Compressed nitrogen foam fire extinguishing system and fire extinguishing parameter self-adaptive control method thereof

Abstract

The invention discloses a compressed nitrogen foam fire extinguishing system and a fire extinguishing parameter self-adaptive control method thereof. The compressed nitrogen foam fire extinguishing system comprises a fire behavior analysis module which is used for preprocessing a transformer substation fire video acquired by a video image acquisition module, and judging a fire type and a combustion position of a transformer substation by using a convolutional neural network model which is trained in advance; the main control module is used for generating a dynamic control strategy according to the fire type and the combustion position and adjusting the opening degree of the medium adjusting valve; and the fire monitor control module adjusts the movement direction of the fire monitor in real time according to the control signal of the main control module, automatically adjusts the spraying angle and the spraying range of the fire extinguishing medium, and ensures that the falling point of the fire extinguishing medium coincides with the position of the fire source. The fire type and the combustion position of the transformer substation are judged by monitoring video images, the output proportion of the fire extinguishing medium is dynamically adjusted, the fire extinguishing efficiency is greatly improved, and the cost performance of a compressed nitrogen foam fire extinguishing system is improved.

Description

Technical Field

[0001] This invention relates to the field of substation fire extinguishing technology, specifically a compressed nitrogen foam fire extinguishing system and an adaptive control method for its fire extinguishing parameters. Background Technology

[0002] Compressed air foam fire extinguishing systems, based on fire alarm signals fed back by the fire alarm control panel, confirm the alarm situation and then sequentially start the air compressor, foam pump, and fire water pump, opening the control valves of the corresponding zones so that the foam water spray can be used to extinguish the fire through the fire monitor. They are commonly used in fire protection projects for large buildings.

[0003] Existing compressed air foam fire extinguishing systems have significant technical limitations and cannot achieve the goal of efficient and intelligent fire suppression: First, the system parameters are set in a fixed manner, which cannot be intelligently adjusted according to the actual working conditions of the fire source (such as fire type and burning location), nor can the spray pattern of the fire monitor be dynamically adjusted; Second, due to the air compressor, conventional systems will mix a large amount of fresh oxygen into the foam during the foam generation process, which contradicts the core fire extinguishing mechanism of foam fire extinguishing systems that isolates oxygen and affects the fire extinguishing efficiency. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies in efficiently and intelligently extinguishing fires, this invention provides a compressed nitrogen foam fire extinguishing system and its adaptive control method for fire extinguishing parameters. This system uses a pre-trained convolutional neural network model to analyze substation fire video images in real time to accurately identify the fire type and location. Based on the fire type and location, it dynamically adjusts the mixing ratio of nitrogen and water / foam liquid, ensuring that the fire extinguishing parameters are adapted to the fire situation in real time. This improves the foam coverage and isolation effect while significantly enhancing fire extinguishing response speed and resource utilization through a real-time fire monitoring and feedback mechanism.

[0005] Therefore, the present invention adopts the following technical solution: a compressed nitrogen foam fire extinguishing system, which includes a fire extinguishing medium generating device, a fire extinguishing medium releasing module, and a fire intelligent detection and control module; the fire intelligent detection and control module includes a video image acquisition module, a fire analysis module, a main control module, and a fire monitor control module;

[0006] The fire extinguishing medium generating device includes fire water delivery pipeline, foam liquid delivery pipeline and nitrogen delivery pipeline. Each of the three delivery pipelines is equipped with a medium regulating valve and a medium flow meter, which are connected to the main control module through signal lines.

[0007] The extinguishing medium release module includes at least one foam fire monitor connected to the outlet of a nitrogen-foam solution mixer, and the foam fire monitor is connected to the fire monitor control module via a signal line;

[0008] The fire analysis module preprocesses the substation fire video images acquired by the video image acquisition module, locates suspicious target areas, extracts features by the area changes, temperature changes and thermal radiation changes of the suspicious target areas, and uses a pre-trained convolutional neural network model to determine the substation fire type and burning location (i.e., the coordinates of the flame center).

[0009] The main control module generates a dynamic control strategy based on the fire type and combustion location, adjusts the opening of the medium regulating valves on the fire water delivery pipeline, foam liquid delivery pipeline and nitrogen delivery pipeline, adjusts the output ratio of water, foam liquid and compressed nitrogen in the extinguishing medium, and then readjusts the opening of the medium regulating valves according to the fire situation after a period of time.

