinets and energy storage warehouses, the environment is complex and often accompanied by numerous non-fire-related interference factors such as dust, ventilation disturbances, and normal heat accumulation from equipment. These factors can easily lead to misjudgments, delayed judgments, or oversensitivity to data from a particular mode in the fusion triggering logic. More seriously, when an individual sensor malfunctions or its data deviates significantly, the entire system may trigger incorrectly due to dependency imbalance, or fail to trigger in a real fire situation, severely weakening the reliability of automatic fire suppression. Therefore, how to achieve both accurate and timely triggering of thermal aerosol fire suppression in complex and ever-changing field environments has become a core problem that urgently needs to be solved in existing technologies. Summary of the Invention

[0005] To address the issues of misjudgment, delayed judgment, or unreliable triggering in existing multimodal data fusion-based automatic triggering technologies for thermal aerosol fire extinguishing systems, which are prone to problems in complex field environments due to data heterogeneity and interference, this invention aims to provide an automatic triggering control system for thermal aerosol fire extinguishing devices. The specific technical solution adopted is as follows:

[0006] One embodiment of the present invention provides an automatic triggering control system for a thermal aerosol fire extinguishing device, including a memory and a processor. The processor executes a computer program stored in the memory to achieve the following process:

[0007] Acquire multiple types of monitoring data of the thermal aerosol fire extinguishing device during the current preset time period; the multiple types of monitoring data include at least temperature data, smoke concentration data, and electrical data, and the electrical data includes at least current and voltage.

[0008] Based on the analysis of the changing trends of the same type of monitoring data at each time point, the degree of deviation of the trend of each type of monitoring data at each time point is determined.

[0009] Determine the fire risk level of each type of monitoring data at each time point, and combine the trend deviation anomaly level to obtain the attention necessity of each type of monitoring data at each time point;

[0010] Based on the correlation between each type of monitoring data and its comparison type monitoring data, and the consistency of fire risk, combined with the aforementioned level of attention necessity, the contribution of each type of monitoring data to fire extinguishing trigger control at each moment is determined.

[0011] The contribution level is used as the judgment weight to determine the fire judgment index at each moment, and the thermal aerosol fire extinguishing device is automatically triggered and controlled.

[0012] Furthermore, the step of analyzing the changing trend of the same type of monitoring data based on the multi-type monitoring data at each time moment, and determining the degree of trend deviation anomaly of each type of monitoring data at each time moment, includes:

[0013] For the same type of monitoring data, the first trend change factor for each time moment is determined based on the difference between the maximum and minimum monitoring data in the time neighborhood of each time moment;

[0014] The second trend change factor for each moment is determined based on the difference between the instantaneous rate of change and the local average rate of change of the monitoring data at each moment.

[0015] Based on the analysis of the first and second trend change factors at each time point, the deviation of the current type of monitoring data change trend from the historical trend benchmark is determined, and the degree of trend deviation anomaly of the current type of monitoring data at each time point is determined.

[0016] Further, obtaining the local average rate of change includes:

[0017] For each time moment, acquire all monitoring data within the time neighborhood of the current time moment, and perform linear fitting on all monitoring data within the time neighborhood to obtain the monitoring fitting line for the current time moment;

[0018] Determine the slope of the monitoring fitted line at the current moment, and use the slope of the monitoring fitted line as the local average rate of change.

[0019] Furthermore, determining the degree of trend deviation anomaly of the current type of monitoring data at each time point includes:

[0020] The first and second trend change factors at the current moment are fused to obtain the trend change performance at the current moment.

[0021] Establish a historical trend benchmark; the historical trend benchmark is at least used to characterize the trend benchmark value under historical stable operating conditions;

[0022] Based on the difference between the current trend change performance and the historical trend benchmark, determine the degree of trend deviation anomaly of the current type of monitoring data at the current moment.

[0023] Furthermore, determining the fire risk level of each type of monitoring data at each time moment includes:

[0024] For each type, the first fire risk factor of the monitoring data at each moment is determined based on the proximity of the monitoring data at each moment to the warning threshold of the corresponding type.

[0025] Based on the local average rate of change corresponding to the monitoring data at each time moment, the second fire risk factor of the monitoring data at each time moment is determined;

[0026] The first and second fire risk factors of the monitoring data at the same time are weighted and summed to obtain the fire risk level of the corresponding type of monitoring data at each time.

[0027] Furthermore, obtaining the fire risk level of the corresponding type of monitoring data at each moment includes:

[0028] For each moment of the monitoring data, a weight is set for the first fire risk factor, denoted as the first weight; and a weight is set for the second fire risk factor, denoted as the second weight.

[0029] Based on the first and second weights, the first fire risk factor and the second fire risk factor are weighted and summed to obtain the fire risk level at the corresponding time.

[0030] Furthermore, based on the correlation and consistency of fire risk between each type of monitoring data and its comparative type of monitoring data, and in conjunction with the degree of attention required, the contribution of each type of monitoring data to the fire extinguishing trigger control at each time point is determined, including:

[0031] Use any type as the target type, and use other types besides the target type as the comparison type;

[0032] Based on the correlation between the monitoring data of the target type and the monitoring data of each comparison type, and the consistency of fire risk, the early warning credibility of the target type monitoring data at each time point is determined;

[0033] Based on the early warning reliability, fire risk, and necessity of attention of the target type monitoring data at each moment, the contribution of the target type monitoring data to the fire extinguishing trigger control at each moment is determined.

[0034] Furthermore, determining the early warning reliability of the target type monitoring data at each time step includes:

[0035] The correlation between the target type and each comparison type is determined based on the monitoring data change curves of the target type and each comparison type; the monitoring data change curves are obtained by curve fitting of monitoring data of the same type at multiple time points.

[0036] Based on the similarity between the fire risk levels of the target type and each comparison type at the same time, determine the risk level consistency index between the target type and each comparison type at each time.

[0037] The credibility of the early warning data of the target type at each time point is determined based on the correlation between the target type and each comparison type, as well as the consistency index of the risk level at each time point.

[0038] Furthermore, the determination of the contribution of target type monitoring data to fire extinguishing trigger control at each moment, based on the early warning reliability, fire risk, and necessity of attention of the target type monitoring data at each moment, includes:

[0039] The fire risk level of the target type monitoring data at each time moment is normalized, and the normalized value is used as the sensitivity at the corresponding time moment.

[0040] The early warning reliability, sensitivity, and necessity of attention of target type monitoring data at the same time are fused to determine the contribution of target type monitoring data to fire extinguishing trigger control at the corresponding time.

