Method and system for intelligently monitoring unsafe behaviors of personnel in thermal power plant operation

Abstract

The invention provides an intelligent monitoring method and system for unsafe behaviors of personnel in thermal power plant operation, and relates to the technical field of intelligent monitoring and safety management, and the method comprises the steps: carrying out the risk feature recognition of personnel in a first monitoring image set, and carrying out the evaluation to obtain a plurality of real-time risk coefficients; determining a plurality of marked persons and a plurality of video coverage devices, and calculating to obtain a plurality of risk coefficient deviations; frequency compensation coefficients and personnel identification precision are determined according to risk coefficient deviation analysis, image extraction is performed based on a plurality of frequency compensation coefficients, and monitoring is performed according to a plurality of personnel identification precision configuration optimization feature extraction channels. The technical problems that a traditional safety monitoring method lacks a dynamic adjustment mechanism and cannot adapt to a complex and changeable boiler operation environment, consequently, monitoring resource configuration is unreasonable, and delay, missing detection and false alarm are prone to occurring can be solved, the rationality and scientificity of monitoring resource configuration can be improved, and efficient, flexible and accurate safety monitoring is achieved.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring and safety management technology, and in particular to a method and system for intelligently monitoring unsafe behaviors of personnel in thermal power plant operations. Background Art

[0002] In the boiler operation area of ​​a thermal power plant, personnel safety management has always been one of the key issues to ensure the safe and stable operation of power production. The boiler operation area usually has high temperature, high pressure, smoke and other dangerous factors, and the operating environment is complex and changeable. In this environment, traditional personnel safety monitoring methods face many challenges, especially in terms of monitoring resource allocation and monitoring efficiency.

[0003] At present, traditional safety monitoring methods mainly rely on fixed monitoring schemes, usually using preset monitoring frequencies and unified monitoring strategies. This approach usually does not take into account the real-time changes and dynamic characteristics of the boiler operating environment. Therefore, it cannot adapt to the complexity of different operating areas and operator behaviors. Summary of the invention

[0004] The purpose of this application is to provide an intelligent monitoring method and system for unsafe behaviors of personnel in thermal power plant operations, so as to solve the technical problems that the traditional personnel safety monitoring method in the boiler operation area of ​​a thermal power plant lacks a dynamic adjustment mechanism and cannot adapt to the complex and changeable boiler operating environment, resulting in unreasonable allocation of monitoring resources and prone to delays, missed detections and false alarms.

[0005] In view of the above problems, the present application provides a method and system for intelligently monitoring unsafe behaviors of personnel in thermal power plant operations.

[0006] In a first aspect, the present application provides a method for intelligently monitoring unsafe behaviors of personnel in thermal power plant operations, which is implemented by an intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations, including: performing regional segmentation and risk assessment on the boiler operation area of ​​the target thermal power plant, and determining a number of operation sub-areas and a number of operation risk coefficients; setting a monitoring frequency according to the several operation risk coefficients, generating a first monitoring frequency distribution, and performing image extraction on the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set; performing risk feature identification on the operating personnel in the first monitoring image set through a feature extraction channel, and obtaining a plurality of real-time risk coefficients based on the evaluation of the risk feature identification results; if the real-time risk coefficient is greater than a predetermined risk threshold, marking the operating personnel, determining a number of marked personnel and a number of video coverage devices, and calculating a number of risk coefficient deviations; determining a number of frequency compensation coefficients and a number of personnel recognition accuracies based on the analysis of the several risk coefficient deviations, performing image extraction on the several video coverage devices based on the several frequency compensation coefficients, configuring a number of optimized feature extraction channels according to the several personnel recognition accuracies, and performing unsafe behavior monitoring of personnel in the boiler operation area.

[0007] In the second aspect, the present application also provides an intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations, which is used to execute an intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations as described in the first aspect, including: a regional risk assessment module, which is used to perform regional segmentation and risk assessment on the boiler operation area of ​​the target thermal power plant, and determine a number of operation sub-areas and a number of operation risk coefficients; a first monitoring frequency setting module, which is used to set the monitoring frequency according to the several operation risk coefficients, generate a first monitoring frequency distribution, and extract images from the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set; a risk feature recognition module, which is used to identify the first monitoring image set through a feature extraction channel. The operating personnel perform risk feature identification, and obtain multiple real-time risk coefficients based on the risk feature identification results; the risk coefficient deviation calculation module is used to mark the operating personnel if the real-time risk coefficient is greater than the predetermined risk threshold, determine a number of marked personnel and a number of video coverage devices, and calculate a number of risk factor deviations; the unsafe behavior monitoring module is used to determine a number of frequency compensation coefficients and a number of personnel recognition accuracies based on the risk factor deviation analysis, perform image extraction on the number of video coverage devices based on the frequency compensation coefficients, configure a number of optimized feature extraction channels according to the number of personnel recognition accuracies, and perform unsafe behavior monitoring of personnel in the boiler operating area.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] By performing regional segmentation and risk assessment on the boiler operation area of ​​the target thermal power plant, a number of operation sub-areas and a number of operation risk coefficients are determined; the monitoring frequency is set according to the several operation risk coefficients, a first monitoring frequency distribution is generated, and images are extracted from the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set; risk characteristics of the operators in the first monitoring image set are identified through a feature extraction channel, and a plurality of real-time risk coefficients are obtained based on the evaluation of the risk characteristic identification results; if the real-time risk coefficient is greater than a predetermined risk threshold, the operator is marked, a number of marked personnel and a number of video coverage devices are determined, and a number of risk coefficient deviations are calculated; a number of frequency compensation coefficients and a number of personnel recognition accuracies are determined according to the analysis of the several risk coefficient deviations, image extraction is performed on the several video coverage devices based on the several frequency compensation coefficients, and a number of optimized feature extraction channels are configured according to the several personnel recognition accuracies to perform unsafe behavior monitoring of personnel in the boiler operation area. In other words, by setting an appropriate monitoring plan based on the real-time working environment and risk conditions, the rationality and scientificity of monitoring resource allocation can be improved, thereby achieving efficient, flexible and accurate safety monitoring, significantly improving the efficiency and quality of personnel safety monitoring, and improving overall operational safety.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0012] Figure 1 A flowchart of an intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations is provided for this application;

[0013] Figure 2 The present invention is a structural schematic diagram of an intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations.