[0010] The fire monitor control module adjusts the movement direction of the fire monitor in real time according to the control signal from the main control module, and automatically adjusts the spray angle and spray range of the extinguishing medium to ensure that the landing point of the extinguishing medium coincides with the location of the fire source until the fire is extinguished.

[0011] Furthermore, the extinguishing medium is a mixture of nitrogen, foam liquid, and water, wherein the gas-liquid ratio of nitrogen flow rate to mixture flow rate is 7-10:1, the mixture flow rate is the sum of water flow rate and foam liquid flow rate, and the mixing ratio of foam liquid flow rate to mixture flow rate is 3-3.9:100.

[0012] This invention also provides an adaptive control method for the extinguishing parameters of the above-mentioned compressed nitrogen foam fire extinguishing system, comprising:

[0013] A multi-position binocular camera is used to collect fire video images and temperature information, which are then uploaded to the fire analysis module along with the temperature sensing cable signal. After the fire analysis module makes a comprehensive judgment on the fire type and burning location, the main control module first sets the fire extinguishing parameters, and then sends instructions to the fire extinguishing medium generating device and the fire extinguishing medium releasing module to release the fire extinguishing medium with appropriate fire extinguishing parameters.

[0014] During the adjustment of the fire monitor's movement, the fire analysis module feeds back changes in fire type, combustion location, and fire combustion stage to the main control module. The main control module then sends valve opening information to the extinguishing medium generating device, readjusting the valve openings of the water regulating valve, foam liquid regulating valve, and nitrogen regulating valve, thereby changing the water-foam liquid mixing ratio and the gas-liquid ratio to ensure they remain within the effective range.

[0015] Furthermore, the fire extinguishing parameters include fire water flow rate, the mixing ratio of water flow rate / (water flow rate + foam liquid flow rate), and the gas-liquid ratio of nitrogen flow rate / (water flow rate + foam liquid flow rate).

[0016] Furthermore, instructions are sent to the extinguishing medium generating device, specifically including:

[0017] Adjust the valve opening of the fire water supply regulating valve on the fire water delivery pipeline according to the set fire water flow rate to regulate the fire water outflow rate;

[0018] Based on the outflow rate of fire-fighting water and according to the set water-foam liquid mixing ratio, the valve opening of the foam liquid regulating valve on the foam liquid delivery pipeline is adjusted to regulate the foam liquid flow rate.

[0019] Based on the fire water flow rate and foam liquid flow rate, and according to the set gas-liquid ratio, adjust the opening of the nitrogen regulating valve on the nitrogen delivery pipeline to regulate the nitrogen flow rate.

[0020] Furthermore, instructions are sent to the extinguishing medium release module, specifically including:

[0021] The horizontal rotation and elevation angle of the foam fire monitor are dynamically adjusted according to the location of the flame center; the spray speed, spray range and spray flow rate of the foam fire monitor are dynamically adjusted according to the temperature changes at the fire scene.

[0022] Furthermore, a fire dataset covering multiple scenarios of power equipment was constructed, encompassing fire video images of transformer fires, high-resistance fires, dry-type reactor fires, cabinet fires, and cable trench fires. A multi-level classification system was used for structured annotation: fire types were labeled based on equipment type, and transformer fires were further subdivided into oil sump fires, surface fires, oil conservator fires, radiator fires, jet fires, and internal fires. Center coordinate box localization and type labeling were implemented for flame areas, while hierarchical labeling rules were established. Time-series frame-level annotations were added to dynamic combustion features, and pixel-level labeling of thermal imaging data was used to supplement high-temperature areas, ultimately forming a dual-modal dataset.

[0023] Furthermore, the convolutional neural network model is the EfficientNet-B3 composite scaling model, and its specific construction process is as follows:

[0024] The model employs EfficientNet-B3 as its backbone network, and uses a composite scaling method to jointly optimize the network depth, width, and input resolution. ResNet residual structures are embedded in Stages 3-5 of the model, with "1×1 convolution for dimensionality reduction → 3×3 convolution for spatial feature extraction → 1×1 convolution for dimensionality enhancement" as the core module. Skip connections are used to preserve shallow details. The model uses multi-task outputs, including global average pooling and fully connected layer fire bounding boxes. A feature pyramid fuses multi-scale information to classify the flame size and fire type of substation fires.

