Electrochemical energy storage power station fire hazard early warning detection prevention and control system and prevention and control method

Abstract

The invention discloses an electrochemical energy storage power station fire early warning detection prevention and control system and a prevention and control method. According to the system, an anti-explosion camera, a hydrogen and carbon monoxide smoke temperature sensing composite detector, a battery module patch type temperature sensor, a battery module voltage transformer, a gas fire extinguishing agent conveying pipeline and a nozzle are installed in a protection area; a fire alarm controller, a manual control box, a light alarm and a sound alarm are installed on the outer side of the protection area, a gas fire extinguishing agent storage bottle, a starting bottle and components of the starting bottle are placed in the fire extinguishing agent storage room, and the gas fire extinguishing agent storage bottle is connected with a gas fire extinguishing agent conveying pipeline through a collecting pipe. According to the system, a video image intelligent identification technology is combined with a detector installed on site to judge the danger degree, comprehensive judgment is carried out in combination with big data of long-term operation of a power station battery module and an artificial intelligence algorithm during judgment, and corresponding equipment can be automatically or manually controlled when it is detected that the electrochemical energy storage power station is in danger. And the fire hazard of the electrochemical energy storage power station is controlled and eliminated.

Description

Technical Field

[0001] The present invention relates to the technical field of fire alarm for electrochemical energy storage power stations, and particularly to a fire early warning detection and prevention and control system and a prevention and control method for an electrochemical energy storage power station. Background Art

[0002] With the continuous and rapid development of the energy storage industry, the number and scale of energy storage projects are both rising rapidly. Fire protection, as the last line of defense for energy storage safety, is particularly crucial. However, the increase in the number and scale of energy storage has made it difficult for many existing fire protection technologies and measures to meet the rapid development needs of the energy storage industry; especially the fire alarm requirements for electrochemical energy storage power stations. Fires in chemical energy storage power stations are difficult to predict, develop rapidly, and often accompany phenomena such as the explosion of flammable gases during the fire process. Especially when a fire occurs, the consequences are very serious, bringing irreparable losses to human life and property; and the existing fire early warning detection methods and fire prevention and control methods can no longer meet the safety production requirements of electrochemical energy storage power stations; the main reasons for fires or explosions in electrochemical energy storage power stations include four aspects: battery body factors, external excitation factors, operating environment factors, and management system factors, etc.; from the analysis of the causes of fire and explosion accidents in global electrochemical energy storage power stations in the past 10 years, the main cause of fire and explosion is due to battery body factors, and the root cause of battery body combustion and explosion lies in battery thermal runaway. The main reasons for inducing battery thermal runaway are, first, battery short circuit resulting in a sharp rise in electrolyte temperature, and second, electrical and thermal shocks outside the battery. Among them, the voltage inconsistency between battery clusters causes circulating current during operation, leading to battery thermal runaway; for this reason, the applicant proposes a fire early warning detection and prevention and control system for an electrochemical energy storage power station based on the characteristics of fires occurring in electrochemical energy storage power stations, which can well solve the problems of fire prevention and control for electrochemical energy storage power stations. Summary of the Invention

[0003] To solve the above technical problems, the present invention proposes a fire early warning detection and prevention and control system and a prevention and control method for an electrochemical energy storage power station. By combining video image intelligent recognition technology with detectors installed on site to determine the degree of danger, and also combining big data and artificial intelligence algorithms of long-term operation of the power station battery modules for comprehensive determination during the determination. When a fire danger occurs in the electrochemical energy storage power station, corresponding equipment can be automatically or manually controlled to eliminate and control the fire danger of the electrochemical energy storage power station, and avoid the occurrence of fire accidents.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] An electrochemical energy storage power station fire warning detection and prevention and control system includes a gas fire extinguishing agent storage bottle, a gas fire extinguishing agent storage bottle frame, a manifold, a liquid flow check valve, a hose, an air flow check valve, a bottle head valve, a starting pipeline, a pressure signaler, a safety valve, a selector valve, a signal feedback line, a solenoid valve, a starting bottle, a starting bottle frame, a fire alarm controller, a control line, a manual control box, a light alarm, a sound alarm, a nozzle, an explosion-proof camera, hydrogen, a combined smoke and temperature detector for carbon monoxide, a battery module patch type temperature sensor, a battery module voltage transformer, and a gas fire extinguishing agent delivery pipeline. It is characterized in that: in the protected area of the electrochemical energy storage power station fire warning detection and prevention and control system, an explosion-proof camera and a combined smoke and temperature detector for hydrogen and carbon monoxide are arranged at the top according to the specification requirements; a battery module patch type temperature sensor and a battery module voltage transformer are arranged on the battery pack equipment; a fire alarm controller, a manual control box, a light alarm, and a sound alarm are installed outside the protected area; the manual control box, the light alarm, the sound alarm, the explosion-proof camera, the combined smoke and temperature detector for hydrogen and carbon monoxide, the battery module patch type temperature sensor, and the battery module voltage transformer are connected to the fire alarm controller through a control line; in the fire extinguishing agent storage room of the protected area of the electrochemical energy storage power station fire warning detection and prevention and control system, a gas fire extinguishing agent storage bottle is placed, the gas fire extinguishing agent storage bottle is installed and fixed through a gas fire extinguishing agent storage bottle frame, a bottle head valve is installed on the top of the gas fire extinguishing agent storage bottle, a hose and a starting pipeline are installed on the bottle head valve, a liquid flow check valve is installed on the upper part of the hose and connected to the manifold, a liquid flow check valve is arranged on the starting pipeline, and an air flow check valve is installed on the manifold between the bottle head valves of the gas fire extinguishing agent storage bottles; the manifold is connected to the gas fire extinguishing agent delivery pipeline, and a pressure signaler, a safety valve, and a selector valve are installed on the gas fire extinguishing agent delivery pipeline; the pressure signaler and the selector valve are connected to the fire alarm controller through a signal feedback line; a starting bottle group is installed outside the gas fire extinguishing agent storage bottle, the starting bottle group is fixed through a starting bottle frame, a solenoid valve is installed on the upper part of the starting bottle, the solenoid valve is connected to the starting pipeline, the starting pipe is connected to the selector valve, and the solenoid valve is connected to the fire alarm controller through a control line; the gas fire extinguishing agent delivery pipeline is arranged at the top of the protected area according to the specification, and nozzles are arranged on the gas fire extinguishing agent delivery pipeline as required.

