Power inspection behavior compliance evaluation method and system, medium and equipment

Abstract

The invention discloses a compliance evaluation method, system, medium and equipment for electric power inspection behaviors, and belongs to the technical field of operation and maintenance intelligentization of an electric power system, and the method comprises the steps: obtaining the track data of an inspector, and preset inspection standard data; calculating a time deviation score and a spatial similarity score according to the trajectory data and the inspection standard data; mapping the trajectory data into a three-dimensional grid, generating a space-time cube, extracting space-time features in the space-time cube through a preset convolutional neural network model, and outputting a space-time score; wherein the spatio-temporal characteristics comprise a motion characteristic, a behavior mode characteristic and a path time sequence mode characteristic; and performing weighted fusion on the time deviation score, the spatial similarity score and the space-time score to obtain an inspection behavior compliance evaluation result corresponding to the inspection personnel. Therefore, through implementation of the method, the problems of spatial-temporal feature splitting and insufficient single-dimension scoring precision during power inspection compliance evaluation in the prior art can be solved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent power system operation and maintenance technology, and relates to a compliance assessment method, system, medium and equipment for power inspection behavior. Background Technology

[0002] Power grid inspection is a crucial link in ensuring the safe and stable operation of power grid equipment. Currently, it is mainly carried out manually on-site to inspect equipment and observe for anomalies to ensure the normal operation of the power system. Inspection personnel need to check key areas within the substation, such as the main transformer area, high-voltage switchyard, and GIS room, according to a predetermined route, record equipment status, and complete procedures such as clocking in and out.

[0003] In existing technologies, inspection work heavily relies on human experience, making it difficult to unify the discipline and operational standards of inspection personnel. This can lead to some employees engaging in non-compliant inspections. Furthermore, substation managers need to manually coordinate and evaluate large amounts of inspection data, often relying on manual comparison of whether trajectories cover preset inspection points or simple threshold judgments (such as whether arrival occurs within a specified time). These methods focus only on isolated indicators of a single time or spatial dimension, failing to capture dynamic behavioral patterns during the inspection process (such as the order of access to critical equipment, abnormal dwell time, etc.). This makes the evaluation results susceptible to random biases and difficult to reflect the deeper compliance logic of inspection behavior. Summary of the Invention

[0004] This application provides a method, system, medium, and equipment for compliance assessment of power inspection activities, which can solve the problems of fragmented spatiotemporal characteristics and insufficient accuracy of single-dimensional scoring in the existing technology for power inspection compliance assessment.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a compliance assessment method for power inspection activities, comprising:

[0006] Acquire the trajectory data of the inspection personnel, as well as the preset inspection standard data;

[0007] Based on the trajectory data and inspection standard data, calculate the time deviation score and spatial similarity score;

[0008] The trajectory data is mapped onto a three-dimensional mesh to generate a spatiotemporal cube, and the spatiotemporal features are extracted from it using a preset convolutional neural network model to output a spatiotemporal score; wherein, the spatiotemporal features include motion features, behavioral pattern features, and path temporal pattern features; the motion features include trajectory direction change features, the behavioral pattern features include equipment inspection dwell time features, and the path temporal pattern features include equipment access sequence features;

[0009] The time deviation score, spatial similarity score, and spatiotemporal score are weighted and fused to obtain the compliance assessment result of the inspection behavior of the inspection personnel.

[0010] Compared with existing technologies, the embodiments of this application have the following beneficial effects: By acquiring the trajectory data of inspection personnel and the preset inspection standard data, it provides basic data support for multi-dimensional analysis of compliance assessment; based on the trajectory data and inspection standard data, it calculates time deviation scores and spatial similarity scores, respectively quantifying the temporal accuracy of inspection behavior from the time dimension and evaluating path matching degree from the spatial dimension, thus initially solving the limitations of single-dimensional assessment; it maps the trajectory data to a three-dimensional mesh to generate a spatiotemporal cube, realizing the structured integration of spatiotemporal information of inspection behavior, and extracts motion features (such as trajectory direction change features), behavior pattern features (such as equipment inspection dwell time features), and path temporal pattern features (such as equipment access order features, which can detect whether key equipment is accessed in the prescribed order) through a preset convolutional neural network model. By detecting abnormal loops or backtracking, and capturing spatiotemporal patterns at different levels through multi-scale convolutional kernels, this method integrates the correlation information between spatial location, time duration, and access logic (such as the compliance pattern of "accessing devices in the prescribed order without abnormal backtracking"), thus solving the problem of difficulty in identifying complex access logic anomalies in traditional assessments. The method weightedly fuses time deviation scores, spatial similarity scores, and spatiotemporal scores to obtain the assessment results. The synergistic effect of multi-dimensional indicators enhances the comprehensiveness of the assessment. Through the synergistic combination of the above technical features, this method effectively solves the problems of fragmented spatiotemporal features leading to one-sided assessments and insufficient accuracy of single-dimensional scoring in existing technologies. It achieves a leap from "point verification" to "full-process semantic understanding" of inspection behavior, enabling a more comprehensive and accurate reflection of the compliance of inspection behavior and providing managers with a more reliable assessment basis.

