Intelligent power plant monitoring integrated system and method based on industrial Internet of Things

Abstract

The invention discloses an intelligent power plant monitoring integrated system and method based on the industrial Internet of Things, and relates to the technical field of power plant safety monitoring. According to the intelligent power plant monitoring integration method based on the industrial Internet of Things, image data of operators in each monitoring area in a power plant are obtained and input into a pre-trained operation identification model for identification analysis, so that operation feature vectors are obtained, and operation feature indexes are analyzed; according to the method, structured recognition is carried out on the image of the operator through the convolutional neural network model, the attitude features, the equipment wearing features and the behavior action features are extracted, and the model is combined to recognize confidence scores, so that the recognition accuracy of the operator is improved. A unified operation feature vector is constructed, a preset scoring rule is utilized to perform weighted calculation on each feature, and an operation feature index is output and can directly reflect the risk level of the current personnel operation state.

Description

Technical Field

[0001] The present invention relates to the technical field of power plant safety monitoring technology, and specifically to a smart power plant monitoring integration system and method based on the Industrial Internet of Things. Background Art

[0002] With the continuous improvement of industrial automation and informatization, smart power plants, as the core form of the new generation of energy management and safe production, are gradually developing towards digitalization, networking and intelligence. As high-risk, high-load industrial sites, power plants have complex operating environments and intensive personnel operations. They involve multiple dangerous factors such as high temperature, high pressure, high altitude, strong noise, and flammable and explosive materials, which puts higher requirements on safety management and risk monitoring.

[0003] At present, the safety monitoring systems of power plants mostly rely on traditional cameras for image monitoring, or use wearable devices to realize the positioning and status perception of workers. Although these methods have achieved basic monitoring of the working environment and workers to a certain extent, there are still problems such as single monitoring means, scattered information, isolated data, and untimely response to abnormalities. It is difficult to meet the real-time safety management needs under the complex operating scenarios of power plants. At the same time, the existing system has weak linkage analysis capabilities for personnel work behavior and environmental risk factors, and lacks deep intelligent identification and risk assessment mechanisms, which limits the intelligence level and management efficiency of smart power plant safety monitoring systems.

[0004] Prior art, such as the invention patent application with announcement number CN114070985A, discloses a smart power plant safety monitoring system, which includes a working environment monitoring module. The output end of the working environment monitoring module is connected to a data receiving module via a data line signal, and the other input ends of the data receiving module are connected to a worker safety equipment wearing monitoring module, an environmental noise monitoring module, an environmental dust monitoring module, and a data collection and forwarding module via data line signals. This smart power plant safety monitoring system not only has the function of monitoring the physical health and working environment of workers, but also has the function of recording and monitoring the operating steps of workers, thereby improving the scope and effectiveness of the smart power plant safety monitoring system, improving the safety of workers working in the power plant, avoiding situations that endanger the physical health of workers, and improving the safety of power plant operations.

[0005] Based on the above solution, it is found that the limitations of existing technologies include at least the following problems: most existing smart power plant safety monitoring systems only use cameras or wearable devices to collect images and make basic behavioral judgments, lack deep intelligent understanding of the image data of operators, and find it difficult to accurately extract and structure multi-dimensional operating features such as operating posture, equipment wearing status, and subtle behavioral movements. At the same time, the abnormal judgment logic usually relies on manual settings or retrospective backtracking, and lacks real-time and individual difference adaptability, resulting in difficulty in timely identifying high-risk behaviors and early intervention in potential safety incidents during actual operation, thereby affecting the overall safety response efficiency and the intelligence level of early warning. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides a smart power plant monitoring integration system and method based on the Industrial Internet of Things, which solves the problems in the existing technology of imprecise operator behavior recognition, non-real-time abnormal judgment, and unintelligent early warning response.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A smart power plant monitoring integration method based on industrial Internet of Things, comprising the following steps: obtaining image data of operators in each monitoring area in the power plant, wherein the operator image data includes pixel values of a number of pixel points; inputting the operator image data of each monitoring area in the power plant into a pre-trained operation recognition model for recognition and analysis, obtaining the operation feature vector of the operators in each monitoring area in the power plant, and analyzing the operation feature index of the operators in each monitoring area in the power plant; judging and analyzing the operation feature index of the operators in each monitoring area in the power plant with a preset operation feature evaluation interval, and marking the operators whose operation feature index is outside the preset operation feature evaluation interval as operation abnormalities, and sending an abnormality alarm.

[0008] Furthermore, the operation recognition model is specifically a convolutional neural network model, which includes an input layer, a backbone network layer, a feature fusion layer, and an output detection layer. The operation feature vector includes posture features, equipment wearing features, behavioral action features, and confidence scores.

[0009] Furthermore, the specific steps for obtaining the operation feature vector of the operating personnel in each monitoring area of the power plant are as follows: in the input layer of the convolutional neural network model, the image data of the operating personnel in each monitoring area of the power plant are formatted and tensor converted respectively, and a standard image tensor is output; in the backbone network layer of the convolutional neural network model, the image tensor generated by the input layer is received, and the multi-scale basic visual features are extracted to generate a feature map; in the feature fusion layer of the convolutional neural network model, the multi-layer feature map output by the backbone network layer is received, and multi-scale feature fusion processing is performed; in the output detection layer of the convolutional neural network model, the target category, posture state and behavior confidence of the operating personnel are identified based on the fused feature map, and the operation feature vector is output.

[0010] Furthermore, the convolutional neural network model adopts a two-stage pre-training strategy, including a general feature learning stage and a power plant scenario fine-tuning training stage.

[0011] Furthermore, the specific steps for analyzing the operation characteristic index of the operating personnel in each monitoring area of the power plant are as follows: based on the preset scoring rules, the posture characteristics, equipment wearing characteristics, and behavioral action characteristics in the operation characteristic vector of the operating personnel in each monitoring area of the power plant are scored and analyzed to obtain the posture characteristic score, equipment wearing characteristic score, and behavioral action characteristic score of the operating personnel in each monitoring area of the power plant; the confidence score of the operating personnel in each monitoring area of the power plant is comprehensively analyzed in combination with the corresponding posture characteristic score, equipment wearing characteristic score, and behavioral action characteristic score of the operating personnel to obtain the operation characteristic index of the operating personnel in each monitoring area of the power plant.