[0041] Furthermore, the step of using the contribution as a judgment weight to determine the fire judgment index at each moment includes:

[0042] The risk of fire ignition for each type of monitoring data is obtained at each moment; the risk is calculated in real time using a machine learning algorithm.

[0043] Based on the judgment weight of each type of monitoring data at each time, the fire risk of different types of monitoring data at each time is weighted and averaged to obtain the fire judgment index at each time.

[0044] The present invention has the following beneficial effects:

[0045] This invention first acquires multiple types of monitoring data, directly addressing the problem in existing technologies where data sources are either single or incomplete, failing to fully reflect complex fire situations. By acquiring multiple types of monitoring data, a comprehensive and multi-dimensional fire characteristic perception system is established, laying the data foundation for subsequent intelligent analysis. Second, it determines the degree of trend deviation anomaly of each type of monitoring data at each moment. This introduces time-series trend analysis to quantify the degree of deviation between the current trend and historical normal patterns. This addresses the problems of existing instantaneous threshold-based judgment methods being slow to respond to slowly evolving fires and unable to effectively distinguish between continuous anomalies and transient interference. By determining the degree of trend deviation anomaly, the detection sensitivity for early, slowly developing fires is significantly improved, while the anti-interference capability against instantaneous interference fluctuations is enhanced, facilitating the extraction of more reliable fire characteristics from a time dimension. Next, the necessity of attention for each type of monitoring data at each moment is determined. This integrates two orthogonal dimensions: the current state of danger and the degree of abnormality in behavioral patterns. This aims to address the problem of a single dimension in the assessment of abnormal parameter states in existing technologies. By determining the necessity of attention, a more refined and robust assessment of single-parameter risk is achieved. This allows the system to focus on parameters approaching the danger threshold, as well as those with low absolute values ​​but evolving in an abnormal pattern. This provides a more scientific basis for prioritizing parameters at the current moment, offering crucial input for subsequent weight allocation. Then, the contribution of each type of monitoring data to fire suppression triggering control at each moment is determined. This is achieved by constructing a cross-validation mechanism, using other types of data to verify the reliability and consistency of the target type of data. This is key to solving core problems such as unstable multimodal data fusion, susceptibility to interference from individual sensor anomalies, and imbalanced triggering logic. Determining the contribution effectively identifies and reduces the decision weight of isolated abnormal signals caused by dust, ventilation disturbances, local heat sources, etc., enhances false negative protection, enables dynamic adaptation, avoids overall system failure, and significantly improves system robustness.

[0046] Finally, the contribution level is used as the judgment weight to determine the fire judgment index at each moment for automatic triggering control. This realizes a fusion decision based on dynamic credibility weight, which aims to replace the simple and fixed data fusion or voting strategy in the prior art and solve the problem of misjudgment or delayed judgment caused by unreasonable weight. Through this invention, the final triggering decision is the most credible and urgent evidence-driven intelligent decision, which fundamentally guarantees the accuracy and timeliness of the triggering action and transforms intelligent analysis into reliable control action. Attached Figure Description

[0047] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 The following is an execution flowchart of an automatic triggering control system for a thermal aerosol fire extinguishing device, according to an embodiment of the present invention.

[0049] Figure 2 This is a flowchart illustrating the steps for determining the degree of trend deviation anomaly of each type of monitoring data at each time point in an embodiment of the present invention.

[0050] Figure 3 This is a flowchart illustrating the steps for determining the contribution of each type of monitoring data to the fire extinguishing trigger control at each moment in an embodiment of the present invention.

[0051] Figure 4 This is a flowchart of the fire extinguishing closed-loop control in an embodiment of the present invention. Detailed Implementation

[0052] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the technical solution proposed according to the present invention are described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0054] One embodiment of the present invention provides an automatic triggering control system for a thermal aerosol fire extinguishing device, including a memory and a processor, wherein the processor executes a computer program stored in the memory to achieve the following process:

[0055] Acquire multiple types of monitoring data of the thermal aerosol fire extinguishing device during the current preset time period. The multiple types of monitoring data include at least temperature data, smoke concentration data, and electrical data. The electrical data includes at least current and voltage.

[0056] Based on the analysis of the changing trends of the same type of monitoring data at each time point, the degree of deviation of the trend of each type of monitoring data at each time point is determined.

[0057] Determine the fire risk level of each type of monitoring data at each time point, and combine the trend deviation anomaly level to obtain the attention necessity of each type of monitoring data at each time point;

[0058] Based on the correlation between each type of monitoring data and its comparison type monitoring data, and the consistency of fire risk, combined with the aforementioned level of attention necessity, the contribution of each type of monitoring data to fire extinguishing trigger control at each moment is determined.

[0059] The contribution level is used as the judgment weight to determine the fire judgment index at each moment, and the thermal aerosol fire extinguishing device is automatically triggered and controlled.

[0060] The following is a detailed explanation of each of the above steps:

[0061] refer to Figure 1 The diagram illustrates an execution flowchart of an automatic triggering control system for a thermal aerosol fire extinguishing device according to an embodiment of the present invention, comprising:

[0062] S1, acquire multiple types of monitoring data of the thermal aerosol fire extinguishing device during the current preset time period.

[0063] Here, various types of monitoring data are used to characterize the environmental conditions and equipment operation status corresponding to the thermal aerosol fire extinguishing device, and are fundamental data that can reflect the development trend of the fire. These various types of monitoring data include at least temperature data, smoke concentration data, and electrical data, with the electrical data including at least current and voltage.

[0064] In this embodiment, multiple types of sensors are deployed within the deployment area of ​​the thermal aerosol fire extinguishing device to collect real-time data on environmental conditions and equipment operation. The deployment area includes, but is not limited to, electrical cabinets, energy storage compartments, distribution boxes, and engine rooms.

[0065] Specifically, temperature sensors are deployed to record the air temperature in small, enclosed spaces at a high sampling frequency. For example, thermistors are used to collect the air temperature inside electrical cabinets to obtain temperature data at each moment within the current preset time period. Photoelectric smoke sensors are deployed to monitor smoke changes to obtain smoke concentration data at each moment within the current preset time period. Inside electrical cabinets, energy storage power supply compartments, or distribution boxes, current transformers, voltage sampling modules, etc., are arranged in corresponding positions according to the busbar layout, cable routing, and location of heat points to monitor electrical data such as current and voltage in real time and obtain electrical data at each moment within the current preset time period.