[0011] In some embodiments of the first aspect of this application, calculating the time deviation score based on the trajectory data and inspection standard data includes:

[0012] Extract the arrival time and dwell time of the inspection personnel at each inspection point from the trajectory data to obtain the first time series;

[0013] Extract the standard arrival time and standard dwell time corresponding to each inspection point from the inspection standard data to obtain the second time series;

[0014] Based on the first time series and the second time series, the original time deviation corresponding to each inspection point is calculated and matched with the preset time gradient thresholds to obtain the gradient score corresponding to each inspection point, and the gradient score sequence is output; wherein, each of the time gradient thresholds is determined based on the priority of each inspection point;

[0015] According to the preset DTW algorithm, the gradient score sequence is calculated using DTW scoring, which is used as the time deviation score.

[0016] Compared to existing technologies, the above embodiments have the following advantages: Arrival and dwell times are extracted from trajectory data to form a first time series, while corresponding times are extracted from standard data to form a second time series, laying a data foundation for time deviation calculation; the original time deviation is calculated and matched with a time gradient threshold determined based on inspection point priority to obtain a gradient score sequence, enabling the time deviation score to reflect the differences in importance of different inspection points; the gradient score sequence is scored using a preset DTW algorithm as the time deviation score. The DTW algorithm can handle the nonlinear alignment problem of time series, further improving the accuracy of time dimension scoring and alleviating the shortcomings of single-time-dimensional evaluation.

[0017] In some embodiments of the first aspect of this application, calculating the spatial similarity score based on the trajectory data and inspection standard data includes:

[0018] The actual spatial coordinate point sequence in the trajectory data is aligned and denoised with the standard spatial coordinate point sequence in the inspection standard data.

[0019] Based on the processed sequence of actual spatial coordinate points and the sequence of standard spatial coordinate points, the corresponding Hausdorff distance is calculated according to the preset Hausdorff distance algorithm and linearly mapped to the spatial similarity score.

[0020] Compared with existing technologies, the above embodiments have the following beneficial effects: aligning and denoising the actual and standard spatial coordinate point sequences reduces data noise interference and improves the accuracy of spatial assessment; based on the processed coordinate point sequence, the distance is calculated using the Hausdorff distance algorithm and linearly mapped to a spatial similarity score. The Hausdorff distance quantifies the maximum mismatch between point sets, and the linear mapping enhances the intuitiveness of the score, effectively improving the reliability of spatial dimension scoring and alleviating the limitations of single spatial dimension assessment.

[0021] In some embodiments of the first aspect of this application, mapping the trajectory data onto a three-dimensional mesh to generate a spacetime cube includes:

[0022] A grid spatial coordinate system is constructed using the X and Y axes as geographic spatial coordinates and the Z axis as the time axis.

[0023] The trajectory data is mapped to the grid space coordinate system and converted into the corresponding three-dimensional tensor to obtain the spacetime cube.

[0024] Compared with existing technologies, the above embodiments have the following beneficial effects: a grid spatial coordinate system is constructed with the X-axis and Y-axis as geographic spatial coordinates and the Z-axis as the time axis to achieve unified modeling of spatiotemporal dimensions; trajectory data is mapped to this coordinate system and converted into a three-dimensional tensor to generate a spatiotemporal cube, and dynamic trajectory is transformed into a data structure that integrates spatiotemporal information, providing a foundation for subsequent convolutional neural network extraction of fused features, breaking the traditional mode of separating spatiotemporal features in evaluation.

[0025] In some embodiments of the first aspect of this application, the step of extracting spatiotemporal features and outputting a spatiotemporal score through a preset convolutional neural network model includes:

[0026] The spatiotemporal cube is convolved to extract motion features, behavior pattern features, and path temporal pattern features;

[0027] The motion features, behavior pattern features, and path temporal pattern features are concatenated and fused, and mapped to the output node through a fully connected layer. After normalization by the sigmoid function, a score is predicted, and the corresponding spatiotemporal score is output. Specifically, when the predicted task type is a classification task, the compliance probability is output as the spatiotemporal score, and when the task type is a regression task, the compliance degree score is output as the spatiotemporal score.

[0028] Compared to existing technologies, the above embodiments have the following beneficial effects: Convolution processing is performed on the spatiotemporal cube to extract motion features, behavioral pattern features, and path temporal pattern features. Motion features achieve dynamic anomaly detection through trajectory direction and speed changes; behavioral pattern features ensure operational sufficiency through device dwell time; and path temporal pattern features verify the access sequence of key devices to avoid process omissions. The motion features, behavioral pattern features, and path temporal pattern features are then concatenated and fused to integrate the correlation information between spatial location, time duration, and access logic. This information is then mapped to the output node through a fully connected layer and normalized using the sigmoid function before predicting a score. When the predicted task type is a classification task, the compliance probability is output as the spatiotemporal score to meet the rapid decision-making needs in binary classification scenarios. When the task type is a regression task, the compliance degree score is output as the spatiotemporal score to adapt to refined evaluation scenarios. Through differentiated output design based on task type, the model can dynamically adapt to the needs of different evaluation scenarios, improving the model's practicality and scenario adaptability.