[0014] Description of reference numerals:

[0015] A regional risk assessment module 11, a first monitoring frequency setting module 12, a risk feature identification module 13, a risk coefficient deviation calculation module 14, and an unsafe behavior monitoring module 15. DETAILED DESCRIPTION

[0016] This application provides an intelligent monitoring method and system for unsafe behaviors of personnel in thermal power plant operations, which solves the problem that the traditional personnel safety monitoring method in the boiler operation area of ​​a thermal power plant lacks a dynamic adjustment mechanism and cannot adapt to the complex and changeable boiler operating environment, resulting in unreasonable allocation of monitoring resources and prone to delays, missed detections and false alarms. By setting a corresponding monitoring plan based on the real-time operating environment and risk conditions, the rationality and scientificity of monitoring resource allocation can be improved, thereby achieving efficient, flexible and accurate safety monitoring, significantly improving the efficiency and quality of personnel safety monitoring, and improving overall operational safety.

[0017] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0018] For example, please refer to the attached Figure 1 The present application provides a method for intelligently monitoring unsafe behaviors of personnel in thermal power plant operations, which is applied to an intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations, and specifically includes the following steps:

[0019] S100: Perform regional segmentation and risk assessment on the boiler operation area of ​​the target thermal power plant, and determine a number of operation sub-areas and a number of operation risk coefficients.

[0020] Furthermore, step S100 of the present application also includes:

[0021] S110: Acquire a video monitoring network of a target thermal power plant, wherein the video monitoring network includes a plurality of image acquisition devices; S120: Segment the boiler operation area according to the plurality of image acquisition devices to determine a plurality of operation sub-areas.

[0022] Specifically, in thermal power plants, the boiler operation area is usually highly complex and dangerous. The operating environment not only involves physical factors such as high temperature and pressure, but may also be accompanied by risks such as smoke and radiation. These factors make traditional personnel safety monitoring methods unable to effectively adapt to the dynamically changing operating environment. Therefore, the use of video surveillance technology to monitor the boiler operation area in real time has become an important means to improve the level of safety management.

[0023] First, the video monitoring network of the target thermal power plant is obtained. The video monitoring network includes several image acquisition devices. As an important part of the safety monitoring of thermal power plants, the video monitoring network is usually composed of several image acquisition devices (such as cameras, thermal imaging devices, infrared sensors, etc.). Through reasonable equipment layout, it can cover different areas of the boiler operation area and realize real-time monitoring and safety management of operators. These devices transmit video image data to the central monitoring system in real time, providing a comprehensive monitoring view for safety managers. The video monitoring network consists of multiple image acquisition devices. The installation and layout of these devices need to be reasonably planned according to the spatial characteristics of the boiler operation area. Each image acquisition device can usually cover a specific part of the boiler area, such as the operating table, maintenance area, and danger zone.

[0024] Next, the boiler operating area is segmented according to the several image acquisition devices, that is, the boiler operating area is segmented according to the coverage range of each image acquisition device, and the coverage area of ​​each image acquisition device is determined, such as dividing the boiler operating area into several uniform grids, each grid area is monitored by one or more image acquisition devices, and several operating sub-areas are determined, wherein each operating sub-area corresponds to at least one image acquisition device.

[0025] S130: Performing an operation risk assessment on the plurality of operation sub-areas to determine a plurality of operation risk coefficients.

[0026] Furthermore, step S130 of the present application also includes:

[0027] S131: query the operation log of the target thermal power plant, count the proportion of unsafe behaviors of personnel in the several operation sub-areas, and determine several risk behavior ratios; S132: use the several risk behavior ratios as the several operation risk coefficients.

[0028] Specifically, the operation log of the target thermal power plant is queried. The operation log of the thermal power plant contains detailed operation records, including the behavior of the operators, operation time, equipment status, on-site conditions, etc. These logs can be obtained through the factory's safety management system or equipment monitoring system; then, based on the operation log, the proportion of unsafe behaviors of personnel in the several operation sub-areas is counted. Unsafe behaviors include not wearing protective equipment, violating operating procedures, and performing non-standard operations in high temperature and high pressure areas, etc., to obtain several risk behavior ratios, which are the ratio of the number of unsafe behaviors to the total number of operations; finally, the several risk behavior ratios are used as the several operation risk coefficients. The operation risk coefficient represents the safety risk level of the area. The higher the risk coefficient, the more serious the safety hazard in the area, and it requires priority attention and more safety monitoring measures.

[0029] S200: Setting a monitoring frequency according to the plurality of operation risk factors, generating a first monitoring frequency distribution, and extracting images from a video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set.

[0030] Furthermore, step S200 of the present application also includes:

[0031] S210: Input the plurality of operation risk coefficients into a coefficient-frequency comparison library for matching, and output a plurality of monitoring frequencies, wherein the risk coefficient and the monitoring frequency are positively correlated.