[0066] The length of the current preset time period must cover the typical cycle of early fire evolution to provide a statistically significant data basis for trend analysis. The sampling frequency must satisfy the Nyquist sampling theorem, meaning it must be twice the highest possible frequency of change of the monitored physical quantity under fire conditions to ensure distortion-free capture of rapid changes in key signals. Different types of monitoring data can use a consistent sampling frequency or a sampling frequency that matches their signal characteristics. The selection of the time neighborhood (used to calculate local features) is the result of a trade-off between smoothing random noise and maintaining sensitivity to real changes.

[0067] Preferably, the specific values ​​of the aforementioned time parameters can be optimized based on historical data and experiments of the target scenario, and can be adaptively calibrated after deployment. Those skilled in the art can make reasonable settings under the guidance of the above principles.

[0068] It should be noted that, in order to facilitate subsequent data analysis of monitoring data of different modes, all collected data underwent preprocessing operations such as timestamp synchronization, noise filtering, and normalization, and the monitoring data analyzed later are all preprocessed monitoring data.

[0069] Thus, this embodiment has obtained multiple types of monitoring data of the thermal aerosol fire extinguishing device during the current preset time period.

[0070] In the embodiments of this invention, to ensure the stability of the calculation and the clarity of the physical meaning, all the following calculation formulas follow the following principles: In all formulas involving fractional calculations, if the denominator has the possibility of being zero, a very small positive constant, such as 0.01, is added to the denominator to prevent calculation interruption; to eliminate the problem caused by inconsistent units of different physical quantities, all intermediate calculation indices are processed into dimensionless relative values ​​through mathematical methods such as normalization, usually mapped to the interval [0,1]. Among them, the typical normalization method is maximum-minimum value normalization.

[0071] S2, based on the multi-type monitoring data at each time point, analyze the changing trend of the same type of monitoring data, and determine the degree of deviation of the trend of each type of monitoring data from the anomaly at each time point.

[0072] In typical application scenarios such as electrical cabinets and energy storage boxes, many fires originate from a slow evolution process, such as gradual temperature rise of equipment, localized heat accumulation, latent arcing, or early insulation breakdown. Traditional judgment methods based on fixed instantaneous thresholds are difficult to sensitively capture such weak early signals, easily leading to response delays. At the same time, this method is susceptible to non-fire factors such as environmental interference, internal power disturbances, normal load changes, or ventilation fluctuations, which may cause false alarms or even false triggering of fire suppression, reducing the overall reliability of the system.

[0073] To address the aforementioned issues, step S2 of this embodiment emphasizes dynamic trend analysis of key fire monitoring parameters. This method aims to characterize the evolution trajectory of a fire (i.e., from its initial state to its intensity) over a temporal dimension, effectively distinguishing the early trends of a real fire from normal operational fluctuations, equipment noise, or transient interference. By quantitatively assessing the degree of anomaly in parameter change trends, the system can significantly improve its sensitivity to initial fire signals and enhance its anti-interference capabilities, thereby achieving more reliable and precise fire suppression triggering control in complex environments.

[0074] Based on the above analysis, the trend deviation anomaly can be determined by comparing the changes in monitoring data during the current preset time period with preset historical benchmark values. It comprehensively reflects the anomaly of different types of monitoring data within a recent time window, including the magnitude, speed, persistence of changes, and the degree of deviation from historical stable operating conditions. It can serve as an important basis for enhancing early fire identification sensitivity and suppressing false triggering. A larger trend deviation anomaly indicates a more significant deviation of the monitoring parameter's change trend from its normal or expected behavior pattern, and is more likely to indicate the existence of an atypical abnormal process, i.e., a significant anomaly.

[0075] As an exemplary implementation, taking any type of monitoring data as an example, the above-mentioned determination of the trend deviation anomaly of each type of monitoring data at each time point can be achieved through... Figure 2 The steps shown are to be implemented as follows:

[0076] S21. Determine the first trend change factor for each time step based on the difference between the maximum and minimum monitoring data within the time neighborhood of each time step.

[0077] Before the fire suppression system is activated, monitoring data, such as temperature data, shows a significant upward trend. The changing trends of monitoring data are crucial for determining the likelihood of a fire; therefore, it is necessary to analyze the real-time trends of the monitoring data.

[0078] Here, the first trend change factor is used at least to characterize the severity of changes in the monitoring data, that is, the magnitude of the fluctuation range of the monitoring data within a given short time window. The larger the fluctuation range, the more drastic the fluctuations in the monitoring data during that period, which may indicate instability of the heat source or accelerated energy release.

[0079] In this embodiment, the difference between the maximum and minimum values ​​in the time neighborhood at the same moment is calculated, the difference between the maximum and minimum values ​​is normalized, and the normalized value is used as the first trend change factor at the corresponding moment.

[0080] The difference can be normalized by using the maximum monitoring data; the time neighborhood can be set to the 10 moments before the current moment, where the current moment is any moment within the current preset time period. For the monitoring data of the 10 moments before the current preset time period, its time neighborhood is the data of all historical moments from itself to the initial moment.

[0081] S22, based on the difference between the instantaneous rate of change and the local average rate of change of the monitoring data at each moment, determine the second trend change factor at each moment.

[0082] Here, the second trend change factor is used at least to characterize the persistence of the trend in the monitored data, i.e., whether the abnormal trend is a short-term fluctuation or a continuous development. While the instantaneous rate of change reflects the speed of change at the current moment, it may be greatly affected by noise. The local average rate of change, on the other hand, smooths out noise and reflects the overall direction and average rate of change over that time period. By comparing the difference between the instantaneous rate of change and the local average rate of change, the consistency or stability of the current change is assessed. If the difference between the instantaneous rate of change and the local average rate of change is small, it indicates that the trend is stable and persistent; if the difference is large, it may be due to noise interference or a trend reversal.

[0083] In this embodiment, the instantaneous rate of change is obtained by dividing the difference between the monitoring data at the current moment and the monitoring data at the previous moment on the fitted curve by the sampling time interval. All monitoring data within the time neighborhood of the current moment are acquired, and a straight line is fitted to all monitoring data within the time neighborhood to obtain the monitoring fitted line at the current moment. The slope of the monitoring fitted line at the current moment is determined, and the slope of the monitoring fitted line is used as the local average rate of change.

[0084] One approach is to use the least squares method to perform linear fitting on all monitoring data within the time neighborhood. Of course, other methods can also be used to achieve linear fitting, and no specific limitations are made here.

[0085] In this embodiment, the absolute value of the difference between the instantaneous rate of change and the local average rate of change of the monitoring data at the same moment is calculated; the absolute value of the difference between the instantaneous rate of change and the local average rate of change is normalized, and the normalized value is used as the second trend change factor at the corresponding moment.

