Substation operation risk identification method based on multi-view video and high-precision positioning

Abstract

The application relates to a substation operation risk identification method based on multi-view video and high-precision positioning. Video data of a work site is obtained through multiple fixed-view cameras and mobile video acquisition equipment; three-dimensional space coordinates of operators and equipment are obtained in real time by using a high-precision positioning system; then a three-dimensional digital twin model of substation equipment is established, and a dynamic scene is reconstructed based on multi-view video streams to generate a real-time three-dimensional scene of the work site; then a spatiotemporal fusion algorithm is used to fuse the positioning data and the three-dimensional scene to generate a dynamic digital portrait of the operator; a risk assessment algorithm is used to analyze the operator's behavior and position in real time based on preset risk rules, a risk assessment value is calculated, and a feedback mechanism is set to continuously optimize the positioning and scene reconstruction accuracy. The application realizes efficient, accurate and real-time identification and early warning of substation operation risks, effectively improving the safety management level of the work site.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation operation monitoring, and particularly relates to a substation operation risk identification method based on multi-view video and high-precision positioning. BACKGROUND

[0002] In the process of substation operation, the traditional operation risk identification method mainly relies on manual observation and experience judgment. However, manual observation is easily affected by factors such as personnel fatigue and inattention, and has intermittent and subjective problems, and it is difficult to continuously and objectively monitor the operation site. Experience judgment lacks quantitative basis, and it is difficult to accurately assess the risk degree for complex substation environment and operation process. At the same time, some relatively simple monitoring methods, such as single-point displacement sensor or infrared detection, can only provide limited local information and cannot fully reflect the complex situation of the operation site.

[0003] In addition, the traditional method has a monitoring blind area for the equipment shielding area (such as the rear of the transformer, the gap between the switch cabinet, etc.), and it is difficult to effectively identify and warn potential risks. In terms of positioning accuracy, the traditional algorithm mainly analyzes based on two-dimensional images, and it is difficult to accurately quantify the spatial distance between the operation personnel and the live equipment, resulting in inaccurate and timely risk assessment, which cannot meet the high-precision and high-reliability risk identification requirements of substation operation.

[0004] Therefore, it is necessary for those skilled in the art to design a risk identification technology that can effectively identify risks without dead angles through monitoring. SUMMARY

[0005] To solve the above problems, the present application provides a substation operation risk identification method based on multi-view video and high-precision positioning, which overcomes the problems of shielding blind area and low positioning accuracy of traditional monitoring methods, realizes efficient, accurate and real-time identification and warning of substation operation risk, and effectively improves the safety management level of the operation site.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is as follows: a substation operation risk identification method based on multi-view video and high-precision positioning is provided, comprising the following steps:

[0007] Synchronously acquiring multiple operation site videos through multiple fixed-view cameras and mobile video acquisition devices deployed in the substation operation area;

[0008] Determining the two-dimensional coordinates of the operation personnel by wearing a work card with positioning function, and determining the three-dimensional spatial coordinates of the operation personnel and equipment in combination with the distances between the operation personnel and multiple monitoring in the multiple operation site videos;

[0009] A three-dimensional digital twin model of the substation equipment is established, and a dynamic scene is reconstructed based on the multi-path work site video to generate a real-time three-dimensional scene of the work site;

[0010] The three-dimensional spatial coordinates of the workers and equipment are spatiotemporally fused with the real-time three-dimensional scene of the work site by a spatiotemporal fusion algorithm to generate a real-time dynamic digital portrait of the workers;

[0011] Based on preset risk rules including a safety distance threshold, a device exclusion zone range, and a high-risk action mode, a risk assessment value is calculated in real time by a risk assessment algorithm for the behavior and position of the workers, and the risk assessment value is used to determine whether the workers currently have a risky work behavior;

[0012] When the risk assessment value is greater than a preset threshold, it is determined that the workers have a risky work behavior, including that the work behavior is risky or the position is in a dangerous area, a multi-level warning signal is generated and pushed to a monitoring terminal.

[0013] Preferably, the method for constructing the three-dimensional digital twin model of the substation equipment includes accurate modeling of the geometric shape, size, position, and electrical connection relationship of the substation equipment, ensuring consistency with the actual substation equipment.

