A new energy operation safety supervision method and system

Abstract

The present application relates to a kind of new energy operation safety supervision method and system, comprising: data acquisition is carried out by multiple types of internet of things sensing terminal, obtain multi-source data;Multi-source data includes: personnel physiological index data, personnel face image data, personnel operation positioning data and field operation video data;Multi-source data is fused and analyzed, and personnel emotion recognition, health assessment and risk identification are carried out based on fusion analysis result;Remote control is carried out according to risk assessment result and personnel operation positioning data.The present application fuses personnel physiological index, facial expression and operation behavior and other multi-source data fusion assessment personnel emotion and health, identifies personnel high risk and violation behavior, and is remotely controlled by multiple means, finally realizes higher risk identification accuracy and whole-process safety control ability.

Description

Technical Field

[0001] This invention relates to the technical field of safety supervision, and more specifically, to a method and system for safety supervision of new energy operations. Background Technology

[0002] The new energy (such as wind power, photovoltaic, pumped storage) power generation industry is characterized by a large number of power stations, wide spatial distribution, and high operational risks. Taking a new energy regional power generation company as an example, it has dozens of power stations under its jurisdiction, distributed in areas of tens of thousands or even hundreds of thousands of square kilometers. High-risk operations include power-related maintenance, lifting and hoisting, and high-altitude operations. These characteristics make it difficult to supervise the safe production of new energy regional companies.

[0003] Currently, safety supervision often employs a combination of on-site personnel supervision and remote video monitoring. Video monitoring involves deploying fixed or mobile surveillance cameras on-site, with manual review of the video feed via back-end software. However, relying solely on video monitoring makes it difficult to identify potential hazards such as changes in workers' emotions and physical condition. Limited by camera deployment costs and monitoring angles, video monitoring cannot cover all work scenarios. Furthermore, back-end supervisors often have to view hundreds or even thousands of monitor feeds in real time, resulting in high workloads. When violations are detected, there is a lack of effective and direct means for remote intervention and control. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for safety supervision of new energy operations, addressing the problems existing in the prior art.

[0005] The technical solution adopted by this invention to solve its technical problem is: constructing a method for safety supervision of new energy operations, comprising: Step S1: Collect data through multiple IoT sensing terminals to obtain multi-source data; the multi-source data includes: personnel physiological index data, personnel facial image data, personnel operation positioning data, and on-site operation video data; Step S2: Perform fusion analysis on the multi-source data, and perform personnel emotion recognition, health assessment and risk identification based on the fusion analysis results; Step S3: Perform remote control based on the risk assessment results and the personnel's work location data.

[0006] In the new energy operation safety supervision method of the present invention, the step of collecting data through multiple types of IoT sensing terminals to obtain multi-source data includes: Before the operation, facial image data of the workers is collected using a camera; Before the operation, the physiological indicators of the workers are collected by a health check machine. During the operation, the physiological indicators of the workers are collected by a smart bracelet to obtain the physiological indicator data of the workers. During the operation, the location of the workers is collected in real time through smart wristbands to obtain the workers' work location data; During the operation, on-site operation videos are collected using drones, network cameras, smart safety helmets, and law enforcement recorders to obtain the on-site operation video data.

[0007] In the new energy operation safety supervision method of the present invention, step S2 specifically includes: Image preprocessing is performed on facial image data to obtain preprocessed image data; An object detection algorithm is used to extract emotion-related visual features from the preprocessed image data; The visual features are used as input to a deep learning model for mapping to obtain corresponding emotion categories, and negative emotions of individuals are identified based on these emotion categories.

[0008] In the new energy operation safety supervision method of the present invention, step S2 specifically includes: Call upon the occupational health risk assessment model; The occupational health risk assessment model integrates pre-job physiological data and in-job physiological data to comprehensively assess an individual's health status.

[0009] In the new energy operation safety supervision method of the present invention, the occupational health risk assessment model includes: a single indicator analysis module and a multi-indicator fusion assessment module; The single-index analysis module is used for: Single-indicator evaluation is performed based on the personnel's physiological index data to obtain the evaluation result of a single indicator. Based on the evaluation results of each individual indicator, output the personnel health risk assessment results one by one; The multi-indicator fusion evaluation module is used for: Obtain a risk assessment indicator system that integrates multiple indicators for evaluation; Based on the personnel physiological index data and the risk assessment index system, and combined with the risk assessment formula, a multi-indicator integrated assessment score is obtained. The health risk assessment results for individuals are output based on the multi-indicator fusion assessment score.

