Power plant safety monitoring system based on artificial intelligence

Abstract

The invention belongs to the technical field of power plant safety monitoring, and discloses a power plant safety monitoring system based on artificial intelligence. Comprising the following modules: a data acquisition module used for acquiring environment data and equipment data of all monitoring areas in a power plant in real time; according to the invention, through setting the fire occurrence remaining time prediction model and the support score, the fire prediction is promoted from simple probability evaluation to accurate time prediction, and the time interval from the current moment to the fire in the unit area is accurately predicted. The prediction of the time dimension enables the implementation of prevention and control measures to be more targeted, on this basis, a support score is established, state parameters such as the battery electric quantity, the fire extinguishing agent remaining amount, the historical reliability and the arrival time of the robot are comprehensively considered, the optimal scheduling of the fire extinguishing robot is realized, and the intelligent scheduling method improves the utilization rate of the robot and reduces the labor intensity of the robot. And meanwhile, prevention and control deployment is completed before a fire disaster occurs.

Description

Technical Field

[0001] This invention relates to the field of power plant safety monitoring technology, and more specifically, to an artificial intelligence-based power plant safety monitoring system. Background Technology

[0002] As a core component of the national energy infrastructure, the safe operation of power plants is related to social stability and economic development. With the expansion of power plant scale and the increasing complexity of equipment, traditional safety monitoring systems can no longer meet the safety management needs of modern power plants.

[0003] Existing technologies lack the ability to quantitatively assess and predict fire risks. Most systems can only provide alarms after the fact and cannot provide effective early warnings before a fire occurs. By the time open flames or high temperatures are detected, the fire has often already grown to a considerable size, missing the best opportunity for control. In terms of emergency resource scheduling, existing systems are seriously inadequate. Traditional fire-fighting robot scheduling often adopts fixed plans or simple shortest path principles, lacking comprehensive consideration of robot status and environmental conditions. When multiple areas issue warnings simultaneously, the system cannot optimize scheduling based on robot status parameters such as battery power, remaining extinguishing agent, and reliability. In addition, existing scheduling systems often ignore time constraints, which may result in robots arriving at the scene only after the fire has occurred, completely negating the significance of early warning and prevention. Summary of the Invention

[0004] To address the problems in the background art, this invention proposes an artificial intelligence-based power plant safety monitoring system.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a power plant safety monitoring system based on artificial intelligence, comprising the following modules: The data acquisition module is used to collect environmental and equipment data from all monitored areas within the power plant in real time. The risk assessment module is used to generate environmental risk scores and equipment risk scores for all monitored areas based on the collected environmental and equipment data, generate a total risk score based on the environmental and equipment risk scores, and mark all monitored areas as risk areas of different levels based on the total risk score. The fire probability prediction module is used to acquire historical fire probability data and build a fire probability prediction model, and use the fire probability prediction model to predict the fire probability in high-risk areas. The fire remaining time prediction module is used to acquire historical fire remaining time data and build a fire remaining time prediction model. The evaluation and scheduling module is used to predict the remaining time of fire in a unit area using a fire remaining time prediction model, and to obtain the available fire-fighting robots that meet the remaining time of fire. It establishes a support score to evaluate all available fire-fighting robots, and schedules an appropriate number of fire-fighting robots to provide support based on the support score.

[0006] Furthermore, the area within the power plant is divided into multiple monitoring zones of equal size, with monitoring points set up within each zone. These monitoring points obtain environmental and equipment data through monitoring equipment. The environmental data includes the concentration of combustible / toxic gases, smoke concentration, ambient temperature, ambient humidity, and dust concentration within the monitoring area; The equipment data includes the vibration spectrum, abnormal acoustic signature, surface temperature, and current anomaly of the equipment within the monitoring area. The concentration of flammable / toxic gases is obtained through gas sensors; Smoke concentration is obtained through a smoke sensor; Ambient temperature and humidity are obtained through temperature and humidity sensors; Dust concentration is obtained through a dust sensor; The vibration spectrum is obtained through a vibration sensor; Abnormal voiceprints are obtained through acoustic sensors, and abnormal noise is identified through FFT analysis; Surface temperature was obtained using an infrared thermal imager; The current anomaly is obtained through a current sensor, and the deviation of the current from the rated value is calculated.

[0007] Further processing of environmental and equipment data is performed: Normalize the concentrations of flammable / toxic gases:

[0008] In the formula, To normalize the concentration of flammable / toxic gases, This represents the measured concentration of flammable / toxic gases. and These are the lower and upper limits for the concentration of flammable / toxic gases; Normalize the smoke concentration:

[0009] In the formula, To normalize the smoke concentration, This represents the measured smoke concentration. and These are the lower and upper limits of smoke concentration; The ambient temperature was normalized.

[0010] In the formula, To normalize the ambient temperature, The actual ambient temperature. and These are the lower and upper limits of the ambient temperature; Normalize the ambient humidity:

[0011] In the formula, To normalize ambient humidity, The measured ambient humidity and These are the lower and upper limits of ambient humidity. The dust concentration was normalized.

[0012] In the formula, To normalize dust concentration, This represents the measured dust concentration. and These are the lower and upper limits of dust concentration; The vibration spectrum was normalized.