[0032] Specifically, first, a coefficient-frequency comparison library is constructed, which is used to map the operation risk coefficient to a specific monitoring frequency. Generally, there is a positive correlation between the operation risk coefficient and the monitoring frequency, that is, the higher the risk, the higher the monitoring frequency. The main function of the comparison library is to correspond the risk coefficient of each area to the preset monitoring frequency. The comparison library matches different risk coefficient intervals with corresponding monitoring frequencies. Generally speaking, areas with higher risk coefficients should be set with higher monitoring frequencies so as to timely discover potential safety problems. Then, the several operation risk coefficients are input into the coefficient-frequency comparison library for matching. According to the matching results of the comparison library, the monitoring frequency of each operation sub-area is output. These frequencies will indicate how frequently each area needs to perform image extraction and monitoring; and several monitoring frequencies are obtained. By matching the operation risk coefficient with the monitoring frequency, accurate, dynamic and effective safety monitoring of the boiler operation area of ​​a thermal power plant can be achieved; this method can reasonably configure monitoring resources, ensure that high-risk areas are adequately monitored, reduce resource waste in low-risk areas, and improve the efficiency and response speed of the entire monitoring system. This dynamic frequency adjustment mechanism not only improves the quality of personnel safety monitoring, but also helps to achieve more refined safety management.

[0033] S220: Mapping and distributing the plurality of monitoring frequencies according to a plurality of image acquisition devices to generate a first monitoring frequency distribution.

[0034] Furthermore, step S220 of the present application also includes:

[0035] S221: According to predetermined environmental interference factors, interference identification is performed regularly on the plurality of operation sub-areas to determine a plurality of environmental interference coefficients, wherein the predetermined environmental interference factors include at least light intensity and smoke concentration.

[0036] Furthermore, step S221 of the present application also includes:

[0037] S2211: Under a predetermined update node, interference factors are collected for the several operation sub-areas according to predetermined environmental interference factors to determine several light intensities and several smoke concentrations; S2212: Several environmental interference coefficients are calculated based on the several light intensities and several smoke concentrations, wherein the environmental interference coefficient is negatively correlated with the light intensity and positively correlated with the smoke concentration.

[0038] Specifically, when conducting personnel safety monitoring in the boiler operation area of ​​a thermal power plant, the impact of environmental interference factors (such as light intensity and smoke concentration) on the monitoring system cannot be ignored. These interference factors may affect the camera's image quality, image recognition accuracy, and the overall safety monitoring effect. In order to improve the accuracy and reliability of the monitoring system, dynamic adjustments need to be made based on these environmental factors.

[0039] First, a predetermined update node is configured, that is, the environmental interference data of the operating area is obtained and updated regularly or on demand. These update nodes can be set based on the actual operating conditions, such as every hour or every day; then, under the predetermined update node, interference factors are collected for the several operating sub-areas according to the predetermined environmental interference factors (light intensity and smoke concentration), wherein the light intensity can be obtained by installing a light sensor (such as a photometer, an ambient light sensor, etc.) to obtain the light intensity in the operating area. The change in light intensity will directly affect the image quality of the image acquisition device (such as a camera, an infrared sensor, etc.), especially in low light or strong light environments, the monitoring effect may be limited; the smoke concentration can be collected by installing a smoke sensor or a gas detector. These sensors can monitor the smoke concentration in the boiler area in real time, thereby affecting the working effect of the image acquisition device. When the smoke concentration is high, it may cause blurred images or unclear transmission signals; obtain several light intensities and several smoke concentrations.

[0040] Then, based on the several light intensities and several smoke concentrations, several environmental interference coefficients are calculated. The environmental interference coefficients can reflect the impact of environmental factors on the monitoring effect, especially the monitoring needs under unstable or harsh environmental conditions. The environmental interference coefficient is negatively correlated with the light intensity, that is, the higher the light intensity, the lower the interference coefficient; the lower the light intensity, the higher the interference coefficient; and positively correlated with the smoke concentration, that is, the higher the smoke concentration, the higher the environmental interference coefficient. When the smoke concentration is high, the image quality will be reduced and even the working efficiency of the monitoring equipment will be affected. Thick smoke or gas will block the line of sight and increase the recognition difficulty of the monitoring equipment; the environmental interference coefficient can be obtained by setting weights and weighting the light intensity and smoke concentration.

[0041] S222: configuring a plurality of first frequency compensation coefficients according to the plurality of environmental interference coefficients, and performing periodic compensation update on the first monitoring frequency distribution, wherein the first frequency compensation coefficient is positively correlated with the environmental interference coefficient.

[0042] Specifically, a number of first frequency compensation coefficients are then configured according to the several environmental interference coefficients, wherein the first frequency compensation coefficient is positively correlated with the environmental interference coefficient, that is, in order to cope with different environmental interferences, the system needs to introduce a compensation mechanism to adjust the monitoring frequency, and the first frequency compensation coefficient is used to dynamically adjust the monitoring frequency according to the environmental interference coefficient. According to actual conditions, the first frequency compensation coefficient is positively correlated with the environmental interference coefficient, that is, the larger the environmental interference coefficient, the larger the frequency compensation coefficient, thereby increasing the monitoring frequency; for example, if the environmental interference is small (such as high light intensity and low smoke concentration), the frequency compensation coefficient is small, and the monitoring frequency can be maintained at a low level; if the environmental interference is large (such as low light intensity and high smoke concentration), the frequency compensation coefficient is large, and the monitoring frequency needs to be increased to ensure the validity of the image data and the monitoring accuracy.