[0029] Secondly, the present invention also provides a compliance assessment system for power inspection activities, including: a data acquisition module, a scoring calculation module, a model processing module, and a result output module;

[0030] The data acquisition module is used to acquire the trajectory data of the inspection personnel and the preset inspection standard data.

[0031] The scoring calculation module is used to calculate the time deviation score and spatial similarity score based on the trajectory data and inspection standard data.

[0032] The model processing module is used to map the trajectory data onto a three-dimensional mesh to generate a spatiotemporal cube, and extract the spatiotemporal features therein through a preset convolutional neural network model and output a spatiotemporal score; wherein, the spatiotemporal features include motion features, behavioral pattern features and path temporal pattern features; the motion features include trajectory direction change features, the behavioral pattern features include equipment inspection dwell time features, and the path temporal pattern features include equipment access sequence features;

[0033] The result output module is used to weight and fuse the time deviation score, spatial similarity score, and spatiotemporal score to obtain the compliance assessment result of the inspection behavior of the inspection personnel.

[0034] Compared with existing technologies, the above embodiments of this application have the following beneficial effects: By acquiring the trajectory data of inspection personnel and the preset inspection standard data, it provides basic data support for multi-dimensional analysis of compliance assessment; based on the trajectory data and inspection standard data, it calculates time deviation scores and spatial similarity scores, respectively quantifying the temporal accuracy of inspection behavior from the time dimension and evaluating the path matching degree from the spatial dimension, thus initially solving the limitations of single-dimensional assessment; it maps the trajectory data to a three-dimensional mesh to generate a spatiotemporal cube, realizing the structured integration of spatiotemporal information of inspection behavior, and extracts motion features (such as trajectory direction change features), behavior pattern features (such as equipment inspection dwell time features), and path temporal pattern features (such as equipment access sequence features, which can detect whether key equipment is accessed in the prescribed order) through a preset convolutional neural network model. This method detects abnormal loops or backtracking, and captures spatiotemporal patterns at different levels through multi-scale convolutional kernels. This integrates the correlation information between spatial location, time length, and access logic (such as the compliance pattern of "accessing devices in the prescribed order without abnormal backtracking"), solving the problem of difficulty in identifying complex access logic anomalies in traditional assessments. The weighted fusion of time deviation score, spatial similarity score, and spatiotemporal score yields the assessment result. The comprehensiveness of the assessment is improved through the synergistic effect of multi-dimensional indicators. Through the synergistic combination of the above technical features, this method effectively solves the problems of one-sided assessment and insufficient accuracy of single-dimensional scoring caused by the fragmentation of spatiotemporal features in existing technologies. It realizes the leap from "point verification" to "full-process semantic understanding" of inspection behavior, and can more comprehensively and accurately reflect the compliance of inspection behavior, providing managers with a more reliable assessment basis.

[0035] In some embodiments of the second aspect of this application, the scoring calculation module includes: a first sequence extraction unit, a second sequence extraction unit, a score calculation unit, and a time deviation calculation unit;

[0036] The first sequence extraction unit is used to extract the arrival time and stay time of the inspection personnel at each inspection point from the trajectory data to obtain a first time sequence.

[0037] The second sequence extraction unit is used to extract the standard arrival time and standard dwell time corresponding to each inspection point from the inspection standard data to obtain a second time series;

[0038] The score calculation unit is used to calculate the original time deviation corresponding to each inspection point based on the first time series and the second time series, and match it with the preset time gradient thresholds to obtain the gradient score corresponding to each inspection point, and output the gradient score sequence; wherein, each of the time gradient thresholds is determined based on the priority of each inspection point;

[0039] The time deviation calculation unit is used to perform DTW score calculation on the gradient score sequence according to the preset DTW algorithm, and use it as the time deviation score.

[0040] Compared to existing technologies, the above embodiments have the following advantages: Arrival and dwell times are extracted from trajectory data to form a first time series, while corresponding times are extracted from standard data to form a second time series, laying a data foundation for time deviation calculation; the original time deviation is calculated and matched with a time gradient threshold determined based on inspection point priority to obtain a gradient score sequence, enabling the time deviation score to reflect the differences in importance of different inspection points; the gradient score sequence is scored using a preset DTW algorithm as the time deviation score. The DTW algorithm can handle the nonlinear alignment problem of time series, further improving the accuracy of time dimension scoring and alleviating the shortcomings of single-time-dimensional evaluation.

[0041] In some embodiments of the second aspect of this application, the scoring calculation module further includes: a preprocessing unit and a spatial similarity calculation unit;

[0042] The preprocessing unit is used to align and denoise the actual spatial coordinate point sequence in the trajectory data with the standard spatial coordinate point sequence in the inspection standard data.

[0043] The spatial similarity calculation unit is used to calculate the corresponding Hausdorff distance based on the processed sequence of actual spatial coordinate points and the sequence of standard spatial coordinate points, according to a preset Hausdorff distance algorithm, and linearly map it to the spatial similarity score.