[0025] Labeled fire image data was used as the training set and input into the EfficientNet-B3 composite scaling model for training. During training, the model's weights and biases were adjusted using the gradient descent optimization algorithm to minimize the loss function. For the substation scenario, the input layer fused visible light and infrared thermal imaging dual-modal data to suppress false detections in high-heat areas of equipment, and robustness was improved by simulating occlusion and dynamic smoke data enhancement.

[0026] After receiving new fire video images, the trained EfficientNet-B3 composite scaling model performs bimodal fusion preprocessing on the input data. During forward propagation, the model outputs the fire type, flame size, and combustion stage in parallel. The prediction results are filtered by nonmaximum suppression and confidence thresholding, triggering fire-fighting strategy matching. If the fire is identified as an internal transformer fire, the main control module is linked to dynamically adjust the valve openings of the water, foam liquid, and nitrogen regulating valves according to the fire type, while simultaneously controlling the fire monitor angle to be aimed at the base of the flame in real time.

[0027] Furthermore, a multi-stage valve-controlled fire extinguishing mechanism is adopted. The main control module dynamically matches the fire type and combustion location, and adjusts the opening signals of the foam liquid regulating valve and nitrogen regulating valve in real time. By sending instructions to the fire extinguishing medium control cabinet, it precisely controls the mixing ratio and gas-liquid ratio of the three media, water, foam and nitrogen, with the priority order being water → foam → nitrogen. For different equipment fire types, the valve opening adopts a three-level combination strategy of high, medium and low. For transformer fire types, the main control module refines the control logic according to the depth of the combustion location.

[0028] Furthermore, based on the real-time flame position (x,y,w,h), the fire monitor is driven to aim at the flame root projection point with an accuracy of ±0.5° in real time to suppress the spread of fire; according to the fire type, combustion location and combustion stage, a dynamic priority algorithm is used to adjust the opening combination of water, foam and nitrogen regulating valves, and simultaneously optimize the gas-liquid mixing ratio and foam coverage density; Kalman filtering is used to eliminate positioning jitter caused by smoke obstruction, ensuring that the fire extinguishing medium spray trajectory overlaps with the core area of ​​the flame by more than 95%.

[0029] The present invention has the following beneficial effects: The present invention uses a fire intelligent detection and control module to automatically identify the fire type and the burning location of important equipment in the monitoring scene, adjust the movement direction of the fire monitor head in real time, and automatically adjust parameters such as the spray speed, spray angle and spray range of the extinguishing medium. For different fire types and burning locations, the valve opening adjustment signal is adjusted to dynamically output the extinguishing medium with the gas-liquid ratio, thereby saving fire extinguishing costs and improving fire extinguishing efficiency.

[0030] This invention improves the insulation effect of foam coverage while significantly enhancing the fire extinguishing response speed and resource utilization through a real-time fire monitoring and feedback mechanism. It not only improves fire extinguishing efficiency but also optimizes the system's economic indicators, forming a more practical and valuable intelligent fire extinguishing solution. Attached Figure Description

[0031] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of a compressed nitrogen foam fire extinguishing system according to the present invention;

[0033] Figure 2 This is a diagram showing the composition of the intelligent fire detection and control module of the present invention;

[0034] Figure 1 The components are: 1. Fire water tank, 2. Fire pump, 3. Foam liquid tank, 4. Foam liquid pressure relief valve, 5. Foam pump.

[0035] 6. High-pressure gas cylinder, 7. Third check valve, 8. High-pressure pressure reducing valve, 9. Foam liquid regulating valve, 10. Foam liquid pressure gauge, 11.

[0036] 12. Foam flow meter, 13. Nitrogen regulating valve, 14. Gas pressure gauge, 15. Liquid-liquid mixer, 16. Gas flow meter.

[0037] 17. Gas-liquid mixer; 18. Pressure gauge; 19. Water pressure relief valve; 20. First check valve; 21. Water regulating valve; 22. Water pressure gauge; 23. Water flow meter; 26. Fire monitor; 28. Binocular camera; 39. Intelligent fire detection and control module; 30. Second check valve. Detailed Implementation

[0038] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0039] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0040] Example 1

[0041] This embodiment provides a compressed nitrogen foam fire extinguishing system, which consists of a fire extinguishing medium generating device, a fire extinguishing medium release module, and a fire intelligent detection and control module 31. Figure 2 As shown, the intelligent fire detection and control module consists of a video image acquisition module, a fire analysis module, a main control module, and a fire monitor control module.