[0012] Furthermore, for each monitoring area in the power plant, a wind speed risk analysis is performed in real time, and the specific steps are as follows: continuously obtain the regional real-time wind speed values of each monitoring area in the power plant at several time points, and perform comprehensive analysis to obtain the regional wind speed average of each monitoring area in the power plant; obtain the safe wind speed threshold, wind speed danger threshold, turbulent wind speed index, and environmental impact index of each monitoring area in the power plant, and input them into the preset wind speed risk analysis model together with the regional wind speed average of the corresponding monitoring area for risk analysis to obtain the wind speed risk assessment index of each monitoring area in the power plant; judge and analyze the wind speed risk assessment index of each monitoring area in the power plant together with the corresponding preset wind speed risk assessment interval, and regard the monitoring area whose wind speed risk assessment index is outside the preset wind speed risk assessment interval as a wind speed abnormality monitoring area, and send a wind speed risk alarm.

[0013] Furthermore, the wind speed risk analysis model is specifically as follows: Among them, FxP iis the wind speed risk assessment index of the i-th monitoring area in the power plant, QfJ i is the regional wind speed mean of the ith monitoring area in the power plant, FaQ i is the safe wind speed threshold of the ith monitoring area in the power plant, μ1 is the medium-risk wind speed gain coefficient stored in the database, TzS i is the turbulent wind speed index of the i-th monitoring area in the power plant, μ2 is the medium-risk turbulent wind speed influence coefficient stored in the database, HyX i is the environmental impact index of the i-th monitoring area in the power plant, μ3 is the environmental impact coefficient stored in the database, FwX i is the wind speed danger threshold of the ith monitoring area in the power plant, ω1 is the high-risk wind speed gain coefficient stored in the database, ω2 is the high-risk turbulent wind speed influence coefficient stored in the database, is the high-risk penalty coefficient stored in the database, i=1, 2, 3,…, i0, i0 is the number of monitoring areas.

[0014] Furthermore, for each monitoring area in the power plant, a dust risk analysis is performed in real time, and the specific steps are as follows: continuously obtain the regional real-time dust concentration values of each monitoring area in the power plant at several time points, and perform comprehensive analysis to obtain the regional dust concentration average of each monitoring area in the power plant; obtain the safe dust concentration threshold, dust concentration danger threshold, operation duration, and air quality index of each monitoring area in the power plant, and input them into the preset dust risk analysis model together with the regional dust concentration average of the corresponding monitoring area for risk analysis to obtain the dust risk assessment index of each monitoring area in the power plant; judge and analyze the dust risk assessment index of each monitoring area in the power plant together with the corresponding preset dust risk assessment interval, and regard the monitoring area whose dust risk assessment index is outside the preset dust risk assessment interval as a dust abnormality monitoring area, and send a dust risk alarm.

[0015] Furthermore, for each monitoring area in the power plant, a noise risk analysis is performed in real time, and the specific steps are as follows: continuously obtain the regional real-time noise values of each monitoring area in the power plant at several time points, and perform a comprehensive analysis to obtain the regional noise mean of each monitoring area in the power plant; obtain the regional noise parameter value and the regional noise maximum tolerance value of each monitoring area in the power plant, and perform a comprehensive analysis in combination with the regional noise mean of the corresponding area to obtain the noise risk assessment index of each monitoring area in the power plant; judge and analyze the noise risk assessment index of each monitoring area in the power plant with the corresponding preset noise risk assessment interval, and regard the monitoring area whose noise risk assessment index is outside the preset noise risk assessment interval as a noise abnormality monitoring area, and send a noise risk alarm.

[0016] An integrated smart power plant monitoring system based on the industrial internet of things comprises: an image data acquisition unit for acquiring operator image data for each monitoring area in the power plant, the operator image data comprising pixel values of a plurality of pixels; an identification and analysis unit for inputting the operator image data for each monitoring area in the power plant into a pre-trained operation recognition model for identification and analysis, thereby obtaining an operation feature vector for the operator in each monitoring area in the power plant; an operation feature analysis unit for analyzing the operation feature index of the operator in each monitoring area in the power plant based on the operation feature vector of the operator in each monitoring area in the power plant; a judgment unit for comparing the operation feature index of the operator in each monitoring area in the power plant with a preset operation feature evaluation interval, marking an operator whose operation feature index falls outside the preset operation feature evaluation interval as having an operation abnormality and sending an abnormality alarm; a wind speed risk analysis unit for performing real-time wind speed risk analysis for each monitoring area in the power plant; a dust risk analysis unit for performing real-time dust risk analysis for each monitoring area in the power plant; and a noise risk analysis unit for performing real-time noise risk analysis for each monitoring area in the power plant.

[0017] The present invention has the following beneficial effects:

[0018] (1) This smart power plant monitoring integration method based on the industrial Internet of Things uses a convolutional neural network model to perform structured recognition on the images of operating personnel, extract posture features, such as bending over, squatting, etc., equipment wearing features, such as whether a helmet or seat belt is worn, etc., and behavioral action features, such as using a mobile phone, smoking, etc., and combines the model recognition confidence score to construct a unified operation feature vector. Subsequently, the various features are weighted and calculated using the preset scoring rules to output the operation feature index. This index can directly reflect the risk level of the current operating status of the personnel. Compared with the existing method that relies on image playback or fixed template recognition, the personnel behavior risk index generation can be completed in milliseconds, supporting high-frequency recognition and real-time warning, significantly improving the granularity of abnormal recognition and the real-time response of the system.

[0019] (2) This smart power plant monitoring integration method based on the industrial Internet of Things constructs a zoning risk index model for three typical high-risk environmental factors: wind speed, dust, and noise. For example, the wind speed risk model integrates multiple parameters such as the regional wind speed average, safe wind speed threshold, turbulence index, and environmental impact index, distinguishes the influence mechanism of medium and high risk weights, and forms a dynamic risk response calculation. The dust risk model introduces air quality index and operation time adjustment, and the noise risk comprehensively evaluates the intensity deviation of the parameter value and the maximum tolerance value. Compared with the traditional single threshold alarm method, it has strong logical coupling and physical rationality, and can dynamically adjust the evaluation basis to ensure that the evaluation results are highly consistent with the changes in the on-site environment, significantly improving the monitoring accuracy and risk sensitivity.

[0020] (3) The integrated monitoring method of smart power plants based on the industrial Internet of Things uses the monitoring area as the minimum control unit to perform fusion analysis on image recognition and environmental perception data. Each monitoring area independently collects image frames and data such as wind speed, dust, and noise. The images of the operators are processed by the model to generate feature vectors in real time and calculate the operation feature index. The environmental factors are simultaneously input into the respective risk models to generate environmental indexes. The preset evaluation interval is used to determine whether each index is abnormal. Once it crosses the boundary, the area is automatically marked as a high-risk area and a linkage warning is triggered. Compared with the existing system that is difficult to achieve regional granular risk isolation, this method supports simultaneous monitoring of multiple people and multiple areas and personalized warnings, which greatly improves the system's ability to handle concurrent operations and cross-risk events in multiple areas, ensuring the efficiency, accuracy, and controllability of the monitoring system.