[0044] Compared with existing technologies, the above embodiments have the following beneficial effects: aligning and denoising the actual and standard spatial coordinate point sequences reduces data noise interference and improves the accuracy of spatial assessment; based on the processed coordinate point sequence, the distance is calculated using the Hausdorff distance algorithm and linearly mapped to a spatial similarity score. The Hausdorff distance quantifies the maximum mismatch between point sets, and the linear mapping enhances the intuitiveness of the score, effectively improving the reliability of spatial dimension scoring and alleviating the limitations of single spatial dimension assessment.

[0045] Thirdly, the present invention also provides a compliance assessment device for power inspection activities, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded onto the processor, it implements the steps of any one of the compliance assessment methods for power inspection activities of the present invention.

[0046] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the compliance assessment method for any power inspection behavior of the present invention. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating a compliance assessment method for power inspection activities provided in some embodiments of the present invention.

[0048] Figure 2 : This is a schematic diagram of the structure of a compliance assessment system for power inspection behavior provided in some embodiments of the present invention.

[0049] Figure 3 : This is a structural diagram of a compliance assessment device for power inspection behavior provided in some embodiments of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1:

[0052] Please refer to Figure 1 To address the problems of fragmented spatiotemporal features and insufficient accuracy of single-dimensional scoring in existing technologies for power inspection compliance assessment, an embodiment of the present invention provides a method for assessing the compliance of power inspection behavior, comprising steps S1 to S6:

[0053] Step S1: Obtain the trajectory data of the inspection personnel and the preset inspection standard data.

[0054] In practice, data collection utilizes GPS or BeiDou positioning to record the real-time location of inspection personnel, as well as sensor data from the inspection equipment to confirm whether the equipment status has been correctly checked, or RFID / NFC check-in records. Standard inspection data includes the location of each inspection point, estimated arrival time, estimated dwell time, and inspection route. The dwell time at different inspection points can be set differently. The inspection route needs to cover all key points requiring inspection, which may include: main transformer area, high-voltage switchyard (AIS), GIS room (gas-insulated substation), control and protection room, and auxiliary equipment area. Furthermore, when developing standardized time and route data, artificial intelligence models can be used to preprocess historical inspection data before further manual adjustments, saving time.

[0055] Step S2: Calculate the time deviation score and spatial similarity score based on the trajectory data and inspection standard data.

[0056] Furthermore, the step S2 of calculating the time deviation score can be implemented through the following preferred embodiments, including steps S21-S24, as follows:

[0057] S21: Extract the arrival time and dwell time of the inspection personnel at each inspection point from the trajectory data to obtain the first time series;

[0058] S22: Extract the standard arrival time and standard dwell time corresponding to each inspection point from the inspection standard data to obtain the second time series;

[0059] S23: Based on the first time series and the second time series, calculate the original time deviation corresponding to each inspection point, and match it with the preset time gradient thresholds to obtain the gradient score corresponding to each inspection point, and output the gradient score sequence; wherein, each of the time gradient thresholds is determined based on the priority of each inspection point;

[0060] S24: According to the preset DTW algorithm, perform DTW scoring on the gradient score sequence and use it as the time deviation score.

[0061] In practice, the smaller the deviation between the actual time and the preset standardized time, the higher the time deviation score. The time gradient threshold is determined in advance based on the priority of the inspection points. For example, if the priority of the critical inspection points is higher, the corresponding time gradient threshold (i.e., the allowable deviation range) can be set smaller.

[0062] In this preferred embodiment, arrival time and dwell time are extracted from trajectory data to form a first time series, and corresponding times are extracted from standard data to form a second time series, laying the data foundation for time deviation calculation. The original time deviation is calculated and matched with the time gradient threshold determined based on the priority of inspection points to obtain a gradient score sequence, so that the time deviation score can reflect the difference in importance of different inspection points. The gradient score sequence is scored using a preset DTW algorithm as the time deviation score. The DTW algorithm can handle the nonlinear alignment problem of time series, further improving the accuracy of time dimension scoring and alleviating the shortcomings of single time dimension evaluation.

[0063] Furthermore, the step S2, which calculates the spatial similarity score, can be implemented through the following preferred embodiments, including steps S25-S26, as follows:

[0064] S25: Align and denoise the actual spatial coordinate point sequence in the trajectory data with the standard spatial coordinate point sequence in the inspection standard data;

[0065] S26: Based on the processed sequence of actual spatial coordinate points and the sequence of standard spatial coordinate points, calculate the corresponding Hausdorff distance according to the preset Hausdorff distance algorithm and linearly map it to the spatial similarity score.

[0066] In this preferred embodiment, the actual and standard spatial coordinate point sequences are aligned and denoised to reduce data noise interference and improve the accuracy of spatial assessment. Based on the processed coordinate point sequence, the distance is calculated by the Hausdorff distance algorithm and linearly mapped to a spatial similarity score. The Hausdorff distance quantifies the maximum mismatch between point sets, and the linear mapping enhances the intuitiveness of the score, effectively improving the reliability of the spatial dimension score and alleviating the limitations of single spatial dimension assessment.