[0042] like Figure 1 As shown, the fire extinguishing medium generating device consists of a fire water delivery pipeline, a foam liquid delivery pipeline, and a nitrogen delivery pipeline. The fire water delivery pipeline is equipped with a fire water tank 1, a fire water pump 2, a water pressure relief valve 18, a first check valve 19, a water regulating valve 20, a water pressure gauge 21, and a water flow meter 22. The water regulating valve 20 and the water flow meter 22 are connected to the main control module via signal lines. The foam liquid delivery pipeline is equipped with a foam liquid tank 3, a foam liquid pressure relief valve 4, a foam pump 5, a second check valve 32, a foam liquid regulating valve 9, a foam liquid pressure gauge 10, and a foam flow meter 11. The foam liquid regulating valve 9 and the foam flow meter 11 are connected to the main control module via signal lines. The nitrogen delivery pipeline is equipped with a high-pressure gas cylinder 6, a third check valve 7, a high-pressure pressure reducing valve 8, a nitrogen regulating valve 12, a gas pressure gauge 13, and a gas flow meter 15. The nitrogen regulating valve 12 and the gas flow meter 15 are connected to the main control module via signal lines. A water-foam liquid mixer 14 is provided at the junction of the fire water supply pipeline and the foam liquid supply pipeline. A nitrogen-foam solution mixer 16 is provided at the junction of the foam solution supply pipeline and the nitrogen supply pipeline connected to the outlet of the water-foam liquid mixer 14. A pressure gauge 17 is provided at the outlet of the nitrogen-foam solution mixer 16.

[0043] The fire extinguishing medium release module consists of four foam fire monitors 23, 26, 27, and 30 connected to the outlet of the nitrogen-foam solution mixer. The foam fire monitors are connected to the fire monitor control module via signal lines, and the fire monitor control module is controlled by the main control module. The video image acquisition module consists of binocular cameras, and four binocular cameras 24, 25, 28, and 29 are installed around the transformer.

[0044] The fire analysis module preprocesses the substation fire video images acquired by the video image acquisition module (i.e., binocular cameras) (e.g., grayscale conversion, binarization), locates suspicious target areas, extracts features based on area changes, temperature changes, and thermal radiation changes in these areas, and uses a pre-trained convolutional neural network model to determine the substation fire type and location. The substation fire type refers to the type of electrical equipment involved in the fire, categorized as transformer fires, high-voltage parallel reactor (HRT) fires, dry-type reactor fires, cabinet fires, and cable trench fires. For transformer fires, further subdivisions include oil sump fires, surface fires, oil conservator fires, radiator fires, jet fires, and internal fires.

[0045] Based on the fire type and location identified by the fire analysis module, the main control module adjusts the ratio of water, foam liquid, and compressed nitrogen in the extinguishing medium, and adjusts the opening of the medium regulating valves (water regulating valve, nitrogen regulating valve, and foam liquid regulating valve) on the fire water delivery pipeline, foam liquid delivery pipeline, and nitrogen delivery pipeline. After a period of time (such as about 2 minutes), the opening of the medium regulating valves is readjusted based on the feedback of the fire situation.

[0046] The fire monitor control module adjusts the movement direction of the fire monitor in real time according to the control signal from the main control module, and automatically adjusts the spray angle and spray range of the extinguishing medium to ensure that the landing point of the extinguishing medium coincides with the location of the fire until the fire is extinguished.

[0047] The extinguishing medium is a mixture of nitrogen, foam liquid, and water, wherein the gas-liquid ratio of nitrogen flow rate to mixture flow rate is 7-10:1, the mixture flow rate is the sum of water flow rate and foam liquid flow rate, and the mixing ratio of foam liquid flow rate to mixture flow rate is 3-3.9:100.

[0048] Example 2

[0049] This embodiment provides an adaptive control method for extinguishing parameters of the compressed nitrogen foam fire extinguishing system described in Embodiment 1. It uses a multi-position binocular camera to collect fire video images and temperature information, and uploads them together with the temperature sensing cable signal to the fire analysis module. After the fire analysis module makes a comprehensive judgment on the fire type and the location of the fire, the main control module first sets the extinguishing parameters, and then sends instructions to the extinguishing medium generating device and the extinguishing medium releasing module to release the extinguishing medium with appropriate extinguishing parameters.