[0010] In the new energy operation safety supervision method of the present invention, step S2 specifically includes: Based on the on-site operation video data, a cross-domain small sample anomaly detection algorithm is used to identify violations and obtain the violation identification results.

[0011] In the new energy operation safety supervision method of the present invention, step S3 specifically includes: For individuals identified as exhibiting negative emotions, the system automatically sends an emotional risk warning message to the corresponding management terminal. For individuals assessed as high-risk, the system automatically sends health risk alerts to the corresponding management terminal. For personnel identified as violating regulations, a violation notice will be sent to the on-site management for on-site control.

[0012] This invention also provides a new energy operation safety monitoring system, comprising: Multiple types of IoT sensing terminals are used to collect data and obtain multi-source data; the multi-source data includes: personnel physiological index data, personnel facial image data, personnel operation positioning data, and on-site operation video data; The assessment module is used to perform fusion analysis on the multi-source data and to perform personnel emotion recognition, health assessment and risk identification based on the fusion analysis results; The control module is used for remote control based on the risk assessment results and the personnel's work location data.

[0013] In the new energy operation safety supervision system described in this invention, the various types of IoT sensing terminals include: The image acquisition module is used to collect facial image data of the workers before the operation. The physiological indicator acquisition module is used to collect physiological indicator data of workers before and during the operation, and to obtain the physiological indicator data of the workers. The positioning module is used to collect the location of the workers in real time during the operation and obtain the workers' positioning data. The on-site video acquisition module is used to acquire on-site operation videos during the operation process and obtain the on-site operation video data.

[0014] In the new energy operation safety supervision system described in this invention, the evaluation module includes: A personnel emotion evaluation unit is used to assess emotions based on the personnel's facial image data and identify negative emotions. The personnel health assessment unit is used to comprehensively assess an individual's health status based on the personnel's physiological indicator data; The operation violation identification unit is used to identify violations based on the on-site operation video data using a cross-domain small sample anomaly detection algorithm, and obtain the violation identification result.

[0015] The new energy operation safety supervision method and system of this invention have the following beneficial effects: They include: data collection through multiple types of IoT sensing terminals to obtain multi-source data; the multi-source data includes: personnel physiological indicator data, personnel facial image data, personnel operation positioning data, and on-site operation video data; the multi-source data is fused and analyzed, and personnel emotion recognition, health assessment, and risk identification are performed based on the fusion analysis results; remote control is carried out based on the risk assessment results and personnel operation positioning data. This invention integrates multi-source data such as personnel physiological indicators, facial expressions, and work behavior to assess personnel emotions and health, identify high-risk personnel and violations, and remotely control them through various means, ultimately achieving a high accuracy rate in risk identification and full-process safety control capabilities. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart illustrating the new energy operation safety supervision method provided by the present invention; Figure 2 This is a diagram of the IoT sensing terminal data acquisition and transmission architecture provided by the present invention; Figure 3 This is a schematic diagram of the few-shot learning algorithm based on domain correction adaptation provided by the present invention; Figure 4 This is a logic block diagram of the new energy operation safety supervision system provided by the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] To address the existing problems in the safety supervision of new energy power plants, this invention provides a method and system for the safety supervision of new energy operations. This method and system is based on IoT sensing and multi-source data fusion. It integrates multi-source data such as video, electrical signals, and positioning, and assesses personnel emotions and health to identify high-risk and illegal behaviors. It also enables remote control through various means, ultimately achieving a high accuracy rate in risk identification and full-process safety management capabilities.

[0019] refer to Figure 1 In a preferred embodiment, the new energy operation safety supervision method includes steps S1, S2 and S3.

[0020] Step S1: Collect data through multiple IoT sensing terminals to obtain multi-source data; the multi-source data includes: personnel physiological index data, personnel facial image data, personnel operation positioning data, and on-site operation video data.