[0006] Further, the electrochemical energy storage power station fire warning detection and prevention and control system combines video image intelligent recognition technology through an explosion-proof camera with the feedback signals of a combined smoke and temperature detector for hydrogen and carbon monoxide, a battery module patch type temperature sensor, and a battery module voltage transformer installed on site to determine the degree of danger.

[0007] Further, when determining the risk level of the fire warning detection and prevention system for the electrochemical energy storage power station, big data on the long-term operation of the power station's battery modules and artificial intelligence algorithms are also combined for comprehensive determination.

[0008] Further, the fire warning detection and prevention system for the electrochemical energy storage power station is provided with control modes of "automatic control", "electrical manual control" and "mechanical emergency operation".

[0009] The prevention and control method of the fire warning detection and prevention system for the electrochemical energy storage power station specifically includes the following steps:

[0010] Step 1: Data collection and preprocessing; collect video image data using explosion-proof cameras installed in the protection area of the electrochemical energy storage power station, and preprocess the collected images.

[0011] Step 2: Collect the dataset for YOLOv5 model training; divide the collected image dataset containing fire and non-fire scenarios into a training set, a validation set and a test set according to the ratio of 7:2:1.

[0012] Step 3: Build a YOLOv5 fire warning model; build a YOLOv5 fire warning model containing components such as an input layer, a backbone, a neck and an output layer for feature extraction and target detection.

[0013] Step 4: Train the YOLOv5 fire warning model; use the training set to train the YOLOv5 fire warning model, adopt Adam as the optimizer, update the model parameters to minimize the loss function, monitor the performance of the model through the validation set, and adjust training parameters such as the learning rate according to the loss and accuracy on the validation set.

[0014] Step 5: Fire warning detection; input the preprocessed video image sequence into the trained YOLOv5 fire warning model, and the model outputs prediction results, each of which contains information such as the coordinates and size of the prediction box, the confidence of the target included, and the predicted category.

[0015] Step 6: Determine the risk level of the electrochemical energy storage power station; comprehensively determine the temperature and voltage data of the battery modules in the safety monitoring center, the hydrogen and carbon monoxide smoke and temperature composite detectors, and the fire-related target detection results in the video images to obtain the fire risk level and conduct processing.

[0016] Further, the process of data collection and preprocessing in Step 1 can be expressed as:

[0017] Let the collected video image sequence be:

[0018] I = {I 1 , I 2 , …, In}

[0019] Where: I represents the collected video image sequence;

[0020] n is the number of video frames;

[0021] The collected images are normalized, and the image pixel values are mapped to the interval [0,1]. The processing formula is:

[0022]

[0023] Where: I(x,y) is the pixel value of the original image at the (x,y) coordinate;

[0024] I norm (x,y) is the normalized pixel value.

[0025] Furthermore, the YOLOv5 fire warning model in step three consists of four parts, namely the input layer, backbone, neck, and output layer; specifically, it can be expressed as:

[0026] Input layer: The function of the input layer is data augmentation, adaptive anchor box calculation, and adaptive image scaling;

[0027] Backbone: The backbone is the main network, used to extract information from the picture;

[0028] Neck: The neck is the FPN+PAN structure, used to improve the robustness of the model;

[0029] Output layer: The output layer serves as the network output, making predictions using the previously extracted features;

[0030] The loss function of YOLOv5 consists of localization loss, confidence loss, and classification loss, and is specifically expressed as follows:

[0031] Localization loss:

[0032]

[0033] Where: L is the loss of YOLOv5;

[0034] L loc is the localization loss;

[0035] L conf is the confidence loss;

[0036] L class is the classification loss;

[0037] is the adjusted localization loss weight;

[0038] is the adjusted confidence loss weight;

[0039] is the adjusted classification loss weight, and satisfies

[0040] Confidence loss:

[0041] Confidence loss L conf is used to measure the difference between the confidence that the predicted bounding box contains the target and the true confidence. The formula is:

[0042]

[0043] where: λ conf is the confidence loss weight;

[0044] c i,j is the confidence that the j-th predicted bounding box in the i-th image contains the target;

[0045] is the confidence that the j-th ground truth box in the i-th image contains the target;

[0046] Classification loss:

[0047] Classification loss L class is used to measure the difference between the predicted class and the true class. The formula is:

[0048]

[0049] where: λ class is the classification loss weight;

[0050] p i,j is the predicted class probability of the j-th predicted bounding box in the i-th image;

[0051] is the predicted class probability of the j-th ground truth box in the i-th image.