[0013] In the formula, For the normalized vibration spectrum, This is the measured vibration spectrum. and These represent the lower and upper limits of the vibration spectrum; Abnormal voiceprints are normalized:

[0014] In the formula, To normalize abnormal voiceprints, These are abnormal voiceprints measured in practice. and These are the lower and upper limits for abnormal voiceprints; The surface temperature was normalized.

[0015] In the formula, To normalize the surface temperature, This is the measured surface temperature. and These are the lower and upper limits of the surface temperature; Normalize the current anomaly:

[0016] In the formula, To normalize the current anomaly degree, This represents the measured current anomaly. and These represent the lower and upper limits of the current anomaly.

[0017] Furthermore, the process of generating environmental risk scores and equipment risk scores for the entire monitoring area based on the collected environmental and equipment data includes: Environmental Risk Score:

[0018] In the formula, , , , and These are weighting coefficients, obtained through training based on historical data; Equipment risk score P:

[0019] In the formula, , , and These are weighting coefficients, obtained through training based on historical data.

[0020] Furthermore, the process of generating a total risk score based on the environmental risk score and the equipment risk score, and then marking all monitored areas as risk zones of different levels based on the total risk score, includes: Overall risk score R:

[0021] In the formula and These are weighting coefficients, obtained through training based on historical data; An appropriate scoring threshold is set based on historical risk scoring data, which refers to the set of data on the total risk score of the monitored area in the past. The total risk score of the monitored area is compared with the scoring threshold, and the monitored area is marked as a risk area of ​​different levels. When R≤X, the monitoring area is determined to be a low-risk area; When X≤R≤Y, the monitoring area is determined to be a medium-risk area; When R ≥ Y, the monitoring area is determined to be a high-risk area.

[0022] Furthermore, the process of acquiring historical fire occurrence probability data and constructing a fire occurrence probability prediction model, and then using this model to predict the fire occurrence probability in high-risk areas, includes: The probability of fire occurrence refers to the probability that a fire may occur within the monitored area; Factors influencing the probability of fires in high-risk areas include: environmental risk score of the monitored area, equipment risk score, historical fire frequency, equipment aging index, and personnel activity density. Historical fire frequency refers to the number of fires that have occurred in the monitored area in the past, and is derived from a historical database. The equipment aging index is calculated by dividing the equipment's service life by its lifespan, and is derived from equipment records. Personnel activity density is obtained by calculating the number of people per unit time. Set a monitoring period and obtain historical fire occurrence probability data for a single monitoring area in different monitoring periods. The historical fire occurrence probability data includes the average environmental risk score, equipment risk score, historical fire frequency, equipment aging index, personnel activity density, and historical fire occurrence probability of the single monitoring area in different monitoring periods. Based on the environmental risk score, equipment risk score, historical fire frequency, equipment aging index, personnel activity density and corresponding historical fire probability of the monitoring area in different historical fire probability data, a fire probability prediction set is generated and divided into the first training set and the first test set. Construct a first convolutional neural network, using environmental risk score, equipment risk score, historical fire frequency, equipment aging index and personnel activity density from different historical fire occurrence probability data in the first training set as the input data of the first convolutional neural network, and using the corresponding historical fire occurrence probability in the first training set as the output data of the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network, whose output is less than or equal to the preset first test error threshold, is used as the fire occurrence probability prediction model. At each monitoring cycle, the average environmental risk score, equipment risk score, historical fire frequency, equipment aging index and personnel activity density of each monitoring area within the monitoring cycle are input into the fire probability prediction model to obtain the predicted fire probability of each monitoring area. An appropriate probability threshold is set based on historical probability data, which refers to the data set of past fire occurrence probabilities in the monitored area. The predicted fire occurrence probability of the monitored area is compared with the probability threshold. When the predicted fire occurrence probability is greater than the probability threshold, the monitored area is marked as a high-risk fire area. Patrol personnel and staff need to be reminded to evacuate quickly and relevant departments should be notified in a timely manner. Evacuation routes for patrol personnel and staff to safety exits should be set up, and the evacuation routes should be far away from the high-risk fire area.

[0023] Furthermore, the process of acquiring historical fire time-remaining data and constructing a fire time-remaining prediction model includes: The fire-prone area was evenly divided into multiple unit areas of equal size. The remaining time for the fire refers to the time from this moment until the fire starts within the unit area; Factors affecting the remaining time of a fire in a unit area include the concentration of combustible gas, the rate of temperature change, the aging index of equipment insulation, the degree of material pyrolysis, and the oxygen consumption rate. The concentration of combustible gas is measured in real time by a combustible gas sensor; The rate of temperature change refers to the change in temperature per unit time. It is calculated by measuring temperature data through a temperature sensor and based on historical data. The equipment insulation aging index is derived from historical maintenance data; The degree of material pyrolysis is obtained by detecting volatile organic compounds using a volatile organic compound sensor, analyzing material color changes using multispectral imaging, and combining both results. The oxygen consumption rate was obtained through time-series analysis of oxygen sensor data. Acquire historical fire time remaining data for a single unit area, including combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree, oxygen consumption rate, and historical fire time remaining for that single unit area. Based on the combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree, oxygen consumption rate and corresponding historical fire remaining time data of the corresponding unit area, a fire remaining time prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree and oxygen consumption rate in the remaining time data of different historical fires in the second training set are used as the input data of the second convolutional neural network, and the remaining time of the corresponding historical fires in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network whose output is less than or equal to the preset second test error threshold is used as the fire remaining time prediction model.