[0050] During the adjustment of the fire monitor's movement, the fire analysis module feeds back changes in fire type, combustion location, and fire combustion stage to the main control module. The main control module then sends valve opening information to the extinguishing medium control cabinet, readjusting the valve openings of the water regulating valve, foam liquid regulating valve, and nitrogen regulating valve, thereby changing the water-foam liquid mixing ratio and the gas-liquid ratio to ensure they remain within the effective range.

[0051] The fire extinguishing parameters include fire water flow rate, the mixing ratio of water flow rate / (water flow rate + foam liquid flow rate), and the gas-liquid ratio of nitrogen flow rate / (water flow rate + foam liquid flow rate).

[0052] Sending instructions to the extinguishing medium generating device, specifically including:

[0053] Adjust the valve opening of the fire water supply regulating valve on the fire water delivery pipeline according to the set fire water flow rate to regulate the fire water outflow rate;

[0054] Based on the outflow rate of fire-fighting water and according to the set water-foam liquid mixing ratio, the valve opening of the foam liquid regulating valve on the foam liquid delivery pipeline is adjusted to regulate the foam liquid flow rate.

[0055] Based on the fire water flow rate and foam liquid flow rate, and according to the set gas-liquid ratio, adjust the opening of the nitrogen regulating valve on the nitrogen delivery pipeline to regulate the nitrogen flow rate.

[0056] Sending instructions to the extinguishing medium release module, specifically including:

[0057] The horizontal rotation and elevation angle of the foam fire monitor are dynamically adjusted according to the location of the flame center; the spray speed, spray range and spray flow rate of the foam fire monitor are dynamically adjusted according to the temperature changes at the fire scene.

[0058] The specific steps of the above-mentioned adaptive control method for fire extinguishing parameters are as follows:

[0059] Step S1, Data Collection and Labeling: Construct a fire dataset covering multiple scenarios of power equipment, encompassing fire video images of typical facilities such as transformers, high-voltage reactors, dry-type reactors, cabinets, and cable trenches. A multi-level classification system is used for structured labeling: First, fire samples are labeled based on equipment type (including transformer fires, cable trench fires, cabinet fires, reactor fires, etc.). Transformer fires are further subdivided into six categories: oil sump fire, surface fire, oil conservator fire, radiator fire, jet fire, and internal fire. Center coordinate boxes are used to locate and label the flame area. At the same time, hierarchical labeling rules are established. Time-series frame-level labeling (such as flame jet direction and diffusion speed) is added to dynamic combustion features such as transformer jet fires. Pixel-level labeling of thermal imaging data (temperature threshold > 450℃) is used to supplement high-temperature areas such as oil conservator fires. Finally, a dual-modal dataset containing 12,000 frames of images and 8,000 video segments is formed, with transformer-specific samples accounting for 45%. All data are normalized and associated with metadata according to the COCO dataset standard.

[0060] Step S2, Data Preprocessing: For the dual-modal data characteristics (visible light RGB and infrared thermal imaging) of the substation fire monitoring scenario, the parallax of the two cameras is corrected by feature point matching to ensure the spatial consistency of the same fire source in the two modal images and eliminate the positioning error caused by equipment vibration or installation offset; the images are uniformly scaled to 320×320 resolution, and adaptive histogram equalization is used to enhance the local contrast of the flame area. At the same time, Z-Score normalization (pixel values ​​are normalized to [-1,1]) is used to eliminate the baseline drift interference of the thermal imaging sensor under different lighting conditions. To address the challenge of false flame signals generated by high-heat equipment in substations (such as radiators and cable joints), a dynamic masking algorithm is employed. This algorithm generates a priori region mask based on equipment layout drawings, suppressing pixel responses in non-fire-source high-temperature areas during preprocessing to reduce the model's false detection rate. Furthermore, to adapt to the multi-task input requirements of EfficientNet-B3, the dual-modal data is concatenated into a 6-channel (RGB+Infrared) tensor along the channel dimension. Online data augmentation (random rotation ±15°, Gaussian noise injection, and smoke simulation occlusion) is applied to expand sample diversity and improve the model's generalization ability to complex fire environments. The preprocessed data is finally input into the network in batch tensor form, with a single frame processing time of less than 5ms, meeting the low-latency requirements of real-time monitoring systems.