[0021] In some embodiments, data is collected through multiple types of IoT sensing terminals to obtain multi-source data, including: collecting facial image data of workers through cameras before the operation; collecting pre-operation physiological index data of workers through health check machines before the operation, collecting in-operation physiological index data of workers through smart bracelets during the operation, obtaining personnel physiological index data; collecting the real-time location of workers through smart bracelets during the operation, obtaining personnel operation location data; and collecting on-site operation video through drones, network cameras, smart safety helmets, and law enforcement recorders during the operation, obtaining on-site operation video data.

[0022] Specifically, various IoT sensing terminals are used to collect personnel physiological data, facial image data, personnel work positioning data, and on-site work video data. The collection and transmission of these various data types include... Figure 2 As shown.

[0023] 1) Personnel facial image data acquisition: Before the operation, facial images of the workers are acquired by a camera to obtain personnel facial image data, and the data is transmitted to the system platform through the GB / T28181 protocol.

[0024] 2) Personnel Physiological Data Collection: Personnel physiological data includes pre-job physiological data and in-job physiological data. Pre-job physiological data can be collected before work begins using a medical-grade health checkup machine, including but not limited to: blood pressure, blood oxygen saturation, pulse, simplified electrocardiogram, body temperature, and breath alcohol test. The health checkup machine reports the data to the system platform via HTTP / S (Hypertext Transfer Protocol Secure). In-job physiological data can be collected in real-time during work using a smart bracelet, including key indicators such as body temperature, heart rate, blood pressure, and blood oxygen saturation, and communicated with the system platform via 4G.

[0025] 3) Personnel Positioning Data Collection: During the operation, the location of the workers is collected in real time through smart wristbands to obtain personnel positioning data. The equipment uses GPS and Beidou positioning and displays the data on the system platform.

[0026] 4) On-site operation video data acquisition: During the operation, on-site operation video is acquired through aerial drones, fixed network cameras, mobile smart safety helmets, and law enforcement recorders. Video data is transmitted to the system platform via the GB / T28181 protocol.

[0027] Step S2: Perform fusion analysis on the multi-source data, and perform personnel emotion recognition, health assessment and risk identification based on the fusion analysis results.

[0028] In some embodiments, step S2 specifically includes: performing image preprocessing based on facial image data to obtain preprocessed image data; using an object detection algorithm to extract emotion-related visual features from the preprocessed image data; mapping the visual features as input to a deep learning model to obtain corresponding emotion categories; and identifying negative emotions of the person based on the emotion categories.

[0029] Step S2 further includes: invoking the occupational health risk assessment model; and comprehensively assessing an individual's health status by integrating pre-work and in-work physiological indicator data through the occupational health risk assessment model. The occupational health risk assessment model includes a single-indicator analysis module and a multi-indicator fusion assessment module. The single-indicator analysis module is used to: perform single-indicator assessments based on personnel physiological indicator data to obtain individual indicator assessment results; and output personnel health risk assessment results one by one based on the individual indicator assessment results. The multi-indicator fusion assessment module is used to: obtain the risk assessment indicator system for multi-indicator fusion assessment; calculate the multi-indicator fusion assessment score based on personnel physiological indicator data and the risk assessment indicator system, combined with the risk assessment formula; and output the personnel health risk assessment result based on the multi-indicator fusion assessment score result.

[0030] Step S2 further includes: using a cross-domain small sample anomaly detection algorithm to identify violations based on on-site operation video data, and obtaining the violation identification results.

[0031] After collecting multi-source data in step S1, multi-source data fusion analysis can be performed to integrate the data for personnel emotion recognition, health assessment, and violation identification, thereby identifying risks and potential hazards. Details are as follows: 1) Personnel Emotion Recognition: Based on collected facial image data, video recognition algorithms identify negative emotional states such as sadness, anger, disgust, and fear, helping managers to understand employees' mental state in a timely manner. Before performing emotion recognition, a deep learning model (using existing models) is constructed to establish a correspondence between features and emotion labels. For example, geometric features (the relative position and shape changes of facial features, such as a downturned mouth in anger) correspond to the emotion category of anger. By mapping these features to emotion labels using the model, the corresponding emotion category is identified.