[0024] Furthermore, the process of predicting the remaining fire time in a unit area using a fire remaining time prediction model and obtaining a fire-fighting robot that matches the remaining fire time includes: The fire-fighting robot mentioned above refers to a fire-fighting robot that can reach the area of ​​the unit to be supported within the remaining time of a fire. The unit area awaiting support refers to the unit area where a fire is about to occur; When it is necessary to predict the remaining time of fire in each unit area, the real-time combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree and oxygen consumption rate of each unit area are input into the remaining time of fire prediction model to obtain the predicted remaining time of fire in each unit area. All firefighting robots in the power plant were analyzed. The arrival time of the firefighting robot to the area to be supported was obtained by dividing the walking distance of the firefighting robot from the walking speed of the firefighting robot to the area to be supported. The arrival time was compared with the predicted remaining time of the fire. Firefighting robots whose arrival time was less than the predicted remaining time of the fire were marked as firefighting robots that could be supported, and firefighting robots whose arrival time was greater than the predicted remaining time of the fire were marked as firefighting robots that could not be supported.

[0025] Furthermore, the process of establishing a support rating system to evaluate all available firefighting robots includes: Collect support parameters that can support the firefighting robot, including the firefighting robot's battery level, remaining extinguishing agent, historical failure index, and arrival time; The battery power of the firefighting robot can be obtained directly from its battery management system. The remaining amount of extinguishing agent in the fire-fighting robot is obtained from the agent storage sensor of the fire-fighting robot; The historical failure index is calculated based on the robot's maintenance records and failure history. Processing support parameters: Normalize the remaining battery capacity:

[0026] In the formula, To normalize the battery capacity, This represents the actual remaining battery capacity. Maximum battery capacity; The remaining extinguishing agent was normalized:

[0027] In the formula, This is the normalized residual amount of extinguishing agent. This represents the measured residual amount of extinguishing agent. This is the maximum amount of extinguishing agent; The historical failure index was normalized:

[0028] In the formula, To normalize the historical failure index, This represents the actual number of failures within a year. The maximum number of failures allowed within one year; The arrival time is normalized:

[0029] In the formula, To normalize arrival time, This is the actual arrival time. The maximum arrival time is the estimated remaining time before the fire occurs. Support rating E is:

[0030] In the formula, , , and These are weighting coefficients, obtained through training based on historical data.

[0031] Furthermore, the process of dispatching an appropriate number of firefighting robots for support based on the support score includes: A fire is about to break out in a unit area within a high-risk fire zone. The remaining time of the fire in the unit area is predicted using a fire time prediction model. The number of fire-fighting robots required is determined based on the predicted remaining time. The longer the predicted remaining time, the fewer fire-fighting robots are needed. The shorter the predicted remaining time, the more intense the fire and the more fire-fighting robots are needed. Based on the remaining time before the fire occurs, all firefighting robots in the power plant are marked as either supportable or non-supportable. All supportable firefighting robots are evaluated using a support score and ranked from highest to lowest. Based on the predicted remaining time before the fire occurs, an appropriate number of supportable firefighting robots are selected and marked as rapid support firefighting robots. These rapid support firefighting robots are dispatched to the unit area to carry out fire prevention or firefighting work. The unselected supportable and non-supportable firefighting robots serve as backup firefighting robots, which will take over the role of rapid support firefighting robots after they become unable to continue working, or control the spread of fire outside the unit area.

[0032] The technical effects and advantages of the artificial intelligence-based power plant safety monitoring system of the present invention are as follows: (1) By setting up a fire time prediction model and support score, fire prediction is advanced from simple probability assessment to accurate time prediction. By using key parameters such as combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree and oxygen consumption rate, the time interval from the current moment to the ignition of the unit area is accurately predicted. Compared with the traditional fire early warning system, this time-dimensional prediction makes the implementation of prevention and control measures more targeted. On this basis, a support score is also established, which comprehensively considers the robot's battery power, extinguishing agent balance, historical reliability and arrival time and other status parameters to achieve optimal scheduling of fire-fighting robots. This intelligent scheduling method improves the utilization rate of robots and ensures that prevention and control deployment is completed before the fire occurs.

[0033] (2) By setting environmental risk scores and equipment risk scores, the multi-source data fusion method improves the comprehensiveness of risk assessment compared with traditional single-parameter monitoring. By setting a total risk score and weighted fusion of environmental risk scores and equipment risk scores, the risk level of the monitoring area is accurately classified. This graded early warning mechanism enables safety management personnel to quickly identify high-risk areas, prioritize the allocation of protective resources, and shorten the response time for accident prevention. By setting a fire probability prediction model, the probability of fire in the monitoring area can be predicted, which can play a role in real-time prevention. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0036] Reference Figure 1 An artificial intelligence-based power plant safety monitoring system includes the following modules: The data acquisition module is used to collect environmental and equipment data from all monitored areas within the power plant in real time. The risk assessment module is used to generate environmental risk scores and equipment risk scores for all monitored areas based on the collected environmental and equipment data, generate a total risk score based on the environmental and equipment risk scores, and mark all monitored areas as risk areas of different levels based on the total risk score. The fire probability prediction module is used to acquire historical fire probability data and build a fire probability prediction model, and use the fire probability prediction model to predict the fire probability in high-risk areas. The fire remaining time prediction module is used to acquire historical fire remaining time data and build a fire remaining time prediction model. The evaluation and scheduling module is used to predict the remaining time of fire in a unit area using a fire remaining time prediction model, and to obtain the available fire-fighting robots that meet the remaining time of fire. It establishes a support score to evaluate all available fire-fighting robots, and schedules an appropriate number of fire-fighting robots to provide support based on the support score.

