Substation violation early warning method, device and equipment based on unmanned aerial vehicle inspection

Abstract

The invention provides a substation violation early warning method, device and equipment based on unmanned aerial vehicle inspection, and relates to the technical field of power inspection. The method comprises the following steps: acquiring current environment data and a target work ticket corresponding to a target substation; obtaining violation risk information according to the current environment data and the target work ticket; generating an unmanned aerial vehicle inspection path according to the violation risk information; performing unmanned aerial vehicle inspection based on the unmanned aerial vehicle inspection path to acquire an inspection image; and performing pose recognition on the inspection image, and performing peccancy operation deduction based on the recognized pose to obtain peccancy operation early warning information. According to the invention, the current environment data and the target work ticket are fused to generate the peccancy operation prediction result, and the high-risk peccancy trend is pre-judged. And a targeted unmanned aerial vehicle inspection path is generated based on the peccancy operation prediction result, and invalid inspection is reduced. Pose recognition is carried out on the collected images, illegal operation deduction is further carried out, and the safety consequence and the production influence can be pre-judged.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power inspection, and in particular to a substation illegal operation early warning method, device and equipment based on unmanned aerial vehicle inspection. BACKGROUND

[0002] As the core hub of the power grid, the substation bears the key functions of power transmission and distribution, and its operating environment has the characteristics of high voltage, multiple devices and strong correlation. Illegal operation during operation, even minor deviations, can cause device tripping, regional power outages, and in severe cases, even personnel casualties and systemic power grid failures, which can have a major impact on industrial production, residential life and social stability. Precise and efficient illegal operation identification is a necessary measure to ensure the safe operation of the power grid, reduce personnel risk and reduce economic losses, and is also a core requirement of power safety production control.

[0003] There are two main methods for identifying illegal operations in substations at present. One is the traditional manual inspection combined with video monitoring mode, in which safety officers conduct on-site inspections or review videos through fixed cameras to determine whether there is illegal operation by comparing with the safety requirements of the work ticket. This method relies on human experience and is mainly used in small-scale and low-frequency operation scenarios. The other is the intelligent unmanned aerial vehicle inspection combined with AI identification mode, which collects operation images through unmanned aerial vehicles and uses pre-trained AI models such as target detection and text analysis to match the operation type and area information in the work ticket, achieving automatic identification of illegal operations. Some solutions also incorporate edge computing to improve data processing real-time performance, making it the main choice for medium and large substations.

[0004] However, the existing methods have obvious limitations. First, the response is lagging, whether it is manual or AI identification, it can only be determined after the illegal operation occurs and the images are collected, which cannot predict high-risk illegal operation trends in advance and miss the best intervention opportunity. Second, the identification is single, only whether there is illegal operation can be determined, and safety management lacks foresight. SUMMARY

[0005] The present application provides a substation illegal operation early warning method, device and equipment based on unmanned aerial vehicle inspection to solve the problem of response lag in the current substation illegal operation early warning method and lack of foresight in safety management.

[0006] In a first aspect, the present application provides a substation illegal operation early warning method based on unmanned aerial vehicle inspection, comprising: obtaining current environment data corresponding to a target substation and a target work ticket; obtaining illegal operation risk information corresponding to each operation link in the target work ticket according to the current environment data and the target work ticket; generating an unmanned aerial vehicle inspection path according to the illegal operation risk information corresponding to each operation link in the target work ticket; The unmanned aerial vehicle is used to patrol according to the unmanned aerial vehicle patrol path to collect the patrol image; The pose of the patrol image is recognized, and a violation operation deduction is performed based on the recognized pose to obtain violation operation early warning information.

[0007] In one possible implementation, the violation risk information corresponding to each operation link in the target work ticket includes a violation type existing in each operation link and a risk probability of the violation; and the unmanned aerial vehicle patrol path is generated according to the violation risk information corresponding to each operation link in the target work ticket, including: The detection points are determined according to the violation type existing in each operation link and the risk probability of the violation; The initial priority of each operation link is determined according to the detection points corresponding to each operation link and the time consumption of each operation link in the target work ticket; The initial priority of each operation link is adjusted according to the spatial boundary corresponding to each operation link in the target work ticket to obtain the target priority of each operation link; The unmanned aerial vehicle patrol path is generated according to the spatial boundary corresponding to each operation link and the target priority.

[0008] In one possible implementation, the initial priority of each operation link is adjusted according to the spatial boundary corresponding to each operation link in the target work ticket to obtain the target priority of each operation link, including: The other operation links adjacent to each operation link are determined based on the spatial boundary corresponding to each operation link; Each operation link is sorted according to its corresponding initial priority; The distance corresponding to the operation links adjacent in priority after sorting is determined; The initial priority of each operation link is adjusted based on the distance corresponding to the operation links adjacent in priority after sorting and the other operation links adjacent to each operation link to obtain the target priority of each operation link.

[0009] In one possible implementation, the initial priority of each operation link is adjusted based on the distance corresponding to the operation links adjacent in priority after sorting and the other operation links adjacent to each operation link to obtain the target priority of each operation link, including: The operation links adjacent in priority after sorting are traversed; For each operation link, the actual passing distance between the operation link and the next operation link is calculated; If the actual passing distance is less than or equal to a preset distance threshold, the initial priority of the next operation link corresponding to the operation link is not adjusted; If the actual passing distance is greater than the preset distance threshold, and there is another work link adjacent to the work link, an initial priority level difference between the work link and the other work link is calculated; wherein the other work link does not include the next work link corresponding to the work link; If the initial priority level difference is less than the preset level difference, the initial priority of the other work link whose initial priority level difference is less than the preset level difference is increased by one level, and the initial priority of the next work link corresponding to the work link is decreased by one level; to obtain the target priority of each work link.

[0010] In one possible implementation, according to the spatial boundary and the target priority corresponding to each work link, a UAV inspection path is generated, including: Superimpose the spatial boundary corresponding to each work link on the current environment data to determine the actual accessibility of the detection points included in each work link; According to the target priority corresponding to each work link, the detection points included in each work link are marked; wherein the marking is used to represent whether each detection point is a must-monitor point; Based on the actual accessibility and the marking of the detection points included in each work link, the detection points in each work link are screened to determine the target detection points in each work link; According to the target priority corresponding to each work link, the target detection points in each work link are given priority weights, and a patrol duration prediction value is determined; The target detection points in each work link are first sorted according to the priority weights, and then each target detection point with the same priority weight is sorted and adjusted based on the spatial adjacency relationship to obtain a time sequence; Based on the time sequence, a UAV inspection path is generated.