[0014] Preferably, in the three-dimensional digital twin model of the substation equipment:

[0015] The three-dimensional spatial coordinates of the workers and equipment are converted to a three-dimensional scene coordinate system in which the three-dimensional digital twin model of the substation equipment is located, a rotation matrix and a translation vector from the coordinate system of the positioning system to the scene coordinate system are calculated using known substation equipment position information and coordinate system parameters of the positioning system, and coordinate transformation is performed on each coordinate point in the three-dimensional spatial coordinates.

[0016] Preferably, the spatiotemporal fusion algorithm specifically includes:

[0017] A unified time coordinate system is established, a timestamp is added to each data point in the two-dimensional coordinates and the three-dimensional spatial coordinates, the two-dimensional coordinates are time-interpolated based on the video frame rate of the three-dimensional spatial coordinates, and the two are aligned in the time dimension;

[0018] For each time point, the position of the workers in the three-dimensional scene is calculated according to the transformed coordinates, and a dynamic digital portrait containing the position, posture, and relative position relationship with the surrounding equipment of the workers is generated in combination with the equipment information in the scene.

[0019] More preferably, the Kalman filter algorithm is used to estimate and correct the uncertainty and error in the spatio-temporal fusion process, continuously optimize the position and working state of the workers through prediction and update, improve the fusion accuracy and stability, specifically including:

[0020] According to the working state of the last moment, the working state of the current moment is predicted;

[0021] The dynamic digital portrait is corrected using the actual measurement data of the current moment, thereby obtaining a more accurate fusion result.

[0022] As a preferred embodiment, the implementation steps of the risk assessment algorithm include:

[0023] Risk factor extraction: real-time extraction of working behavior characteristics and position information of workers from the dynamic digital portrait;

[0024] Safety distance assessment: for each worker in the scene, calculate the distance between the worker and the surrounding live equipment, compare the calculated distance with the preset safety distance threshold, if the actual distance is less than the threshold, it is determined that there is a safety distance risk, and the risk degree is quantified according to the degree of distance exceeding the threshold;

[0025] No-go area range judgment: according to the layout of the substation equipment and the safety specification, the equipment no-go area range is pre-marked in the three-dimensional digital twin model, and it is detected in real time whether the worker enters the no-go area, if so, it is determined that there is a risk violation behavior, and the corresponding risk weight is recorded and archived;

[0026] High-risk action recognition: using the pose estimation and action recognition algorithm in video AI technology, the action sequence of the worker is analyzed, and whether there is a high-risk action mode such as climbing and approaching live parts is identified through the trained deep learning model, once the high-risk action is identified, it is determined that there is a risk behavior and the corresponding risk warning is triggered;

[0027] Risk comprehensive assessment: the safety distance risk, no-go area violation risk and high-risk action risk are comprehensively calculated by using the weighted summation method, and the final risk assessment value is obtained, and the weights of different risk factors are dynamically adjusted according to the importance and danger degree of the actual operation on site.

[0028] More preferably, the calculation formula of the risk assessment value is as follows:

[0029]

[0030] Wherein, R represents the risk assessment value, 、 and respectively represent the weights of the safety distance risk, the no-go area violation risk and the high-risk action risk, and satisfy , represents a safety distance risk value, calculated according to the ratio of the actual distance to the safety distance threshold value; represents a forbidden zone violation risk value, taking the value of 1 when the operating personnel enters the forbidden zone, and 0 otherwise; represents a high-risk action risk value, taking the value of 1 when a high-risk action is identified, and 0 otherwise;

[0031] When the risk assessment value is greater than the preset risk threshold value, a warning signal is triggered, and the relevant information of the risk event, including the time, location, personnel and risk type of occurrence, is recorded, and the risk event is fed back to the monitoring terminal, and the management personnel timely take corresponding measures.

[0032] As a preferred, a feedback mechanism is also provided, based on the motion trajectory and posture change information of the operating personnel in the multi-path operating site video, error analysis and compensation are performed on the obtained three-dimensional space coordinates, real-time correction of the positioning data is performed using an adaptive filtering algorithm, and more accurate positioning results are continuously repaired.