[0037] It should be further explained that, in the specific implementation process, the area within the power plant is divided into multiple monitoring areas of equal size. Monitoring points are set up in the monitoring areas, and the monitoring points obtain environmental data and equipment data through monitoring equipment. The environmental data includes the concentration of combustible / toxic gases, smoke concentration, ambient temperature, ambient humidity, and dust concentration within the monitoring area; The equipment data includes the vibration spectrum, abnormal acoustic signature, surface temperature, and current anomaly of the equipment within the monitoring area. The concentration of flammable / toxic gases is obtained through gas sensors, reflecting the risk of explosion or poisoning; Smoke concentration is obtained through smoke sensors, reflecting early signs of a fire. Ambient temperature and humidity are obtained through temperature and humidity sensors. High temperatures may cause equipment to overheat, and high humidity may cause electrical short circuits. Dust concentration is obtained through dust sensors; dust accumulation may cause explosions or fires. The vibration spectrum is obtained through vibration sensors, reflecting mechanical faults in the equipment; Abnormal sound signatures are obtained through acoustic sensors, and abnormal noise is identified through FFT analysis, reflecting internal equipment faults. Surface temperature is obtained using an infrared thermal imager and reflects equipment overheating. Current anomaly is obtained through a current sensor, and the deviation between the current and the rated value is calculated to reflect electrical faults.

[0038] It should be further explained that environmental and equipment data are processed during the specific implementation process: Normalize the concentrations of flammable / toxic gases:

[0039] In the formula, To normalize the concentration of flammable / toxic gases, This represents the measured concentration of flammable / toxic gases. and These are the lower and upper limits for the concentration of flammable / toxic gases, specifically 0 and 1000 respectively. Normalize the smoke concentration:

[0040] In the formula, To normalize the smoke concentration, This represents the measured smoke concentration. and These represent the lower and upper limits of smoke concentration, specifically 0 and 1000 respectively. The ambient temperature was normalized.

[0041] In the formula, To normalize the ambient temperature, The actual ambient temperature. and These represent the lower and upper limits of the ambient temperature, specifically 0 and 100, respectively. Normalize the ambient humidity:

[0042] In the formula, To normalize ambient humidity, The measured ambient humidity and These represent the lower and upper limits of ambient humidity, specifically 0 and 100% respectively. The dust concentration was normalized.

[0043] In the formula, To normalize dust concentration, This represents the measured dust concentration. and These represent the lower and upper limits of dust concentration, specifically 0 and 500, respectively. The vibration spectrum was normalized.

[0044] In the formula, For the normalized vibration spectrum, This is the measured vibration spectrum. and These represent the lower and upper limits of the vibration spectrum, specifically 0 and 10, respectively. Abnormal voiceprints are normalized:

[0045] In the formula, To normalize abnormal voiceprints, These are abnormal voiceprints measured in practice. and These are the lower and upper limits for abnormal voiceprints, specifically 0 and 120 respectively; The surface temperature was normalized.

[0046] In the formula, To normalize the surface temperature, This is the measured surface temperature. and These are the lower and upper limits of the surface temperature, specifically 0 and 200, respectively; Normalize the current anomaly:

[0047] In the formula, To normalize the current anomaly degree, This represents the measured current anomaly. and These represent the lower and upper limits of the current anomaly, specifically 0 and 50 respectively.

[0048] It should be further explained that, in the specific implementation process, the process of generating environmental risk scores and equipment risk scores for the entire monitoring area based on the collected environmental and equipment data includes: Environmental Risk Score:

[0049] In the formula, , , , and These are weighting coefficients, obtained from training on historical data, and set to 0.3, 0.2, 0.2, 0.15, and 0.15 respectively. If the environmental data for a certain monitoring area is: It is 500. It is 500. It is 50. It is 10. If the value is 300, then the environmental risk score S for the monitoring area is 0.575. Equipment risk score P:

[0050] In the formula, , , and These are weighting coefficients, derived from historical data training, and set to 0.3, 0.3, 0.2, and 0.2 respectively. If the equipment data for a certain monitoring area is: F = 2, V = 30, t = 100, I = 10, then the equipment risk score P for the equipment in that monitoring area is 0.275.

[0051] It should be further explained that, in the specific implementation process, the process of generating a total risk score based on environmental risk scores and equipment risk scores, and then marking all monitored areas as risk zones of different levels based on the total risk score, includes: Overall risk score R:

[0052] In the formula and These are weighting coefficients, derived from historical data training, and set to 0.6 and 0.4 respectively; If the environmental risk score S of a certain monitoring area is 0.575 and the equipment risk score P is 0.275, then the total risk score R of the monitoring area is 0.455. An appropriate scoring threshold is set based on historical risk scoring data, which refers to the set of data on the total risk score of the monitored area in the past. The total risk score of the monitored area is compared with the scoring threshold, and the monitored area is marked as a risk area of ​​different levels. When R≤X, the monitoring area is determined to be a low-risk area; When X≤R≤Y, the monitoring area is determined to be a medium-risk area; When R ≥ Y, the monitoring area is determined to be a high-risk area; If X is 0.4 and Y is 0.6, then the monitoring area with a total risk score R of 0.455 is identified as a medium-risk area.