[0086] S23, based on the first trend change factor and the second trend change factor at each time point, analyze the deviation of the current type of monitoring data change trend from the historical trend benchmark, and determine the degree of trend deviation anomaly of the current type of monitoring data at each time point.

[0087] A single, absolute trend measure (such as the rate of change) can still be affected by cyclical fluctuations in normal equipment operation. Therefore, the final judgment needs to be based on a comparison with historical trend benchmarks. Here, the historical trend benchmark is used at least to characterize the trend benchmark value under historical stable operating conditions, representing the typical change patterns and fluctuation ranges of different types of monitoring data under normal stable operation without fire.

[0088] By quantifying the degree of trend deviation anomaly, we can determine the extent, in terms of its intensity and persistence, of the statistical deviation from historical normal conditions in the changing behavior of the current type of monitoring data. Specifically, we compare the two dimensions mentioned above (the current trend performance characterized by the combined first and second trend change factors) with the historical trend benchmark of this type of monitoring data under similar operating conditions to determine the degree of trend deviation anomaly.

[0089] As an exemplary implementation, step S23 described above can be achieved through the following steps:

[0090] The first step is to perform data fusion processing on the first and second trend change factors at the current moment to obtain the trend change performance at the current moment.

[0091] The first trend change factor directly responds to the rapid changes in monitoring data within a short period of time, forming a large range of changes. It is used to quantify the magnitude and severity of the changes, which is the basic dimension for judging whether the situation is serious. The second trend change factor directly responds to the requirement that the data changes within a relatively neighborhood range still maintain a large range of changes. By comparing the instantaneous and average change rates, the stability and consistency of the trend can be quantified, and random fluctuations can be eliminated.

[0092] In this embodiment, the product of the first trend change factor and the second trend change factor at the current moment is calculated, and the product of the two trend change factors is used as the trend change performance at the current moment.

[0093] As an example, the formula for calculating the trend change performance of the i-th type of monitoring data at time j can be:

[0094] In the formula, This indicates the degree of trend change performance of the i-th type of monitoring data at time j. This represents the difference between the maximum and minimum values ​​of the i-th type of monitoring data within its time neighborhood at time j. This represents the monitoring data of the i-th type of monitoring data within the time neighborhood of the j-th time point, and max represents the function for finding the maximum value. Represents the first trend change factor of the i-th type of monitoring data at time j; norm represents the normalization function. This represents the instantaneous rate of change of the i-th type of monitoring data at time j. This represents the local average rate of change of the i-th type of monitoring data at time j. This represents the second trend change factor of the i-th type of monitoring data at time j.

[0095] In the formula for calculating the degree of trend change, to avoid the extreme case where the denominator of the fraction is zero, a non-zero constant, such as 0.01, can be added to the denominator.

[0096] The second step is to set a benchmark for historical trends.

[0097] In this embodiment, during the equipment commissioning phase or normal operation, multiple historical time periods representing stable equipment operating conditions are collected, excluding fires or obvious anomalies. For each type of monitoring data within each historical time period, the trend change performance at each moment is calculated according to the method described in step S23. Finally, all calculated historical trend change performance values ​​are statistically analyzed, and their average or median is used as the historical trend benchmark for that type of monitoring data. Each type of monitoring data has its corresponding historical trend benchmark, which can be stored in the system for online real-time comparison.

[0098] The third step is to determine the degree of trend deviation anomaly of the current type of monitoring data at the current moment based on the difference between the current trend performance and the historical trend benchmark.

[0099] The greater the anomaly of the trend deviation, the stronger the indication that the current parameter changes are not only drastic and continuous, but also significantly deviate from the historical pattern of normal equipment operation. This suggests a higher probability that the changes are due to abnormal processes such as the early evolution of a fire, rather than normal fluctuations or disturbances in operating conditions. Therefore, determining the anomaly of the trend deviation provides a crucial and reliable input for subsequent calculations and for achieving reliable triggering.

[0100] In this embodiment, the absolute value of the difference between the current trend change performance and the historical trend benchmark is calculated as the numerator of the ratio, the historical trend benchmark is used as the denominator of the ratio, and the ratio is used as the trend deviation anomaly of the current type of monitoring data at the current moment.

[0101] As an example, the formula for calculating the trend deviation anomaly of the i-th type of monitoring data at time j can be:

[0102] In the formula, This indicates the degree of trend deviation anomaly of the i-th type of monitoring data at time j. This indicates the degree of trend deviation anomaly of the i-th type of monitoring data at time j. Indicating a baseline for historical trends, This represents the function for finding the absolute value.

[0103] The trend deviation anomaly determined by step S2 above can be used to analyze whether the current parameter changes may originate from potential early fire evolution (such as slow temperature rise, local heat accumulation, hidden electric arc, etc.) rather than normal operating condition fluctuations (such as load changes, ventilation disturbances, periodic temperature fluctuations) or random noise in the context of fire identification, and provide a reliable parameter basis for the subsequent quantification of the necessity of attention.

[0104] S3 determines the fire risk level of each type of monitoring data at each time point, and combines the trend deviation anomaly level to obtain the attention necessity of each type of monitoring data at each time point.

[0105] In the automatic triggering control of thermal aerosol fire extinguishing systems, before conducting contribution analysis on multi-source monitoring data, it is first necessary to assess the necessity of attention for each type of monitoring data at each moment. Early fire identification relies not only on the instantaneous absolute values ​​of parameters but also on the anomalies in their changing trends. Therefore, the necessity of attention can be quantified by combining two dimensions: fire risk level and trend deviation anomaly, to determine the indicative value and urgency of different types of monitoring data at each moment.

[0106] As an exemplary implementation, determining the fire risk level of each type of monitoring data at each time point includes:

[0107] The first step is to determine the primary fire risk factor for each type of monitoring data based on how close the monitoring data at each moment is to the warning threshold of the corresponding type.

[0108] The occurrence and development of a fire inevitably cause different types of monitoring data to exceed their normal range. Warning thresholds are physical quantities characterizing the boundaries of a safe state, set based on extensive experimental and historical data. The closer the monitoring data is to or exceeds the warning threshold for the corresponding type, the further the current environmental state deviates from the safe range, and the higher the static probability of a fire. Therefore, the first fire risk factor here is a static assessment based on the proximity of the current state.

[0109] In this embodiment, the ratio of the monitoring data at each moment to the corresponding type of early warning threshold is determined, and the ratio is used as the first fire risk factor of the monitoring data at the corresponding moment.

[0110] The second step is to determine the second fire risk factor for the monitoring data at each time point based on the local average rate of change corresponding to the monitoring data at each time point.