[0067] Step S3: Map the trajectory data onto a three-dimensional mesh to generate a spatiotemporal cube, and extract the spatiotemporal features from it using a preset convolutional neural network model and output a spatiotemporal score;

[0068] The spatiotemporal features include motion features, behavioral pattern features, and path temporal pattern features; the motion features include trajectory direction change features, the behavioral pattern features include equipment inspection dwell time features, and the path temporal pattern features include equipment access sequence features.

[0069] Furthermore, the spacetime cube can be obtained through the following preferred embodiments, including steps S31-S32, as follows:

[0070] S31: Construct a grid spatial coordinate system using the X-axis and Y-axis as geographic spatial coordinates and the Z-axis as the time axis;

[0071] S32: Map the trajectory data to the grid space coordinate system, convert it into the corresponding three-dimensional tensor, and obtain the spacetime cube.

[0072] In practice, the X and Y axes represent geographic spatial coordinates, such as dividing the substation area into a grid (e.g., a 10m × 10m grid); the Z axis (time axis) divides time into discrete intervals (e.g., a slice per minute). Each grid cell value can represent the activity intensity within that grid during that time period (e.g., dwell time, number of passes), or can be represented in binary as whether the cell was passed. Thus, the data for each inspection task can be represented as a three-dimensional array (width × height × time depth).

[0073] In this preferred embodiment, a grid spatial coordinate system is constructed with the X-axis and Y-axis as geographic spatial coordinates and the Z-axis as the time axis to achieve unified modeling of spatiotemporal dimensions. The trajectory data is mapped to this coordinate system and converted into a three-dimensional tensor to generate a spatiotemporal cube. The dynamic trajectory is transformed into a data structure that integrates spatiotemporal information, providing a foundation for the subsequent extraction of fused features by the convolutional neural network and breaking the traditional model of separating spatiotemporal features in evaluation.

[0074] Furthermore, the output spatiotemporal score in step S3 can be implemented through the following preferred embodiments, including steps S33-S34, as follows:

[0075] S33: Perform convolution processing on the spatiotemporal cube to extract motion features, behavior pattern features, and path temporal pattern features;

[0076] S34: The motion features, behavior pattern features, and path temporal pattern features are concatenated and fused, and mapped to the output node through a fully connected layer. After normalization by the sigmoid function, the score is predicted and the corresponding spatiotemporal score is output. When the predicted task type is a classification task, the compliance probability is output as the spatiotemporal score. When the task type is a regression task, the compliance degree score is output as the spatiotemporal score.

[0077] In practical implementation, after constructing the spatiotemporal cube, it is input into a 3D CNN model. The 3D CNN model captures both spatial and temporal features simultaneously through convolution operations in three-dimensional space. During convolution, the 3D convolution kernel slides across the three-dimensional space (x, y, t), calculating a weighted sum of data within a small cube at each step. For example, a 3×3×3 convolution kernel covers a 3×3 grid area spatially and three consecutive time slices temporally. Extracted feature types include motion features such as edges and corners (in both space and time), including changes in trajectory direction and speed; behavioral pattern features, such as lingering in front of a device or moving in a corridor; and path temporal pattern features, such as the overall inspection path pattern, whether key equipment is accessed in the prescribed order, and whether there are abnormal loops or backtracking. After feature extraction, the model's fully connected layers combine these features to ultimately output a spatiotemporal score. The specific process includes:

[0078] Feature mapping: Convolutional layers and pooling layers progressively transform the original spatiotemporal cube into a high-level feature representation;

[0079] Feature fusion: At a deeper level of the network, feature representations simultaneously incorporate spatial and temporal information. For example, a feature representation might activate the pattern of "staying in the transformer area for more than 5 minutes" (simultaneously identifying spatial location and time duration).

[0080] Score Prediction: Fully connected layers predict scores based on these fused features. For classification tasks (compliant / non-compliant), the sigmoid function is used to output probabilities; for regression tasks (scores from 0 to 1), the sigmoid function can also be used. For example, consider an inspection task: the standard route is A→B→C (A, B, and C are the locations of key equipment), while the actual trajectory is A→C→B (skipping B and going to C first). In a spatiotemporal cube, this trajectory will show that the time at equipment C is earlier than that at B on the timeline. 3D CNNs can capture this anomaly in order (because the temporal order can be detected by the convolutional kernels), thus giving a low score.

[0081] In this preferred embodiment, the spatiotemporal cube is convolved to extract motion features, behavioral pattern features, and path temporal pattern features. Motion features utilize trajectory direction and speed changes for dynamic anomaly detection; behavioral pattern features ensure operational sufficiency through device dwell time; and path temporal pattern features verify the access sequence of key devices to avoid process omissions. The motion features, behavioral pattern features, and path temporal pattern features are then concatenated and fused to integrate the correlation information between spatial location, time duration, and access logic. This information is then mapped to the output node through a fully connected layer and normalized using a sigmoid function before a score is predicted. When the predicted task type is a classification task, the compliance probability is output as the spatiotemporal score to meet the rapid decision-making needs in binary classification scenarios. When the task type is a regression task, the compliance degree score is output as the spatiotemporal score to adapt to refined evaluation scenarios. This differentiated output design based on task type allows the model to dynamically adapt to the needs of different evaluation scenarios, improving the model's practicality and scenario adaptability.