[0011] In one possible implementation, based on the recognized poses, a violation operation deduction is performed to obtain violation operation warning information, including: Determine whether the number of violation poses included in the recognized poses is greater than 1; If not, the violation poses are matched with the preset rule library to obtain the violation operation type and the violation operation consequence; If yes, based on the violation poses, a violation operation deduction is performed to obtain the violation operation warning information.

[0012] In one possible implementation, based on the violation poses, a violation operation deduction is performed to obtain violation operation warning information, including: Extract the pose attributes of each violation pose; According to the pose attributes of each violation pose, the correlation strength between each violation pose is calculated; Clustering is performed based on the correlation strength between each violation pose to obtain multiple chain violation scenarios; Based on the pose attributes of the violation poses included in each chain violation scenario, the causal probability obtained by causal relationship deduction of the violation poses is superimposed to obtain the violation operation deduction result. Based on the results of the violation operation simulation, a warning message for violation operation is obtained.

[0013] In one possible implementation, based on current environmental data and the target work order, the violation risk information corresponding to each operational step in the target work order is obtained, including: Semantic recognition is performed on the target work order to obtain semantic features; Semantic features are segmented to determine the safety specifications corresponding to each operational step; By using the safety regulations corresponding to each work step as matrix rows and the current environmental data as matrix columns, a risk assessment matrix corresponding to each work step is obtained. Based on the risk assessment matrix corresponding to each work step, calculate the standard risk value corresponding to each safety standard in each work step; Based on the standard risk value corresponding to each safety standard in each work process, calculate the comprehensive risk value corresponding to each work process; Work processes with a comprehensive risk value greater than a preset comprehensive threshold are classified as high-risk work processes, while work processes with a comprehensive risk value not greater than the preset comprehensive threshold are classified as low-risk work processes. For each high-risk operation, safety regulations with a standard risk value greater than a first preset threshold are identified as high-risk regulations to determine their corresponding violation types; and based on the standard risk value of each high-risk regulation, the probability of a violation occurring is determined. For each low-risk operation, safety standards with a standard risk value greater than the second preset threshold are identified as low-risk standards to determine their corresponding violation types. The first preset threshold is greater than the second preset threshold.

[0014] Secondly, embodiments of the present invention provide a substation violation early warning device based on unmanned aerial vehicle (UAV) inspection, comprising: The acquisition module is used to acquire the current environmental data and target work order corresponding to the target substation; The prediction module is used to obtain the violation risk information corresponding to each operation step in the target work order based on the current environmental data and the target work order; The generation module is used to generate drone inspection paths based on the violation risk information corresponding to each operation step in the target work order; The data acquisition module is used to perform drone inspections based on the drone inspection path in order to acquire inspection images. The early warning module is used to perform pose recognition on the inspection images and to deduce the violation operation based on the recognized pose, so as to obtain the early warning information of the violation operation.

[0015] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0016] In this embodiment of the invention, by integrating current environmental data with the target work order to generate violation risk information corresponding to each operational stage in the target work order, high-risk violation trends can be predicted in advance. This overcomes the limitation of traditional methods that only determine violations after they occur, providing accurate basis for early intervention and significantly improving the effectiveness of intervention. Based on the violation risk information corresponding to each operational stage in the target work order, targeted drone inspection paths are generated, focusing inspections on high-risk scenarios and reducing ineffective inspections. Pose recognition is performed on the collected images, and based on the recognized poses, further deductions of violation operations are made. This not only clarifies the violation itself but also predicts its safety consequences and production impact, solving the problem of traditional methods that only determine whether a violation has occurred, and providing forward-looking decision support for safety management. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the implementation of the substation violation early warning method based on drone inspection provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the substation violation early warning device based on drone inspection provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0018] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart illustrating the implementation of a substation violation early warning method based on drone inspection, as provided in an embodiment of the present invention. Figure 1 As shown, the method may include: Step 110: Obtain the current environmental data and target work order corresponding to the target substation.

[0020] Current environmental data is crucial for reflecting the real-time status of the target substation during operations and for assessing its impact on operational safety and the risk of violations. It encompasses two types of information: physical environment and equipment status. Physical environment information reflects the natural conditions of the work area, while equipment status information reflects the real-time operating conditions of the equipment associated with the work. This data needs to be collected in real time and accurately correspond to the work area of ​​the target substation to provide a basis for subsequent analysis of the degree to which safety regulations are affected by environmental interference.

[0021] A target work order is a formal document issued for a specific operation at a target substation. It contains all the execution requirements and constraints of the operation, and clarifies the core contents such as the division of work links, safety specifications for each link, work sequence, and personnel qualification requirements.

[0022] Step 120: Based on the current environmental data and the target work order, obtain the violation risk information corresponding to each operation step in the target work order.

[0023] Each operational step in the target work order serves as the basic unit for carrying out violation risk information. Each operational step corresponds to specific operational content and safety requirements. Violation risk information includes the type of violation and the probability of risk for each operational step. It is determined by combining the interference analysis of current environmental data on safety regulations and the comprehensive risk assessment of the operational steps. This reflects both the impact of the environment on violations and the risk characteristics of the operational steps themselves, providing precise guidance for the subsequent generation of UAV inspection paths.

[0024] Specifically, the target work order is first structured using a text semantic recognition model. This involves extracting a complete sequence of work steps from the work order content, clarifying the operational theme and execution order of each step. Then, for each work step, the corresponding safety regulations are extracted. These regulations cover operational standards, personnel protection requirements, equipment usage guidelines, and other content, establishing a correspondence between work steps and safety regulations.

[0025] Based on the current environmental data and the target work order after semantic recognition, a corresponding risk assessment matrix is ​​constructed.

[0026] Analyze the risk assessment matrix to determine the violation risk information corresponding to each operational step in the target work order.

[0027] Step 130: Generate the drone inspection path based on the violation risk information corresponding to each operation step in the target work order; To ensure that the generated drone inspection path takes into account both risk priority and spatial rationality, ensuring that high-risk links are inspected first, while reducing invalid drone flights by optimizing the order of inspection points, and avoiding equipment obstruction and no-fly zones, thus ensuring the safety and efficiency of the inspection process, this embodiment determines the drone inspection path by combining the violation risk information corresponding to each operation link in the target work order.

[0028] First, based on the violation risk information corresponding to each work step in the target work order, the key monitoring targets for each step can be identified. Then, by referring to the risk probability of each work step and the time taken for each step in the target work order, the priority of each monitoring target can be comprehensively determined. Steps with higher risk probabilities have higher initial priority; if the risk probabilities are similar, steps with shorter time are prioritized to ensure that high-risk steps that are prone to rapid violations are inspected first within a limited time, laying the foundation for subsequent priority adjustments.