[0052] Furthermore, the fire warning detection in the fifth step can be specifically expressed as:

[0053] 1) Input the preprocessed video image sequence I norm into the trained YOLOv5 model, and the model outputs the prediction result R = {R 1 , R 2 , …, R n}, where each R i contains information such as the predicted bounding box coordinates, size, confidence of containing the target, and predicted class;

[0054] 2) Screen and process the prediction results, and set the confidence threshold τ conf , and only retain the prediction results with a confidence greater than τ conf . Let the result after screening be R filtered = {R filtered1 , R filtered2 , …, R filteredn}, where m ≤ n;

[0055] 3) Determine whether it is a fire-related target according to the predicted category, predict the fire-related features, and set the fire-related target detection result as R fire = {R fire1 , R fire2 ,..., R firek}, where k ≤ m.

[0056] Furthermore, the specific process of fire risk level determination and processing in step six is as follows:

[0057] 1) Set the temperature threshold of the battery module as T threshold , the voltage threshold as V threshold , the hydrogen threshold as H threshold , and the carbon monoxide threshold as C threshold . If the temperature of all battery modules is less than the threshold, the voltage is less than the threshold, the hydrogen is less than the threshold, and the carbon monoxide is less than the threshold, and no fire-related targets are detected in the video image, it is determined as "system safe", and the system operates normally;

[0058] 2) Fire-related targets are detected in the video image. Combining the data of the battery module temperature, voltage, hydrogen concentration, and carbon dioxide concentration to judge that a fire may occur, it is determined as "system fire". The safety monitoring center directly opens the selection valve and container valve of the area where the alarm battery module is located, and then releases the fire extinguishing agent until the fire is extinguished;

[0059] 3) Fire-related targets are detected in the video image, but combining the data of the battery module temperature, voltage, hydrogen concentration, and carbon dioxide concentration to judge that the battery module will not undergo thermal runaway resulting in a fire, it is determined as "system warning". The safety monitoring center uses electrical isolation to isolate the alarm battery module from the energy storage power station, and then incorporates it into the energy storage power station after detecting that the safety of the alarm battery module meets the requirements.

[0060] Furthermore, the positioning loss L loc is used to measure the position difference between the predicted bounding box and the ground truth bounding box, and its formula is:

[0061]

[0062] where: N is the number of images;

[0063] M is the number of predicted bounding boxes in each image;

[0064] λ loc is the localization loss weight;

[0065] (x i,j ,y i,j ,w i,j ,h i,j ) are the coordinates and dimensions of the j-th predicted bounding box in the i-th image, where x i,j and y i,j are the coordinates, and w i,j and h i,j are the width and height;

[0066] are the coordinates and dimensions of the j-th ground truth bounding box in the i-th image, and are the coordinates, and are the width and height.

[0067] The benefits brought by this application are:

[0068] 1. The fire early warning detection and prevention and control system for electrochemical energy storage power stations is designed and manufactured according to the characteristics of fire occurrence in electrochemical energy storage power stations, and has good fire early recognition ability and control ability;

[0069] 2. The fire early warning detection and prevention and control system for electrochemical energy storage power stations combines video image intelligent recognition technology with detectors installed on site to determine the degree of danger, with high recognition and accuracy;

[0070] 3. When determining the fire danger, the fire early warning detection and prevention and control system for electrochemical energy storage power stations combines the big data of long-term operation of the power station battery modules and artificial intelligence algorithms for comprehensive determination, reducing errors and avoiding misjudgment;

[0071] 4. The fire early warning detection and prevention and control system for electrochemical energy storage power stations has "automatic control", "electrical manual control" and "mechanical emergency operation", and multiple control methods ensure the stable and reliable operation of the system.

[0072] 5. The fire early warning detection and prevention and control system for electrochemical energy storage power stations uses YOLOv5 as the fire early warning model, which includes components such as the input layer, backbone, neck, and output layer. This architecture design is beneficial to feature extraction and target detection, and improves the accuracy of early warning. Brief Description of the Drawings

[0073] Figure 1 is the structural layout schematic diagram of the present invention;

[0074] Figure 2 is the flow chart of the video image intelligent recognition technology of the present invention;

[0075] Figure 3 This is the fire warning and identification diagram of the present invention;

[0076] Figure 4 This is the schematic diagram of the system control flow of the present invention.

[0077] The markings in the figure are: 1. Gas fire extinguishing agent storage bottle; 2. Gas fire extinguishing agent storage bottle frame; 3. Manifold; 4. Liquid flow check valve; 5. Hose; 6. Air flow check valve; 7. Bottle head valve; 8. Starting pipeline; 9. Pressure signaler; 10. Safety valve; 11. Selector valve; 12. Signal feedback line; 13. Solenoid valve; 14. Starting bottle; 15. Starting bottle frame; 16. Fire alarm controller; 17. Control line; 18. Manual control box; 19. Optical alarm; 20. Acoustic alarm; 21. Nozzle; 22. Explosion-proof camera; 23. Hydrogen and carbon monoxide smoke and temperature composite detector; 24. Battery module patch type temperature sensor; 25. Battery module voltage transformer; 26. Gas fire extinguishing agent delivery pipeline. Specific embodiments