[0082] Step S4: Weightedly fuse the time deviation score, spatial similarity score, and spatiotemporal score to obtain the compliance assessment result of the inspection behavior corresponding to the inspection personnel.

[0083] In practical implementation, it can also be combined with rule modification items, for example:

[0084] DTW quantifies time deviation, for example: DTW = 0.3 → "Inspection time is delayed by 18 minutes compared to the plan";

[0085] Hausdorff quantifies the degree of spatial deviation. For example, Hausdorff = 50m → "maximum path deviation of 50 meters".

[0086] The model provides complex pattern recognition, such as identifying nonlinear relationships: "The dwell time in the main transformer area should be shortened under high temperature weather" and "The surge arrester should be checked immediately after the GIS room inspection", and detects unknown abnormal patterns, such as "The confidence level of abnormal detour pattern detection is 87%".

[0087] The final weighted evaluation result (score) can be calculated using the following formula: Final Score = 0.4 × DTW Score + 0.3 × Hausdorff Score + 0.3 × CNN Score + Rule Correction Items (which can be based on industry standards and safety rules, for example, based on safety rules:

[0088] If the surge arrester is not checked during a thunderstorm, the final score will be 0, which is equivalent to a veto.

[0089] In addition, when using this solution, a 3D model of the substation can be built, standard inspection routes can be drawn in the 3D model, employee inspection routes that need to be collected in real time and evaluated can be dynamically drawn in the 3D model, and the output compliance assessment results can be displayed along the employee inspection routes.

[0090] Alternatively, a connection can be established with employees' mobile devices to provide real-time feedback on assessment results. When an employee's inspection score falls below a threshold, a notification message is sent to them, along with a standardized inspection route (including time and location). The method of sending the notification message (such as voice, beep, vibration, and text) and the timing are adjusted based on the employee's current real-time status. If the employee is currently inspecting equipment with a high risk factor, the notification message should be sent later to avoid affecting their attention.

[0091] In summary, compared with the prior art, the above embodiments of this application have the following beneficial effects: By acquiring the trajectory data of inspection personnel and the preset inspection standard data, it provides basic data support for multi-dimensional analysis of compliance assessment; based on the trajectory data and inspection standard data, it calculates time deviation scores and spatial similarity scores, respectively quantifying the temporal accuracy of inspection behavior from the time dimension and evaluating the path matching degree from the spatial dimension, thus initially solving the limitations of single-dimensional assessment; it maps the trajectory data to a three-dimensional mesh to generate a spatiotemporal cube, realizing the structured integration of spatiotemporal information of inspection behavior, and extracts motion features (such as trajectory direction change features), behavior pattern features (such as equipment inspection dwell time features), and path temporal pattern features (such as equipment access sequence features, which can detect whether key equipment is accessed in the prescribed order) through a preset convolutional neural network model. This method, by capturing spatiotemporal patterns at different levels through multi-scale convolutional kernels (e.g., whether there are abnormal loops or backtracks in the equipment), integrates the correlation information of spatial location, time length, and access logic (e.g., the compliance pattern of "accessing equipment in the prescribed order without abnormal backtracks"), thus solving the problem of difficulty in identifying complex access logic anomalies in traditional assessments. The method obtains the assessment result by weighted fusion of time deviation scores, spatial similarity scores, and spatiotemporal scores, improving the comprehensiveness of the assessment through the synergistic effect of multi-dimensional indicators. Through the synergistic combination of the above technical features, this method effectively solves the problems of fragmented spatiotemporal features leading to one-sided assessments and insufficient accuracy of single-dimensional scoring in existing technologies. It achieves a leap from "point verification" to "full-process semantic understanding" of inspection behavior, enabling a more comprehensive and accurate reflection of the compliance of inspection behavior and providing managers with a more reliable assessment basis.

[0092] Example 2:

[0093] Please refer to Figure 2Based on the same inventive concept, the present invention discloses a compliance assessment system for power inspection behavior, comprising: a data acquisition module M1, a scoring calculation module M2, a model processing module M3, and a result output module M4;

[0094] The data acquisition module M1 is used to acquire the trajectory data of the inspection personnel and the preset inspection standard data.

[0095] The scoring calculation module M2 is used to calculate the time deviation score and spatial similarity score based on the trajectory data and inspection standard data.

[0096] Furthermore, the scoring calculation module M2 includes: a first sequence extraction unit, a second sequence extraction unit, a score calculation unit, and a time deviation calculation unit;

[0097] The first sequence extraction unit is used to extract the arrival time and stay time of the inspection personnel at each inspection point from the trajectory data to obtain a first time sequence.