[0033] The beneficial effects of the present application are that the present application provides a substation operation risk identification method based on multi-view video and high-precision positioning, which can effectively solve the deficiencies in the traditional technology. First, through the multi-view camera and the mobile video acquisition device, the real-time monitoring of the substation operation site is realized in all directions and without dead angles, effectively eliminating the equipment shielding blind area, and ensuring the integrity of the monitoring information. Secondly, the three-dimensional space coordinates of the operating personnel and the equipment are obtained in real time by using the high-precision positioning system, combined with the three-dimensional digital twin model and the space-time fusion algorithm, the accurate three-dimensional reconstruction and dynamic digital portrait of the operating site are realized, the quantization accuracy of the spatial relationship between personnel and equipment is significantly improved, and the limitations of traditional two-dimensional image analysis in distance judgment are overcome. Thirdly, based on the preset risk rules and risk assessment algorithm, the behavior and position of the operating personnel can be analyzed in real time, the risks such as safety distance violation, forbidden zone intrusion and high-risk action can be quickly and accurately identified, and the multi-level warning function is realized, which provides timely and effective risk prompt for the operating personnel and the management personnel, and enhances the on-site emergency disposal ability. Finally, through the feedback correction mechanism, the positioning data and the scene reconstruction results are continuously optimized, the adaptability and stability of the system to the complex operating environment are improved, and the long-term reliability of the risk identification is ensured. In summary, the present application significantly improves the efficiency, accuracy and intelligent level of the substation operation risk identification, effectively safeguards the safety of the operating personnel and the stable operation of the substation. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is the flow chart of the substation operation risk identification method based on multi-view video and high-precision positioning of the present application. DETAILED DESCRIPTION

[0035] Referring to Figure 1 As shown in the drawings, the application provides a substation operation risk identification method based on multi-view video and high-precision positioning, including the following steps:

[0036] Through the deployment of multiple fixed-view cameras and mobile video acquisition devices in the substation operation area, multiple operation site videos are synchronously acquired;

[0037] In this embodiment, taking a 220kV substation operation area as an example, multiple fixed-view cameras are arranged, respectively installed on the columns around the substation, with a height of about 3 meters, to ensure that different angles of the entire operation area can be covered. At the same time, mobile video acquisition devices with real-time transmission function are provided for the operation personnel, installed on the safety helmets of the operation personnel, so as to acquire the site video from the first perspective of the operation personnel.

[0038] In addition, a high-precision positioning system is installed in the substation, including a positioning base station and a positioning tag. The positioning base station is fixed on multiple known position points of the substation to form a positioning network; the operation personnel wear the positioning tag to send wireless signals to the positioning base station in real time to obtain their own three-dimensional space coordinates.

[0039] The fixed-view cameras and the mobile video acquisition devices are synchronously started to acquire multiple operation site video data at a rate of 30 frames per second, including the operation process of the operation personnel, the running state of the equipment, and the site environment, etc. These video data are transmitted in real time to the data acquisition server on site for storage and preprocessing.

[0040] The high-precision positioning system collects the three-dimensional space coordinate data of the operation personnel and the equipment at a frequency of 10 times per second. The positioning base station receives the signals sent by the positioning tag, calculates and records the coordinate position (x, y, z) of each positioning tag in the three-dimensional space by using the ultra-wideband (UWB) positioning technology, and transmits the data to the data acquisition server.

[0041] The two-dimensional coordinates of the operation personnel are determined by wearing the positioning badge, and the three-dimensional space coordinates of the operation personnel and the equipment are determined by combining the distance between the operation personnel and multiple monitors in the multiple operation site videos;

[0042] A three-dimensional digital twin model of the substation equipment is established, and a dynamic scene reconstruction is performed based on the multiple operation site videos to generate a real-time three-dimensional scene of the operation site;

[0043] Based on the design drawings and equipment parameters of the substation, a three-dimensional digital twin model of the substation equipment is constructed in computer software, accurately describing the geometric shape, size, position, and electrical connection relationship of the equipment, etc., ensuring high consistency with the actual substation equipment. For example, for the main transformer model, its length, width, and height are set to 3.5 meters, 2 meters, and 2.2 meters respectively according to the actual size, accurately simulating its appearance and position.