[0061] Step S3, construct the convolutional neural network model: EfficientNet-B3 is used as the backbone network. A composite scaling method is used to jointly optimize the network depth (coefficient 1.2), width (coefficient 1.4), and input resolution (320×320), improving the performance of small flame targets (e.g., 0.5m). 2 While enhancing fire detection capabilities, it maintains low computational costs for edge devices. A ResNet residual structure is embedded in Stages 3-5 of EfficientNet-B3, with a core module of "1×1 convolution dimensionality reduction → 3×3 convolution spatial feature extraction → 1×1 convolution dimensionality increase." Skip connections preserve shallow details, mitigating the gradient vanishing problem in deep networks and improving the localization accuracy of fire source center coordinates (x, y, w, h). The model employs multi-task outputs, primarily including: global average pooling and fully connected layer fire bounding boxes (IoU reaching 92.3%); and a feature pyramid fusing multi-scale information to classify the flame size of substation fires (small <1m). 2 Medium size 1-5m 2 Large size > 5m 2 ) and fire type (main transformer fire, high voltage reactor fire, cabinet fire, cable trench fire).

[0062] Step S4, Model Training: Annotated fire image data is used as the training set and input into the convolutional neural network model for training. During training, the model's weights and biases are adjusted using a gradient descent optimization algorithm to minimize the loss function. For the substation scenario, the input layer fuses visible light and infrared thermal imaging dual-modal data to suppress false detections in high-heat areas of equipment, and robustness is improved through data enhancement simulating occlusion and dynamic smoke.

[0063] Step S5, Model Application: After receiving a new fire image, the trained EfficientNet-B3 model first performs dual-modal fusion preprocessing on the input data: aligning the visible light (RGB) and infrared thermal imaging images and scaling them to 320×320 resolution; and eliminating spectral interference from high-heat equipment (such as radiators and cable joints) in the substation background through normalization. During forward propagation, the model outputs fire type classification (e.g., transformer oil sump fire, cable trench fire), flame size (small / medium / large), and combustion stage (initial / spreading / intense) in parallel. After the prediction results are filtered by non-maximum suppression (NMS) and confidence thresholding, fire-fighting strategy matching is triggered. If the fire is identified as an internal transformer fire (such as an oil conservator fire or a jet fire), the fire monitor system will be activated to dynamically adjust the opening of the water, foam liquid, and nitrogen valves according to the location of the fire (for example, high-grade nitrogen will be used to suppress splashing in jet fires, and low-grade foam will be used to cover the oil surface in oil conservator fires). At the same time, the angle of the fire monitor will be controlled to be aimed at the base of the flames in real time to ensure that the compressed nitrogen foam extinguishing medium accurately covers the core area of ​​the fire source and meets the real-time requirements of substation fire emergency response.

[0064] Step S6, Multi-stage Valve-Controlled Fire Extinguishing: In the multi-stage valve-controlled fire extinguishing mechanism, the main control module dynamically matches the fire type and combustion scenario, adjusting the opening signals of the foam liquid outlet valve and nitrogen valve in real time. By sending instructions to the fire extinguishing medium generating device, it precisely controls the mixing ratio (priority order: water → foam → nitrogen) and gas-liquid ratio of the three media (water, foam, and nitrogen). For different equipment fire types, the valve opening adopts a three-level combination strategy: high-resistance fires correspond to (medium, medium, medium), dry-type reactor fires to (medium, medium, high), cabinet fires to (low, low, medium), and cable trench fires to (medium, medium, high). Specifically targeting transformer fires, the main control module refines the control logic based on the depth of the fire's location: oil sump fires and jet fires activate the full high-voltage mode (high, high, high); surface fires and internal fires use (medium, medium, high); and oil conservator fires and radiator fires switch to the lowest medium intensity (low, low, low). This tiered strategy, through coupled analysis of extinguishing medium intensity and combustion characteristics (such as the high-temperature heat capacity of the oil conservator and the kinetic energy impact of the jet), achieves multi-stage coordination of nitrogen coverage suppression, foam penetration cooling, and water flow cooling protection, forming an adaptive closed-loop control link from fire source location to medium ratio. Specific levels are shown in Table 1.