[0078] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0079] As Figure 1 shown, a fire warning, detection and prevention and control system for an electrochemical energy storage power station is shown, including a gas fire extinguishing agent storage bottle 1, a gas fire extinguishing agent storage bottle frame 2, a manifold 3, a liquid flow check valve 4, a hose 5, an air flow check valve 6, a bottle head valve 7, a starting pipeline 8, a pressure signaler 9, a safety valve 10, a selector valve 11, a signal feedback line 12, a solenoid valve 13, a starting bottle 14, a starting bottle frame 15, a fire alarm controller 16, a control line 17, a manual control box 18, an optical alarm 19, an acoustic alarm 20, a nozzle 21, an explosion-proof camera 22, a hydrogen and carbon monoxide smoke and temperature composite detector 23, a battery module patch type temperature sensor 24, a battery module voltage transformer 25 and a gas fire extinguishing agent delivery pipeline 26, as Figure 1As shown: In the protected area of the electrochemical energy storage power station fire warning detection and prevention and control system, an explosion-proof camera 22 and a hydrogen and carbon monoxide smoke and temperature composite detector 23 are arranged at the top according to the specification requirements; their arrangement is reasonably distributed at the top of the protected area according to the detection requirements of the equipment, and the corresponding pipelines are pre-buried during the civil engineering construction; on the battery pack equipment, a battery module patch type temperature sensor 24 and a battery module voltage transformer 25 are arranged. The operating temperature and voltage of the battery pack are monitored through the battery module patch type temperature sensor 24 and the battery module voltage transformer 25. When abnormalities are found, signals are sent out in time to notify the management personnel for handling; a fire alarm controller 16, a manual control box 18, a light alarm 19 and a sound alarm 20 are installed outside the protected area; through the manual control box 18, the equipment in the protected area can be controlled outside the protected area, and relevant necessary measures can be taken to prevent fires from occurring. The installed light alarm 19 and sound alarm 20 can timely remind relevant personnel to evacuate in time to avoid injuries caused by fires. The manual control box 18, light alarm 19, sound alarm 20, explosion-proof camera 22, hydrogen and carbon monoxide smoke and temperature composite detector 23, battery module patch type temperature sensor 24 and battery module voltage transformer 25 are connected to the fire alarm controller 24 through a control line 17, and the control line 17 is set according to different communication requirements; in the fire extinguishing agent storage room in the protected area of the electrochemical energy storage power station fire warning detection and prevention and control system, a gas fire extinguishing agent storage bottle 1 is placed. The gas fire extinguishing agent storage bottle 1 is installed and fixed through a gas fire extinguishing agent storage bottle frame 2. The number of groups of the gas fire extinguishing agent storage bottles 1 is determined according to the size of the protected area, and multiple protected areas can be protected. A bottle head valve 7 is installed on the top of the gas fire extinguishing agent storage bottle 1. A hose 5 and a starting pipeline 8 are installed on the bottle head valve 7. A liquid flow check valve 4 is installed on the upper part of the hose 5 and is connected to a manifold 3. A liquid flow check valve 4 is provided on the starting pipeline 8. An air flow check valve 6 is installed on the manifold 3 between the bottle head valves 7 of the gas fire extinguishing agent storage bottles 1; the manifold 3 is connected to a gas fire extinguishing agent delivery pipeline 26. A pressure signaler 9, a safety valve 10 and a selector valve 11 are installed on the gas fire extinguishing agent delivery pipeline 26. The pressure signaler 9 monitors the pressure of the pipeline in real time to avoid damage to the pipeline due to excessive pressure. The safety valve 10 releases pressure when the pipeline pressure is too high to ensure the safe operation of the gas fire extinguishing agent delivery pipeline 26; the pressure signaler 9 and the selector valve 11 are connected to the fire alarm controller 16 through a signal feedback line 12; 14 groups of starting bottles 14 are installed outside the gas fire extinguishing agent storage bottle 1. The 14 groups of starting bottles 14 are fixed through a starting bottle frame 15. An electromagnetic valve 13 is installed on the upper part of the starting bottle 14. The electromagnetic valve 13 is connected to the starting pipeline 8. The starting pipe 8 is connected to the selector valve 11. The electromagnetic valve 13 is connected to the fire alarm controller 16 through a control line 17;The shown gas fire extinguishing agent conveying pipeline 26 is laid out at the top of the protected area according to the specifications, and the gas fire extinguishing agent conveying pipeline 26 is made and constructed of seamless steel pipe; nozzles 21 are arranged on the installed gas fire extinguishing agent conveying pipeline 26 according to requirements.

[0080] The shown fire warning detection and prevention and control system for an electrochemical energy storage power station combines the feedback signals of a hydrogen and carbon monoxide smoke and temperature composite detector 23, a battery module patch type temperature sensor 24, and a battery module voltage transformer 25 installed on site with the video image intelligent recognition technology by means of an explosion-proof camera 22 to determine the degree of danger; when determining the degree of danger of the shown fire warning detection and prevention and control system for an electrochemical energy storage power station, it also combines the big data of the long-term operation of the power station battery module and artificial intelligence algorithms for comprehensive determination; the shown fire warning detection and prevention and control system for an electrochemical energy storage power station is provided with control modes of "automatic control", "electrical manual control", and "mechanical emergency operation".