[0043] Finally, the first monitoring frequency distribution is periodically compensated and updated according to the several first frequency compensation coefficients. The frequency compensation coefficient is regularly updated to adapt to the dynamic changes of the environment. According to the real-time changes of the environmental interference coefficient (such as changes in light intensity, fluctuations in smoke concentration), the frequency compensation update can be performed at predetermined update nodes (such as every hour, every shift, or during specific work activities). Through this dynamic frequency compensation mechanism, the system can flexibly adjust the monitoring frequency according to real-time environmental conditions (such as changes in light intensity, changes in smoke concentration, etc.). A reasonable frequency compensation mechanism can avoid excessive use of monitoring resources when the environment is relatively stable, reduce the burden on the monitoring system, and improve overall efficiency, thereby ensuring that the monitoring system can perform tasks efficiently and accurately.

[0044] S230: Extracting images from a plurality of image acquisition devices in the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set.

[0045] Specifically, according to the first monitoring frequency distribution, images are extracted from several image acquisition devices in the video monitoring network. That is, for areas with higher monitoring frequencies (such as high-risk areas or areas with greater environmental interference), the image acquisition devices will extract more image frames per second to ensure that relevant information can be captured in a timely manner when high-frequency events occur; for areas with lower risks and less environmental interference, the image acquisition devices will acquire images at a lower frequency to save resources and avoid excessive monitoring; and the first monitoring image set is obtained.

[0046] By controlling the image acquisition devices in the video surveillance network according to the first monitoring frequency distribution, the monitoring strategy can be flexibly adjusted under different risks and environmental conditions to ensure that high-risk areas are fully monitored while optimizing the efficiency of monitoring resource utilization. The final generated first monitoring image set provides an accurate data basis for subsequent security monitoring, abnormal behavior detection and accident prevention.

[0047] S300: Identify risk characteristics of the operators in the first monitoring image set through a feature extraction channel, and obtain multiple real-time risk coefficients based on the risk characteristic identification results.

[0048] Furthermore, step S300 of the present application also includes:

[0049] S310: Pre-training a feature extraction channel, wherein the feature extraction channel is constructed based on a convolutional neural network.

[0050] Furthermore, step S310 of the present application also includes:

[0051] S311: Query the boiler area operation log to obtain a sample personnel image set, a sample moving direction set and a sample warning distance set; S312: Use the sample personnel image set, the sample moving direction set and the sample warning distance set as a sample data set, and divide the sample data set into Q equal parts, select Q times with replacement to obtain a first training set, iterate and select Q times to obtain Q training sets, where Q is an integer greater than 10; S313: Take the sample personnel image as input, and the sample moving direction and the sample warning distance as output, and supervise the convolutional neural network through the Q training sets until convergence, obtain Q feature extraction branches, and construct a feature extraction channel based on the fusion of the Q feature extraction branches.

[0052] Specifically, query the boiler area operation log, collect sample personnel image set, sample movement direction set and sample warning distance set, where the movement direction is close to the warning area or away from the warning area; the warning distance is the distance between the operator and the boiler equipment or dangerous area. Then the sample personnel image set, sample movement direction set and sample warning distance set are used as sample data sets, and the sample data set is divided into Q parts, where Q is an integer greater than 10, and the value of Q can be set according to actual needs, such as 20, to obtain Q data sets, and select Q times with replacement from the Q data sets to obtain the first training set, and iterate and select Q times to obtain Q training sets.

[0053] Then, the sample personnel image is used as input, and the sample moving direction and the sample warning distance are used as output. The convolutional neural network is supervised and trained through the Q training sets, that is, the convolutional neural network (CNN) is used for training. The CNN can effectively extract spatial features from the image data and can recognize important patterns in the image; by performing supervised learning on the Q training sets, the CNN outputs the corresponding moving direction and warning distance according to the input image, and the loss function adopts the mean square error (MSE) or cross entropy loss to optimize the model parameters until the model converges; the trained Q feature extraction branches are obtained, and a feature extraction channel is constructed according to the fusion of the Q feature extraction branches.

[0054] S320: performing risk feature identification on the operating personnel in the first monitoring image set through a feature extraction channel to obtain a plurality of risk feature identification results, wherein each risk feature identification result includes a warning area interval distance, a personnel movement direction, and a personnel movement speed, wherein the personnel movement direction is approaching the warning area or moving away from the warning area;

[0055] Furthermore, step S320 of the present application also includes:

[0056] S321: Randomly select a predetermined number of feature extraction branches in the feature extraction channel to identify risk characteristics of the workers in the first monitoring image set, and obtain multiple warning area interval distances and multiple personnel movement directions, where the predetermined number defaults to 1; S322: If the warning area interval distance is less than the predetermined distance threshold, and the personnel movement direction is close to the warning area, then retrieve the historical risk feature sequence in the predetermined historical time zone, and perform personnel movement feature analysis based on the historical risk feature sequence to obtain the personnel movement rate.

[0057] Specifically, first, a predetermined number of feature extraction branches are randomly selected in the feature extraction channel, wherein the predetermined number defaults to 1, that is, the initial selection number is 1, which reduces computing power consumption and improves analysis efficiency. Only when necessary, the predetermined number can be adjusted (such as increasing to 2 or more) to improve the robustness and accuracy of the model; that is, any one of the Q feature extraction branches is randomly selected, and the risk characteristics of the operators in the first monitoring image set are identified through the selected feature extraction branch to obtain multiple warning area intervals and multiple personnel movement directions. If the warning area interval distance is less than the predetermined distance threshold (based on the actual operation scene setting), and the personnel movement direction is close to the warning area, then the historical risk feature sequence in the predetermined historical time zone (such as the last 1 minute) is retrieved, and based on the retrieved historical risk feature sequence, the movement characteristics of the operators are analyzed and their movement rate is calculated. The movement rate is a measure of the speed at which personnel move to the dangerous area. People with faster speeds may indicate an emergency, and it is necessary to increase the monitoring frequency or take other safety measures to obtain the personnel movement rate.