[0044] Using the multi-view video stream in the data acquisition server, a dynamic scene reconstruction is performed through computer vision algorithms. First, feature extraction and matching are performed on the video frames of multiple views to identify the common feature points in the video. Then, based on these feature points, the internal and external parameters of the camera are calculated using algorithms such as bundle adjustment, and the three-dimensional structure of the scene is recovered. At the same time, the three-dimensional digital twin model is registered and fused with the reconstructed three-dimensional scene to generate a real-time three-dimensional scene of the work site containing equipment and workers. In this process, according to the feature matching and depth information in the video, the position and posture information of the workers in the three-dimensional scene is updated in real time, ensuring high consistency between the three-dimensional scene and the actual work site.

[0045] The three-dimensional spatial coordinates of the workers and equipment are spatio-temporally fused with the real-time three-dimensional scene of the work site through a spatio-temporal fusion algorithm to generate a real-time dynamic digital portrait of the workers;

[0046] The spatio-temporal fusion algorithm specifically includes:

[0047] A unified time coordinate system is established by adding a timestamp to each data point in the two-dimensional coordinates and three-dimensional spatial coordinates, and the two-dimensional coordinates are time-interpolated based on the video frame rate of the three-dimensional spatial coordinates, so that they are aligned in the time dimension.

[0048] For each time point, the position of the worker in the three-dimensional scene is calculated based on the transformed coordinates, and combined with the equipment information in the scene, a dynamic digital portrait containing the position, posture, and relative position relationship with the surrounding equipment of the worker is generated, specifically including:

[0049] Time synchronization: accurate timestamps are added to the three-dimensional spatial coordinate data obtained by the high-precision positioning system and the multi-view video data. Based on the video frame rate, the positioning data is time-interpolated to align the positioning data and video data in the time dimension.

[0050] Coordinate transformation: based on the known camera position and posture information and the coordinate system parameters of the positioning system, the rotation matrix and translation vector from the positioning system coordinate system to the three-dimensional scene coordinate system are calculated. The three-dimensional spatial coordinates in the positioning data are converted to the three-dimensional scene coordinate system through the coordinate transformation formula.

[0051] Feature matching: Extract feature points of equipment and workers from the three-dimensional scene, such as the edges and corners of equipment, and the joint points of workers. At the same time, according to the positioning data, predict the possible position range of the workers in the three-dimensional scene, narrow the search area of feature matching, and improve the matching efficiency.

[0052] Fusion calculation: Fuse the time-aligned and coordinate-transformed positioning data with the three-dimensional scene data. For each time point, determine the accurate position of the worker in the three-dimensional scene according to the transformed coordinates, and combine with the equipment information in the scene to generate a dynamic digital portrait containing the worker's position, posture, and relative position relationship with the surrounding equipment. For example, through the fused data, the distance between the worker and the live equipment, whether the worker's action posture conforms to the safety specification, and other information can be clearly seen.

[0053] And use Kalman filtering algorithm to estimate and correct the uncertainty and error in the spatio-temporal fusion process, continuously optimize the worker's position and working state through prediction and update, improve the fusion accuracy and stability, including:

[0054] According to the working state of the last time, predict the working state of the current time;

[0055] Use the actual measurement data of the current time to correct the dynamic digital portrait, so as to get more accurate fusion results, as follows:

[0056] First, initialize the state vector and covariance matrix of Kalman filtering. The state vector contains the position and velocity estimates of the worker, and the covariance matrix represents the uncertainty of these estimates.

[0057] Prediction step

[0058] Use the state transition model and the state estimate of the last time to predict the state of the current time. The prediction formula is:

[0059]

[0060] Where, is the prior state estimate of the current time, F is the state transition matrix, is the posterior state estimate of the last time.

[0061] At the same time, update the prior estimate covariance matrix:

[0062]

[0063] Where, is the prior estimate covariance matrix, Q is the process noise covariance matrix.

[0064] Update Step: Compare the current time's measurement data (such as three-dimensional coordinates from high-precision positioning systems and pose estimates from video analysis) with the predicted state, and adjust the predicted state using the Kalman gain to obtain a more accurate estimate. Calculate the Kalman gain:

[0065]

[0066] where, is the Kalman gain, H is the observation matrix, and R is the measurement noise covariance matrix.