[0111] The early stages of a fire are often a process of gradually increasing energy or material release rates. The local average rate of change reflects the average speed and direction of change in monitored data at the current stage. A sustained and large positive rate of change (such as a rapid rise in temperature or a rapid increase in smoke concentration) indicates that the potential fire is dynamically developing, and energy accumulation or chemical reactions are accelerating, even if the current absolute value may not yet have reached the threshold. Therefore, the second fire risk factor here is a dynamic assessment based on the trend of acceleration.

[0112] In this embodiment, the local average rate of change corresponding to the monitoring data at each time point is dimensionless, and the dimensionless value is used as the second fire risk factor of the monitoring data at the corresponding time point. Dimensionlessness can be achieved using a norm function or by using the ratio of the local average rate of change to a preset maximum reasonable rate of change. The preset maximum reasonable rate of change needs to be determined through historical data or a physical model; for example, the maximum safe temperature rise rate might be 10°C / minute.

[0113] The third step is to perform a weighted summation of the first and second fire risk factors of the monitoring data at the same time to obtain the fire risk level of the corresponding type of monitoring data at each time.

[0114] Real fire risk is the result of the combined effect of the current dangerous state and the future deterioration trend. Integrating the primary and secondary fire risk factors through weighted summation aligns with the thought process of comprehensive judgment. The weighting process itself is a strategy choice, allowing for adjustments to the emphasis on the current dangerous state and the future deterioration trend based on the characteristics of different parameters and the focus of the application scenario.

[0115] In this embodiment, for each monitoring data moment, a weight is set for the first fire risk factor, denoted as the first weight; and a weight is set for the second fire risk factor, denoted as the second weight. Based on the first weight and the second weight, the first fire risk factor and the second fire risk factor are weighted and summed to obtain the fire risk level at the corresponding moment.

[0116] As an example, the formula for calculating the fire risk level of the i-th type of monitoring data at time j can be:

[0117] In the formula, This represents the fire risk level of the i-th type of monitoring data at time j. Indicates the first weight. This represents the value of the i-th type of monitoring data at time j. This represents the warning threshold for the i-th type of monitoring data. This represents the first fire risk factor of the i-th type of monitoring data at time j. Indicates the second weight. This represents the local average rate of change of the i-th type of monitoring data at time j. This represents the maximum reasonable rate of change corresponding to the i-th type of monitoring data. This represents the second fire risk factor of the i-th type of monitoring data at time j.

[0118] In the formula for calculating fire risk, the first fire risk factor, even if a fire has just occurred, if its intensity is sufficient to cause a certain monitoring data to quickly reach near the threshold, must be given the highest level of attention immediately. It is the most basic and direct line of defense for ensuring safety. The second fire risk factor is specifically used to quantify early fire signals that are not high in absolute value but are growing rapidly. It is a key supplement to the first fire risk factor and solves the problem of delayed response of traditional threshold methods to the early stages of slow-accumulating fires and rapid-breaking fires.

[0119] The weight used to balance the importance of the current value and the changing trend can be set to 0.7. All weights used for weighted calculation in this embodiment can be initially set based on expert experience and optimized during the system debugging phase using a test dataset containing various typical fire and non-fire interference scenarios. The optimization method can employ grid search or automated optimization algorithms, aiming to make the final fire determination index respond fastest and strongest in real fire scenarios, and weakest in interference scenarios, thereby optimizing the overall system performance.

[0120] Statistical analysis was conducted on the warning thresholds of different types of monitoring data, and the point with the highest true alarm rate within an acceptable false alarm rate was selected as the threshold. The threshold can be a fixed value pre-set based on relevant industry safety standards (such as GB 50116-2013 "Code for Design of Automatic Fire Alarm Systems") or through statistical analysis of a large amount of experimental data. For example, for the temperature parameter inside an electrical cabinet, It can be set to 85℃; for smoke concentration, It can be set to 0.1 mg / m³.

[0121] It should be noted that the fire risk level determined through weighted fusion can increase the weight of the primary fire risk factor in scenarios with extremely high absolute safety requirements, ensuring that any situation approaching the threshold is given high priority. Conversely, in scenarios with high early warning requirements, the weight of the secondary fire risk factor can be increased to more sensitively capture growth trends. Furthermore, weighting can effectively prevent the susceptibility of a single indicator to interference. For example, a momentary high value may cause a sharp increase in the primary fire risk factor, but if the secondary fire risk factor is very low, the weighted overall risk level will not be abnormally high, thus suppressing misjudgments.

[0122] After obtaining the fire risk level of each type of monitoring data at each time point, the degree of attention required for each type of monitoring data at each time point is obtained by combining the trend deviation anomaly.

[0123] In the automatic triggering control of thermal aerosol fire extinguishing devices, priority should be given to situations where there are abnormal trend deviations, and where the fire risk is high according to the analysis of a certain type of monitoring data at a certain moment. This indicates that the monitoring data of that type at the corresponding moment may be the cause of the fire or be greatly affected by the fire, and therefore requires high attention.

[0124] In this embodiment, the fire risk level and trend deviation anomaly of the same type of monitoring data at the same time are weighted and summed to obtain the attention necessity level at the corresponding time. For example, it can be calculated according to the following formula:

[0125] In the formula, This represents the degree of attention required for the i-th type of monitoring data at time j, where norm represents the normalization function. This represents the fire risk level of the i-th type of monitoring data at time j. This indicates the degree of trend deviation anomaly of the i-th type of monitoring data at time j. and For example, preset weights It equals 0.6. It equals 0.4, and .

[0126] Of course, implementers can also determine the level of attention required at a given time by directly calculating the product of the fire risk level and the degree of trend deviation anomaly of the same type of monitoring data at the same time.

[0127] By calculating the degree of attention required in step S3 above, the system can dynamically identify the monitoring parameters that require the most attention at the current moment, providing an important basis for subsequent contribution allocation, thereby achieving more accurate and adaptive early fire identification and trigger control in complex environments.

[0128] S4. Based on the correlation between each type of monitoring data and its comparative type of monitoring data, and the consistency of fire risk, combined with the degree of attention required, determine the contribution of each type of monitoring data to the fire extinguishing trigger control at each moment.

[0129] In the automatic triggering control of thermal aerosol fire extinguishing systems, monitoring data from a single sensor may be distorted due to its own characteristics, installation location, environmental interference, or accidental malfunctions. Therefore, determining the importance of monitoring data at a given moment in decision-making requires not only analyzing the performance of the monitoring data itself (i.e., focusing on its necessity), but also analyzing different types of monitoring data to determine the correlation and consistency of fire risk between a particular type and its comparative types.