[0081] As Figure 2 shown, the shown is a flow chart of video image intelligent recognition technology, and its specific process steps are as follows:

[0082] Step 1: Data collection and preprocessing; use the explosion-proof camera set in the protected area of the electrochemical energy storage power station to collect video image data, and preprocess the collected images; the process of its data collection and preprocessing can be expressed as:

[0083] Let the collected video image sequence be:

[0084] I = {I 1 , I 2 , …, I n}

[0085] Among them: I represents the collected video image sequence;

[0086] n is the number of video frames;

[0087] Perform normalization processing on the collected images, map the image pixel values to the interval [0, 1], and the processing formula is:

[0088]

[0089] Among them: I(x, y) is the pixel value of the original image at the (x, y) coordinate;

[0090] I norm (x, y) is the pixel value after normalization;

[0091] Step 2: Collect the data set for YOLOv5 model training; divide the collected image data set containing fire and non-fire scenes into a training set, a validation set, and a test set according to the ratio of 7:2:1;

[0092] Step 3: Build the YOLOv5 fire warning model; build a YOLOv5 fire warning model containing components such as an input layer, backbone, neck, and output layer for feature extraction and object detection; specifically, it can be expressed as:

[0093] Input layer: The function of the input layer is data augmentation, adaptive anchor box calculation, and adaptive image scaling;

[0094] Backbone: The backbone is the main network for extracting information from images;

[0095] Neck: The neck is the FPN+PAN structure for improving the robustness of the model;

[0096] Output layer: The output layer serves as the network output and makes predictions using the previously extracted features;

[0097] The loss function of YOLOv5 consists of localization loss, confidence loss, and classification loss, and is specifically expressed as follows:

[0098] Localization loss:

[0099]

[0100] Where: L is the loss of YOLOv5;

[0101] L loc is the localization loss;

[0102] L conf is the confidence loss;

[0103] L class is the classification loss;

[0104] is the adjusted localization loss weight;

[0105] is the adjusted confidence loss weight;

[0106] is the adjusted classification loss weight, and satisfies

[0107] The localization loss L loc is used to measure the position difference between the predicted box and the ground truth box, and its formula is:

[0108]

[0109] Where: N is the number of images;

[0110] M is the number of predicted boxes in each image;

[0111] λ loc is the localization loss weight;

[0112] (x i,j , y i,j , w i,j , h i,j ) are the coordinates and dimensions of the j-th predicted bounding box in the i-th image. x i,j and y i,j are the coordinates, and w i,j and h i,j are the width and height;

[0113] are the coordinates and dimensions of the j-th ground truth bounding box in the i-th image, and are the coordinates, and are the width and height

[0114] Confidence loss:

[0115] The confidence loss L conf is used to measure the difference between the confidence that the predicted bounding box contains the object and the true confidence. The formula is:

[0116]

[0117] where: λ conf is the confidence loss weight;

[0118] c i,j is the confidence that the j-th predicted bounding box in the i-th image contains the object;

[0119] is the confidence that the j-th ground truth bounding box in the i-th image contains the object;

[0120] Classification loss:

[0121] The classification loss L class is used to measure the difference between the predicted class and the true class. The formula is:

[0122]

[0123] where: λ class is the classification loss weight;

[0124] p i,j is the predicted class probability of the j-th predicted bounding box in the i-th image;

[0125] is the predicted class probability of the j-th ground truth bounding box in the i-th image.

[0126] Step 4: Train the YOLOv5 fire warning model; use the training set to train the YOLOv5 fire warning model, adopt Adam as the optimizer, update the model parameters to minimize the loss function, monitor the performance of the model through the validation set, and adjust the training parameters such as the learning rate according to the loss and accuracy on the validation set;

[0127] Step 5: Fire warning detection; input the preprocessed video image sequence into the trained YOLOv5 fire warning model, and the model outputs the prediction results, where each result contains information such as the coordinates and size of the prediction box, the confidence of the included target, and the predicted category; specifically, it can be expressed as:

[0128] 1) Input the preprocessed video image sequence I norm into the trained YOLOv5 model, and the model outputs the prediction results R = {R 1 , R 2 , …, R n}, where each R i contains information such as the coordinates and size of the prediction box, the confidence of the included target, and the predicted category;

[0129] 2) Screen and process the prediction results, set the confidence threshold τ conf , and only retain the prediction results with a confidence greater than τ conf . Let the screened results be R filtered = {R filtered1 , R filtered2 ,..., R filteredn}, where m ≤ n;

[0130] 3) Judge whether it is a fire-related target according to the predicted category, and predict the fire-related features. Let the fire-related target detection results be R fire = {R fire1 , R fire2 ,..., R firek}, where k ≤ m;

[0131] Step 6: Determine the danger level of the electrochemical energy storage power station; comprehensively determine the temperature, voltage data of the battery modules in the safety monitoring center, the hydrogen and carbon monoxide smoke temperature composite detectors, and the fire-related target detection results in the video images to obtain the fire danger level and conduct processing; the specific process is as follows:

[0132] 1) Set the temperature threshold of the battery module as T threshold , the voltage threshold as V threshold , the hydrogen threshold as H threshold , and the carbon monoxide threshold as C threshold, if the temperatures of all battery modules are less than the threshold, the voltages are less than the threshold, the hydrogen content is less than the threshold, the carbon monoxide content is less than the threshold, and no fire-related targets are detected in the video image, it is determined as "system safe", and the system operates normally;