[0053] It should be further explained that, in the specific implementation process, the process of obtaining historical fire occurrence probability data and constructing a fire occurrence probability prediction model, and using the fire occurrence probability prediction model to predict the fire occurrence probability in high-risk areas, includes: The probability of fire occurrence refers to the probability that a fire may occur within the monitored area; Factors influencing the probability of fires in high-risk areas include: environmental risk score of the monitored area, equipment risk score, historical fire frequency, equipment aging index, and personnel activity density. Historical fire frequency refers to the number of fires that have occurred in the monitored area in the past, and is derived from a historical database. The equipment aging index is calculated by dividing the equipment's service life by its lifespan, and is derived from equipment records. Personnel activity density is obtained by calculating the number of people per unit time. Set a monitoring period and obtain historical fire occurrence probability data for a single monitoring area in different monitoring periods. The historical fire occurrence probability data includes the average environmental risk score, equipment risk score, historical fire frequency, equipment aging index, personnel activity density, and historical fire occurrence probability of the single monitoring area in different monitoring periods. Based on the environmental risk score, equipment risk score, historical fire frequency, equipment aging index, personnel activity density and corresponding historical fire probability of the monitoring area in different historical fire probability data, a fire probability prediction set is generated and divided into the first training set and the first test set. Construct a first convolutional neural network, using environmental risk score, equipment risk score, historical fire frequency, equipment aging index and personnel activity density from different historical fire occurrence probability data in the first training set as the input data of the first convolutional neural network, and using the corresponding historical fire occurrence probability in the first training set as the output data of the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network, whose output is less than or equal to the preset first test error threshold, is used as the fire occurrence probability prediction model. At each monitoring cycle, the average environmental risk score, equipment risk score, historical fire frequency, equipment aging index and personnel activity density of each monitoring area within the monitoring cycle are input into the fire probability prediction model to obtain the predicted fire probability of each monitoring area. An appropriate probability threshold is set based on historical probability data, which refers to the data set of past fire occurrence probabilities in the monitored area. The predicted fire occurrence probability of the monitored area is compared with the probability threshold. When the predicted fire occurrence probability is greater than the probability threshold, the monitored area is marked as a high-risk fire area. Patrol personnel and staff need to be reminded to evacuate quickly and relevant departments should be notified in a timely manner. Evacuation routes for patrol personnel and staff to safety exits should be set up, and the evacuation routes should be far away from high-risk fire areas. In an embodiment of the present invention, the predicted fire probability of all monitored areas is obtained through a fire probability prediction model. The predicted fire probability is related to environmental risk score, equipment risk score, historical fire frequency, equipment aging index and personnel activity density. The level of the environmental risk score directly affects the probability of a fire. The higher the environmental risk score, the better the ignition conditions provided by the environment, and the greater the probability of a fire. Therefore, the environmental risk score is positively correlated with the probability of a fire. The level of equipment risk score directly affects the predicted probability of fire occurrence. The higher the equipment risk score, the better the ignition conditions provided by the equipment, and the greater the predicted probability of fire occurrence. Therefore, equipment risk score and fire occurrence probability are positively correlated. The frequency of historical fires directly affects the probability of a predicted fire. The higher the frequency of historical fires, the easier it is for a fire to start. Therefore, the frequency of historical fires is positively correlated with the probability of a fire. The level of the equipment aging index directly affects the predicted probability of fire occurrence. The higher the equipment aging index, the worse the condition of the equipment, and the greater the predicted probability of fire occurrence. Therefore, the equipment aging index is positively correlated with the probability of fire occurrence. The density of human activity directly affects the probability of a fire. The higher the density of human activity, the greater the risk of fire due to human error, and the higher the probability of a fire. Therefore, the density of human activity is positively correlated with the probability of a fire.

[0054] It should be further explained that, in the specific implementation process, the process of obtaining historical fire time-remaining data and constructing a fire time-remaining prediction model includes: The fire-prone area was evenly divided into multiple unit areas of equal size. The remaining time for the fire refers to the time from this moment until the fire starts within the unit area; Factors affecting the remaining time of a fire in a unit area include the concentration of combustible gas, the rate of temperature change, the aging index of equipment insulation, the degree of material pyrolysis, and the oxygen consumption rate. The concentration of combustible gas is measured in real time by a combustible gas sensor; The rate of temperature change refers to the change in temperature per unit time. It is calculated by measuring temperature data through a temperature sensor and based on historical data. The equipment insulation aging index is derived from historical maintenance data; The degree of material pyrolysis is obtained by detecting volatile organic compounds using a volatile organic compound sensor, analyzing material color changes using multispectral imaging, and combining both results. The oxygen consumption rate was obtained through time-series analysis of oxygen sensor data. Acquire historical fire time remaining data for a single unit area, including combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree, oxygen consumption rate, and historical fire time remaining for that single unit area. Based on the combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree, oxygen consumption rate and corresponding historical fire remaining time data of the corresponding unit area, a fire remaining time prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree and oxygen consumption rate in the remaining time data of different historical fires in the second training set are used as the input data of the second convolutional neural network, and the remaining time of the corresponding historical fires in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network whose output is less than or equal to the preset second test error threshold is used as the fire remaining time prediction model.