[0130] As an exemplary implementation, the above-described determination of the contribution of each type of monitoring data to the fire suppression triggering control at each moment can be achieved through... Figure 3 The steps shown are to be implemented as follows:

[0131] S41, take any type as the target type and other types other than the target type as the comparison type.

[0132] In this embodiment, taking any type as an example, we analyze the correlation between this type and monitoring data of other types besides itself. Here, any type is taken as the target type, and other types besides the target type are taken as the comparison type.

[0133] S42, based on the correlation between the monitoring data of the target type and the monitoring data of each comparison type and the consistency of fire risk, determine the early warning credibility of the target type monitoring data at each time point.

[0134] In enclosed environments such as electrical distribution boxes, there are correlations between various types of sensor data. For example, temperature and electrical data can be correlated; when the load is high, electrical data fluctuates drastically, and the temperature also rises. When a type of monitoring data shows a close correlation with other types of monitoring data at the same time, and analysis reveals similar fire risk values, it indicates that this type of monitoring data is highly reliable, unlikely to be a false positive due to interference from other factors, and has strong reference value for triggering control, meaning the early warning is highly credible.

[0135] The first step is to determine the correlation between the target type and each comparison type based on the monitoring data change curves of the target type and each comparison type.

[0136] A real fire is a physicochemical process that simultaneously or sequentially affects multiple environmental parameters. Therefore, monitoring data from different types of fire events should theoretically exhibit an inherent statistical correlation or temporal correlation in their change curves. For example, a sustained rise in temperature may be accompanied by an increase in smoke concentration and the release of specific gases, and these changes should exhibit a certain degree of synchronicity or a fixed lag relationship in their trends. Therefore, the correlation here is used to examine the synergistic patterns of changes in different types of monitoring data.

[0137] In this embodiment, curve fitting is first performed on the monitoring data of all times within a preset time period under the same type to obtain the monitoring data change curve of each type; then, based on the monitoring data change curves of the target type and each comparison type, the correlation between the target type and each comparison type is calculated.

[0138] Curve fitting can be achieved using the least squares method. The correlation can be calculated using time series similarity measurement methods known in the art. These include, but are not limited to: calculating the Pearson correlation coefficient or Spearman rank correlation coefficient after resampling and aligning the sequences; evaluating the morphological similarity of sequences using dynamic time warping algorithms; measuring statistical dependence by calculating mutual information; or comparing their spectral characteristics in the frequency domain. Those skilled in the art can choose an appropriate method based on the characteristics of the specific data (such as linearity or the presence of phase differences).

[0139] It should be noted that the correlation degree determined in this embodiment is a positive number.

[0140] The second step is to determine the risk consistency index between the target type and each comparison type at each time point based on the similarity between the fire risk levels of the target type and each comparison type at the same time point.

[0141] When analyzing the reliability of early warnings, it is necessary not only to analyze the similarity in the shape of the change curves of different types of monitoring data, but also to pay attention to whether the fire risk level quantified by different types of monitoring data at the same time is consistent. Therefore, the risk level consistency index here is used to achieve cross-validation of multiple types of monitoring data.

[0142] In this embodiment, the absolute value of the difference between the fire risk level of the target type and each comparison type at the same time is calculated, and the absolute value of the difference is normalized by negative correlation. The normalized value is used as the consistency index of the risk level between the target type and each comparison type at the corresponding time.

[0143] Among them, exp(-) can be used to achieve negative correlation normalization, and the value range of the risk consistency index is between 0 and 1.

[0144] The third step is to determine the credibility of the early warning data of the target type at each time point based on the correlation between the target type and each comparison type and the consistency index of the risk level at each time point.

[0145] As an example, the formula for calculating the early warning reliability of the i-th type of monitoring data at time j can be:

[0146] In the formula, This represents the confidence level of the early warning for the i-th type of monitoring data at time j, where N represents the number of monitoring data types. This represents the correlation between the changes in monitoring data of type i and type j. This represents the fire risk level of the i-th type of monitoring data at time j. This represents the fire risk level of the nth type of comparison monitoring data at time j. The function represents the absolute value function, exp represents the exponential function with the natural constant as the base, and exp(-) is used to normalize negative correlations. This represents the consistency index of risk level between the i-th type and the n-th comparison type at time j.

[0147] In the formula for calculating the credibility of early warning, the correlation can be used to assess whether the changes in monitoring data of the target type and each comparison type follow the expected fire physics model. A high correlation means that the change in the monitoring data of the i-th type is consistent with the change pattern of the monitoring data of the comparison type, and its credibility in indicating the fire is enhanced. A low correlation may indicate that the anomaly in the monitoring data of the i-th type is an isolated event caused by local interference or sensor failure, and its credibility should be reduced.

[0148] If, at a certain moment, the monitoring data of type i shows a high risk, and at the same time, most of the comparative monitoring data also show a medium to high risk, forming a corroborating relationship, the weight of the monitoring data of type i at that moment should be significantly increased in this judgment. Conversely, if, at a certain moment, the monitoring data of type i shows a high risk, while most of the comparative monitoring data show a low risk, it indicates a conflict. In this case, it should be suspected that the monitoring data of type i at that moment is being interfered with or is experiencing a false alarm, and its contribution should be reduced to avoid the system being misled by a single abnormal signal.

[0149] S43, based on the early warning reliability, fire risk and attention necessity of the target type monitoring data at each moment, determine the contribution of the target type monitoring data to the fire extinguishing trigger control at each moment.

[0150] Thermal aerosol fire extinguishing devices activate solid aerosol generators via electric igniters. The triggering chain response is slow; failure to trigger in a confined electrical environment may lead to device malfunction and greater damage. Therefore, rapid and accurate identification of early fire signals is crucial. When a certain type of monitoring data shows a high fire risk value at a given moment, it indicates its significant role in early fire identification and high sensitivity to fire indications. Therefore, to further improve the accuracy and reliability of contribution values, the sensitivity determined based on fire risk level should be considered alongside the credibility and necessity of early warnings.

[0151] The first step is to normalize the fire risk level of the target type monitoring data at each time point, and use the normalized value as the sensitivity at the corresponding time point.

[0152] In this embodiment, the maximum-minimum method can be used to normalize the fire risk level of the target type monitoring data at each time step to obtain the sensitivity of the target type monitoring data at each time step.

[0153] The second step involves data fusion processing of the early warning reliability, sensitivity, and necessity of attention of target type monitoring data at the same time to determine the contribution of target type monitoring data to fire extinguishing trigger control at the corresponding time.