[0029] Based on the determined priorities and the spatial boundaries corresponding to each operational stage, the corresponding path generation algorithm is used to generate the UAV inspection path.

[0030] Step 140: Conduct drone inspections based on the drone inspection path to collect inspection images; The preset drone inspection path is imported into the drone control system. The system automatically analyzes the flight nodes, stopping positions and inspection duration in the path, and associates them with the spatial layout information of the target substation to plan the flight altitude and obstacle avoidance parameters for the drone, ensuring that the flight process can avoid equipment obstruction and no-fly zones.

[0031] After taking off from the preset launch point, the drone flies sequentially to the target inspection points of each operational stage, following the inspection path analyzed by the control system. During flight, the navigation system locates the drone's position in real time and compares it with the preset path. If any deviation occurs, the flight direction is automatically adjusted to ensure accurate arrival at each inspection point. Upon arrival at the inspection point, the drone stays for the preset inspection duration while adjusting its flight attitude to ensure that the image acquisition equipment is directly facing the monitored object, guaranteeing that the acquisition angle and distance meet the image clarity requirements.

[0032] During its stay at each inspection point, the drone's onboard image acquisition equipment continuously collects inspection images at a preset frequency. During acquisition, the equipment parameters are automatically adjusted according to ambient lighting conditions to ensure the images clearly show details of the monitored objects, such as equipment operating status and personnel wearing protective gear. The acquired images are transmitted in real-time to the ground data processing terminal via a communication module. If the signal is weak in an area, the images are temporarily stored on the drone's local storage device and retransmitted once the signal is restored, preventing image data loss.

[0033] Step 150: Perform pose recognition on the inspection image, and infer the violation operation based on the recognized pose to obtain the violation operation warning information.

[0034] In this embodiment, the target objects of pose recognition include the body posture of the operator, the operation action, and the operating posture of the equipment. By recognizing these objects, it can be determined whether their poses meet the safety specifications, and poses that may be in violation can be screened out, providing initial analysis objects for subsequent deduction of violations.

[0035] Based on the identified pose, combined with power safety regulations and operational logic, the process of analyzing the possible violations and chain reactions caused by the pose is analyzed. The different situations of a single violation pose and multiple related violation poses are distinguished. Through logical deduction and risk assessment, the types of violations and the possible safety consequences are identified.

[0036] The results of violation simulations are integrated and categorized according to risk level, which is determined by the severity of the potential consequences of the violation. Warning information must include the operational stage where the violation occurred, the specific type of violation, the potential safety consequences and production impact, and the corresponding risk level. Simultaneously, the warning information is linked to the time and location information recorded during drone inspections to form structured warning content. This content is then pushed in real-time to the on-site safety control terminal and the back-end management system via a communication module, ensuring that staff can promptly receive warning information and take targeted intervention measures.

[0037] In summary, the method provided by this invention integrates current environmental data with the target work order to generate violation risk information corresponding to each operational stage in the target work order. This enables proactive prediction of high-risk violation trends, overcoming the limitation of traditional methods that only determine violations after they occur. It provides precise evidence for early intervention, significantly improving intervention effectiveness. Based on the violation risk information corresponding to each operational stage in the target work order, targeted drone inspection paths are generated, focusing inspections on high-risk scenarios and reducing ineffective inspections. Pose recognition is performed on the collected images, and further violation operation deduction is based on the recognized poses. This not only clarifies the violation itself but also predicts its safety consequences and production impact, solving the problem of traditional methods that only determine whether a violation has occurred. This provides forward-looking decision support for safety management.

[0038] In an optional embodiment, step 120, which involves obtaining violation risk information corresponding to each operational stage in the target work order based on the current environmental data and the target work order, may include: Semantic recognition is performed on the target work order to obtain semantic features.

[0039] Semantic features are segmented to determine the safety standards corresponding to each operational step.

[0040] By using the safety regulations corresponding to each operational step as the rows of a matrix and the current environmental data as the columns, a risk assessment matrix corresponding to each operational step is obtained.

[0041] Based on the risk assessment matrix corresponding to each work step, calculate the standard risk value corresponding to each safety standard in each work step.

[0042] Based on the standard risk value corresponding to each safety standard in each work process, calculate the comprehensive risk value corresponding to each work process.

[0043] Work processes with a comprehensive risk value greater than a preset comprehensive threshold are classified as high-risk work processes, while work processes with a comprehensive risk value not greater than the preset comprehensive threshold are classified as low-risk work processes.

[0044] For each high-risk operation, safety standards with a standard risk value greater than a first preset threshold are identified as high-risk standards to determine their corresponding violation types; and based on the standard risk value of each high-risk standard, the probability of a violation occurring is determined.

[0045] For each low-risk operation, safety standards with a standard risk value greater than the second preset threshold are identified as low-risk standards to determine their corresponding violation types.

[0046] The first preset threshold is greater than the second preset threshold.

[0047] This invention is based on the universally accepted safety work procedures in the power industry, clearly defining the specific requirements of safety regulations, the criteria for determining violation types, and the assessment principles for risk consequences. This ensures that the matching of violation types and risk assessments comply with the industry's safety control bottom line. When conducting risk assessments, the invention follows the power industry's risk assessment standards, clarifying the construction method of the risk matrix, the rules for setting weights, and the logic for calculating risk values. This ensures the standardization of the risk assessment process and the objectivity of the results, avoiding biases caused by subjective judgments.

[0048] Accordingly, key information, such as descriptions of work processes and statements of safety requirements, is extracted from the text using a power-specific semantic model and transformed into semantic features that can be used for subsequent analysis. These semantic features include the logical relationships of the work process and the constraints of safety regulations. These features clearly distinguish different work processes and their corresponding safety requirements, ensuring that the subsequent classification results accurately reflect the actual statements on the work order.

[0049] Based on the workflow logic in semantic features, the target work order is divided into multiple consecutive work steps, clarifying the operational theme and execution sequence of each step. For each work step, the safety requirements directly related to it are selected from the semantic features to form a set of safety specifications for that step, ensuring that each work step has clear safety constraints.

[0050] Using the set of safety regulations for each work stage as the rows of a matrix, and categorizing current environmental data by dimensions such as physical environment and equipment status as the columns, a risk assessment matrix is ​​constructed for each work stage. The value of each cell in the matrix is ​​based on power industry risk assessment standards, reflecting the degree of interference of the corresponding environmental data on the implementation of safety regulations. Based on the risk assessment matrix and the importance weights of the safety regulations, the regulatory risk value of each safety regulation under each work stage is calculated. The weight settings must reflect the differences in the safety impact of different types of regulations. For each work stage, the comprehensive risk value of that stage is obtained by integrating the regulatory risk values ​​of all safety regulations under it, comprehensively reflecting the overall violation risk level of the stage.