[0098] The second sequence extraction unit is used to extract the standard arrival time and standard dwell time corresponding to each inspection point from the inspection standard data to obtain a second time series;

[0099] The score calculation unit is used to calculate the original time deviation corresponding to each inspection point based on the first time series and the second time series, and match it with the preset time gradient thresholds to obtain the gradient score corresponding to each inspection point, and output the gradient score sequence; wherein, each of the time gradient thresholds is determined based on the priority of each inspection point;

[0100] The time deviation calculation unit is used to perform DTW score calculation on the gradient score sequence according to the preset DTW algorithm, and use it as the time deviation score.

[0101] In this preferred embodiment, arrival time and dwell time are extracted from trajectory data to form a first time series, and corresponding times are extracted from standard data to form a second time series, laying the data foundation for time deviation calculation. The original time deviation is calculated and matched with the time gradient threshold determined based on the priority of inspection points to obtain a gradient score sequence, so that the time deviation score can reflect the difference in importance of different inspection points. The gradient score sequence is scored using a preset DTW algorithm as the time deviation score. The DTW algorithm can handle the nonlinear alignment problem of time series, further improving the accuracy of time dimension scoring and alleviating the shortcomings of single time dimension evaluation.

[0102] Furthermore, the scoring calculation module M2 also includes: a preprocessing unit and a spatial similarity calculation unit;

[0103] The preprocessing unit is used to align and denoise the actual spatial coordinate point sequence in the trajectory data with the standard spatial coordinate point sequence in the inspection standard data.

[0104] The spatial similarity calculation unit is used to calculate the corresponding Hausdorff distance based on the processed sequence of actual spatial coordinate points and the sequence of standard spatial coordinate points, according to a preset Hausdorff distance algorithm, and linearly map it to the spatial similarity score.

[0105] In this preferred embodiment, the actual and standard spatial coordinate point sequences are aligned and denoised to reduce data noise interference and improve the accuracy of spatial assessment. Based on the processed coordinate point sequence, the distance is calculated by the Hausdorff distance algorithm and linearly mapped to a spatial similarity score. The Hausdorff distance quantifies the maximum mismatch between point sets, and the linear mapping enhances the intuitiveness of the score, effectively improving the reliability of the spatial dimension score and alleviating the limitations of single spatial dimension assessment.

[0106] The model processing module M3 is used to map the trajectory data onto a three-dimensional mesh, generate a spatiotemporal cube, and extract the spatiotemporal features therein through a preset convolutional neural network model and output a spatiotemporal score; wherein, the spatiotemporal features include motion features, behavioral pattern features, and path temporal pattern features; the motion features include trajectory direction change features, the behavioral pattern features include equipment inspection dwell time features, and the path temporal pattern features include equipment access sequence features.

[0107] Furthermore, the model processing module M3 includes: a coordinate system construction unit and a mapping unit;

[0108] The coordinate system construction unit is used to construct a grid spatial coordinate system with the X-axis and Y-axis as geographic spatial coordinates and the Z-axis as the time axis.

[0109] The mapping unit is used to map the trajectory data to the grid space coordinate system, convert it into a corresponding three-dimensional tensor, and obtain the spacetime cube.

[0110] In this preferred embodiment, a grid spatial coordinate system is constructed with the X-axis and Y-axis as geographic spatial coordinates and the Z-axis as the time axis to achieve unified modeling of spatiotemporal dimensions. The trajectory data is mapped to this coordinate system and converted into a three-dimensional tensor to generate a spatiotemporal cube. The dynamic trajectory is transformed into a data structure that integrates spatiotemporal information, providing a foundation for the subsequent extraction of fused features by the convolutional neural network and breaking the traditional model of separating spatiotemporal features in evaluation.

[0111] Furthermore, the model processing module M3 also includes: a convolution unit and a scoring output unit;

[0112] The convolutional unit is used to perform convolution processing on the spatiotemporal cube to extract motion features, behavior pattern features, and path temporal pattern features.

[0113] The scoring output unit is used to concatenate and fuse the motion features, behavior pattern features, and path temporal pattern features, and map them to the output node through a fully connected layer. After normalization by the sigmoid function, the unit predicts the score and outputs the corresponding spatiotemporal score. Specifically, when the predicted task type is a classification task, the compliance probability is output as the spatiotemporal score, and when the task type is a regression task, the compliance degree score is output as the spatiotemporal score.

[0114] In this preferred embodiment, the spatiotemporal cube is convolved to extract motion features, behavioral pattern features, and path temporal pattern features. Motion features utilize trajectory direction and speed changes for dynamic anomaly detection; behavioral pattern features ensure operational sufficiency through device dwell time; and path temporal pattern features verify the access sequence of key devices to avoid process omissions. The motion features, behavioral pattern features, and path temporal pattern features are then concatenated and fused to integrate the correlation information between spatial location, time duration, and access logic. This information is then mapped to the output node through a fully connected layer and normalized using a sigmoid function before a score is predicted. When the predicted task type is a classification task, the compliance probability is output as the spatiotemporal score to meet the rapid decision-making needs in binary classification scenarios. When the task type is a regression task, the compliance degree score is output as the spatiotemporal score to adapt to refined evaluation scenarios. This differentiated output design based on task type allows the model to dynamically adapt to the needs of different evaluation scenarios, improving the model's practicality and scenario adaptability.