[0065] Table 1. Query Ratio of Fire Types and Mixed Output Media

[0066]

[0067]

[0068] Step S7, Fire Situation Linkage Intelligent Control: Based on the real-time flame position (x, y, w, h), drive the fire monitor to aim at the flame root projection point with an accuracy of ±0.5° (three-dimensional spatial coordinate deviation <10cm) in real time to suppress the spread of fire. According to the fire type, fire location, and combustion stage, use a dynamic priority algorithm to adjust the combination of water, foam, and nitrogen valve openings, and simultaneously optimize the gas-liquid mixing ratio (7:1~10:1) and foam coverage density (≥0.5kg / m³). 2 Kalman filtering eliminates positioning jitter caused by smoke obstruction, ensuring that the extinguishing medium spray trajectory overlaps with the core area of ​​the flame by more than 95%.

[0069] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A compressed nitrogen foam fire extinguishing system, characterized in that, It includes a fire extinguishing medium generating device, a fire extinguishing medium release module, and a fire intelligent detection and control module; the fire intelligent detection and control module includes a video image acquisition module, a fire analysis module, a main control module, and a fire monitor control module; The fire extinguishing medium generating device includes fire water delivery pipeline, foam liquid delivery pipeline and nitrogen delivery pipeline. Each of the three delivery pipelines is equipped with a medium regulating valve and a medium flow meter, which are connected to the main control module through signal lines. The extinguishing medium release module includes at least one foam fire monitor connected to the outlet of the nitrogen-foam solution mixer, and the foam fire monitor is connected to the fire monitor control module via a signal line. The fire analysis module preprocesses the substation fire video images acquired by the video image acquisition module, locates suspicious target areas, extracts features by the area changes, temperature changes and thermal radiation changes of the suspicious target areas, and uses a pre-trained convolutional neural network model to determine the substation fire type and burning location. The main control module generates a dynamic control strategy based on the fire type and combustion location, adjusts the opening of the medium regulating valves on the fire water delivery pipeline, foam liquid delivery pipeline and nitrogen delivery pipeline, adjusts the output ratio of water, foam liquid and compressed nitrogen in the extinguishing medium, and then readjusts the opening of the medium regulating valves according to the fire situation after a period of time. The fire monitor control module adjusts the movement direction of the fire monitor in real time according to the control signal from the main control module, and automatically adjusts the spray angle and spray range of the extinguishing medium to ensure that the landing point of the extinguishing medium coincides with the location of the fire source.

2. The compressed nitrogen foam fire extinguishing system according to claim 1, characterized in that, The extinguishing medium is a mixture of nitrogen, foam liquid, and water, wherein the gas-liquid ratio of nitrogen flow rate to mixture flow rate is 7-10:1, the mixture flow rate is the sum of water flow rate and foam liquid flow rate, and the mixing ratio of foam liquid flow rate to mixture flow rate is 3-3.9:

100.

3. The adaptive control method for extinguishing parameters of the compressed nitrogen foam fire extinguishing system according to claim 1 or 2, characterized in that, include: A multi-position binocular camera is used to collect fire video images and temperature information, which are then uploaded to the fire analysis module along with the temperature sensing cable signal. After the fire analysis module makes a comprehensive judgment on the fire type and burning location, the main control module first sets the fire extinguishing parameters, and then sends instructions to the fire extinguishing medium generating device and the fire extinguishing medium releasing module to release the fire extinguishing medium with appropriate fire extinguishing parameters. During the adjustment of the fire monitor's movement, the fire analysis module feeds back changes in fire type, combustion location, and fire combustion stage to the main control module. The main control module then sends valve opening information to the extinguishing medium control cabinet, readjusting the valve openings of the water regulating valve, foam liquid regulating valve, and nitrogen regulating valve, thereby changing the water-foam liquid mixing ratio and the gas-liquid ratio to ensure they remain within the effective range.

4. The adaptive control method for fire extinguishing parameters according to claim 3, characterized in that, The fire extinguishing parameters include fire water flow rate, the mixing ratio of water flow rate / (water flow rate + foam liquid flow rate), and the gas-liquid ratio of nitrogen flow rate / (water flow rate + foam liquid flow rate).

5. The adaptive control method for fire extinguishing parameters according to claim 4, characterized in that, Sending instructions to the extinguishing medium generating device, specifically including: Adjust the valve opening of the fire water supply regulating valve on the fire water delivery pipeline according to the set fire water flow rate to regulate the fire water outflow rate; Based on the outflow rate of fire-fighting water and according to the set water-foam liquid mixing ratio, the valve opening of the foam liquid regulating valve on the foam liquid delivery pipeline is adjusted to regulate the foam liquid flow rate. Based on the fire water flow rate and foam liquid flow rate, and according to the set gas-liquid ratio, adjust the opening of the nitrogen regulating valve on the nitrogen delivery pipeline to regulate the nitrogen flow rate.