[0051] The overall risk value of each work step is compared with a preset threshold. Work steps with an overall risk value exceeding the preset threshold are classified as high-risk, while those below are classified as low-risk. This initial assessment of the risk level of each work step provides a grading basis for subsequent standard screening and violation analysis. The overall risk value is set based on power industry safety management standards, ensuring that the risk grading results meet actual safety control requirements.

[0052] For high-risk operational procedures, the standard risk value of their safety regulations is compared with a first preset threshold to identify high-risk regulations whose standard risk value exceeds the threshold. The correspondence between regulations and violation types in the power safety work procedures is then used to match high-risk violation types to the high-risk regulations. Based on the standard risk value of the high-risk regulations and the deterministic nature of interference in the current environmental data, the probability of occurrence for each high-risk violation type is calculated and determined.

[0053] For low-risk work processes, the standard risk value of their safety regulations is compared with a second preset threshold to filter out low-risk regulations whose standard risk value exceeds the threshold. Similarly, by referring to the power safety work regulations, the low-risk violation types corresponding to the low-risk regulations are matched to form violation information for low-risk processes, without the need to calculate the probability of high-risk violations.

[0054] Through the above process, complete violation risk information corresponding to each operation in the target work order is finally obtained, including risk level, violation type and violation probability of high-risk links, providing accurate basis for subsequent drone inspection path planning.

[0055] In an optional embodiment, the violation risk information corresponding to each operational stage in the target work order includes the type of violation and the probability of violation occurring in each operational stage; step 130, generating the UAV inspection path based on the violation risk information corresponding to each operational stage in the target work order, may include: Step 131: Determine the detection points based on the types of violations present in each work process and the probability of violations occurring.

[0056] Step 132: Determine the initial priority of each work step based on the detection points corresponding to each work step and the time consumption of each work step in the target work order.

[0057] Step 133: Adjust the initial priority of each work step according to the spatial boundary corresponding to each work step in the target work order to obtain the target priority of each work step.

[0058] Step 134: Generate the UAV inspection path based on the spatial boundaries and target priorities corresponding to each operation stage.

[0059] For each operational stage, the corresponding violation type is first analyzed to clarify the target of monitoring. For example, violations related to equipment operation should focus on key operating parts of the equipment, while violations related to personnel protection should focus on areas where personnel are active. The density and layout of detection points are adjusted based on the probability of violation occurrence. For stages with high risk, the number of detection points is increased to ensure multi-directional and multi-angle coverage of the monitored targets; for stages with low risk, the number of detection points is appropriately reduced, retaining only core monitoring locations. Simultaneously, detection points must avoid areas obstructed by equipment or with strong interference to ensure that the drone can successfully acquire clear images, providing reliable data for subsequent violation identification.

[0060] The number of detection points for each work step is counted. More detection points indicate a higher monitoring demand for that step, and thus carry greater weight in priority determination. The initial priority is determined by combining the time consumed by each work step in the target work order. Specifically: steps with high risk probability and many detection points are prioritized; if the risk probability and number of detection points are similar across different steps, steps with shorter time consumption are prioritized to avoid delays in the inspection of high-risk, short-time steps due to resource consumption by long-time steps, ensuring that high-risk scenarios are monitored in a timely manner.

[0061] Retrieve the spatial boundaries corresponding to each operational step in the target work order, clarifying the physical scope of each step and other adjacent operational steps. After sorting the steps according to their initial priority, calculate the actual travel distance between adjacent steps in the sort. If the actual travel distance is too large, and there are other steps adjacent to the current step but not immediately adjacent in the sort, calculate the initial priority difference between the current step and that adjacent step. If the priority difference is small, it indicates that the risk levels of the two are similar. In this case, increase the priority of the adjacent step and decrease the priority of the next step in the original sort. Through this adjustment, spatially adjacent steps are more closely connected in the inspection sort, reducing the time and distance costs of UAVs flying across areas.

[0062] By combining the spatial boundaries of each operational stage with target priorities, the inspection sequence of each stage is first determined according to target priority. For each operational stage, based on the distribution of detection points within its spatial boundaries, an inspection sub-path is planned to ensure that detection points are covered in an orderly manner. Then, the sub-paths of each stage are connected in sequence according to priority. Considering the flight characteristics of drones and the layout of substation equipment, the transition routes between stages are optimized, prioritizing wide passages and unobstructed areas for flight, and maintaining a safe distance from high-voltage equipment and buildings. Based on the monitoring complexity of each detection point, a reasonable inspection time is allocated to each detection point, and the dwell time of the drone at the detection point is clearly defined. Ultimately, a complete, efficient, and safe drone inspection path is formed, which covers all detection points while focusing on high-risk stages according to priority.

[0063] In an optional embodiment, step 133, which adjusts the initial priority of each work step according to the spatial boundary corresponding to each work step in the target work order to obtain the target priority of each work step, may include: Based on the spatial boundaries corresponding to each work step, other adjacent work steps are determined.

[0064] Each task step is sorted according to its corresponding initial priority.

[0065] Determine the distances between adjacent work steps of each priority level after sorting.

[0066] Based on the distances between adjacent work steps after sorting, and other work steps adjacent to each work step, the initial priority of each work step is adjusted to obtain the target priority of each work step.

[0067] In this embodiment, the adjacent other operation links refer to other operation links that are physically connected to or close to the current operation link on the spatial boundary. Their determination is based on the spatial range definition of each link, reflecting the distribution and relationship of operation links in the physical space of the substation, and is a key reference for optimizing inspection sequencing and reducing invalid flight.

[0068] To determine the target priority of each work step, the spatial boundary information corresponding to each work step is first retrieved. Combined with the physical layout and equipment distribution map of the target substation, the spatial positional relationship of each step is determined through spatial overlay analysis. If the spatial boundary of one work step overlaps with the spatial boundary of another work step, or if the shortest straight-line distance between the spatial boundaries of the two steps is within a reasonable range, then the other work step is determined as an adjacent work step of the current step, forming a list of adjacent relationships for each step.

[0069] Then, the initial priority assessment results of all work steps are collected, and the work steps are sorted in descending order according to their priority to form an initial priority sequence. During the sorting process, if multiple work steps have the same initial priority, they are further sorted according to the number of corresponding detection points or the work time, so as to ensure that the initial sequence highlights both risk priority and basic execution logic.

[0070] Subsequently, based on the digital model of the substation's spatial layout, the spatial boundary coordinates of two adjacent work processes in the initial priority sequence were extracted. Taking into account actual scenario factors such as the distribution of passageways within the substation and equipment obstructions, the shortest path length that the drone can traverse between the two processes was calculated. This length represents the distance between adjacent work processes after prioritization, providing a quantitative basis for determining whether priority adjustments are necessary.