[0133] 2) If fire-related targets are detected in the video image, and it is judged that a fire may occur by combining the data of battery module temperature, voltage, hydrogen concentration and carbon dioxide concentration, it is determined as "system fire", and the safety monitoring center directly opens the selection valve and container valve of the area where the alarm battery module is located, and then releases the fire extinguishing agent until the fire is extinguished;

[0134] 3) If fire-related targets are detected in the video image, but it is judged that the battery module will not experience thermal runaway and cause a fire by combining the data of battery module temperature, voltage, hydrogen concentration and carbon dioxide concentration, it is determined as "system warning", and the safety monitoring center uses electrical isolation to isolate the alarm battery module from the energy storage power station. After detecting that the safety of the alarm battery module meets the requirements, it is reconnected to the energy storage power station, as Figure 3 shown.

[0135] As Figure 4As shown in the figure, it is the control flow chart of the fire warning detection and prevention and control system for an electrochemical energy storage power station. In each protected area, according to the specifications, hydrogen and carbon monoxide smoke and temperature composite detectors for energy storage power stations need to be set, and a patch temperature sensor and a voltage transformer are set outside each battery module to measure the temperature and voltage of each battery module in real time, providing early warning information for thermal runaway caused by the increase in the temperature and voltage of the battery body. At the same time, according to the type and characteristics of the battery body, such as lithium iron phosphate battery or ternary lithium battery, reasonable temperature and voltage thresholds are set. During normal operation, the battery module temperature detector and the battery module voltage detector transmit the detected battery module temperature and voltage to the fire alarm controller in real time. When the values detected by all detectors are lower than the set thresholds, it is considered that the battery module, that is, the energy storage power station, is safe, and thus the temperature and voltage signals of the battery module are continuously collected at regular intervals. When the temperature and voltage of any battery module exceed the set thresholds, the fire alarm controller issues an alarm signal and transmits the alarm signal to the safety monitoring center in real time. The safety monitoring center combines the temperature and voltage signals of the battery module with the real-time fire video of the explosion-proof camera in the area where the alarm battery module is located, and uses video image intelligent recognition technology to comprehensively determine the degree of danger. When the safety monitoring center determines the degree of danger of the electrochemical energy storage power station, it combines the big data of the long-term operation of the power station battery module and artificial intelligence algorithms for comprehensive determination. When the safety monitoring center determines that the area where the alarm battery module is located is due to external factors that cause the temperature or voltage of a certain battery module to suddenly increase and cause an alarm, and the battery module and the energy storage power station are safe after the external factors are removed, it is determined as "system safe", and the system still operates normally. When the safety monitoring center determines that a fire has occurred or is about to occur in the area where the alarm battery module is located, it is determined as "system fire". The safety monitoring center directly opens the selection valve and the container valve in the area where the alarm battery module is located, and then releases the fire extinguishing agent until the fire is extinguished. When the safety monitoring center determines that it is only because the temperature or voltage of a certain battery module increases, causing the battery module temperature and voltage detector to alarm, and there will be no thermal runaway of the battery module resulting in a fire, it is determined as "system warning". The safety monitoring center can use the method of electrical isolation to isolate the alarm battery module from the energy storage power station, and then incorporate it into the energy storage power station after detecting that the safety of the alarm battery module meets the requirements. In this way, the safety warning function of the energy storage power station is achieved.

[0136] The electrochemical energy storage power station fire protection system shown is in the "automatic control" state; when a fire breaks out in the protected area, the hydrogen and carbon monoxide smoke and temperature composite detectors in the energy storage power station send out fire signals. After being confirmed by the alarm controller and the safety monitoring center, the fire extinguishing controller emits audible and visual alarm signals, and at the same time issues linkage commands. The relevant equipment is linked. After a delay time of 0 - 30S, a fire extinguishing command is issued to open the solenoid valve to release the driving gas. The driving gas opens the corresponding selector valve and bottle valve through the starting pipeline, releases the fire extinguishing agent, and implements fire extinguishing. When the fire extinguishing agent is released, the external spray display light in the protected area lights up to play a warning role.

[0137] As shown, when the electrochemical energy storage power station fire protection system cannot be automatically controlled, the system adopts on-site "electrical manual control" and "mechanical emergency operation" control methods; that is, when a fire breaks out in the protected area, press the "start" button on the manual control box or controller or manually open the electromagnetic drive container valve. The fire extinguishing controller emits audible and visual alarm signals, and at the same time issues linkage commands. The relevant equipment is linked. After a delay time of 0 - 30S, a fire extinguishing command is issued to open the solenoid valve to release the starting gas. The starting gas opens the corresponding selector valve and bottle valve through the starting pipeline, releases the fire extinguishing agent, and implements fire extinguishing. When the fire extinguishing agent is released, the external spray display light in the protected area lights up to play a warning role.

[0138] The above is only a preferred embodiment of the present invention, and it is not a limitation of the present invention in any other form. Any modification or equivalent change made according to the technical essence of the present invention still belongs to the scope protected by the present invention.