[0154] Necessity of attention is a fundamental parameter in contribution calculation, reflecting the intensity and urgency of anomalies in monitored data. Monitoring data with high necessity of attention, relative sensitivity, strong correlation with other types of monitoring data, and good risk consistency will receive the highest contribution, becoming the dominant basis for triggering decisions. Monitoring data with high necessity of attention and relative sensitivity, but lacking correlation with other types of parameters or exhibiting inconsistent risk, will have its contribution moderately suppressed, and the system will treat it with caution. Monitoring data with low necessity of attention, even if it is correlated with certain types of monitoring data at the same time, will naturally have a lower contribution, avoiding irrelevant fluctuations from interfering with decision-making.

[0155] In this embodiment, a weighted fusion method is used to process the warning reliability, sensitivity, and attention necessity of the target type monitoring data at the same time. The resulting fusion value is used as the contribution of the target type monitoring data to the fire extinguishing trigger control at the corresponding time. Since attention necessity is the basic parameter for contribution calculation, the weight of attention necessity can be set to be relatively large. The sum of the preset weights of warning reliability, sensitivity, and attention necessity is 1.

[0156] Of course, implementers can also use other data fusion methods to process the data on the credibility, sensitivity and necessity of attention of the early warning, such as multiplication.

[0157] The contribution determined by step S4 above not only solves the technical bottleneck of effective fusion of multimodal data, but also brings multiple significant benefits such as reduced false alarm rate, reduced false alarm rate, enhanced early warning capability, and improved system environmental adaptability. This is conducive to achieving a comprehensive breakthrough in the accuracy, timeliness and reliability of fire identification and automatic triggering in complex real-world scenarios.

[0158] S5 uses the contribution as the judgment weight to determine the fire judgment index at each moment and automatically triggers the thermal aerosol fire extinguishing device.

[0159] In the automatic triggering control of thermal aerosol fire extinguishing devices, a single parameter is insufficient to fully reflect the fire situation. It is necessary to integrate monitoring data from each type at the same time into a unified fire judgment index to achieve automated and executable triggering decisions. By using the contribution of each type of monitoring data as a weighting criterion, key types of indicators are given greater influence, making the comprehensive judgment result more accurately reflect the severity of the actual fire. This allows for rapid crossing of the trigger threshold in the early stages of a fire, enabling timely release of thermal aerosol for extinguishing, avoiding false alarms caused by fluctuations in a single characteristic, and ensuring high reliability of the system even under complex field conditions.

[0160] The first step is to obtain the fire hazard level for each type of monitoring data at each moment.

[0161] Here, the level of danger is calculated in real time using machine learning algorithms.

[0162] In this embodiment, a pre-trained machine learning model is used to predict the probability of a fire, and the SHAP algorithm is used to interpret the risk of a fire caused by each type of monitoring data at each moment.

[0163] Specifically, all types of monitoring data at the current moment are used as input and fed into the pre-trained core prediction model. The model outputs an overall fire probability value. Subsequently, the SHAP algorithm is used to interpret this prediction and calculate the contribution of each type of monitoring data at each moment to the fire probability value, which is taken as the degree of risk of causing a fire.

[0164] The construction and training of the core prediction model involves pre-collecting a large historical dataset containing various types of monitoring data and corresponding labels (0 or 1) indicating whether a fire has occurred at that time. This historical dataset is then used to train a machine learning classification model, such as a gradient boosting decision tree, random forest, or neural network model. After training, the model can output a value in the range [0,1] based on real-time input of various types of monitoring data, representing the probability of a fire occurring at the current moment.

[0165] The SHAP (SHapley Additive exPlanations) algorithm is a tool for interpreting the predictions of any machine learning model. In this embodiment, the SHAP algorithm is not used to directly calculate hazard levels, but rather to post-hoc interpret the trained core prediction model. Specifically, after real-time monitoring data is input into the trained model to obtain a fire probability, the SHAP algorithm can analyze and calculate the contribution of each input monitoring data value (such as the current temperature value) to the final fire probability. This contribution is defined as the degree to which the monitoring data poses a fire risk at that moment.

[0166] The second step is to calculate the fire risk of different types of monitoring data at each time point by weighting and averaging the data based on the judgment weight of each type of monitoring data at each time point, so as to obtain the fire judgment index at each time point.

[0167] As an example, the formula for calculating the fire determination index at time j can be:

[0168] In the formula, This represents the fire detection index at time j, and N represents the number of types of monitoring data. This represents the judgment weight of the i-th type of monitoring data at time j. This represents the value of the i-th type of monitoring data at time j. This represents the risk of a fire caused by the i-th type of monitoring data at time j.

[0169] After obtaining the fire assessment indicators at each moment, the thermal aerosol fire extinguishing device is automatically triggered and controlled.

[0170] In this embodiment, when the fire determination index exceeds the preset trigger threshold for the first time at a certain moment, the system does not trigger immediately, but enters a pending trigger confirmation state. In this pending trigger confirmation state, the system will continuously monitor the fire determination index at subsequent moments. If the fire determination index continues to exceed the preset trigger threshold within the next preset confirmation time, the system determines that the fire is valid, and the trigger control module immediately sends a start command to the thermal aerosol extinguishing device, specifically causing the internal igniter or drive mechanism to act instantly and rapidly release highly efficient fire extinguishing aerosol substances, so that the extinguishing agent forms a uniformly dispersed fire extinguishing particle group in the space, covering and suppressing the combustion reaction. If the fire determination index does not exceed the preset trigger threshold within the preset confirmation time, the system cancels the pending trigger state and returns to the normal monitoring mode.

[0171] The trigger threshold is determined based on the analysis of historical fire simulation data. It is calculated by plotting receiver operation characteristic curves to find the point with the highest true alarm rate within an acceptable false alarm rate; this threshold can be set to 0.8. The preset confirmation time is set to strike a balance between response timeliness and decision reliability. Its value should be greater than the duration of typical electrical interference or environmental fluctuations, but much less than the critical time for fire to become uncontrollable. The specific value can be determined based on the statistical distribution of historical interference signals, such as setting it to 3 seconds.

[0172] It should be noted that the triggering mechanism with time confirmation can effectively filter out instantaneous signal pulses or brief interference, significantly improving the reliability of triggering decisions.

[0173] The fire extinguishing closed-loop control flowchart is as follows: Figure 4 As shown, Figure 4 In this process, the sensing layer monitors key parameters in real time through multimodal sensors; the data processing layer performs noise reduction and normalization on the collected data; the fire determination layer analyzes the multimodal data and integrates it with the contribution to obtain fire determination indicators; the trigger decision layer compares the fire determination indicators with the trigger threshold and outputs a trigger signal; and the execution layer executes the fire extinguishing operation through the igniter inside the thermal aerosol fire extinguishing device.