[0055] It should be further explained that, in the specific implementation process, the process of predicting the remaining time of fire in a unit area using a fire remaining time prediction model and obtaining a fire-fighting robot that matches the remaining time of fire includes: The fire-fighting robot mentioned above refers to a fire-fighting robot that can reach the area of ​​the unit to be supported within the remaining time of a fire. The unit area awaiting support refers to the unit area where a fire is about to occur; When it is necessary to predict the remaining time of fire in each unit area, the real-time combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree and oxygen consumption rate of each unit area are input into the remaining time of fire prediction model to obtain the predicted remaining time of fire in each unit area. All firefighting robots in the power plant were analyzed. The arrival time of the firefighting robot to the area to be supported was obtained by dividing the walking distance of the firefighting robot from the walking speed of the firefighting robot to the area to be supported. The arrival time was compared with the predicted remaining time of the fire. Firefighting robots whose arrival time was less than the predicted remaining time of the fire were marked as firefighting robots that could be supported, and firefighting robots whose arrival time was greater than the predicted remaining time of the fire were marked as firefighting robots that could not be supported.

[0056] It should be further explained that, in the specific implementation process, the process of establishing a support rating system to evaluate all available firefighting robots includes: Collect support parameters that can support the firefighting robot, including the firefighting robot's battery level, remaining extinguishing agent, historical failure index, and arrival time; The battery power of the firefighting robot can be obtained directly from its battery management system. The remaining amount of extinguishing agent in the fire-fighting robot is obtained from the agent storage sensor of the fire-fighting robot; The historical failure index is calculated based on the robot's maintenance records and failure history. Processing support parameters: Normalize the remaining battery capacity:

[0057] In the formula, To normalize the battery capacity, This represents the actual remaining battery capacity. This is the maximum battery capacity, specifically 100. The remaining extinguishing agent was normalized:

[0058] In the formula, This is the normalized residual amount of extinguishing agent. This represents the measured residual amount of extinguishing agent. This is the maximum amount of extinguishing agent, specifically 100. The historical failure index was normalized:

[0059] In the formula, To normalize the historical failure index, This represents the actual number of failures within a year. The maximum number of allowed failures within one year is 10. The arrival time is normalized:

[0060] In the formula, To normalize arrival time, This is the actual arrival time. The maximum arrival time, which is the estimated remaining time before the fire occurs, is set to 10 minutes. Support rating E is:

[0061] In the formula, , , and These are weighting coefficients, derived from training on historical data, and set to 0.2, 0.3, 0.2, and 0.3 respectively. If a certain support parameter for a firefighting robot is: It is 80. It is 80. It is 2. If the score is 2, then the support rating E for the firefighting robot is 0.8.

[0062] It should be further explained that, in the specific implementation process, the process of dispatching an appropriate number of firefighting robots for support based on the support score includes: A fire is about to break out in a unit area within a high-risk fire zone. The remaining time of the fire in the unit area is predicted using a fire time prediction model. The number of fire-fighting robots required is determined based on the predicted remaining time. The longer the predicted remaining time, the fewer fire-fighting robots are needed. The shorter the predicted remaining time, the more intense the fire and the more fire-fighting robots are needed. Based on the remaining time before the fire occurs, all firefighting robots in the power plant are marked as either supportable or non-supportable. All supportable firefighting robots are evaluated using a support score and ranked from highest to lowest. Based on the predicted remaining time before the fire occurs, an appropriate number of supportable firefighting robots are selected and marked as rapid support firefighting robots. These rapid support firefighting robots are dispatched to the unit area to carry out fire prevention or firefighting work. The unselected supportable and non-supportable firefighting robots serve as backup firefighting robots, which will take over the role of rapid support firefighting robots after they become unable to continue working, or control the spread of fire outside the unit area.

[0063] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0064] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power plant safety monitoring system based on artificial intelligence, characterized in that, Includes the following modules: The data acquisition module is used to collect environmental and equipment data from all monitored areas within the power plant in real time. The risk assessment module is used to generate environmental risk scores and equipment risk scores for all monitored areas based on the collected environmental and equipment data, generate a total risk score based on the environmental and equipment risk scores, and mark all monitored areas as risk areas of different levels based on the total risk score. The fire probability prediction module is used to acquire historical fire probability data and build a fire probability prediction model, and use the fire probability prediction model to predict the fire probability in high-risk areas. The fire remaining time prediction module is used to acquire historical fire remaining time data and build a fire remaining time prediction model. The evaluation and scheduling module is used to predict the remaining time of fire in a unit area using a fire remaining time prediction model, and to obtain the available fire-fighting robots that meet the remaining time of fire. It establishes a support score to evaluate all available fire-fighting robots, and schedules an appropriate number of fire-fighting robots to provide support based on the support score.

2. The power plant safety monitoring system based on artificial intelligence according to claim 1, characterized in that, The area within the power plant is divided into multiple monitoring zones of equal size. Monitoring points are set up within each monitoring zone, and environmental and equipment data are obtained from the monitoring points through monitoring equipment. The environmental data includes the concentration of combustible / toxic gases, smoke concentration, ambient temperature, ambient humidity, and dust concentration within the monitoring area; The equipment data includes the vibration spectrum, abnormal acoustic signature, surface temperature, and current anomaly of the equipment within the monitoring area. The concentration of flammable / toxic gases is obtained through gas sensors; Smoke concentration is obtained through a smoke sensor; Ambient temperature and humidity are obtained through temperature and humidity sensors; Dust concentration is obtained through a dust sensor; The vibration spectrum is obtained through a vibration sensor; Abnormal voiceprints are obtained through acoustic sensors, and abnormal noise is identified through FFT analysis; Surface temperature was obtained using an infrared thermal imager; The current anomaly is obtained through a current sensor, and the deviation of the current from the rated value is calculated.