Claims

1. A fire early warning detection and prevention system for an electrochemical energy storage power station, comprising a gas fire extinguishing agent storage bottle (1), a gas fire extinguishing agent storage bottle frame (2), a collecting pipe (3), a liquid flow check valve (4), a hose (5), an air flow check valve (6), a bottle head valve (7), a starting pipeline (8), a pressure signal device (9), a safety valve (10), a selection valve (11), a signal feedback circuit (12), a solenoid valve (13), a starting bottle (14), a starting bottle frame (15), a fire alarm controller (16), a control circuit (17), a manual control box (18), a light alarm (19), an acoustic alarm (20), a nozzle (21), an explosion-proof camera (22), a hydrogen, carbon monoxide smoke temperature sensing composite detector (23), a battery module patch temperature sensor (24), a battery module voltage transformer (25) and a gas fire extinguishing agent delivery pipeline (26), characterized in that: The electrochemical energy storage power station fire early warning detection and control system is provided with an explosion-proof camera (22) and a hydrogen, carbon monoxide smoke temperature sensing composite detector (23) on the top of the protection area according to the specification requirements; a battery module surface-mount temperature sensor (24) and a battery module voltage transformer (25) are provided on the battery pack equipment; a fire alarm controller (16), a manual control box (18), a light alarm (19) and a sound alarm (20) are installed outside the protection area; the manual control box (18), the light alarm (19), the sound alarm (20), the explosion-proof camera (22), the hydrogen, carbon monoxide smoke temperature sensing composite detector (23) are provided on the top of the protection area according to the specification requirements; a battery module surface-mount temperature sensor (24) and a battery module voltage transformer (25) are provided on the battery pack equipment; a fire alarm controller (16), a manual control box (18), a light alarm (19), and a sound alarm (20) are installed outside the protection area The carbon smoke temperature sensing composite detector (23), the battery module patch temperature sensor (24) and the battery module voltage transformer (25) are connected to the fire alarm controller (24) through a control line (17); the electrochemical energy storage power station fire early warning detection and prevention and control system places a gas fire extinguishing agent storage bottle (1) in the fire extinguishing agent storage room of the protection area, the gas fire extinguishing agent storage bottle (1) is installed and fixed through a gas fire extinguishing agent storage bottle frame (2), the top of the gas fire extinguishing agent storage bottle (1) is installed with a bottle head valve (7), the bottle head valve (7) is installed with a hose (5) and a starting pipe (8), the upper part of the hose (5) A liquid flow check valve (4) is installed and connected to the manifold (3); the start-up pipeline (8) is provided with a liquid flow check valve (4); a gas flow check valve (6) is installed on the manifold (3) between the bottle head valve (7) of the gas fire extinguishing agent storage bottle (1); the manifold (3) is connected to the gas fire extinguishing agent delivery pipeline (26); a pressure signal device (9), a safety valve (10) and a selection valve (11) are installed on the gas fire extinguishing agent delivery pipeline (26); the pressure signal device (9) and the selection valve (11) are connected to the fire alarm controller (16) via a signal feedback line (12) The gas fire extinguishing agent storage bottle (1) is externally mounted with a starter bottle (14) group, the starter bottle (14) group is fixed by a starter bottle frame (15), a solenoid valve (13) is mounted on the top of the starter bottle (14), the solenoid valve (13) is connected to a starter pipe (8), the starter pipe (8) is connected to a selector valve (11), and the solenoid valve (13) is connected to a fire alarm controller (16) via a control line (17); the gas fire extinguishing agent delivery pipe (26) is laid out at the top of the protection zone according to the specification, and a nozzle (21) is arranged on the gas fire extinguishing agent delivery pipe (26) as required.

2. According to claim 1, a fire early warning detection and prevention system for an electrochemical energy storage power station is characterized in that: The electrochemical energy storage power station fire early warning detection and prevention system uses an explosion-proof camera (22) to combine video image intelligent recognition technology with feedback signals from a hydrogen, carbon monoxide smoke temperature sensing composite detector (23), a battery module SMD temperature sensor (24) and a battery module voltage transformer (25) installed on site to determine the degree of danger.

3. The electrochemical energy storage power station fire early warning detection and prevention system according to claim 1, characterized in that: The fire warning detection and prevention system of the electrochemical energy storage power station also combines the big data of the long-term operation of the power station battery module and the artificial intelligence algorithm for comprehensive judgment when determining the degree of danger.

4. The electrochemical energy storage power station fire early warning detection and prevention system according to claim 1, characterized in that: The electrochemical energy storage power station fire early warning detection and prevention system is provided with "automatic control", "electrical manual control" and "mechanical emergency operation" control modes.

5. The prevention and control method of the electrochemical energy storage power station fire early warning detection and control system according to any one of claims 1 to 4, characterized in that: The specific steps are: Step 1: Data collection and preprocessing: using an explosion-proof camera installed in the protection area of ​​the electrochemical energy storage power station to collect video image data, and preprocessing the collected images; Step 2: Collect the dataset for YOLOv5 model training; divide the collected image dataset containing fire and non-fire scenes into training set, validation set and test set in a ratio of 7:2:1; Step 3: Build a YOLOv5 fire warning model; build a YOLOv5 fire warning model containing input layer, backbone, neck and output layer components for feature extraction and target detection; Step 4: Train the YOLOv5 fire warning model. Use the training set to train the YOLOv5 fire warning model, use Adam as the optimizer, update the model parameters to minimize the loss function, monitor the performance of the model through the validation set, and adjust the training parameters such as the learning rate according to the loss and accuracy on the validation set. Step 5: Fire early warning detection; The preprocessed video image sequence is input into the trained YOLOv5 fire warning model, and the model outputs the prediction results, each of which contains information such as the prediction box coordinates, size, confidence of the included target, and prediction category; Step 6: Determine the hazard level of the electrochemical energy storage power station; comprehensively determine the temperature and voltage data of the battery module of the safety monitoring center, the hydrogen, carbon monoxide smoke temperature sensing composite detector, and the fire-related target detection results in the video image to obtain the fire hazard level and perform processing.