[0174] In summary, this invention provides an automatic triggering control system for a thermal aerosol fire extinguishing device. This system analyzes the sensitivity of key fire parameters reaching the thermal aerosol fire extinguishing trigger point, quantifying the indicative ability of each key fire parameter to the actual fire situation, thereby obtaining the contribution of different types of monitoring data to the triggering decision. Based on a parameter contribution-driven triggering mechanism, the system can more accurately identify the evolution trend of early fires, avoiding misjudgments caused by single thresholds or single sensors, and achieving early response to real hazardous situations. Simultaneously, it significantly improves the accuracy and robustness of the thermal aerosol fire extinguishing device's triggering decision, shortens the triggering response time, reduces missed and false triggers, ensures the extinguishing agent is released at the optimal time, thereby more effectively controlling the fire and improving overall fire extinguishing reliability.

[0175] The above-described 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 scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An automatic triggering control system for a thermal aerosol fire extinguishing device, characterized in that, Includes a memory and a processor, the processor executing a computer program stored in the memory to perform the following process: Acquire multiple types of monitoring data of the thermal aerosol fire extinguishing device during the current preset time period; the multiple types of monitoring data include at least temperature data, smoke concentration data, and electrical data, and the electrical data includes at least current and voltage. Based on the analysis of the changing trends of the same type of monitoring data at each time point, the degree of deviation of the trend of each type of monitoring data at each time point is determined. Determine the fire risk level of each type of monitoring data at each time point, and combine the trend deviation anomaly level to obtain the attention necessity of each type of monitoring data at each time point; Based on the correlation between each type of monitoring data and its comparison type monitoring data, and the consistency of fire risk, combined with the aforementioned level of attention necessity, the contribution of each type of monitoring data to fire extinguishing trigger control at each moment is determined. The contribution level is used as the judgment weight to determine the fire judgment index at each moment, and the thermal aerosol fire extinguishing device is automatically triggered and controlled.

2. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 1, characterized in that, The step of analyzing the changing trend of the same type of monitoring data based on the multi-type monitoring data at each time moment, and determining the degree of trend deviation anomaly of each type of monitoring data at each time moment, includes: For the same type of monitoring data, the first trend change factor for each time moment is determined based on the difference between the maximum and minimum monitoring data in the time neighborhood of each time moment; The second trend change factor for each moment is determined based on the difference between the instantaneous rate of change and the local average rate of change of the monitoring data at each moment. Based on the analysis of the first and second trend change factors at each time point, the deviation of the current type of monitoring data change trend from the historical trend benchmark is determined, and the degree of trend deviation anomaly of the current type of monitoring data at each time point is determined.

3. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 2, characterized in that, Obtaining the local average rate of change includes: For each time moment, acquire all monitoring data within the time neighborhood of the current time moment, and perform linear fitting on all monitoring data within the time neighborhood to obtain the monitoring fitting line for the current time moment; Determine the slope of the monitoring fitted line at the current moment, and use the slope of the monitoring fitted line as the local average rate of change.

4. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 2, characterized in that, The determination of the trend deviation anomaly of the current type of monitoring data at each time point includes: The first and second trend change factors at the current moment are fused to obtain the trend change performance at the current moment. Establish a historical trend benchmark; the historical trend benchmark is at least used to characterize the trend benchmark value under historical stable operating conditions; Based on the difference between the current trend change performance and the historical trend benchmark, determine the degree of trend deviation anomaly of the current type of monitoring data at the current moment.

5. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 3, characterized in that, The determination of the fire risk level for each type of monitoring data at each time point includes: For each type, the first fire risk factor of the monitoring data at each moment is determined based on the proximity of the monitoring data at each moment to the warning threshold of the corresponding type. Based on the local average rate of change corresponding to the monitoring data at each time moment, the second fire risk factor of the monitoring data at each time moment is determined; The first and second fire risk factors of the monitoring data at the same time are weighted and summed to obtain the fire risk level of the corresponding type of monitoring data at each time.

6. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 5, characterized in that, The obtained fire risk level of the corresponding type of monitoring data at each moment includes: For each moment of the monitoring data, a weight is set for the first fire risk factor, denoted as the first weight; and a weight is set for the second fire risk factor, denoted as the second weight. Based on the first and second weights, the first fire risk factor and the second fire risk factor are weighted and summed to obtain the fire risk level at the corresponding time.

7. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 1, characterized in that, Based on the correlation and consistency of fire risk between each type of monitoring data and its comparative type, and in conjunction with the degree of attention required, the contribution of each type of monitoring data to fire extinguishing trigger control at each time point is determined, including: Use any type as the target type, and use other types besides the target type as the comparison type; Based on the correlation between the monitoring data of the target type and the monitoring data of each comparison type, and the consistency of fire risk, the early warning credibility of the target type monitoring data at each time point is determined; Based on the early warning reliability, fire risk, and necessity of attention of the target type monitoring data at each moment, the contribution of the target type monitoring data to the fire extinguishing trigger control at each moment is determined.

8. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 7, characterized in that, The determination of the early warning reliability of the target type monitoring data at each time step includes: The correlation between the target type and each comparison type is determined based on the monitoring data change curves of the target type and each comparison type; the monitoring data change curves are obtained by curve fitting of monitoring data of the same type at multiple time points. Based on the similarity between the fire risk levels of the target type and each comparison type at the same time, determine the risk level consistency index between the target type and each comparison type at each time. The credibility of the early warning data of the target type at each time point is determined based on the correlation between the target type and each comparison type, as well as the consistency index of the risk level at each time point.

9. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 7, characterized in that, The contribution of target type monitoring data to fire extinguishing trigger control at each moment is determined based on the early warning reliability, fire risk, and necessity of attention of the target type monitoring data at each moment, including: The fire risk level of the target type monitoring data at each time moment is normalized, and the normalized value is used as the sensitivity at the corresponding time moment. The early warning reliability, sensitivity, and necessity of attention of target type monitoring data at the same time are fused to determine the contribution of target type monitoring data to fire extinguishing trigger control at the corresponding time.

10. The automatic triggering control system for a thermal aerosol fire extinguishing device according to claim 1, characterized in that, The step of using the contribution as a judgment weight to determine the fire judgment index at each moment includes: The risk of fire ignition for each type of monitoring data is obtained at each moment; the risk is calculated in real time using a machine learning algorithm. Based on the judgment weight of each type of monitoring data at each time, the fire risk of different types of monitoring data at each time is weighted and averaged to obtain the fire judgment index at each time.

Images