3. The power plant safety monitoring system based on artificial intelligence according to claim 2, characterized in that, Processing environmental and equipment data: Normalize the concentrations of flammable / toxic gases: In the formula, To normalize the concentration of flammable / toxic gases, This represents the measured concentration of flammable / toxic gases. and These are the lower and upper limits for the concentration of flammable / toxic gases; Normalize the smoke concentration: In the formula, To normalize the smoke concentration, This represents the measured smoke concentration. and These are the lower and upper limits of smoke concentration; The ambient temperature was normalized. In the formula, To normalize the ambient temperature, The actual ambient temperature. and These are the lower and upper limits of the ambient temperature; Normalize the ambient humidity: In the formula, To normalize ambient humidity, The measured ambient humidity and These are the lower and upper limits of ambient humidity. The dust concentration was normalized. In the formula, To normalize dust concentration, This represents the measured dust concentration. and These are the lower and upper limits of dust concentration; The vibration spectrum was normalized. In the formula, For the normalized vibration spectrum, This is the measured vibration spectrum. and These represent the lower and upper limits of the vibration spectrum; Abnormal voiceprints are normalized: In the formula, To normalize abnormal voiceprints, These are abnormal voiceprints measured in practice. and These are the lower and upper limits for abnormal voiceprints; The surface temperature was normalized. In the formula, To normalize the surface temperature, This is the measured surface temperature. and These are the lower and upper limits of the surface temperature; Normalize the current anomaly: In the formula, To normalize the current anomaly degree, This represents the measured current anomaly. and These represent the lower and upper limits of the current anomaly.

4. The power plant safety monitoring system based on artificial intelligence according to claim 3, characterized in that, The process of generating environmental risk scores and equipment risk scores for the entire monitoring area based on the collected environmental and equipment data includes: Environmental Risk Score: In the formula, , , , and These are weighting coefficients, obtained through training based on historical data; Equipment risk score P: In the formula, , , and These are weighting coefficients, obtained through training based on historical data.

5. The power plant safety monitoring system based on artificial intelligence according to claim 4, characterized in that, The process of generating a total risk score based on environmental risk scores and equipment risk scores, and then marking all monitored areas as risk zones of different levels based on the total risk score, includes: Overall risk score R: In the formula and These are weighting coefficients, obtained through training based on historical data; An appropriate scoring threshold is set based on historical risk scoring data, which refers to the set of data on the total risk score of the monitored area in the past. The total risk score of the monitored area is compared with the scoring threshold, and the monitored area is marked as a risk area of ​​different levels. When R≤X, the monitoring area is determined to be a low-risk area; When X≤R≤Y, the monitoring area is determined to be a medium-risk area; When R ≥ Y, the monitoring area is determined to be a high-risk area.

6. The power plant safety monitoring system based on artificial intelligence according to claim 5, characterized in that, The process of acquiring historical fire probability data and constructing a fire probability prediction model, and then using that model to predict the probability of fires in high-risk areas, includes: The probability of fire occurrence refers to the probability that a fire may occur within the monitored area; Factors influencing the probability of fires in high-risk areas include: environmental risk score of the monitored area, equipment risk score, historical fire frequency, equipment aging index, and personnel activity density. Historical fire frequency refers to the number of fires that have occurred in the monitored area in the past, and is derived from a historical database. The equipment aging index is calculated by dividing the equipment's service life by its lifespan, and is derived from equipment records. Personnel activity density is obtained by calculating the number of people per unit time. Set a monitoring period and obtain historical fire occurrence probability data for a single monitoring area in different monitoring periods. The historical fire occurrence probability data includes the average environmental risk score, equipment risk score, historical fire frequency, equipment aging index, personnel activity density, and historical fire occurrence probability of the single monitoring area in different monitoring periods. Based on the environmental risk score, equipment risk score, historical fire frequency, equipment aging index, personnel activity density and corresponding historical fire probability of the monitoring area in different historical fire probability data, a fire probability prediction set is generated and divided into the first training set and the first test set. Construct a first convolutional neural network, using environmental risk score, equipment risk score, historical fire frequency, equipment aging index and personnel activity density from different historical fire occurrence probability data in the first training set as the input data of the first convolutional neural network, and using the corresponding historical fire occurrence probability in the first training set as the output data of the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network, whose output is less than or equal to the preset first test error threshold, is used as the fire occurrence probability prediction model. At each monitoring cycle, the average environmental risk score, equipment risk score, historical fire frequency, equipment aging index and personnel activity density of each monitoring area within the monitoring cycle are input into the fire probability prediction model to obtain the predicted fire probability of each monitoring area. An appropriate probability threshold is set based on historical probability data, which refers to the data set of past fire occurrence probabilities in the monitored area. The predicted fire occurrence probability of the monitored area is compared with the probability threshold. When the predicted fire occurrence probability is greater than the probability threshold, the monitored area is marked as a high-risk fire area. Patrol personnel and staff need to be reminded to evacuate quickly and relevant departments should be notified in a timely manner. Evacuation routes for patrol personnel and staff to safety exits should be set up, and the evacuation routes should be far away from the high-risk fire area.