[0058] S330: Through the expert system, the risk assessment of the operating personnel is performed according to the interval distance of the warning area, the movement direction of the personnel and the movement speed of the personnel, and multiple real-time risk coefficients are obtained.

[0059] Specifically, the expert system is finally used to evaluate the risk of the operators according to the distance between the warning areas, the direction of personnel movement and the speed of personnel movement. The expert system is a computer program that simulates the expert decision-making process. It can analyze the input data according to the expert experience and rules and give corresponding judgments. In this system, the expert system is used to evaluate the safety risk according to the behavioral characteristics of the operators (such as distance, direction, and movement speed), and then obtain the real-time risk coefficient; among them, the distance between the warning areas is small, the coefficient is large, and the risk is high; if the distance is large, the coefficient is small. The coefficient towards the dangerous area is high, and the coefficient away from the dangerous area is low. The personnel movement speed is too fast, the coefficient is high, indicating a greater risk; multiple real-time risk coefficients are obtained.

[0060] S400: If the real-time risk factor is greater than a predetermined risk threshold, the operator is marked, a number of marked personnel and a number of video coverage devices are determined, and a number of risk factor deviations are calculated.

[0061] Furthermore, step S400 of the present application also includes:

[0062] S410: If the real-time risk coefficient is greater than the predetermined risk threshold, the real-time risk coefficient is marked as an overflow risk coefficient; S420: Randomly select a first overflow risk coefficient, and based on the predetermined risk threshold, calculate the deviation between the first overflow risk coefficient and the predetermined risk threshold, obtain a first risk coefficient deviation, and add it to the several risk coefficient deviations.

[0063] Specifically, a predetermined risk threshold is configured, and the predetermined risk threshold can be set based on the actual operation scenario; if the real-time risk coefficient is greater than the predetermined risk threshold, it means that the operator may be at a higher safety risk, then the operator is marked, a number of marked personnel are determined, and a number of video coverage devices are determined, that is, according to the position and activity of the operator, the appropriate camera or monitoring equipment is automatically selected for image acquisition to ensure real-time tracking of the person's behavior. The real-time risk coefficient is further marked as an overflow risk coefficient, and multiple overflow risk coefficients are determined; and any one of the multiple overflow risk coefficients is randomly selected as the first overflow risk coefficient; the deviation of the first overflow risk coefficient from the predetermined risk threshold is further calculated to obtain a first risk coefficient deviation, which is added to the several risk coefficient deviations.

[0064] S500: Determine a number of frequency compensation coefficients and a number of personnel recognition accuracies according to the deviation analysis of the several risk coefficients, perform image extraction on the several video coverage devices based on the several frequency compensation coefficients, configure a number of optimized feature extraction channels according to the several personnel recognition accuracies, and perform unsafe behavior monitoring of personnel in the boiler operation area.

[0065] Furthermore, step S500 of the present application also includes:

[0066] S510: Calculate the ratios of the several risk coefficient deviations to the historical maximum risk coefficient deviation respectively to obtain several risk deviation ratios; S520: Configure several second frequency compensation coefficients based on the several risk deviation ratios, wherein the second frequency compensation coefficient is positively correlated with the risk deviation ratio.

[0067] Specifically, the ratios of the several risk coefficient deviations to the historical maximum risk coefficient deviation (which can be set according to historical records) are calculated respectively to obtain several risk deviation ratios; then, according to the several risk deviation ratios, several second frequency compensation coefficients are configured. The function of the second frequency compensation coefficient is to dynamically adjust the image acquisition frequency according to the changes in real-time risks to ensure that the monitoring of high-risk areas is more accurate. Among them, the second frequency compensation coefficient is positively correlated with the risk deviation ratio, that is, the larger the risk deviation ratio, the larger the compensation coefficient, which means that the monitoring frequency of the area will increase, thereby improving the monitoring accuracy. For example, if the risk deviation ratio of a certain area is higher (indicating a higher risk), a larger compensation coefficient will be configured, thereby increasing the image acquisition frequency of the area; if the risk deviation ratio is lower, the compensation coefficient will be smaller, thereby reducing the monitoring frequency of the area and saving computing power.

[0068] S530: Multiply the several risk deviation ratios by Q and round them to the integer to obtain several personnel recognition accuracies, where the personnel recognition accuracy is the number of selected feature extraction branches.

[0069] Furthermore, step S530 of the present application also includes:

[0070] S531: Multiply the several risk deviation ratios by Q and round them up to obtain several initial branch selection numbers; S532: Obtain several environmental interference coefficients of several video coverage devices in the current time zone, and calculate the ratios of several environmental interference coefficients to the standard environmental interference coefficients to obtain several selection compensation coefficients; S533: Compensate and round up the several initial branch selection numbers according to the several selection compensation coefficients to obtain several branch selection numbers, which are set as several personnel recognition accuracies.