[0032] When performing personnel emotion recognition, image preprocessing is first carried out to balance the lighting and convert the grayscale of the images to address the loss of facial details caused by backlighting and low light, while also eliminating interference from the background and makeup. Secondly, an object detection algorithm is used to extract core visual features related to emotions from the image. These core visual features include geometric features (relative positions and shape changes of facial features, such as the downward turn of the mouth when angry) and texture features (texture changes produced by facial muscle movements, such as forehead wrinkles when frowning). These features are then mapped using a deep learning model to obtain the corresponding emotion category. By recognizing personnel emotions before work commences, negative emotions can be identified in a timely manner, automatically triggering control procedures on the system platform.

[0033] 2) Personnel health assessment: The properties and trends of physiological parameters contain rich information about human physiology and pathology. This invention utilizes a constructed occupational health risk assessment model to comprehensively evaluate an individual's health status based on the occupational health risk assessment model, by integrating physiological indicator data (including physiological indicator data before and during work) collected by health examination machines and smart bracelets, for single or multiple indicators of mechanistic physiological parameters.

[0034] In this embodiment of the invention, the occupational health risk assessment model includes a single-indicator analysis module and a multi-indicator fusion assessment module. The single-indicator analysis module performs single-indicator analysis, providing risk warnings for each individual indicator. The multi-indicator fusion assessment module performs a multi-indicator fusion assessment, adding other factors to the single-indicator analysis for further evaluation. Both modules can be used simultaneously.

[0035] Single indicator analysis: It primarily analyzes indicators that directly reflect health status, such as blood pressure (systolic and diastolic), blood oxygen, heart rate, body temperature, and body fat (BMI). Table 1 below is a prompt table for single-indicator analysis of personnel health using the single-indicator analysis module.

[0036] Table 1. Analysis Tips for Individual Health Indicators As shown in Table 1, the single-indicator analysis module can directly output risk warnings based on the collected physiological indicator data (including physiological indicator data before and during the work). For example, when BMI < 18.5, a risk warning of mild abnormality is output; when heart rate is between 101 and 102, a risk warning of mild abnormality is output.

[0037] Multi-indicator integrated evaluation: The multi-indicator fusion assessment module integrates multiple indicators for fusion assessment and provides a risk assessment indicator system, as detailed in Table 2.

[0038] Table 2. Personnel Health Risk Assessment Index System Specifically, as shown in Table 2, the multi-indicator fusion assessment module uses the risk assessment indicator system in Table 2 and the risk assessment formula to comprehensively calculate and obtain the multi-indicator fusion assessment score. Finally, it outputs the personnel health risk assessment result based on the score. The risk assessment formula is as follows: R=0.32C1+0.18C2+0.5C3(1); In the formula, R represents the scoring result.

[0039] Risk warnings are output based on the calculated R value. Among them, (0-20) represents low risk, (21-57) represents medium risk, and >58 represents high risk.

[0040] Furthermore, during the multi-indicator fusion assessment, if the R-value calculated according to Table 2 and the risk assessment formula falls within the three ranges mentioned above, a corresponding risk warning will be output. Specifically, in the risk assessment indicator system of Table 2, important indicators such as exceeding the alcohol test limit and clear occupational contraindications are directly assessed as high risk. For example, if the alcohol test exceeds the limit, it is directly assessed as high risk, and in this case, it is not necessary to calculate the R-value according to the assessment formula.

[0041] 3) Identification of violations of work regulations: Specifically, based on on-site operation videos collected by drones, cameras, law enforcement recorders, smart safety helmets, etc., trained visual recognition algorithms are used to identify violations such as personnel walking around the work area while looking at their mobile phones, using safety belts with the belts hanging low and using them high, and unauthorized entry into crane and hoisting warning areas.

[0042] It should be noted that violations related to safety in new energy production are low-frequency scenarios, resulting in a limited number of real images being collected. Therefore, training sample data for each category is extremely scarce, and there are cross-domain data distribution differences between the data and the pre-training data of large models. To address this issue, this invention employs a cross-domain small-sample anomaly detection algorithm. This algorithm relies on only a small number of labeled samples to achieve cross-domain detection training of the model. In specific applications, the overall solution comprises two main parts: a domain perturbation module and a domain correction module. The domain perturbation module uses both local and global methods to Gaussian-noise the statistics of the input feature map, thereby simulating the generation of images with different styles to increase the generalization ability for images from different domains. The domain correction module uses a domain adapter, taking the perturbation feature map as input, to generate a correction factor, which is used to correct the noisy statistics to the training domain. Figure 3 As shown, the domain adapter consists of several convolutional layers and has the characteristics of small number of parameters and easy training. Under small sample settings, it can effectively reduce the risk of overfitting. Figure 3The perturbation module in P is the domain perturbation module, and the correction module is the domain correction module. local : Local style perturbation probability; P global : Global style perturbation probability; AdaIn (Adaptive Instance Normalization): Adaptive instance normalization; μ, σ: Feature mean and feature standard deviation; μ a σ a Mean of the average style, standard deviation of the average style; μ rect σr ect : Corrected mean, corrected standard deviation; p(x) normal distribution sampling probability.