7. The power plant safety monitoring system based on artificial intelligence according to claim 6, characterized in that, The process of obtaining historical fire time remaining data and building a fire time remaining prediction model includes: The fire-prone area was evenly divided into multiple unit areas of equal size. The remaining time for the fire refers to the time from this moment until the fire starts within the unit area; Factors affecting the remaining time of a fire in a unit area include the concentration of combustible gas, the rate of temperature change, the aging index of equipment insulation, the degree of material pyrolysis, and the oxygen consumption rate. The concentration of combustible gas is measured in real time by a combustible gas sensor; The rate of temperature change refers to the change in temperature per unit time. It is calculated by measuring temperature data through a temperature sensor and based on historical data. The equipment insulation aging index is derived from historical maintenance data; The degree of material pyrolysis is obtained by detecting volatile organic compounds using a volatile organic compound sensor, analyzing material color changes using multispectral imaging, and combining both results. The oxygen consumption rate was obtained through time-series analysis of oxygen sensor data. Acquire historical fire time remaining data for a single unit area, including combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree, oxygen consumption rate, and historical fire time remaining for that single unit area. Based on the combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree, oxygen consumption rate and corresponding historical fire remaining time data of the corresponding unit area, a fire remaining time prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree and oxygen consumption rate in the remaining time data of different historical fires in the second training set are used as the input data of the second convolutional neural network, and the remaining time of the corresponding historical fires in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network whose output is less than or equal to the preset second test error threshold is used as the fire remaining time prediction model.

8. The power plant safety monitoring system based on artificial intelligence according to claim 7, characterized in that, The process of predicting the remaining time of fire in a unit area using a fire time-of-flight prediction model and obtaining a fire-fighting robot that matches the remaining time of fire includes: The fire-fighting robot mentioned above refers to a fire-fighting robot that can reach the area of ​​the unit to be supported within the remaining time of a fire. The unit area awaiting support refers to the unit area where a fire is about to occur; When it is necessary to predict the remaining time of fire in each unit area, the real-time combustible gas concentration, temperature change rate, equipment insulation aging index, material pyrolysis degree and oxygen consumption rate of each unit area are input into the remaining time of fire prediction model to obtain the predicted remaining time of fire in each unit area. All firefighting robots in the power plant were analyzed. The arrival time of the firefighting robot to the area to be supported was obtained by dividing the walking distance of the firefighting robot from the walking speed of the firefighting robot to the area to be supported. The arrival time was compared with the predicted remaining time of the fire. Firefighting robots whose arrival time was less than the predicted remaining time of the fire were marked as firefighting robots that could be supported, and firefighting robots whose arrival time was greater than the predicted remaining time of the fire were marked as firefighting robots that could not be supported.

9. The power plant safety monitoring system based on artificial intelligence according to claim 8, characterized in that, The process of establishing a support rating system to evaluate all available firefighting robots includes: Collect support parameters that can support the firefighting robot, including the firefighting robot's battery level, remaining extinguishing agent, historical failure index, and arrival time; The battery power of the firefighting robot can be obtained directly from its battery management system. The remaining amount of extinguishing agent in the fire-fighting robot is obtained from the agent storage sensor of the fire-fighting robot; The historical failure index is calculated based on the robot's maintenance records and failure history. Processing support parameters: Normalize the remaining battery capacity: In the formula, To normalize the battery capacity, This represents the actual remaining battery capacity. Maximum battery capacity; The remaining extinguishing agent was normalized: In the formula, This is the normalized residual amount of extinguishing agent. This represents the measured residual amount of extinguishing agent. This is the maximum amount of extinguishing agent; The historical failure index was normalized: In the formula, To normalize the historical failure index, This represents the actual number of failures within a year. The maximum number of failures allowed within one year; The arrival time is normalized: In the formula, To normalize arrival time, This is the actual arrival time. The maximum arrival time is the estimated remaining time before the fire occurs. Support rating E is: In the formula, , , and These are weighting coefficients, obtained through training based on historical data.

10. The power plant safety monitoring system based on artificial intelligence according to claim 9, characterized in that, The process of dispatching an appropriate number of firefighting robots to provide support based on the support score includes: A fire is about to break out in a unit area within a high-risk fire zone. The remaining time of the fire in the unit area is predicted using a fire time prediction model. The number of fire-fighting robots required is determined based on the predicted remaining time. The longer the predicted remaining time, the fewer fire-fighting robots are needed. The shorter the predicted remaining time, the more intense the fire and the more fire-fighting robots are needed. Based on the remaining time before the fire occurs, all firefighting robots in the power plant are marked as either supportable or non-supportable. All supportable firefighting robots are evaluated using a support score and ranked from highest to lowest. Based on the predicted remaining time before the fire occurs, an appropriate number of supportable firefighting robots are selected and marked as rapid support firefighting robots. These rapid support firefighting robots are dispatched to the unit area to carry out fire prevention or firefighting work. The unselected supportable and non-supportable firefighting robots serve as backup firefighting robots, which will take over the role of rapid support firefighting robots after they become unable to continue working, or control the spread of fire outside the unit area.

Images