[0021] (4) The smart power plant monitoring integration system based on the industrial Internet of Things is structurally divided according to functional modules, including image data acquisition unit, recognition and analysis unit, operation feature analysis unit, judgment unit and multiple environmental risk analysis units. The functional units communicate with each other through standardized data interfaces, and have good module independence and composability. The system architecture supports deployment and operation in different plant areas or distributed edge computing nodes, and can flexibly configure computing resources according to actual scenarios to achieve closed-loop control from data collection, intelligent identification to early warning response. Compared with traditional centralized, single-function monitoring systems, it is easier to expand and has more efficient operation and maintenance, and is particularly suitable for smart power plant scenarios with large regional spans and complex operation types.

[0022] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a smart power plant monitoring integration method based on the Industrial Internet of Things of the present invention.

[0024] Figure 2This is a flowchart of the specific steps for obtaining the operation feature vectors of operators in each monitoring area of a power plant in an integrated method for smart power plant monitoring based on the industrial Internet of Things of the present invention.

[0025] Figure 3 This is an example diagram of regional real-time wind speed data in a certain monitoring area in the smart power plant monitoring integration method based on the industrial Internet of Things of the present invention.

[0026] Figure 4 This is an example diagram of regional real-time dust concentration data in a monitoring area in an integrated method for smart power plant monitoring based on the industrial Internet of Things of the present invention.

[0027] Figure 5 This is an example diagram of regional real-time noise data of a certain monitoring area in the integrated method for smart power plant monitoring based on the industrial Internet of Things of the present invention.

[0028] Figure 6 This is a block diagram of a smart power plant monitoring integrated system based on the Industrial Internet of Things in the present invention. DETAILED DESCRIPTION

[0029] See also Figure 1 , an embodiment of the present invention provides a technical solution: a smart power plant monitoring integration method based on the industrial Internet of Things, comprising the following steps: obtaining image data of operators in each monitoring area in the power plant, the operator image data including pixel values of a plurality of pixel points; inputting the operator image data of each monitoring area in the power plant into a pre-trained operation recognition model for recognition analysis, obtaining the operation feature vector of the operators in each monitoring area in the power plant, and analyzing the operation feature index of the operators in each monitoring area in the power plant; judging and analyzing the operation feature index of the operators in each monitoring area in the power plant with a preset operation feature evaluation interval, marking the operators whose operation feature index is outside the preset operation feature evaluation interval as operation abnormalities, and sending an abnormality alarm to relevant staff.

[0030] The job recognition model is specifically a convolutional neural network model, which includes an input layer, a backbone network layer, a feature fusion layer, and an output detection layer. The job feature vector includes posture features, equipment wearing features, behavioral action features, and confidence scores.

[0031] Specifically, if Figure 2As shown in FIG, the specific steps for obtaining the operation feature vector of the operator in each monitoring area of the power plant are as follows: in the input layer of the convolutional neural network model, the image data of the operator in each monitoring area of the power plant are formatted and converted into tensors respectively, and a standard image tensor is output. The input layer includes an image size adjustment module, a pixel normalization module, and a tensor format conversion module. The function of this layer is to receive the image data of the operator in each monitoring area and perform unified processing on the image: scaling the image to the model specified size (such as 416×416 pixels), normalizing the pixel value from the 0255 range to the 01 range, and converting the image format from the standard height×width×channel (H×W ×C) format is converted into the tensor format (C×H×W) required by the model, and the standard image tensor is output for subsequent feature extraction; in the backbone network layer of the convolutional neural network model, the image tensor generated by the input layer is received, and the multi-scale basic visual features are extracted to generate feature maps. The backbone network layer includes multiple groups of convolution units, activation functions (such as ReLU or SiLU), normalization layers (such as BatchNorm) and residual structure modules (such as ResBlock or CSPBlock). This layer is responsible for multi-level feature extraction of the input image tensor, obtaining underlying visual features such as edges, textures, and structures through convolution operations, and using residual connections to retain key words. semantic information, thereby generating multi-scale, multi-semantic feature maps, providing rich feature expressions for target detection; in the feature fusion layer of the convolutional neural network model, the multi-layer feature maps output by the backbone network layer are received and multi-scale feature fusion processing is performed. The feature fusion layer includes a feature pyramid network (FPN) structure and a path aggregation network (PAN) structure. The function of this layer is to receive feature maps of different levels output by the backbone network, fuse and enhance semantic information of different scales, and realize the compatible recognition ability of small and large targets through the top-down and bottom-up path connection mechanism, thereby improving the detection robustness and regional positioning accuracy of complex operation behaviors; in the convolutional neural network In the output detection layer of the model, the target category, posture state and behavior confidence of the operator are identified based on the fused feature map, and the operation feature vector is output. The output detection layer mainly includes a multi-scale prediction branch, a target classification branch, a bounding box regression branch and a behavior confidence calculation unit. Based on the fused feature map, this layer performs position detection, category recognition and posture state judgment on the operator target in the image, and outputs the corresponding recognition category label (such as not wearing a safety helmet, using a mobile phone, etc.), behavioral action features, posture type and corresponding confidence score. Finally, the above recognition results are structured and represented as an operation feature vector as the input for the subsequent operation feature index calculation.

[0032] In this implementation, by constructing a convolutional neural network structure with clear layers and clear functions, efficient and accurate analysis of images of operating personnel is achieved. The input layer unifies and standardizes the image format to ensure data consistency; the backbone network layer extracts visual features in multiple layers to retain key semantic information; the feature fusion layer integrates multi-scale features to improve the model's adaptability to targets of different sizes; the output detection layer realizes target classification, position positioning and behavior confidence calculation, and finally generates an operation feature vector containing multi-dimensional information such as posture, equipment status, and behavioral actions. This method realizes an automated and intelligent process from image input to feature structured output, providing a reliable data basis for subsequent risk assessment and anomaly determination.

[0033] Specifically, the convolutional neural network model adopts a two-stage pre-training strategy, including a general feature learning stage and a power plant scenario fine-tuning training stage. The specific steps are as follows:

[0034] General feature learning stage (first stage pre-training):

[0035] In this stage, public image recognition datasets (such as COCO, OpenImages, and PascalVOC) are used to conduct basic training on the model. The training objectives include person detection, safety equipment recognition, common human movements (such as standing, squatting, raising hands, bending over, etc.) and target classification. By training the model with large-scale multi-scene data, it is equipped with good target perception and general image understanding capabilities, and obtains basic feature extraction capabilities.