[0043] Domain feature perturbation based on statistical features (i.e., domain perturbation module): In low-to-mid-level convolutional output feature maps, the statistical values ​​along the channel dimension can characterize the latent style of the image. Therefore, to allow the model to see as many style images as possible and effectively reduce the domain gap between different domains, this invention proposes to perturb the statistical values ​​of low-to-mid-level feature maps by adding Gaussian noise within a certain range to simulate different images. Specifically, firstly, the feature maps of the convolutional network are extracted, and the statistical mean and variance vectors along the channel dimension of the feature maps are calculated respectively. Two Gaussian noise vectors are then randomly generated accordingly. The noise is added to the statistical mean and variance vectors of the image, thus completing the perturbation of the image style. The specific mathematical expression is as follows: (2); in, It is the feature map of the original image; These are the height and width of the feature map, respectively. These are the statistical mean and variance of the feature map, respectively. These are two randomly generated Gaussian noises; ζ is a very small constant, usually 10. -5 Or 10 -8 This is used to prevent instability of standard numerical values.

[0044] The Adaptive Instance Normalization (AdaIn) method from style transfer is used, with the perturbed style as the target style, to perform style transfer on the original image. This yields perturbed feature maps of different simulated domains, which are then input into the subsequent adapter to train it. The specific mathematical expression is as follows: (3); in, It is the feature map after the perturbation; , : Mean and standard deviation of source domain features; , : The mean and standard deviation of the perturbed domain features.

[0045] The adapter-based domain correction module corrects the perturbed simulated domain feature maps to the distribution of the original domain image, allowing well-trained models to process them. This module consists of two parts.

[0046] (1) Generating correction factors from the domain adapter. The domain adapter consists of three convolutional layers. After the simulated feature map is input into the domain adapter, a temporary feature map with unchanged size and channel dimension is obtained. The mean and variance vector of this temporary feature map in the channel dimension are calculated respectively and used as correction factors.

[0047] (2) The statistical mean and variance of the simulated feature map are added to the two correction factors respectively to obtain the corrected statistical vector. Then, the corrected vector is used as the target style to perform a new AdaIn operation on the simulated feature map to obtain the final corrected feature map. This feature map replaces the original training feature map and is input into the next convolutional layer.

[0048] Step S3: Conduct remote management and control based on the risk assessment results and personnel operation location data.

[0049] In some embodiments, step S3 specifically includes: for personnel identified as having negative emotions, the system automatically sends emotional risk warning information to the corresponding management terminal; for personnel with a high health risk assessment, the system automatically sends health risk warning information to the corresponding management terminal; for personnel identified as having violated regulations, a violation warning is sent to the on-site management for on-site control.

[0050] In practical applications, control measures are automatically triggered for high-risk areas and potential on-site hazards identified in the assessment. These include: 1) For personnel identified as being in a negative emotional state, the system automatically sends an emotional risk warning message (such as an SMS reminder via mobile phone or smart bracelet) to the corresponding management terminal (such as the management terminal of the direct supervisor of their organization) to remind the supervisor to pay attention to the employee's emotional state.

[0051] 2) For personnel assessed as high-risk, the system automatically sends messages to the high-risk personnel and their team leaders to remind them of the health risks. This can be done via SMS reminders sent through mobile phones, smart bracelets, etc.

[0052] 3) For identified violations, the information is first sent to back-end personnel for verification. After verification, management is conducted via voice communication with on-site safety monitoring personnel in the violation area. This can be achieved through on-site execution recorders or smart safety helmets. Alternatively, reminders can be sent to the workers involved in the violations via vibration from their smart wristbands.

[0053] refer to Figure 4 The present invention also provides a new energy operation safety monitoring system.