[0071] Specifically, on the other hand, the several risk deviation ratios are multiplied by Q and rounded to obtain several initial branch selection numbers; then, several environmental interference coefficients of several video coverage devices in the current time zone are obtained, and the ratios of several environmental interference coefficients to standard environmental interference coefficients are calculated to obtain several selection compensation coefficients. Finally, the several initial branch selection numbers are compensated and rounded according to the several selection compensation coefficients, that is, the personnel recognition accuracy of each operation area (i.e., the number of branch selections) will be adjusted according to dynamic factors such as environmental interference and risk deviation. By increasing the number of branches, the system can improve the monitoring accuracy of personnel behavior in complex or highly disturbed environments, and obtain several branch selection numbers, which are set as several personnel recognition accuracies. The level of recognition accuracy directly affects the number of feature extraction branches, thereby affecting the monitoring system's ability to detect operating personnel. If the risk deviation ratio of a certain area is high, the personnel recognition accuracy of the area will be high, and the system will increase the number of feature extraction branches, so that the system's behavior recognition of the area is more accurate; for areas with low risk deviation ratios, the personnel recognition accuracy is low, the number of feature extraction branches is small, and the consumption of computing resources is correspondingly reduced.

[0072] In summary, the intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations provided by this application has the following technical effects:

[0073] By performing regional segmentation and risk assessment on the boiler operation area of ​​the target thermal power plant, a number of operation sub-areas and a number of operation risk coefficients are determined; the monitoring frequency is set according to the several operation risk coefficients, a first monitoring frequency distribution is generated, and images are extracted from the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set; risk characteristics of the operators in the first monitoring image set are identified through a feature extraction channel, and a plurality of real-time risk coefficients are obtained based on the evaluation of the risk characteristic identification results; if the real-time risk coefficient is greater than a predetermined risk threshold, the operator is marked, a number of marked personnel and a number of video coverage devices are determined, and a number of risk coefficient deviations are calculated; a number of frequency compensation coefficients and a number of personnel recognition accuracies are determined according to the analysis of the several risk coefficient deviations, image extraction is performed on the several video coverage devices based on the several frequency compensation coefficients, and a number of optimized feature extraction channels are configured according to the several personnel recognition accuracies to perform unsafe behavior monitoring of personnel in the boiler operation area. In other words, by setting an appropriate monitoring plan based on the real-time working environment and risk conditions, the rationality and scientificity of monitoring resource allocation can be improved, thereby achieving efficient, flexible and accurate safety monitoring, significantly improving the efficiency and quality of personnel safety monitoring, and improving overall operational safety.

[0074] Embodiment 2: Based on the same invention concept as the method for intelligently monitoring unsafe behaviors of personnel in thermal power plant operations in the aforementioned embodiment, the present application also provides an intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations, see the attached Figure 2 , including: a regional risk assessment module 11, which is used to perform regional segmentation and risk assessment on the boiler operation area of ​​the target thermal power plant, and determine a number of operation sub-areas and a number of operation risk coefficients; a first monitoring frequency setting module 12, which is used to set the monitoring frequency according to the several operation risk coefficients, generate a first monitoring frequency distribution, and extract images from the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set; a risk feature identification module 13, which is used to identify the risk features of the operators in the first monitoring image set through a feature extraction channel, and obtain multiple real-time risk coefficients according to the risk feature identification result evaluation; a risk factor deviation calculation module 14, which is used to mark the operators if the real-time risk coefficient is greater than a predetermined risk threshold, determine a number of marked operators and a number of video coverage devices, and calculate a number of risk factor deviations; an unsafe behavior monitoring module 15, which is used to determine a number of frequency compensation coefficients and a number of personnel recognition accuracies according to the several risk factor deviations, extract images from the several video coverage devices based on the several frequency compensation coefficients, configure a number of optimized feature extraction channels according to the several personnel recognition accuracies, and perform unsafe behavior monitoring of personnel in the boiler operation area.

[0075] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: obtain a video monitoring network of a target thermal power plant, wherein the video monitoring network includes a plurality of image acquisition devices; segment the boiler operating area according to the plurality of image acquisition devices to determine a plurality of operating sub-areas; perform an operating risk assessment on the plurality of operating sub-areas to determine a plurality of operating risk coefficients.

[0076] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: query the operation log of the target thermal power plant, count the proportions of unsafe behaviors of personnel in the several operation sub-areas, determine several risk behavior ratios; and use the several risk behavior ratios as the several operation risk coefficients.

[0077] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: input the several operation risk coefficients into a coefficient-frequency comparison library for matching, and output several monitoring frequencies, wherein the risk coefficient and the monitoring frequency are positively correlated; map and distribute the several monitoring frequencies according to several image acquisition devices to generate a first monitoring frequency distribution; extract images from several image acquisition devices in the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set.

[0078] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: regularly identify interference on the several operating sub-areas according to predetermined environmental interference factors, and determine several environmental interference coefficients, wherein the predetermined environmental interference factors include at least light intensity and smoke concentration; configure several first frequency compensation coefficients according to the several environmental interference coefficients, and regularly compensate and update the first monitoring frequency distribution, wherein the first frequency compensation coefficient is positively correlated with the environmental interference coefficient.

[0079] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: at a predetermined update node, collect interference factors for the several operation sub-areas according to predetermined environmental interference factors, and determine several light intensities and several smoke concentrations; calculate several environmental interference coefficients based on the several light intensities and several smoke concentrations, wherein the environmental interference coefficient is negatively correlated with the light intensity and positively correlated with the smoke concentration.

[0080] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used for: pre-training a feature extraction channel, wherein the feature extraction channel is constructed based on a convolutional neural network; performing risk feature identification on the operating personnel in the first monitoring image set through the feature extraction channel to obtain a number of risk feature identification results, wherein each risk feature identification result includes the warning area interval distance, the personnel movement direction and the personnel movement rate, wherein the personnel movement direction is approaching the warning area or moving away from the warning area; through an expert system, performing operating personnel risk assessment based on the warning area interval distance, the personnel movement direction and the personnel movement rate to obtain multiple real-time risk coefficients.