[0036] The training loss functions include:

[0037] Classification loss (such as cross entropy);

[0038] Bounding box regression loss (such as IoU or CIoU);

[0039] Confidence loss (e.g., object existence prediction error).

[0040] The training optimizer is SGD or Adam, and the learning rate is dynamically adjusted using cosine annealing or periodic decay strategy.

[0041] Power plant scenario fine-tuning training phase (second phase pre-training):

[0042] After completing general training, the model is fine-tuned using an image dataset collected in a power plant operating environment. This dataset consists of manually annotated images of workers, with annotations including work behavior categories (such as using a mobile phone or smoking), posture states (such as squatting or bending over), safety equipment wearing conditions (such as whether a helmet or seat belt is worn, etc.), and behavior location boxes.

[0043] Through fine-tuning training, the model's general recognition capabilities can be further adapted to the actual characteristics of power plant operation scenarios, thereby improving the model's recognition accuracy and robustness for specific behaviors.

[0044] Fine-tuning uses a smaller learning rate (such as 1×10 -4 ), avoid covering the original features;

[0045] Data enhancement methods (such as brightness perturbation, mirror flipping, and partial occlusion) can be added to improve the model's adaptability to complex scenes.

[0046] After training is completed, the optimal model weights are selected based on the indicators on the validation set (such as mAP, Precision, and Recall), and deployed to the power plant edge computing device or AI analysis server to support real-time analysis and feature extraction of input operator image data.

[0047] In this implementation plan, a two-stage pre-training strategy is used to significantly improve the recognition accuracy and adaptability of the model in actual power plant scenarios. The general feature learning stage uses large-scale public data sets to enable the model to have basic target recognition and action understanding capabilities; while the power plant scenario fine-tuning stage uses labeled local operation image data to conduct targeted training on the model to enhance the model's recognition effect on specific behaviors (such as not wearing a safety helmet, using a mobile phone, etc.). This strategy not only avoids the waste of resources caused by training from scratch, but also effectively improves the robustness and generalization ability of the model in complex and high-risk operating environments, providing a strong guarantee for stable operation and efficient recognition after system deployment.

[0048] Specifically, the specific steps for analyzing the operation characteristic index of the operating personnel in each monitoring area of the power plant are as follows: based on the preset scoring rules, the posture characteristics, equipment wearing characteristics, and behavioral action characteristics in the operation characteristic vector of the operating personnel in each monitoring area of the power plant are scored and analyzed to obtain the posture characteristic score, equipment wearing characteristic score, and behavioral action characteristic score of the operating personnel in each monitoring area of the power plant; the confidence score of the operating personnel in each monitoring area of the power plant is comprehensively analyzed (i.e., weighted analysis) in combination with the corresponding posture characteristic score, equipment wearing characteristic score, and behavioral action characteristic score of the operating personnel to obtain the operation characteristic index of the operating personnel in each monitoring area of the power plant.

[0049] The preset scoring rules include but are not limited to the following examples:

[0050] Posture feature scoring rules:

[0051] Normal standing (upper body upright, joint distribution consistent with standing posture): 0, indicating no risk;

[0052] Bending over (upper body leaning forward angle > 40°): 1, indicating unstable posture and potential occupational disease risks;

[0053] Squatting (knee flexion > 90°, foot distance narrowed): 2, indicating poor balance and risk of falling;

[0054] Prone or low-position work (torso close to the ground, limited space for movement): 3, indicating high risk, with increased risk of fatigue and difficulty breathing;

[0055] Momentary loss of balance (shaking / falling, drastic changes in posture and joints, center of gravity moving out of the normal range): 4, indicating extremely high risk, possible falling or loss of control;

[0056] Equipment wearing feature scoring rules:

[0057] The helmet is worn correctly (the helmet is detected and overlaps with the head position): 0, indicating compliance;

[0058] Not wearing a helmet (no helmet or the helmet is not on the head area): 3, indicating high risk, head exposure;

[0059] Seat belt worn correctly (seat belt detected and covering the torso): 0, indicating compliance;

[0060] Not wearing a safety belt (no safety belt is detected in the high-altitude working area): 4, indicating extremely high risk and violation of high-altitude working regulations;

[0061] The worn equipment is loose or improperly worn (the equipment identification position deviates from the target area): 2, indicating a risk of misuse;

[0062] Behavioral action feature scoring rules:

[0063] No abnormal behavior detected (no dangerous actions, such as mobile phone / smoking / playing around, etc.): 0, indicating normal behavior;

[0064] Using a mobile phone (detecting a person holding a mobile phone): 3, indicating distraction and potential safety hazards;

[0065] Smoking (smoking posture and smoke characteristics are recognized): 4, indicating that it is strictly prohibited and there is a fire risk;

[0066] Looking at the phone while walking (detecting both "walking" and "holding the phone"): 5, indicating a high-level dangerous behavior that is very likely to cause an accident;

[0067] Unauthorized area operation (cross-matching with area information and finding illegal stops / actions): 4, indicating that the operation has crossed the boundary and there is a risk to equipment / personnel.

[0068] In this implementation plan, by constructing a detailed scoring rule system, a quantitative assessment of the behavioral status of operators and risk level judgment are achieved. The system sets clear scoring standards for posture characteristics, equipment wearing characteristics and behavioral action characteristics, and performs weighted calculations based on the confidence score output by the model to generate an operation characteristic index. Compared with the traditional method of relying solely on a single rule or manual judgment, this method has the advantages of strong interpretability, comprehensive evaluation dimensions, and high risk identification granularity. It can accurately reflect the safety status of operators under different working conditions and provide reliable data support for real-time identification and intelligent early warning of abnormal behaviors.

[0069] Specifically, for each monitoring area in the power plant, wind speed risk analysis is performed in real time, and the specific steps are as follows: continuously obtain the regional real-time wind speed values of each monitoring area in the power plant at several time points, and perform comprehensive analysis (i.e., mean analysis) to obtain the regional wind speed mean of each monitoring area in the power plant; obtain the safe wind speed threshold, wind speed danger threshold, turbulent wind speed index, and environmental impact index of each monitoring area in the power plant, and input them into the preset wind speed risk analysis model together with the regional wind speed mean of the corresponding monitoring area for risk analysis to obtain the wind speed risk assessment index of each monitoring area in the power plant; judge and analyze the wind speed risk assessment index of each monitoring area in the power plant together with the corresponding preset wind speed risk assessment interval, and regard the monitoring area whose wind speed risk assessment index is outside the preset wind speed risk assessment interval as a wind speed abnormality monitoring area, and send a wind speed risk alert to relevant staff.