[0071] Finally, the initial priority sequence is traversed, and the distance between adjacent operational steps in the sorting is checked one by one. If the distance is within a preset reasonable range, it means that the flight connection between the two steps is convenient, and there is no need to adjust the initial priority. If the distance exceeds the reasonable range, and there are other operational steps that are not immediately adjacent in the current step's adjacent relationship list, the initial priority level difference between the current step and that other operational step is calculated. If the level difference is small, it indicates that the risk levels of the two are similar. At this time, the initial priority of the other operational step is increased, while the priority of the next step in the original sorting is decreased, so that spatially adjacent steps are adjacent in the sequence. This adjustment process is repeated until all adjacent steps with high distance intervals are optimized, and finally a target priority that takes into account both risk priority and spatial passage efficiency is formed.

[0072] In an optional embodiment, the initial priority of each task is adjusted based on the distance between adjacent task steps after sorting, and other task steps adjacent to each task step, to obtain the target priority of each task step. This adjustment may include: Traverse the adjacent task steps based on their sorted priorities.

[0073] For each work step, calculate the actual travel distance between that work step and the next work step.

[0074] If the actual travel distance is less than or equal to the preset distance threshold, the initial priority of the next work step corresponding to that work step will not be adjusted.

[0075] If the actual travel distance is greater than the preset distance threshold, and there are other work steps adjacent to this work step, then the initial priority level difference between the adjacent work steps and other work steps is calculated; where, the next work step corresponding to this work step is not included in the other work steps.

[0076] If the initial priority level difference is less than the preset level difference, then the initial priority of other work steps with an initial priority level difference less than the preset level difference will be increased by one level, and the initial priority of the next work step corresponding to that work step will be decreased by one level; so as to obtain the target priority of each work step.

[0077] In this embodiment, the target priority of each work step can be obtained in the following way: Following the initial priority sequence, adjacent work steps are selected sequentially. Starting from the first step in the sequence, each step is processed in turn with the next step until all adjacent steps with the initial priority are traversed, ensuring that each step can accept subsequent distance verification and adjustment judgments.

[0078] For each combination of work processes encountered, based on the current environmental data of the substation, information such as equipment distribution, passageway locations, and no-fly zones within the substation is determined. The drone's flight path is simulated, and the shortest safe passage distance between two processes is calculated, i.e., the actual passage distance. This provides accurate data for subsequent threshold comparisons. For example, a digital layout model of the substation can be used based on the current environmental data, and the actual passage distance can be calculated based on this model.

[0079] The calculated actual travel distance between each operational step and its next operational step is compared with a preset distance threshold. If the actual travel distance is less than or equal to the preset distance threshold, it indicates that the flight connection between the two steps is efficient, and there is no need to adjust the initial priority of the next step in this combination; the original order remains unchanged. If the actual travel distance is greater than the preset distance threshold, the adjacent relationship list of the current operational step is retrieved to check whether there are other operational steps that are not immediately adjacent in the initial sequence. If there are no such other operational steps, the original initial priority remains unchanged; if there are, the initial priority of the other operational step is extracted, and its difference from the initial priority level of the current operational step is calculated.

[0080] The calculated initial priority level difference is compared with the preset level difference. If the level difference is greater than or equal to the preset level difference, it indicates a significant difference in risk levels, making adjustment unsuitable; the original order is maintained. If the level difference is less than the preset level difference, it indicates similar risks; in this case, the initial priority of the other work step is increased by one level, while the initial priority of the next work step in the original order is decreased by one level. After completing a single adjustment, the remaining adjacent combinations of initial priorities are iterated, and the above process is repeated until all combinations are processed, ultimately forming the target priority.

[0081] In an optional embodiment, step 134, which generates the UAV inspection path based on the spatial boundaries and target priorities corresponding to each operational stage, may include: By overlaying the spatial boundaries corresponding to each operational step with the current environmental data, the actual accessibility of the detection points included in each operational step can be determined.

[0082] Based on the target priority corresponding to each operation step, the detection points included in each operation step are marked; the marking is used to indicate whether each detection point is a mandatory monitoring point.

[0083] Based on the actual accessibility and marking of the detection points included in each operation stage, the detection points in each operation stage are screened to determine the target detection points in each operation stage.

[0084] Based on the target priority corresponding to each operation stage, priority weights are assigned to the target detection points in each operation stage, and the predicted inspection time value is determined.

[0085] The target detection points in each operation are first sorted according to their priority weights, and then the target detection points with the same priority weights are sorted and adjusted based on their spatial adjacency to obtain a time sequence.

[0086] Based on the time sequence, an inspection path for the drone is generated.

[0087] In this embodiment, the property that the detection point can be successfully reached by the drone and image acquisition is affected by both the spatial boundary and the current environmental data. Factors such as equipment obstruction, strong electromagnetic interference, and terrain obstacles need to be comprehensively considered. It is necessary to ensure that the detection point has the conditions for inspection in the actual scene. This embodiment uses actual reachability to characterize this condition.

[0088] In other words, spatial boundary data of each operational stage is retrieved and current environmental data is overlaid for analysis. Combining spatial information such as substation equipment layout and terrain features, as well as interference factors in the environmental data, each detection point is assessed for issues such as obstruction, interference, or terrain limitations. If a detection point has no obvious obstacles and the drone can safely fly there, it is determined to be reachable; if there are unavoidable obstacles or strong interference preventing the drone from acquiring images normally, it is determined to be inaccessible.

[0089] Based on the target priority of each operation stage, the detection points within each stage are marked. For operation stages with high target priority, most of the detection points are marked as mandatory monitoring points to ensure that core monitoring needs in high-risk scenarios are not overlooked. For operation stages with low target priority, some non-core detection points can be marked as non-mandatory monitoring points, flexibly adapting to the allocation of inspection resources and clarifying the monitoring priority attribute of each detection point. Operation stages with two or more detection points are designated as mandatory monitoring points.

[0090] Based on the actual accessibility and marking results of the inspection points, inspection points are screened. Inspection points that are accessible and marked as mandatory are prioritized for retention; for accessible inspection points marked as non-mandatory, some key points are selectively retained based on overall inspection resources and time budget; inaccessible inspection points are directly eliminated, regardless of their marking attributes, to avoid ineffective path planning. Through this screening, the target inspection points for each operational stage are finally determined.

[0091] Based on the objective priority of each operational stage, each target detection point is assigned a corresponding priority weight. The higher the objective priority of the stage, the greater the weight of the corresponding target detection point, ensuring that high-risk detection points have an advantage in the ranking. At the same time, considering factors such as the complexity of the monitoring content and the image acquisition accuracy requirements of the target detection points, a predicted inspection time is given for each target detection point. Detection points with complex monitoring objects are allocated a longer dwell time to ensure the quality of data acquisition.