[0054] Specifically, such as Figure 4 As shown, the new energy operation safety monitoring system includes: The multi-type IoT sensing terminal 401 is used for data collection to obtain multi-source data. The multi-source data includes: personnel physiological index data, personnel facial image data, personnel operation positioning data, and on-site operation video data.

[0055] The multi-type IoT sensing terminal 401 includes: an image acquisition module for acquiring facial image data of workers before the operation; a physiological indicator acquisition module for acquiring physiological indicator data of workers before and during the operation to obtain personnel physiological indicator data; a positioning module for acquiring the positioning of workers in real time during the operation to obtain personnel operation positioning data; and a field video acquisition module for acquiring field operation video during the operation to obtain field operation video data.

[0056] Specifically, such as Figure 2 As shown, the image acquisition module may include: a camera; the positioning module may be a smart bracelet; the physiological indicator acquisition module may include: a health checkup machine and a smart bracelet. The on-site video acquisition module may include: an aerial drone, a fixed network camera, a mobile smart safety helmet, and a law enforcement recorder.

[0057] The assessment module 402 is used to perform fusion analysis on multi-source data and to perform human emotion recognition, health assessment and risk identification based on the fusion analysis results.

[0058] In some embodiments, the evaluation module 402 includes: a personnel emotion evaluation unit, used to evaluate emotions based on personnel facial image data and identify negative emotions of personnel; a personnel health evaluation unit, used to comprehensively evaluate the health status of individuals based on personnel physiological indicator data; and a work violation identification unit, used to identify violations based on on-site work video data using a cross-domain small sample anomaly detection algorithm and obtain violation identification results.

[0059] The control module 403 is used for remote control based on risk assessment results and personnel work location data.

[0060] The control module 403 includes: vibration reminder unit (such as smart bracelet, mobile phone, etc.), voice call reminder unit (such as execution recorder, smart safety helmet, etc.), and SMS reminder unit (such as smart bracelet, mobile phone, etc.).

[0061] In practical applications, for personnel identified as being in a negative emotional state, the system automatically sends emotional risk alerts (such as SMS reminders via mobile phones or smart bracelets) to the corresponding management terminal (e.g., the management terminal of their direct supervisor) to remind supervisors to pay attention to employees' emotional states. For personnel with high-risk health assessments, the system automatically sends information to the high-risk personnel and their team leaders to remind them of the health risks, such as SMS reminders via mobile phones or smart bracelets. For identified work violations, the information is first sent to back-end personnel for verification. After verification by back-end personnel, management is conducted via voice communication with on-site safety monitoring personnel in the violation area. This can be achieved through voice communication using on-site execution recorders or smart safety helmets. Alternatively, reminders can be sent to the workers involved in the violations, with vibration alerts sent to their smart bracelets.

[0062] Specifically, the specific coordination and operation process between the various units in the new energy operation safety supervision system can be referred to the above-mentioned new energy operation safety supervision method, and will not be repeated here.

[0063] In response to the characteristics and challenges of the new energy regional company, such as the wide distribution of its sites, numerous high-risk operations, and high personnel pressure, we made full use of existing on-site monitoring videos, smart bracelets, and other sensing devices. Based on the analysis and evaluation needs, we appropriately added necessary IoT sensing devices and integrated multi-source data such as personnel physiological indicators, facial expressions, and work behaviors to conduct risk assessment, thereby achieving higher risk identification accuracy.

[0064] Meanwhile, this invention enables full-process risk control, timely identification of personnel's negative emotions and high health risks before operations, real-time assessment of personnel's health status and violations during operations, and automatic triggering of control processes after risk identification. It can be remotely controlled in real time through various means such as voice calls, SMS reminders, and wristband vibration, which greatly improves the efficiency of remote centralized supervision of dozens of substations by new energy regional companies, and achieves cost reduction and efficiency improvement while enhancing on-site operational safety.

[0065] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0066] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0067] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0068] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They do not limit the scope of protection of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should fall within the scope of the claims of the present invention.

Claims

1. A method for safety supervision of new energy operations, characterized in that, include: Step S1: Collect data through multiple types of IoT sensing terminals to obtain multi-source data; The multi-source data includes: personnel physiological index data, personnel facial image data, personnel work positioning data, and on-site work video data; Step S2: Perform fusion analysis on the multi-source data, and perform personnel emotion recognition, health assessment and risk identification based on the fusion analysis results; Step S3: Perform remote control based on the risk assessment results and the personnel's work location data.