[0070] The specific implementation example of obtaining the regional wind speed average value of a monitoring area in the power plant is as follows. The following parameters are available, including the regional real-time wind speed value (unit: m / s) at five time points. The specific data are shown in Table 1 and Figure 3 As shown:

[0071] Table 1 Examples of regional real-time wind speed data at five time points

[0072]

[0073] By performing mean analysis on the data in Table 1, it is found that the mean wind speed in a certain monitoring area of the power plant is approximately 3.643 m / s.

[0074] Among them, the real-time wind speed value of the area can be measured and obtained through a wind speed sensor.

[0075] The safe wind speed threshold refers to the value below which the operating environment is considered safe and poses no threat to the safety of workers in certain areas of the power plant. Below this threshold, the wind speed usually has no effect on operations, and operations can proceed normally. This value can be determined by analyzing the power plant's historical meteorological data and combining it with the operational requirements of different areas to determine the common wind speed range for the operating environment. Wind speed data can be collected through on-site wind speed sensors or meteorological monitoring stations.

[0076] The wind speed danger threshold refers to the value at which the working environment becomes dangerous when the wind speed reaches or exceeds this value, posing a major threat to the safety of workers. When the wind speed exceeds this threshold, the operation of the power plant will be significantly affected, potentially causing equipment damage and worker safety accidents. This value can be calculated by long-term monitoring of wind speed data and combining it with the climatic conditions around the power plant to determine the standard value of high-risk wind speed. For example, wind speed sensors can be used to collect data in real time, and the wind speed in the power plant area can be counted to find a generally applicable wind speed critical point.

[0077] The specific steps for obtaining the turbulent wind speed index are as follows: obtain the regional wind speed standard deviation of the monitoring area, and perform ratio analysis with the regional wind speed mean to obtain the turbulent wind speed index.

[0078] The specific steps for obtaining the environmental impact index are: obtaining the regional temperature value (which can be measured by a temperature sensor), the regional temperature parameter value, the regional humidity value (which can be measured by a humidity sensor), and the regional humidity parameter value of the monitoring area, and performing comprehensive analysis to obtain the environmental impact index. The calculation formula is as follows: Among them, HyX is the environmental impact index of the monitoring area, QwD is the regional temperature value of the monitoring area, QwC is the regional temperature parameter value of the monitoring area, QsD is the regional humidity value of the monitoring area, QsC is the regional humidity parameter value of the monitoring area, It is the humidity adjustment coefficient stored in the database.

[0079] It should be explained that the humidity adjustment coefficient stored in the database The specific acquisition steps are as follows: It is determined by analyzing the historical temperature and humidity data, equipment operation conditions and environmental changes in different monitoring areas of the power plant. First, these data are collected and processed, and the relationship between humidity changes and environmental factors is analyzed through statistical methods. Then, a humidity adjustment model is established. The humidity adjustment coefficient calculated by the model is stored in the database and updated regularly to ensure the accuracy and timeliness of the risk assessment model.

[0080] The wind speed risk analysis model is as follows: Among them, FxP i is the wind speed risk assessment index of the i-th monitoring area in the power plant, QfJ iis the regional wind speed mean of the ith monitoring area in the power plant, FaQ i is the safe wind speed threshold of the ith monitoring area in the power plant, μ1 is the medium-risk wind speed gain coefficient stored in the database, TzS i is the turbulent wind speed index of the i-th monitoring area in the power plant, μ2 is the medium-risk turbulent wind speed influence coefficient stored in the database, HyX i is the environmental impact index of the i-th monitoring area in the power plant, μ3 is the environmental impact coefficient stored in the database, FwX i is the wind speed danger threshold of the ith monitoring area in the power plant, ω1 is the high-risk wind speed gain coefficient stored in the database, ω2 is the high-risk turbulent wind speed influence coefficient stored in the database, is the high-risk penalty coefficient stored in the database, i=1, 2, 3,…, i0, i0 is the number of monitoring areas.

[0081] It needs to be explained that the specific steps for obtaining the medium-risk wind speed gain coefficient μ1, the medium-risk turbulent wind speed influence coefficient μ2, and the environmental influence coefficient μ3 stored in the database are: by analyzing the historical data of different monitoring areas of the power plant, collecting the change data of environmental parameters such as wind speed, turbulence, humidity, and temperature, and then using regression analysis or multivariate statistical methods, according to the relationship between wind speed and environmental factors, the influence coefficient of the medium-risk stage is calculated. The environmental impact coefficient is determined according to the specific environmental characteristics of each area of the power plant (such as the plant climate), and the corresponding coefficient is obtained through regression analysis of historical monitoring data and environmental factors.

[0082] High-risk wind speed gain coefficient ω1, high-risk turbulent wind speed influence coefficient ω2, high-risk penalty coefficient stored in the database The specific steps for obtaining the data are as follows: by collecting data on wind speed, airflow turbulence and other environmental changes in high-risk areas inside and outside the power plant (such as high-altitude work areas, equipment exposure areas, etc.), and combining them with accident cases or simulation data, analyze the impact of wind speed under high-risk conditions on operational safety, and use nonlinear regression and other methods to calculate the gain coefficient, turbulence influence coefficient and penalty coefficient of wind speed under different risk levels. In this process, the extreme effects of high wind speed and turbulence are taken into account, and the actual operation records and early warning data are used to calibrate these coefficients.

[0083] In this implementation plan, by introducing a multi-dimensional parameter-driven wind speed risk analysis mechanism, the refined perception and dynamic response capabilities of wind speed risks in the power plant operating environment are improved. The system not only collects real-time wind speed values and calculates the average, but also comprehensively introduces a safe wind speed threshold, a wind speed danger threshold, a turbulent wind speed index, and an environmental impact index. The assessment is not only based on the current wind speed level, but also considers the comprehensive impact of wind speed volatility and the combined effects of temperature and humidity on personnel safety. In particular, the medium and high risk gain coefficients and impact coefficients involved in the risk model are modeled and generated based on historical environmental data and accident samples using regression analysis and other means, which enhances the model's adaptability and scientific nature to real scenarios. Compared with the traditional static threshold judgment method, this mechanism can achieve early identification and dynamic warning of unstable meteorological conditions such as high wind speed and strong turbulence, thereby effectively preventing equipment failures and personnel injuries caused by sudden environmental changes, and significantly improving the intelligence and initiative of power plant wind speed risk management.