[0092] First, all target detection points are sorted in descending order according to their priority weight, with detection points with higher weights listed first to ensure that high-risk points are inspected first. For target detection points with the same priority weight, the order is adjusted based on spatial adjacency, grouping detection points that are spatially close together to reduce the flight distance and time consumed by the UAV between different detection points, ultimately forming a time sequence that balances risk priority and spatial efficiency.

[0093] Based on time-series data and considering the flight characteristics of UAVs and the spatial layout of substations, a path optimization algorithm is employed to plan flight routes between target inspection points. Route planning must ensure that the UAVs maintain a safe distance from high-voltage equipment and buildings during flight between inspection points, prioritizing wide passageways and unobstructed areas to avoid detours. The predicted inspection time for each inspection point is incorporated into the path, clarifying the UAV's dwell time at each point. Simultaneously, the locations of takeoff and landing points are optimized to ensure smooth path transitions, ultimately forming a complete, efficient, and safe UAV inspection path.

[0094] In an optional embodiment, step 150, which involves deducing a violation operation based on the identified pose to obtain a violation operation warning, may include: Determine whether the number of illegal poses included in the identified poses is greater than 1.

[0095] If not, the violation pose will be matched with the preset rule base to obtain the violation type and the violation consequences.

[0096] If so, then based on the violation posture, a violation operation simulation is performed to obtain a violation operation warning.

[0097] In this embodiment, the violations mainly include two categories: violations related to operators and violations related to equipment, covering the core abnormal states that may lead to violations during substation operations. Among them, violations related to operators include situations where personnel's operating actions, body postures, and the wearing of protective equipment do not comply with safety regulations. For example, the distance between limbs and high-voltage equipment does not meet safety requirements during operation, incorrect posture of holding tools leading to operational deviations, standing in a dangerous area, exposed limb posture when not wearing protective equipment according to regulations, and actions and postures that violate operating procedures. These postures directly reflect the violations of personnel's operations.

[0098] Equipment-related violations include situations where the operating status, switch positions, and connection methods of the equipment involved in the operation do not comply with safety regulations or operational requirements. For example, equipment may be in a live operating posture instead of being in a shutdown state as required; switches, valves, and other control components may not be in the designated positions; equipment connections may not be installed in a way that conforms to specifications; and equipment may be temporarily placed or fixed in a way that does not meet safety requirements during the operation. These postures reflect abnormal situations in equipment operation or work.

[0099] The system retrieves all erroneous pose data output from the pose recognition stage and determines the total number of erroneous poses through counting. The system automatically compares the total number with a preset judgment standard to determine whether the current scene involves a single erroneous pose or multiple erroneous poses.

[0100] If the number of violation poses is determined to be no more than one, a preset rule base matching process is initiated. The feature data of the single violation pose is compared with the violation pose features stored in the rule base to find feature entries that are completely matched or highly similar. Based on the matched entries, the corresponding violation operation type and violation consequences are directly extracted to ensure rapid and accurate acquisition of core violation information in a single violation scenario.

[0101] If the number of violations is determined to be greater than one, a multi-violation posture simulation process is initiated. First, the core attributes of each violation posture are extracted, including the relevant work stage, the objects involved, and its relationship with surrounding equipment or personnel. Then, based on electrical work logic and safety regulations, the causal relationships and mutual influences between the violations are analyzed. For example, whether a particular violation posture will induce other violations, or whether the combined effect of multiple violations will amplify the risk. Combining the correlation analysis results, a complete description of the violation scenario is formed, clarifying the nature of the overall violation, its core triggering factors, and potential chain reactions.

[0102] Integrate the above matching or deduction results and present the violation warning information in a structured format according to a unified standard. The information must include the work process to which the violation occurred, the type of violation, a description of the specific violation, and the potential safety consequences and production impact. Risk levels are categorized based on the severity of the consequences to ensure the warning information is clear and focused. The warning information is pushed to the on-site safety control terminal and the back-end management system in real time, providing staff with clear intervention directions and helping them to take timely measures to avoid safety hazards.

[0103] In an optional embodiment, step 150, which involves deducing a violation based on the violation pose to obtain a violation warning information, may include: Extract the pose attributes of each illegal pose.

[0104] Calculate the correlation strength between each illegal pose based on its pose attributes.

[0105] Clustering is performed based on the correlation strength between each violation pose to obtain multiple chain violation scenarios.

[0106] Based on the pose attributes of the violation poses included in each chain violation scenario, the causal probability obtained by causal relationship deduction of the violation poses is superimposed to obtain the violation operation deduction result.

[0107] Based on the results of the violation operation simulation, a warning message for violation operation is obtained.

[0108] In this embodiment, for each identified violation pose, its corresponding pose attributes are extracted from the inspection image association data and work background information. The work process to which each violation pose belongs, the equipment or personnel involved, the specific time of occurrence, its spatial location within the substation, and its relative relationship with other surrounding equipment and personnel are clearly defined, forming a structured pose attribute dataset that lays the foundation for subsequent correlation analysis.

[0109] Based on the extracted pose attributes, the similarity between two non-compliant poses is analyzed in terms of operational steps, objects, and time points. Simultaneously, the spatial proximity and functional relevance of the two poses are considered. The correlation strength value between each pair of non-compliant poses is calculated to quantify their degree of correlation, forming a correlation strength matrix for all non-compliant poses.

[0110] Violation poses with values ​​exceeding a threshold in the correlation strength matrix are grouped together for clustering. Violation poses within each cluster are closely related in terms of operational logic, spatial location, or temporal timing, collectively forming a cascading violation scenario. Clustering integrates multiple dispersed violation poses into clearly correlated scenario units, facilitating subsequent targeted analysis of causal relationships and cumulative consequences.

[0111] For each cascading violation scenario, based on the positional attributes of each violation position within the scenario, and in conjunction with power operation safety regulations and process logic, the causal relationships between violation positions are analyzed one by one. It is determined which violation positions are triggering factors and which are subsequent derivative results, clarifying the causal path and calculating the probability of each causal relationship. The probabilities of all causal relationships within the scenario are logically superimposed to comprehensively assess the overall risk level and potential cumulative consequences of the cascading violation scenario, forming the violation operation deduction result.