2. The method for safety supervision of new energy operations according to claim 1, characterized in that, The data collection through multiple types of IoT sensing terminals to obtain multi-source data includes: Before the operation, facial image data of the workers is collected using a camera; Before the operation, the physiological indicators of the workers are collected by a health check machine. During the operation, the physiological indicators of the workers are collected by a smart bracelet to obtain the physiological indicator data of the workers. During the operation, the location of the workers is collected in real time through smart wristbands to obtain the workers' work location data; During the operation, on-site operation videos are collected using drones, network cameras, smart safety helmets, and law enforcement recorders to obtain the on-site operation video data.

3. The method for safety supervision of new energy operations according to claim 1, characterized in that, Step S2 specifically includes: Image preprocessing is performed on facial image data to obtain preprocessed image data; An object detection algorithm is used to extract emotion-related visual features from the preprocessed image data; The visual features are used as input to a deep learning model for mapping to obtain corresponding emotion categories, and negative emotions of individuals are identified based on these emotion categories.

4. The method for safety supervision of new energy operations according to claim 2, characterized in that, Step S2 specifically also includes: Call upon the occupational health risk assessment model; The occupational health risk assessment model integrates pre-job physiological data and in-job physiological data to comprehensively assess an individual's health status.

5. The method for safety supervision of new energy operations according to claim 4, characterized in that, The occupational health risk assessment model includes: a single-indicator analysis module and a multi-indicator fusion assessment module; The single-index analysis module is used for: Single-indicator evaluation is performed based on the personnel's physiological index data to obtain the evaluation result of a single indicator. Based on the evaluation results of each individual indicator, output the personnel health risk assessment results one by one; The multi-indicator fusion evaluation module is used for: Obtain a risk assessment indicator system that integrates multiple indicators for evaluation; Based on the personnel physiological index data and the risk assessment index system, and combined with the risk assessment formula, a multi-indicator integrated assessment score is obtained. The health risk assessment results for individuals are output based on the multi-indicator fusion assessment score.

6. The method for safety supervision of new energy operations according to claim 1, characterized in that, Step S2 specifically includes: Based on the on-site operation video data, a cross-domain small sample anomaly detection algorithm is used to identify violations and obtain the violation identification results.

7. The method for safety supervision of new energy operations according to any one of claims 1-6, characterized in that, Step S3 specifically includes: For individuals identified as exhibiting negative emotions, the system automatically sends an emotional risk warning message to the corresponding management terminal. For individuals assessed as high-risk, the system automatically sends health risk alerts to the corresponding management terminal. For personnel identified as violating regulations, a violation notice will be sent to the on-site management for on-site control.

8. A new energy operation safety monitoring system, characterized in that, include: Multiple types of IoT sensing terminals are used to collect data and obtain multi-source data; The multi-source data includes: personnel physiological index data, personnel facial image data, personnel work positioning data, and on-site work video data; The assessment module is used to perform fusion analysis on the multi-source data and to perform personnel emotion recognition, health assessment and risk identification based on the fusion analysis results; The control module is used for remote control based on the risk assessment results and the personnel's work location data.

9. The new energy operation safety supervision system according to claim 8, characterized in that, The various types of IoT sensing terminals include: The image acquisition module is used to collect facial image data of the workers before the operation. The physiological indicator acquisition module is used to collect physiological indicator data of workers before and during the operation, and to obtain the physiological indicator data of the workers. The positioning module is used to collect the location of the workers in real time during the operation and obtain the workers' positioning data. The on-site video acquisition module is used to acquire on-site operation videos during the operation process and obtain the on-site operation video data.

10. The new energy operation safety supervision system according to claim 8, characterized in that, The evaluation module includes: A personnel emotion evaluation unit is used to assess emotions based on the personnel's facial image data and identify negative emotions. The personnel health assessment unit is used to comprehensively assess an individual's health status based on the personnel's physiological indicator data; The operation violation identification unit is used to identify violations based on the on-site operation video data using a cross-domain small sample anomaly detection algorithm, and obtain the violation identification result.

Images