6. The adaptive control method for fire extinguishing parameters according to claim 4, characterized in that, Sending instructions to the extinguishing medium release module, specifically including: The horizontal rotation and elevation angle of the foam fire monitor are dynamically adjusted according to the location of the flame center; the spray speed, spray range and spray flow rate of the foam fire monitor are dynamically adjusted according to the temperature changes at the fire scene.

7. The adaptive control method for fire extinguishing parameters according to claim 3, characterized in that, A fire dataset covering multiple scenarios of power equipment was constructed, including fire video images of transformer fires, high-resistance fires, dry-type reactor fires, cabinet fires, and cable trench fires. A multi-level classification system was used for structured annotation: fire types were labeled based on equipment type, and transformer fires were further subdivided into oil sump fires, surface fires, oil conservator fires, radiator fires, jet fires, and internal fires. Center coordinate box localization and type labeling were implemented for flame areas, and hierarchical annotation rules were established. Time-series frame-level annotations were added to dynamic combustion features, and pixel-level labeling of thermal imaging data was used to supplement high-temperature areas, ultimately forming a dual-modal dataset.

8. The adaptive control method for fire extinguishing parameters according to claim 7, characterized in that, The convolutional neural network model described is the EfficientNet-B3 composite scaling model, and its specific construction process is as follows: The model employs EfficientNet-B3 as its backbone network, and jointly optimizes the network depth, width, and input resolution using a composite scaling method. ResNet residual structures are embedded in Stages 3-5 of the model, with "1×1 convolution for dimensionality reduction → 3×3 convolution for spatial feature extraction → 1×1 convolution for dimensionality enhancement" as the core module. Skip connections are used to preserve shallow details. The model employs multi-task outputs, including global average pooling and fully connected layer fire bounding boxes. A feature pyramid fuses multi-scale information to classify the flame size and fire type of substation fires. Labeled fire image data was used as the training set and input into the EfficientNet-B3 composite scaling model for training. During training, the model's weights and biases were adjusted using the gradient descent optimization algorithm to minimize the loss function. For the substation scenario, the input layer fused visible light and infrared thermal imaging dual-modal data to suppress false detections in high-heat areas of equipment, and robustness was improved by simulating occlusion and dynamic smoke data enhancement. After receiving new fire video images, the trained EfficientNet-B3 composite scaling model performs bimodal fusion preprocessing on the input data. During forward propagation, the model outputs the fire type, flame size, and combustion stage in parallel. The prediction results are filtered by nonmaximum suppression and confidence thresholding, triggering fire-fighting strategy matching. If the fire is identified as an internal transformer fire, the main control module is linked to dynamically adjust the valve openings of the water, foam liquid, and nitrogen regulating valves according to the fire type, while simultaneously controlling the fire monitor angle to be aimed at the base of the flame in real time.

9. The adaptive control method for fire extinguishing parameters according to claim 8, characterized in that, A multi-stage valve-controlled fire extinguishing mechanism is adopted. The main control module dynamically matches the fire type and the location of the fire, and adjusts the opening signals of the foam liquid regulating valve and the nitrogen regulating valve in real time. By sending instructions to the fire extinguishing medium control cabinet, it precisely controls the mixing ratio and gas-liquid ratio of the three media, water, foam and nitrogen, with the priority order being water → foam → nitrogen. For different equipment fire types, the valve opening adopts a three-level combination strategy of high, medium and low. For transformer fire types, the main control module refines the control logic according to the depth of the fire location.

10. The adaptive control method for fire extinguishing parameters according to claim 8, characterized in that, Based on the real-time flame position (x, y, w, h), the fire monitor is driven to aim at the flame root projection point with an accuracy of ±0.5° in real time to suppress the spread of fire. According to the fire type, combustion location and combustion stage, the opening combination of water, foam and nitrogen regulating valves is adjusted by dynamic priority algorithm, and the gas-liquid ratio, mixing ratio and foam coverage density are optimized simultaneously. Kalman filtering is used to eliminate positioning jitter caused by smoke obstruction, ensuring that the overlap between the fire extinguishing medium spray trajectory and the core area of ​​the flame is more than 95%.

Images