[0067] Update the state estimate using the Kalman gain as follows:

[0068]

[0069] where, is the posterior state estimate at the current time, is the measurement value at the current time.

[0070] Update the posterior estimate covariance matrix as follows:

[0071]

[0072] where I is the identity matrix.

[0073] Kalman filtering continuously adjusts the state estimate and covariance matrix by iteratively performing the prediction and update steps, reducing estimation errors. In substation operation risk assessment, this process is repeated to ensure the accuracy of the worker's position and pose estimates, enhancing the reliability of risk assessment. For example, when the worker approaches a live device, Kalman filtering can accurately reflect this change in time, allowing the risk assessment system to respond quickly.

[0074] Based on the preset risk rules, including safety distance thresholds, device exclusion zones, and high-risk action patterns, the risk assessment value is calculated in real time through the risk assessment algorithm based on the worker's behavior and position, which is used to determine whether the worker currently has risky behavior;

[0075] The preset risk rules are as follows:

[0076] Extract the worker's behavior characteristics and position information from the real-time three-dimensional scene, including action poses (such as arm stretch angles, leg bend degrees, etc.) and relative positions with devices (such as distances from live devices, whether within device operation ranges, etc.).

[0077] ​​According to the safety specification of the substation, set the safety distance threshold of the live equipment. For example, for a 220kV live equipment, the safety distance threshold is set to 1.8 meters. Calculate the Euclidean distance between the worker and the live equipment in real time. If the actual distance is less than the threshold, it is determined that there is a safety distance risk, and according to the degree of distance exceeding the threshold, the safety distance risk value is calculated according to certain quantitative rules. For example, when the actual distance is 1.5 meters, the safety distance risk value can be calculated as (1.8 - 1.5) / 1.8 = 0.167, and the specific quantitative rules can be determined according to the actual situation.

[0078] In the three-dimensional digital twin model, the equipment exclusion zone range is pre-determined, such as setting a circular exclusion zone with a radius of 1 meter around the high-voltage equipment. Real-time detection of whether the worker enters the exclusion zone, if so, record the violation behavior, and set the exclusion zone violation risk value to 1, otherwise 0.

[0079] Using deep learning-based video AI technology, analyze the action sequence of the worker. Train a convolutional neural network (CNN) model, use a large number of video data annotated with high-risk actions (such as climbing, approaching live parts, etc.) to train the model. In actual application, input the real-time worker video frames into the trained CNN model to identify whether there is a high-risk action pattern. Once a high-risk action is identified, trigger the corresponding risk warning, and set the high-risk action risk value to 1, otherwise 0.

[0080] Risk comprehensive evaluation: set the weights of safety distance risk, exclusion zone violation risk and high-risk action risk as 0.5, 0.3 and 0.2 respectively, the actual weights can be dynamically adjusted according to the importance and danger degree of the field operation. According to the risk evaluation value calculation formula:

[0081]

[0082] Where R represents the risk evaluation value, , and respectively represent the weights of safety distance risk, exclusion zone violation risk and high-risk action risk, satisfying , represents the safety distance risk value, calculated according to the ratio of the actual distance to the safety distance threshold; represents the exclusion zone violation risk value, taking the value of 1 when the worker enters the exclusion zone, otherwise 0; represents the high-risk action risk value, taking the value of 1 when a high-risk action is identified, otherwise 0;

[0083] For example, when the safety distance risk value is 0.167, the forbidden zone violation risk value is 0, and the high-risk action risk value is 1, the risk assessment value is as follows:

[0084] R = 0.5 x 0.167 + 0.3 x 0 + 0.2 x 1 = 0.233 + 0 + 0.2 = 0.433.

[0085] When the risk assessment value is greater than the preset risk threshold, an early warning signal is triggered, and relevant information of the risk event, including the occurrence time, location, personnel, and risk type, is recorded and fed back to the monitoring terminal, so that the management personnel can take appropriate measures in a timely manner.

[0086] In this embodiment, the preset risk threshold is 0.5. When R = 0.433, it is less than the preset risk threshold, and the early warning signal is not triggered; when the calculated risk assessment value exceeds the threshold, the early warning signal is triggered immediately.