[0081] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: query the boiler area operation log to obtain a sample personnel image set, a sample moving direction set and a sample warning distance set; use the sample personnel image set, the sample moving direction set and the sample warning distance set as a sample data set, and divide the sample data set into Q equal parts, select Q times with replacement to obtain a first training set, iterate and select Q times to obtain Q training sets, where Q is an integer greater than 10; use sample personnel images as input, and sample moving directions and sample warning distances as outputs, supervise the convolutional neural network through the Q training sets until convergence, obtain Q feature extraction branches, and construct a feature extraction channel based on the fusion of the Q feature extraction branches.

[0082] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: randomly select a predetermined number of feature extraction branches in the feature extraction channel to identify risk characteristics of the operating personnel in the first monitoring image set, and obtain multiple warning area interval distances and multiple personnel movement directions, wherein the predetermined number defaults to 1; if the warning area interval distance is less than a predetermined distance threshold, and the personnel movement direction is close to the warning area, then retrieve the historical risk feature sequence in the predetermined historical time zone, and perform personnel movement feature analysis based on the historical risk feature sequence to obtain the personnel movement rate.

[0083] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: if the real-time risk coefficient is greater than a predetermined risk threshold, mark the real-time risk coefficient as an overflow risk coefficient; randomly select a first overflow risk coefficient, and take the predetermined risk threshold as a benchmark, calculate the deviation between the first overflow risk coefficient and the predetermined risk threshold, obtain a first risk coefficient deviation, and add it to the several risk coefficient deviations.

[0084] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: calculate the ratios of the several risk coefficient deviations to the historical maximum risk coefficient deviations, respectively, to obtain several risk deviation ratios; configure several second frequency compensation coefficients based on the several risk deviation ratios, wherein the second frequency compensation coefficient is positively correlated with the risk deviation ratio; multiply the several risk deviation ratios by Q and round them up, to obtain several personnel identification accuracies, wherein the personnel identification accuracy is the selected number of feature extraction branches.

[0085] Furthermore, the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations is also used to: multiply the several risk deviation ratios by Q and round them up to obtain several initial branch selection numbers; obtain several environmental interference coefficients of several video coverage devices in the current time zone, and calculate the ratios of several environmental interference coefficients to the standard environmental interference coefficients to obtain several selection compensation coefficients; compensate and round up the several initial branch selection numbers according to the several selection compensation coefficients to obtain several branch selection numbers, which are set as several personnel recognition accuracies.

[0086] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations and the specific examples in the aforementioned embodiment 1 are also applicable to an intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations in this embodiment. Through the aforementioned detailed description of the intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations, those skilled in the art can clearly know the intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0087] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0088] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. An intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations, characterized in that: include: Perform regional segmentation and risk assessment on the boiler operation area of ​​the target thermal power plant, and determine several operation sub-areas and several operation risk coefficients; Setting monitoring frequencies according to the plurality of operation risk factors to generate a first monitoring frequency distribution, and extracting images from a video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set; Identify risk characteristics of the operators in the first monitoring image set through a feature extraction channel, and obtain multiple real-time risk coefficients based on the risk characteristic identification results; If the real-time risk factor is greater than a predetermined risk threshold, the operator is marked, a number of marked personnel and a number of video coverage devices are determined, and a number of risk factor deviations are calculated; According to the deviation analysis of the several risk coefficients, several frequency compensation coefficients and several personnel recognition accuracies are determined, and images are extracted for the several video coverage devices based on the several frequency compensation coefficients. According to the several personnel recognition accuracies, several optimized feature extraction channels are configured to perform unsafe behavior monitoring of personnel in the boiler operation area.

2. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 1 is characterized in that: The boiler operation area of ​​the target thermal power plant is segmented and risk assessed to determine several operation sub-areas and several operation risk factors, including: Acquire a video monitoring network of a target thermal power plant, wherein the video monitoring network includes a plurality of image acquisition devices; Segmenting the boiler operation area according to the plurality of image acquisition devices to determine a plurality of operation sub-areas; An operation risk assessment is performed on the several operation sub-areas to determine several operation risk coefficients.

3. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 2 is characterized in that: Conducting an operation risk assessment on the several operation sub-areas and determining several operation risk coefficients, including: Query the operation log of the target thermal power plant, count the proportion of unsafe behaviors of personnel in the several operation sub-areas, and determine several risk behavior proportions; The ratios of the risk behaviors are used as the risk coefficients of the operations.

4. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 2 is characterized in that: The monitoring frequency is set according to the plurality of operation risk coefficients to generate a first monitoring frequency distribution, and images are extracted from the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set, including: Inputting the several operation risk coefficients into the coefficient-frequency comparison library for matching, and outputting several monitoring frequencies, wherein the risk coefficient and the monitoring frequency are positively correlated; Mapping and distributing the plurality of monitoring frequencies according to a plurality of image acquisition devices to generate a first monitoring frequency distribution; Image extraction is performed on a plurality of image acquisition devices in a video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set.

5. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 4 is characterized in that: Generating a first monitoring frequency distribution includes: According to predetermined environmental interference factors, regularly performing interference identification on the plurality of operation sub-areas to determine a plurality of environmental interference coefficients, wherein the predetermined environmental interference factors at least include light intensity and smoke concentration; A plurality of first frequency compensation coefficients are configured according to the plurality of environmental interference coefficients, and the first monitoring frequency distribution is periodically compensated and updated, wherein the first frequency compensation coefficient is positively correlated with the environmental interference coefficient.

6. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 5 is characterized in that: According to the predetermined environmental interference factors, interference identification is performed regularly on the plurality of operation sub-areas to determine a plurality of environmental interference coefficients, including: At a predetermined update node, interference factors are collected for the plurality of operation sub-areas according to predetermined environmental interference factors, and a plurality of light intensities and a plurality of smoke concentrations are determined; A plurality of environmental interference coefficients are calculated based on the plurality of light intensities and the plurality of smoke concentrations, wherein the environmental interference coefficient is negatively correlated with the light intensity and positively correlated with the smoke concentration.

7. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 2 is characterized in that: The risk characteristics of the operators in the first monitoring image set are identified through the feature extraction channel, and multiple real-time risk coefficients are obtained according to the risk characteristic identification results, including: A pre-trained feature extraction channel, wherein the feature extraction channel is constructed based on a convolutional neural network; Perform risk feature identification on the operating personnel in the first monitoring image set through a feature extraction channel to obtain a plurality of risk feature identification results, wherein each risk feature identification result includes a warning area interval distance, a personnel movement direction, and a personnel movement speed, wherein the personnel movement direction is approaching the warning area or moving away from the warning area; Through the expert system, the risk assessment of the operating personnel is carried out according to the interval distance of the warning area, the movement direction of the personnel and the movement speed of the personnel, and multiple real-time risk coefficients are obtained.

8. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 7 is characterized in that: Pre-trained feature extraction pipeline, including: Query the boiler area operation log to obtain the sample personnel image set, sample movement direction set, and sample warning distance set; The sample personnel image set, the sample moving direction set and the sample warning distance set are used as sample data sets, and the sample data sets are equally divided into Q parts, and the first training set is obtained by selecting Q parts with replacement, and the first training set is obtained by iterative selection Q times, where Q is an integer greater than 10; With the sample personnel image as input and the sample moving direction and the sample warning distance as output, the convolutional neural network is supervised and trained through the Q training sets until convergence, to obtain Q feature extraction branches, and a feature extraction channel is constructed according to the fusion of the Q feature extraction branches.

9. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 8 is characterized in that: The risk characteristics of the operators in the first monitoring image set are identified through the feature extraction channel to obtain several risk characteristic identification results, including: Randomly selecting a predetermined number of feature extraction branches in the feature extraction channel to perform risk feature recognition on the operators in the first monitoring image set to obtain multiple warning area interval distances and multiple personnel movement directions, wherein the predetermined number defaults to 1; If the distance between the warning areas is less than the preset distance threshold and the direction of personnel movement is close to the warning area, the historical risk feature sequence in the preset historical time zone is retrieved, and the personnel movement feature analysis is performed based on the historical risk feature sequence to obtain the personnel movement rate.

10. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 8, characterized in that: Several risk factor deviations are calculated, including: If the real-time risk factor is greater than a predetermined risk threshold, the real-time risk factor is marked as an overflow risk factor; A first spillover risk coefficient is randomly selected, and the deviation between the first spillover risk coefficient and the predetermined risk threshold is calculated based on the predetermined risk threshold to obtain a first risk coefficient deviation, which is added to the plurality of risk coefficient deviations.

11. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operation according to claim 10 is characterized in that: Determining a number of frequency compensation coefficients and a number of personnel identification accuracies according to the plurality of risk coefficient deviation analyses includes: Calculating the ratios of the several risk coefficient deviations to the historical maximum risk coefficient deviation respectively to obtain several risk deviation ratios; configuring a plurality of second frequency compensation coefficients based on the plurality of risk deviation ratios, wherein the second frequency compensation coefficients are positively correlated with the risk deviation ratios; The several risk deviation ratios are multiplied by Q and rounded to the integer to obtain several personnel identification accuracies, where the personnel identification accuracy is the number of selected feature extraction branches.

12. The intelligent monitoring method for unsafe behaviors of personnel in thermal power plant operations according to claim 11 is characterized in that: The risk deviation ratios are multiplied by Q and rounded to the integer to obtain several personnel identification accuracies, including: Multiply the risk deviation ratios by Q and round them to the integer to obtain the initial number of branch selections; Obtaining several environmental interference coefficients of several video coverage devices in the current time zone, respectively calculating the ratios of several environmental interference coefficients to the standard environmental interference coefficient, and obtaining several selected compensation coefficients; The initial branch selection quantities are compensated and rounded according to the selection compensation coefficients to obtain a number of branch selection quantities, which are set as a number of personnel recognition accuracies.

13. Intelligent monitoring system for unsafe behaviors of personnel in thermal power plant operations, characterized by: The steps for implementing the method for intelligently monitoring unsafe behaviors of personnel in thermal power plant operations according to any one of claims 1 to 12 include: The regional risk assessment module is used to segment and assess the risk of the boiler operation area of ​​the target thermal power plant, and determine a number of operation sub-areas and a number of operation risk coefficients; A first monitoring frequency setting module, used to set the monitoring frequency according to the plurality of operation risk factors, generate a first monitoring frequency distribution, and extract images from the video monitoring network according to the first monitoring frequency distribution to obtain a first monitoring image set; A risk feature identification module, used to identify risk features of operators in the first monitoring image set through a feature extraction channel, and obtain multiple real-time risk coefficients based on risk feature identification results; A risk coefficient deviation calculation module is used to mark the operator if the real-time risk coefficient is greater than a predetermined risk threshold, determine a number of marked personnel and a number of video coverage devices, and calculate a number of risk coefficient deviations; The unsafe behavior monitoring module is used to determine a number of frequency compensation coefficients and a number of personnel recognition accuracies based on the deviation analysis of the several risk coefficients, perform image extraction on the several video coverage devices based on the several frequency compensation coefficients, configure a number of optimized feature extraction channels according to the several personnel recognition accuracies, and perform unsafe behavior monitoring of personnel in the boiler operation area.

Images