[0084] Specifically, for each monitoring area in the power plant, dust risk analysis is performed in real time, and the specific steps are as follows: continuously obtain the regional real-time dust concentration values of each monitoring area in the power plant at several time points, and perform comprehensive analysis (i.e., mean analysis) to obtain the regional dust concentration mean of each monitoring area in the power plant; obtain the safe dust concentration threshold, dust concentration danger threshold, operation duration, and air quality index of each monitoring area in the power plant, and input them into the preset dust risk analysis model together with the regional dust concentration mean of the corresponding monitoring area for risk analysis to obtain the dust risk assessment index of each monitoring area in the power plant; judge and analyze the dust risk assessment index of each monitoring area in the power plant against the corresponding preset dust risk assessment interval, and regard the monitoring area whose dust risk assessment index is outside the preset dust risk assessment interval as a dust abnormality monitoring area, and send a dust risk alarm to the relevant staff.

[0085] The specific implementation example of obtaining the average dust concentration of a monitoring area in a power plant is as follows. The following parameters are available, including the real-time dust concentration values of the area at five time points (unit: mg / m 3 ), the specific data are shown in Table 2 and Figure 4 As shown:

[0086] Table 2 Examples of regional real-time dust concentration data at five time points

[0087]

[0088]

[0089] The mean value analysis of the data in Table 2 shows that the mean dust concentration in a certain monitoring area of the power plant is approximately: 1.725 / m 3 .

[0090] Among them, the real-time dust concentration value of the area can be measured and obtained through dust sensors, such as laser dust sensors, light scattering dust sensors, optical sensors, etc., and the dust concentration includes but is not limited to the following examples: PM2.5, PM10 and other particulate matter concentrations.

[0091] The duration of an operation indicates the time a worker works in a contaminated environment. The longer the operation time, the greater the risk of exposure to dust. If there are multiple workers in the area, the duration of the operation is the sum of the working hours of multiple workers. It can be calculated by monitoring the working time records of the workers or based on the actual working time of the operation task.

[0092] The safe dust concentration threshold refers to the threshold below which the dust concentration in a power plant's operating environment will not pose a significant risk to personnel health. This value is usually determined based on industry standards (such as occupational health and safety management regulations) and environmental health research. It can be set by consulting national or local occupational health standards and combining the specific conditions of the power plant area (such as ventilation, working hours, etc.).

[0093] The dust concentration hazard threshold refers to the dust concentration in the power plant operating environment. When the dust concentration exceeds this threshold, the working environment poses a significant threat to the health of personnel. When this concentration is reached or exceeded, the power plant needs to take emergency measures, such as limiting working hours, wearing protective equipment or suspending high-risk operations, to reduce the exposure risk of workers. This threshold is usually related to long-term health hazards (such as respiratory diseases, lung lesions, etc.) and is usually set based on safety regulatory requirements and historical data of the power plant.

[0094] The Air Quality Index is a comprehensive indicator used to measure the degree of air pollution. It is usually calculated by standardizing the concentration values of multiple pollutants (such as PM2.5, PM10, CO, NO2, SO2, O3, etc.) to reflect the air pollution level in the area.

[0095] The dust risk analysis model is as follows: Among them, CxP i is the dust risk assessment index of the ith monitoring area in the power plant, QcJ i is the mean dust concentration of the ith monitoring area in the power plant, CaQ i is the safe dust concentration threshold of the ith monitoring area in the power plant, e is a natural constant, which is 2.71 in this embodiment, θ1 is the medium-risk dust concentration gain coefficient stored in the database, ZcX i is the duration of the operation in the i-th monitoring area of the power plant, θ2 is the influence coefficient of the medium-risk operation duration stored in the database, KqZ iis the air quality index of the i-th monitoring area in the power plant, θ3 is the medium-risk air quality impact coefficient stored in the database, CwX i is the dust concentration hazard threshold of the i-th monitoring area in the power plant, η1 is the high-risk dust concentration gain coefficient stored in the database, η2 is the high-risk operation time impact coefficient stored in the database, η3 is the high-risk air quality impact coefficient stored in the database, i = 1, 2, 3, …, i0, i0 is the number of monitoring areas.

[0096] It should be explained that the specific steps for obtaining the medium-risk dust concentration gain coefficient θ1, medium-risk operation time impact coefficient θ2, and medium-risk air quality impact coefficient θ3 stored in the database are as follows: For medium-risk areas, data collection and analysis are based on the environmental impact under routine operation conditions. By collecting long-term stable dust concentration, operation time and air quality data in the daily operation areas of the power plant (such as distribution rooms, control rooms, etc.), basic statistical analysis methods (such as regression analysis) are used to infer the impact coefficients of these factors on operational safety. Since the risks in these areas are relatively low, they mainly rely on daily monitoring data for modeling, and the obtained coefficients reflect the potential impact on health under normal circumstances.

[0097] The specific steps for obtaining the high-risk dust concentration gain coefficient η1, high-risk operation time impact coefficient η2, and high-risk air quality impact coefficient η3 stored in the database are as follows: For high-risk areas (such as high temperature and high dust, poorly ventilated areas or emergency operation areas), data collection focuses on situations with high concentration and high exposure risks. The dust concentration, operation time and air quality impact coefficients in these areas are obtained by analyzing extreme operating environments (such as high temperature and high dust concentration) and historical accident data. More complex statistical analysis methods (such as multivariate regression analysis, machine learning models, etc.) are used to obtain high-risk coefficients based on actual operation records and emergency response data. Since the risks in these areas are greater, the analysis method is more refined and takes into account the impact of various emergencies on health.

[0098] In this implementation plan, by constructing a multi-factor driven dust risk analysis mechanism, dynamic assessment and intelligent early warning of dust hazards in the working environment of each monitoring area of the power plant are realized. The system not only collects dust concentration values in real time and performs mean analysis, but also comprehensively introduces multiple key parameters such as operation duration, air quality index, safety and danger concentration thresholds, and outputs dust risk assessment index based on the preset model. The model distinguishes between medium-risk and high-risk stages, and introduces corresponding gain coefficients and influence coefficients to ensure that the evaluation logic is hierarchical and precise. These coefficients are obtained by modeling daily environmental monitoring data and historical data of high-risk operation scenarios, taking into account the ability to respond to stable scenarios and sudden risks. Compared with the traditional single-dimensional judgment method based on concentration thresholds, this solution can simultaneously consider the degree of pollution, personnel exposure time and overall air pollution level, improve the accuracy of risk identification and the targeted response, and significantly enhance the intelligence level of power plant dust safety management.