[0087] When the risk assessment value is greater than the preset threshold, it is determined that the operation personnel have risk operation behaviors, including that the operation behavior is at risk or the location is in a dangerous area, a multi-level early warning signal is generated and pushed to the monitoring terminal.

[0088] In addition, the early warning signal is divided into multiple levels, and is classified according to the size of the risk assessment value. For example, the risk assessment value between 0.5 and 0.7 is a secondary early warning, prompting the operation personnel to pay attention to safety; above 0.7 is a primary early warning, urgently reminding the operation personnel to stop operation and evacuate the dangerous area. At the same time, the relevant information of the risk event (such as occurrence time, location, personnel, risk type, etc.) is recorded and fed back to the monitoring terminal, so that the management personnel can understand the on-site situation in a timely manner and take appropriate measures. Through analysis of historical risk events, the risk rules and algorithm parameters are continuously optimized to improve the accuracy and reliability of risk assessment.

[0089] In addition, the present application also sets up a feedback mechanism, based on the motion trajectory and posture change information of the operation personnel in the multi-path operation site video, error analysis and compensation are performed on the obtained three-dimensional space coordinates, real-time correction of the positioning data is performed by using an adaptive filtering algorithm, and more accurate positioning results are continuously repaired.

[0090] The specific steps are as follows:

[0091] Video data and three-dimensional scene correction: Compare the multi-task site video data with the real-time three-dimensional scene. Use the texture, edge and other feature information in the video, and adopt the optimization algorithm based on feature matching to correct the geometric structure of the three-dimensional scene. For example, when it is found that the edge of a device in the video deviates from the edge of the model of the device in the three-dimensional scene, according to the feature point matching result in the video, the parameters of the device model in the three-dimensional scene are adjusted to make it more consistent with the shape and position of the actual device, thereby improving the accuracy of scene reconstruction.

[0092] Positioning data correction: Based on the motion trajectory and posture change information of the workers in the multi-view video, the error of the positioning data obtained by the high-precision positioning system is analyzed. The adaptive filtering algorithm is used to estimate the error in the positioning data in real time, and the positioning data is compensated and corrected. For example, when it is found that there is a system deviation in the positioning data in a certain time period, according to the actual motion trajectory of the workers in the video, the coordinate values of the positioning data are adjusted to make the corrected positioning data more accurately reflect the actual position of the workers. By continuously repeating this feedback correction process, the algorithm parameters are optimized to improve the positioning accuracy and scene reconstruction effect, thereby improving the reliability and accuracy of the entire risk identification system.

[0093] The above embodiments only describe the preferred embodiments of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements to the technical solutions of the present application made by ordinary engineering and technical personnel in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A substation operation risk identification method based on multi-view video and high-precision positioning, characterized in that, The method comprises the following steps: Synchronously acquiring multiple operation site videos through multiple fixed-view cameras and mobile video acquisition devices deployed in the operation area of the substation; Determining the two-dimensional coordinates of the operation personnel by having the operation personnel wear a work card with positioning function, and determining the three-dimensional spatial coordinates of the operation personnel and equipment in combination with the distances of the operation personnel in the multiple operation site videos and the multiple monitors; Establishing a three-dimensional digital twin model of the substation equipment, and dynamically reconstructing a scene based on the multiple operation site videos to generate a real-time three-dimensional scene of the operation site; Performing spatio-temporal fusion of the three-dimensional spatial coordinates of the operation personnel and equipment and the real-time three-dimensional scene of the operation site through a spatio-temporal fusion algorithm to generate a real-time dynamic digital portrait of the operation personnel; Based on preset risk rules including a safety distance threshold, a device exclusion zone range and a high-risk action mode, performing real-time calculation on the behavior and position of the operation personnel through a risk assessment algorithm to obtain a risk assessment value, which is used to determine whether the operation personnel currently have a risky operation behavior; When the risk assessment value is greater than a preset threshold, it is determined that the operation personnel have a risky operation behavior, including that the operation behavior is risky or the position is in a dangerous area, and a multi-level early warning signal is generated and pushed to a monitoring terminal; The spatio-temporal fusion algorithm specifically comprises: Establishing a unified time coordinate system, adding a time stamp to each data point in the two-dimensional coordinates and the three-dimensional spatial coordinates, and performing time interpolation on the two-dimensional coordinates based on the video frame rate of the three-dimensional spatial coordinates, so that the two are aligned in the time dimension; For each time point, the position of the operation personnel in the three-dimensional scene is calculated, and the dynamic digital portrait containing the position, posture and relative position relationship with the surrounding equipment of the operation personnel is generated in combination with the equipment information in the scene; A Kalman filtering algorithm is used to estimate and correct the uncertainty and errors in the spatio-temporal fusion process, the position and working state of the operation personnel are continuously optimized through prediction and update, and the fusion accuracy and stability are improved, specifically including: Predicting the working state at the current time according to the working state at the previous time; Using the actual measurement data at the current time to correct the dynamic digital portrait, so as to obtain a more accurate fusion result.