6. The prevention and control method of the electrochemical energy storage power station fire early warning detection and control system according to claim 5, characterized in that: The process of data collection and preprocessing in step 1 is expressed as follows: Assume the acquired video image sequence is: I={I1,I2,…,I n } Where: I represents the acquired video image sequence; n is the number of video frames; The collected image is normalized and the image pixel values ​​are mapped to the [0,1] interval. The processing formula is: Where: I(x,y) is the pixel value of the original image at the (x,y) coordinate; I norm (x,y) is the normalized pixel value.

7. The prevention and control method of the electrochemical energy storage power station fire early warning detection and control system according to claim 5, characterized in that: The YOLOv5 fire warning model in step 3 is composed of four parts, namely the input layer, backbone, neck and output layer; it can be specifically expressed as: Input layer: The functions of the input layer are data enhancement, adaptive anchor box calculation, and adaptive image scaling; Backbone: Backbone is the main network used to extract information from the image; Neck: Neck is a FPN+PAN structure, which is used to improve the robustness of the model; Output layer: The output layer serves as the network output and makes predictions using previously extracted features; The loss function of YOLOv5 consists of positioning loss, confidence loss, and classification loss, which are specifically expressed as follows: Positioning loss: Where: L is the loss of YOLOv5; L loc is the positioning loss; L conf is the confidence loss; L class is the classification loss; is the adjusted positioning loss weight; is the adjusted confidence loss weight; is the adjusted classification loss weight and satisfies Confidence loss: Confidence loss L conf It is used to measure the difference between the confidence that the predicted box contains the target and the actual confidence. The formula is: Where: conf is the confidence loss weight; c i,j is the confidence that the j-th prediction box in the i-th image contains the target; is the confidence that the jth ground truth box in the i-th image contains the target; Classification loss: Classification loss L class Used to measure the difference between the predicted category and the true category. The formula is: Where: class is the classification loss weight; p i,j is the predicted category probability of the jth prediction box in the i-th image; is the predicted category probability of the jth ground-truth box in the i-th image.

8. The prevention and control method of the electrochemical energy storage power station fire early warning detection and control system according to claim 5, characterized in that: The fire warning detection in step 5 is specifically represented as follows: 1) The preprocessed video image sequence I norm Input the trained YOLOv5 model, and the model outputs the prediction result R = {R1, R2, …, R n }, where each R i Contains information such as the predicted box coordinates, size, confidence of the included object, and predicted category; 2) Screen and process the prediction results and set the confidence threshold τ conf , only keep the confidence greater than τ conf The prediction result, assuming the filtered result is R filtered = {R filtered1 ,R filtered2 ,...,R filteredn }, where m≤n; 3) Determine whether it is a fire-related target based on the predicted category, predict fire-related features, and set the fire-related target detection result as R fire = {R fire1 ,R fire2 ,...,R firek }, where k≤m.

9. The prevention and control method of the electrochemical energy storage power station fire early warning detection and control system according to claim 5, characterized in that: The specific process of fire hazard level determination and processing in step 6 is as follows: 1) Assume the battery module temperature threshold is T threshold , the voltage threshold is V threshold , the hydrogen threshold is H threshold , the carbon monoxide threshold is C threshold If the temperature of all battery modules is less than the threshold, the voltage is less than the threshold, the hydrogen is less than the threshold, the carbon monoxide is less than the threshold, and no fire-related targets are detected in the video image, it is determined to be "system safe" and the system is maintained in normal operation; 2) Fire-related targets are detected in the video image, and combined with the battery module temperature, voltage, hydrogen concentration, and carbon dioxide concentration data, it is determined that a fire may occur, and it is determined that "a fire has occurred in the system". The security monitoring center directly opens the selection valve and container valve in the area where the alarm battery module is located, and then releases the fire extinguishing agent until the fire is extinguished; 3) If fire-related targets are detected in the video image, but the battery module temperature, voltage, hydrogen concentration, and carbon dioxide concentration data show that thermal runaway of the battery module will not occur, resulting in a fire, it is determined to be a "system warning". The safety monitoring center uses electrical isolation to isolate the alarming battery module from the energy storage power station, and only incorporates the alarming battery module into the energy storage power station after the safety of the alarming battery module meets the requirements.

10. The prevention and control method of the electrochemical energy storage power station fire early warning detection and control system according to claim 7, characterized in that: The positioning loss L loc It is used to measure the position difference between the predicted box and the real box. The formula is: Where: N is the number of images; M is the number of prediction boxes in each image; λ loc is the positioning loss weight; (x i,j ,y i,j ,w i,j ,h i,j ) are the coordinates and size of the jth prediction box in the ith image, x i,j and i,j is the coordinate, w i,j and h i,j is the width and height; are the coordinates and size of the jth ground truth box in the ith image, and are coordinates, and is the width and height.

Images