[0112] The system integrates the simulation results of violations across all chain-reaction violation scenarios and generates early warning information in a standardized, structured format. Each early warning message must include the operational stage of each chain-reaction violation scenario, the key attributes of the core violation position within the scenario, the triggering mechanism and development path of the violation, the potential safety consequences and production impact, and the overall risk level derived from probability aggregation. The early warning information is pushed in real-time to on-site safety control terminals and the back-end management system, ensuring that staff can quickly grasp the core situation of the violation risk and take precise and effective intervention measures.

[0113] In summary, this invention integrates current environmental data with the target work order to generate violation risk information corresponding to each operational stage in the target work order. This enables proactive prediction of high-risk violation trends, overcoming the limitation of traditional methods that only determine violations after they occur. It provides precise evidence for early intervention, significantly improving intervention effectiveness. Based on the violation risk information corresponding to each operational stage in the target work order, targeted drone inspection paths are generated, focusing inspections on high-risk scenarios and reducing ineffective inspections. Pose recognition is performed on the collected images, and further deduction of violation operations is based on the recognized poses. This not only clarifies the violation itself but also predicts its safety consequences and production impact, solving the problem of traditional methods that only determine whether a violation has occurred. This provides forward-looking decision support for safety management.

[0114] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0115] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0116] Figure 2 The diagram shows a schematic of a substation violation early warning device based on drone inspection provided in an embodiment of the present invention. For ease of explanation, only the parts relevant to the embodiment of the present invention are shown, and are described in detail below: like Figure 2 As shown, the substation violation early warning device 2 based on drone inspection includes: Module 21 is used to acquire the current environmental data and target work order corresponding to the target substation; Prediction module 22 is used to obtain the violation risk information corresponding to each operation link in the target work order based on the current environmental data and the target work order; The generation module 23 is used to generate the drone inspection path based on the violation risk information corresponding to each operation link in the target work order; The acquisition module 24 is used to perform drone inspections based on the drone inspection path to acquire inspection images; The early warning module 25 is used to perform pose recognition on the inspection image and to deduce the violation operation based on the recognized pose to obtain the early warning information of the violation operation.

[0117] In one possible implementation, the violation risk information corresponding to each work step in the target work order includes the type of violation and the probability of violation occurring in each work step; the generation module 23 is specifically used for: The detection points are determined based on the types of violations present in each work process and the probability of such violations occurring. Based on the detection points corresponding to each work step and the time consumption of each work step in the target work order, determine the initial priority of each work step; Based on the spatial boundaries corresponding to each work step in the target work order, the initial priority of each work step is adjusted to obtain the target priority of each work step; Based on the spatial boundaries and target priorities corresponding to each operational stage, a drone inspection path is generated.

[0118] In one possible implementation, module 23 is specifically used for: Based on the spatial boundaries corresponding to each work step, determine the other work steps adjacent to each work step. Sort each task step according to its corresponding initial priority; Determine the distances between adjacent work steps of each priority level after sorting; Based on the distances between adjacent work steps after sorting, and other work steps adjacent to each work step, the initial priority of each work step is adjusted to obtain the target priority of each work step.

[0119] In one possible implementation, module 23 is specifically used for: Traverse adjacent task steps based on their sorted priorities; For each work step, calculate the actual travel distance between that work step and the next work step; If the actual travel distance is less than or equal to the preset distance threshold, the initial priority of the next work step corresponding to this work step will not be adjusted. If the actual travel distance is greater than the preset distance threshold, and there are other work steps adjacent to this work step, then the initial priority level difference between the adjacent work steps and other work steps is calculated; where, the next work step corresponding to this work step is not included in the other work steps. If the initial priority level difference is less than the preset level difference, then the initial priority of other work steps with an initial priority level difference less than the preset level difference will be increased by one level, and the initial priority of the next work step corresponding to that work step will be decreased by one level; so as to obtain the target priority of each work step.

[0120] In one possible implementation, module 23 is specifically used for: The spatial boundaries corresponding to each operation stage are overlaid with the current environmental data to determine the actual accessibility of the detection points included in each operation stage. Based on the target priority corresponding to each operation step, the detection points included in each operation step are marked; the marking is used to indicate whether each detection point is a mandatory monitoring point. Based on the actual accessibility and marking of the detection points included in each operation, the detection points in each operation are screened to determine the target detection points in each operation. Based on the target priority corresponding to each operation, priority weights are assigned to the target detection points in each operation, and the predicted inspection time is determined. The target detection points in each operation are first sorted according to priority weight, and then the target detection points with the same priority weight are sorted and adjusted based on spatial adjacency to obtain a time sequence. Based on the time sequence, an inspection path for the drone is generated.

[0121] In one possible implementation, the early warning module 25 is specifically used for: Determine whether the number of illegal poses included in the identified poses is greater than 1; If not, the violation pose will be matched with the preset rule base to obtain the violation type and the violation consequences; If so, then based on the violation posture, a violation operation simulation is performed to obtain a violation operation warning.

[0122] In one possible implementation, the early warning module 25 is specifically used for: Extract the pose attributes of each illegal pose; Calculate the correlation strength between each illegal pose based on its pose attributes; Clustering is performed based on the correlation strength between each violation pose to obtain multiple chain violation scenarios; Based on the pose attributes of the violation poses included in each chain violation scenario, the causal probability obtained by causal relationship deduction of the violation poses is superimposed to obtain the violation operation deduction result. Based on the results of the violation operation simulation, a warning message for violation operation is obtained.

[0123] In one possible implementation, prediction module 22 is specifically used for: Semantic recognition is performed on the target work order to obtain semantic features; Semantic features are segmented to determine the safety specifications corresponding to each operational step; By using the safety regulations corresponding to each work step as matrix rows and the current environmental data as matrix columns, a risk assessment matrix corresponding to each work step is obtained. Based on the risk assessment matrix corresponding to each work step, calculate the standard risk value corresponding to each safety standard in each work step; Based on the standard risk value corresponding to each safety standard in each work process, calculate the comprehensive risk value corresponding to each work process; Work processes with a comprehensive risk value greater than a preset comprehensive threshold are classified as high-risk work processes, while work processes with a comprehensive risk value not greater than the preset comprehensive threshold are classified as low-risk work processes. For each high-risk operation, safety regulations with a standard risk value greater than a first preset threshold are identified as high-risk regulations to determine their corresponding violation types; and based on the standard risk value of each high-risk regulation, the probability of a violation occurring is determined. For each low-risk operation, safety standards with a standard risk value greater than the second preset threshold are identified as low-risk standards to determine their corresponding violation types. The first preset threshold is greater than the second preset threshold.

[0124] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. For example... Figure 3 As shown, the electronic device 3 of this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. When the processor 30 executes the computer program 32, it implements the steps in the various method embodiments described above. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module / unit in the various device embodiments described above.

[0125] For example, computer program 32 may be divided into one or more modules / units, which are stored in memory 31 and executed by processor 30 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3.