[0099] Specifically, for each monitoring area in the power plant, a noise risk analysis is performed in real time, and the specific steps are as follows: continuously obtain the regional real-time noise values of each monitoring area in the power plant at several time points, and perform a comprehensive analysis (i.e., mean analysis) to obtain the regional noise mean of each monitoring area in the power plant; obtain the regional noise parameter value and the regional noise maximum tolerance value of each monitoring area in the power plant, and perform a comprehensive analysis in combination with the regional noise mean of the corresponding area to obtain the noise risk assessment index of each monitoring area in the power plant; judge and analyze the noise risk assessment index of each monitoring area in the power plant with the corresponding preset noise risk assessment interval, and regard the monitoring area whose noise risk assessment index is outside the preset noise risk assessment interval as a noise abnormality monitoring area, and send a noise risk alert to relevant staff.

[0100] The specific implementation example of obtaining the regional noise mean value of a monitoring area in a power plant is as follows. The following parameters are available, including the regional real-time noise value (unit: dB) at five time points. The specific data are shown in Table 3 and Figure 5 As shown:

[0101] Table 3 Examples of regional real-time noise data at five time points

[0102]

[0103] By performing mean analysis on the data in Table 3, it is found that the mean regional noise value of a certain monitoring area in the power plant is approximately 84.951dB.

[0104] Among them, the real-time noise value of the area can be measured and obtained through noise sensors (such as sound level meters, digital noise sensors, etc.).

[0105] The regional noise parameter value refers to the safe noise level set in the working environment of a power plant, which is usually below 85dB. Below this value, the working environment is considered to have less impact on the health of the workers. It can be adjusted according to the specific needs of the power plant or historical data.

[0106] The maximum regional noise tolerance value refers to the maximum noise intensity that workers can withstand, usually 95dB or higher. If this value is exceeded, the health risks of workers will increase significantly, which may cause serious health problems such as hearing loss. It is usually based on the maximum safe noise value specified in international standards (such as OSHA) and can be adjusted based on the actual operating environment and risk assessment of the power plant.

[0107] The specific formula for calculating the noise risk assessment index for each monitoring area in the power plant is as follows: Among them, ZxP i is the noise risk assessment index of the ith monitoring area in the power plant, ZsJ i is the regional noise mean of the i-th monitoring area in the power plant, ZsC i is the regional noise parameter value of the i-th monitoring area in the power plant, ZsZ i is the maximum regional noise tolerance value of the i-th monitoring area in the power plant, α1 is the noise deviation gain coefficient stored in the database, α2 is the maximum tolerance gain coefficient stored in the database, χ is the noise intensity risk coefficient stored in the database, i = 1, 2, 3, …, i0, i0 is the number of monitoring areas.

[0108] It should be explained that the specific steps for obtaining the noise deviation gain coefficient α1, maximum tolerance gain coefficient α2, and noise intensity risk coefficient χ stored in the database are as follows: the noise deviation gain coefficient is obtained by analyzing the deviation between the noise intensity in each monitoring area of the power plant and the reference value (such as 85dB), and performing regression analysis in combination with historical health data. This coefficient reflects the impact of noise intensity deviating from the reference value on health risks, and it is usually necessary to consider long-term exposure health data for determination; the maximum tolerance gain coefficient is set by studying the sharp impact on the health of workers when the noise intensity approaches or exceeds the maximum tolerance value (such as 95dB). Based on historical accident or health damage records, this coefficient can be obtained through regression analysis. This coefficient usually reflects the sharp increase in health risks when the noise intensity approaches the limit value; the noise intensity risk coefficient is a comprehensive coefficient based on statistical analysis of historical data, reflecting the basic impact of noise intensity on health. It is usually calculated through weighted analysis and regression of health risk data under multiple exposure conditions.

[0109] In this implementation plan, by constructing a noise risk analysis mechanism driven by dynamic parameters and historical health risk data, accurate assessment and intelligent early warning of the noise environment in each monitoring area of the power plant are achieved. The system collects regional noise values in real time and calculates their averages. At the same time, noise parameter values and maximum tolerance values are introduced as reference standards to evaluate the potential threats and extreme risks of current noise to personnel health. The risk assessment model incorporates noise deviation gain coefficients, maximum tolerance gain coefficients and noise intensity risk coefficients, all of which are based on historical data analysis modeling between working environment noise and health damage, fully considering the cumulative damage to hearing and body functions caused by long-term exposure and high-intensity instantaneous noise. Compared with traditional static alarm methods, this method can dynamically reflect the changing trend of noise risks in the area and provide a scientific and quantitative risk index, providing data support for accurate identification and personalized early warning of noise-exceeding areas, effectively improving the scientific nature and execution efficiency of power plant occupational health management.

[0110] See also Figure 6 , an embodiment of the present invention provides a technical solution: a smart power plant monitoring integrated system based on the industrial Internet of Things, comprising: an image data acquisition unit, used to acquire image data of operators in each monitoring area of the power plant, the operator image data including pixel values of a plurality of pixel points; an identification and analysis unit, used to input the image data of operators in each monitoring area of the power plant into a pre-trained operation recognition model for identification and analysis, and obtain the operation feature vector of the operators in each monitoring area of the power plant; an operation feature analysis unit, used to analyze the operation feature vector of the operators in each monitoring area of the power plant based on the operation feature vector of the operators in each monitoring area of the power plant The system comprises an operation characteristic index of the operator in each monitoring area of the power plant, a judgment unit for judging and analyzing the operation characteristic index of the operator in each monitoring area of the power plant with a preset operation characteristic evaluation interval, marking the operator whose operation characteristic index is outside the preset operation characteristic evaluation interval as having an abnormal operation, and sending an abnormality alarm to the relevant staff; a wind speed risk analysis unit for performing real-time wind speed risk analysis on each monitoring area of the power plant; a dust risk analysis unit for performing real-time dust risk analysis on each monitoring area of the power plant; and a noise risk analysis unit for performing real-time noise risk analysis on each monitoring area of the power plant.

[0111] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0112] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A smart power plant monitoring integration method based on industrial Internet of Things, characterized by: The following steps are involved: Acquire operator image data of each monitoring area in the power plant, wherein the operator image data includes pixel values of a plurality of pixel points; The image data of the workers in each monitoring area of the power plant are input into the pre-trained work recognition model for recognition analysis to obtain the work feature vector of the workers in each monitoring area of the power plant, and analyze the work feature index of the workers in each monitoring area of the power plant; The operating characteristic index of the operators in each monitoring area of the power plant is judged and analyzed with the preset operating characteristic evaluation interval, and the operators whose operating characteristic index is outside the preset operating characteristic evaluation interval are marked as operating abnormalities and an abnormality alarm is sent.

2. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 1 is characterized in that: The operation recognition model is specifically a convolutional neural network model, which includes an input layer, a backbone network layer, a feature fusion layer, and an output detection layer. The operation feature vector includes posture features, equipment wearing features, behavioral action features, and confidence scores.

3. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 2 is characterized in that: The specific steps for obtaining the operation feature vector of the operator in each monitoring area of the power plant are as follows: In the input layer of the convolutional neural network model, the operator image data of each monitoring area in the power plant is formatted and converted into tensors, and a standard image tensor is output; In the backbone network layer of the convolutional neural network model, the image tensor generated by the input layer is received and multi-scale basic visual features are extracted to generate feature maps; In the feature fusion layer of the convolutional neural network model, the multi-layer feature maps output by the backbone network layer are received and multi-scale feature fusion processing is performed; In the output detection layer of the convolutional neural network model, the target category, posture state and behavior confidence of the operator are identified based on the fused feature map, and the operation feature vector is output.

4. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 2 is characterized in that: The convolutional neural network model adopts a two-stage pre-training strategy, including a general feature learning stage and a power plant scenario fine-tuning training stage.

5. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 2 is characterized in that: The specific steps for analyzing the operational characteristic index of operators in each monitoring area of the power plant are as follows: Based on the preset scoring rules, the posture features, equipment wearing features, and behavior action features in the operation feature vectors of the operators in each monitoring area of the power plant are scored and analyzed to obtain the posture feature score, equipment wearing feature score, and behavior action feature score of the operators in each monitoring area of the power plant; The confidence scores of the operators in each monitoring area of the power plant are comprehensively analyzed in combination with the corresponding operator's posture feature scores, equipment wearing feature scores, and behavioral action feature scores to obtain the operator's operation feature index in each monitoring area of the power plant.

6. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 1 is characterized in that: For each monitoring area in the power plant, wind speed risk analysis is performed in real time. The specific steps are as follows: Continuously obtain the regional real-time wind speed values of each monitoring area in the power plant at several time points, and conduct comprehensive analysis to obtain the regional average wind speed of each monitoring area in the power plant; Obtain the safe wind speed threshold, wind speed danger threshold, turbulent wind speed index, and environmental impact index for each monitoring area in the power plant, and input them into the preset wind speed risk analysis model together with the regional wind speed average of the corresponding monitoring area for risk analysis to obtain the wind speed risk assessment index for each monitoring area in the power plant; The wind speed risk assessment index of each monitoring area in the power plant is judged and analyzed against the corresponding preset wind speed risk assessment interval, and the monitoring area whose wind speed risk assessment index is outside the preset wind speed risk assessment interval is regarded as a wind speed abnormality monitoring area, and a wind speed risk alert is sent.

7. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 6 is characterized in that: The wind speed risk analysis model is as follows: Among them, FxP i 、QfJ i 、FaQ i 、TzS i 、HyX i 、FwX i They are the wind speed risk assessment index, regional wind speed mean, safe wind speed threshold, turbulent wind speed index, environmental impact index, and wind speed danger threshold of the i-th monitoring area in the power plant, respectively. μ1, μ2, and μ3 are the medium-risk wind speed gain coefficient, medium-risk turbulent wind speed impact coefficient, and environmental impact coefficient stored in the database, respectively. ω1, ω2, and θ are the high-risk wind speed gain coefficient, high-risk turbulent wind speed impact coefficient, and high-risk penalty coefficient stored in the database, respectively. i = 1, 2, 3, …, i0, where i0 is the number of monitoring areas.

8. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 1 is characterized in that: For each monitoring area in the power plant, a dust risk analysis is performed in real time. The specific steps are as follows: Continuously obtain the regional real-time dust concentration values of each monitoring area in the power plant at several time points, and conduct comprehensive analysis to obtain the regional dust concentration mean value of each monitoring area in the power plant; Obtain the safe dust concentration threshold, dust concentration danger threshold, operation duration, and air quality index for each monitoring area in the power plant, and input them into the preset dust risk analysis model together with the regional dust concentration mean of the corresponding monitoring area for risk analysis to obtain the dust risk assessment index for each monitoring area in the power plant; The dust risk assessment index of each monitoring area in the power plant is judged and analyzed against the corresponding preset dust risk assessment interval, and the monitoring area with the dust risk assessment index outside the preset dust risk assessment interval is regarded as a dust abnormality monitoring area, and a dust risk alarm is sent.

9. The smart power plant monitoring integration method based on industrial Internet of Things according to claim 1 is characterized in that: For each monitoring area in the power plant, noise risk analysis is performed in real time. The specific steps are as follows: Continuously obtain the regional real-time noise values of each monitoring area in the power plant at several time points, and conduct comprehensive analysis to obtain the regional noise mean value of each monitoring area in the power plant; Obtain the regional noise parameter value and the regional noise maximum tolerance value of each monitoring area in the power plant, and conduct a comprehensive analysis based on the regional noise mean value of the corresponding area to obtain the noise risk assessment index of each monitoring area in the power plant; The noise risk assessment index of each monitoring area in the power plant is judged and analyzed against the corresponding preset noise risk assessment interval, and the monitoring area whose noise risk assessment index is outside the preset noise risk assessment interval is regarded as a noise abnormality monitoring area, and a noise risk alarm is sent.

10. A smart power plant monitoring integration system based on the industrial Internet of Things, applying the smart power plant monitoring integration method based on the industrial Internet of Things according to any one of claims 1 to 9, characterized in that: include: An image data acquisition unit, configured to acquire operator image data of each monitoring area in the power plant, wherein the operator image data includes pixel values of a plurality of pixel points; An identification and analysis unit is used to input the image data of the operators in each monitoring area of the power plant into a pre-trained operation recognition model for identification and analysis, and obtain an operation feature vector of the operators in each monitoring area of the power plant; an operation characteristic analysis unit, configured to analyze an operation characteristic index of an operator in each monitoring area of the power plant based on an operation characteristic vector of the operator in each monitoring area of the power plant; a judgment unit, configured to judge and analyze the operation characteristic index of the operators in each monitoring area of the power plant against the preset operation characteristic evaluation interval, mark the operators whose operation characteristic index is outside the preset operation characteristic evaluation interval as having abnormal operation, and send an abnormality alarm; Wind speed risk analysis unit, used to perform real-time wind speed risk analysis on each monitoring area in the power plant; Dust risk analysis unit, used to conduct real-time dust risk analysis for each monitoring area in the power plant; The noise risk analysis unit is used to perform real-time noise risk analysis on each monitoring area in the power plant.

Images