2. The method according to claim 1, wherein, The construction method of the three-dimensional digital twin model of the substation equipment comprises accurate modeling of the geometric shape, size, position and electrical connection relationship of the substation equipment, so as to ensure consistency with the actual substation equipment.

3. The method according to claim 1, wherein, In the three-dimensional digital twin model of the substation equipment: The three-dimensional spatial coordinates of the operation personnel and equipment are converted into the three-dimensional scene coordinate system of the three-dimensional digital twin model of the substation equipment, the rotation matrix and translation vector from the coordinate system of the positioning system to the scene coordinate system are calculated by using the known position information of the substation equipment and the coordinate system parameters of the positioning system, and coordinate transformation is performed on each coordinate point in the three-dimensional spatial coordinates.

4. The method according to claim 1, wherein, The implementation steps of the risk assessment algorithm comprise: Risk factor extraction: real-time extraction of the working behavior characteristics and position information of the operation personnel from the dynamic digital portrait; Safety distance evaluation: For each worker in the scene, calculate the distance between him and the surrounding live equipment, compare the calculated distance with the preset safety distance threshold, if the actual distance is less than the threshold, it is determined that there is a safety distance risk, and the risk degree is quantified according to the degree of distance exceeding the threshold; No-go zone range judgment: According to the layout of the substation equipment and the safety specification, the equipment no-go zone range is pre-determined in the three-dimensional digital twin model, and it is detected in real time whether the worker enters the no-go zone, if so, it is determined that there is a risk violation behavior, which is recorded and given a corresponding risk weight; High-risk action recognition: Using the pose estimation and action recognition algorithms in video AI technology, the action sequence of the worker is analyzed, and whether there is a high-risk action mode such as climbing and approaching live parts is recognized through the trained deep learning model, once the high-risk action is recognized, it is determined that there is a risk behavior and the corresponding risk warning is triggered; Risk comprehensive evaluation: The safety distance risk, no-go zone violation risk and high-risk action risk are comprehensively calculated by using the weighted summation method to obtain the final risk evaluation value, and the weights of different risk factors are dynamically adjusted according to the importance and danger degree of the actual operation on site.

5. The method according to claim 4, wherein, The calculation formula of the risk evaluation value is as follows: wherein R represents a risk evaluation value, , and respectively represent the weight of the safe distance risk, the restricted area violation risk and the high-risk action risk, and satisfy , represents a safe distance risk value, which is calculated according to the ratio of the actual distance to the safe distance threshold value; represents a restricted area violation risk value, which is 1 when the operating personnel enters the restricted area, and 0 otherwise; represents a high-risk action risk value, which is 1 when the high-risk action is identified, and 0 otherwise; When the risk evaluation value is greater than the preset risk threshold, the warning signal is triggered, and the related information of the risk event is recorded, including the time, place, personnel and risk type, and the risk event is fed back to the monitoring terminal, and the management personnel take corresponding measures in time.

6. The method according to claim 1, wherein, A feedback mechanism is also provided, based on the motion trajectory and posture change information of the workers in the multi-path operation site video, the three-dimensional space coordinates obtained are analyzed for error and compensated, the positioning data is corrected in real time by using adaptive filtering algorithm, and more accurate positioning results are obtained by continuous repair.

Images