[0126] Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 3 may also include input / output devices, network access devices, buses, etc.

[0127] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0128] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0129] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for early warning of violations in substations based on unmanned aerial vehicle (UAV) inspection, characterized in that, include: Obtain the current environmental data and target work order corresponding to the target substation; Based on the current environmental data and the target work order, obtain the violation risk information corresponding to each work step in the target work order; Based on the violation risk information corresponding to each operation step in the target work order, generate the drone inspection path; The drone inspection is carried out based on the aforementioned drone inspection path to collect inspection images; The inspection image is subjected to pose recognition, and the violation operation is deduced based on the recognized pose to obtain violation operation warning information.

2. The substation violation early warning method based on UAV inspection according to claim 1, characterized in that, The violation risk information corresponding to each work step in the target work order includes the type of violation and the probability of violation occurring in each work step; The step of generating a drone inspection path based on the violation risk information corresponding to each operational stage in the target work order includes: The detection points are determined based on the types of violations present in each work process and the probability of such violations occurring. The initial priority of each work step is determined based on the detection points corresponding to each work step and the time consumption of each work step in the target work order. Based on the spatial boundaries corresponding to each work step in the target work order, the initial priority of each work step is adjusted to obtain the target priority of each work step. Based on the spatial boundaries and target priorities corresponding to each operational stage, a drone inspection path is generated.

3. The substation violation early warning method based on UAV inspection according to claim 2, characterized in that, The step of adjusting the initial priority of each work step according to the spatial boundary corresponding to each work step in the target work order to obtain the target priority of each work step includes: Based on the spatial boundaries corresponding to each work step, determine the other work steps adjacent to each work step. Sort each task step according to its corresponding initial priority; Determine the distances between adjacent work steps of each priority level after sorting; Based on the distances between adjacent work steps after sorting, and other work steps adjacent to each work step, the initial priority of each work step is adjusted to obtain the target priority of each work step.

4. The substation violation early warning method based on UAV inspection according to claim 3, characterized in that, The initial priority of each task is adjusted based on the distances between adjacent task steps after sorting, and other task steps adjacent to each task step, to obtain the target priority of each task step, including: Traverse adjacent task steps based on their sorted priorities; For each work step, calculate the actual travel distance between that work step and the next work step; If the actual travel distance is less than or equal to the preset distance threshold, the initial priority of the next work step corresponding to that work step will not be adjusted. If the actual travel distance is greater than a preset distance threshold, and there are other work steps adjacent to this work step, then the initial priority level difference between the adjacent work steps and other work steps is calculated; wherein, the other work steps do not include the next work step corresponding to this work step; If the initial priority level difference is less than the preset level difference, then the initial priority of other work steps with an initial priority level difference less than the preset level difference is increased by one level, and the initial priority of the next work step corresponding to that work step is decreased by one level; so as to obtain the target priority of each work step.

5. The substation violation early warning method based on UAV inspection according to claim 2, characterized in that, The process of generating UAV inspection paths based on the spatial boundaries and target priorities corresponding to each operational stage includes: The spatial boundaries corresponding to each work step are overlaid with the current environmental data to determine the actual accessibility of the detection points included in each work step; Based on the target priority corresponding to each operation step, the detection points included in each operation step are marked; wherein, the marking is used to indicate whether each detection point is a mandatory monitoring point; Based on the actual accessibility and marking of the detection points included in each operation, the detection points in each operation are screened to determine the target detection points in each operation. Based on the target priority corresponding to each operation, priority weights are assigned to the target detection points in each operation, and the predicted inspection time is determined. The target detection points in each operation are first sorted according to priority weight, and then the target detection points with the same priority weight are sorted and adjusted based on spatial adjacency to obtain a time sequence. Based on the time sequence, an inspection path for the UAV is generated.

6. The substation violation early warning method based on UAV inspection according to claim 1, characterized in that, The process of deducing illegal operations based on the identified pose to obtain illegal operation warning information includes: Determine whether the number of illegal poses included in the identified poses is greater than 1; If not, the violation pose is matched with a preset rule base to obtain the violation type and the violation consequences. If so, then based on the aforementioned violation posture, a violation operation deduction is performed to obtain a violation operation warning message.

7. The substation violation early warning method based on UAV inspection according to claim 6, characterized in that, The deduction of illegal operation based on the illegal posture to obtain illegal operation warning information includes: Extract the pose attributes of each illegal pose; Calculate the correlation strength between each illegal pose based on its pose attributes; Clustering is performed based on the correlation strength between each violation pose to obtain multiple chain violation scenarios; Based on the pose attributes of the violation poses included in each chain violation scenario, the causal probability obtained by causal relationship deduction of the violation poses is superimposed to obtain the violation operation deduction result. Based on the results of the violation operation simulation, a violation operation warning message is obtained.

8. The substation violation early warning method based on UAV inspection according to claim 1, characterized in that, The step of obtaining violation risk information corresponding to each work step in the target work order based on the current environmental data and the target work order includes: Semantic recognition is performed on the target work order to obtain semantic features; The semantic features are segmented to determine the safety specifications corresponding to each operational step; By using the safety standards corresponding to each work step as matrix rows and the current environmental data as matrix columns, a risk assessment matrix corresponding to each work step is obtained. Based on the risk assessment matrix corresponding to each work step, calculate the standard risk value corresponding to each safety standard in each work step; Based on the standard risk value corresponding to each safety standard in each work process, calculate the comprehensive risk value corresponding to each work process; Work processes with a comprehensive risk value greater than a preset comprehensive threshold are classified as high-risk work processes, while work processes with a comprehensive risk value not greater than the preset comprehensive threshold are classified as low-risk work processes. For each high-risk operation, safety regulations with a standard risk value greater than a first preset threshold are identified as high-risk regulations to determine their corresponding violation types; and based on the standard risk value of each high-risk regulation, the probability of a violation occurring is determined. For each low-risk operation, safety standards with a standard risk value greater than the second preset threshold are identified as low-risk standards to determine their corresponding violation types. Wherein, the first preset threshold is greater than the second preset threshold.

9. A substation violation early warning device based on drone inspection, characterized in that, include: The acquisition module is used to acquire the current environmental data and target work order corresponding to the target substation; The prediction module is used to obtain the violation risk information corresponding to each operation step in the target work order based on the current environmental data and the target work order; The generation module is used to generate a drone inspection path based on the violation risk information corresponding to each operation link in the target work order; The data acquisition module is used to perform drone inspections based on the drone inspection path to acquire inspection images. The early warning module is used to perform pose recognition on the inspection image and to deduce the violation operation based on the recognized pose to obtain the early warning information of the violation operation.

10. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.

Images