[0115] The result output module M4 is used to weight and fuse the time deviation score, spatial similarity score and spatiotemporal score to obtain the compliance assessment result of the inspection behavior of the inspection personnel.

[0116] In summary, compared with existing technologies, the embodiments of this application have the following beneficial effects: By acquiring the trajectory data of inspection personnel and preset inspection standard data, it provides basic data support for multi-dimensional analysis of compliance assessment; based on the trajectory data and inspection standard data, it calculates time deviation scores and spatial similarity scores, respectively quantifying the temporal accuracy of inspection behavior from the time dimension and evaluating path matching degree from the spatial dimension, thus initially solving the limitations of single-dimensional assessment; it maps the trajectory data to a three-dimensional mesh to generate a spatiotemporal cube, realizing the structured integration of spatiotemporal information of inspection behavior, and extracts motion features (such as trajectory direction change features), behavior pattern features (such as equipment inspection dwell time features), and path temporal pattern features (such as equipment access order features, which can detect whether key equipment is accessed in the prescribed order) through a preset convolutional neural network model. This method, by using multi-scale convolutional kernels to capture spatiotemporal patterns at different levels (such as whether there are abnormal loops or backtracks), integrates the correlation information between spatial location, time length, and access logic (e.g., the compliance pattern of "accessing equipment in the prescribed order without abnormal backtracks"), thus solving the problem of difficulty in identifying complex access logic anomalies in traditional assessments. The method weightedly fuses time deviation scores, spatial similarity scores, and spatiotemporal scores to obtain the assessment results. The comprehensiveness of the assessment is improved through the synergistic effect of multi-dimensional indicators. Through the synergistic combination of the above technical features, this method effectively solves the problems of fragmented spatiotemporal features leading to one-sided assessments and insufficient accuracy of single-dimensional scoring in existing technologies. It achieves a leap from "point verification" to "full-process semantic understanding" of inspection behavior, enabling a more comprehensive and accurate reflection of the compliance of inspection behavior and providing managers with a more reliable assessment basis.

[0117] Example 3:

[0118] Figure 3 A structural diagram of a compliance assessment device for power line inspection activities, as described in this application, is presented. Figure 3 As shown, the compliance assessment equipment for this power inspection activity may include: processor N1, memory N2, data interface N3, and communication bus N4.

[0119] Wherein: processor N1, memory N2, and data interface N3 communicate with each other through communication bus N4; data interface N3 is used for data communication with other devices such as input devices or output devices; processor N1 is used to execute program N5, which can specifically execute the relevant steps in any of the above-mentioned compliance assessment method embodiments for power inspection behavior.

[0120] Specifically, program N5 may include program code, which includes computer-executable instructions.

[0121] The processor N1 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The compliance assessment device for power inspection activities includes one or more processors, which may be processors of the same type, such as one or more CPUs, or processors of different types, such as one or more CPUs and one or more ASICs.

[0122] Memory N2 is used to store program N5. Memory N2 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage.

[0123] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Furthermore, the embodiments in this application are not directed to any particular programming language.

[0124] Example 4:

[0125] This invention also provides a computer-readable storage medium storing at least one executable instruction. When the executable instruction is run on a power inspection compliance assessment device / system, it causes the power inspection compliance assessment device / system to perform a power inspection compliance assessment method from any of the above-described method embodiments.

[0126] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. Similarly, for the purpose of simplification and aiding understanding of one or more aspects of the invention, in the above description of exemplary embodiments of this application, various features of the embodiments are sometimes grouped together in a single embodiment, figure, or description thereof. The claims, which follow the detailed description, are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0127] Those skilled in the art will understand that the modules in the device of the embodiment can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiment can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components, except that at least some of such features and / or processes or units are mutually exclusive.

Claims

1. A compliance assessment method for power grid inspection activities, characterized in that, include: Acquire the trajectory data of the inspection personnel, as well as the preset inspection standard data; Based on the trajectory data and inspection standard data, calculate the time deviation score and spatial similarity score; The trajectory data is mapped onto a three-dimensional mesh to generate a spatiotemporal cube, and the spatiotemporal features are extracted from it using a preset convolutional neural network model to output a spatiotemporal score; wherein, the spatiotemporal features include motion features, behavioral pattern features, and path temporal pattern features; the motion features include trajectory direction change features, the behavioral pattern features include equipment inspection dwell time features, and the path temporal pattern features include equipment access sequence features; The time deviation score, spatial similarity score, and spatiotemporal score are weighted and fused to obtain the compliance assessment result of the inspection behavior of the inspection personnel.

2. The compliance assessment method for power inspection activities as described in claim 1, characterized in that, The step of calculating the time deviation score based on the trajectory data and inspection standard data includes: Extract the arrival time and dwell time of the inspection personnel at each inspection point from the trajectory data to obtain the first time series; Extract the standard arrival time and standard dwell time corresponding to each inspection point from the inspection standard data to obtain the second time series; Based on the first time series and the second time series, the original time deviation corresponding to each inspection point is calculated and matched with the preset time gradient thresholds to obtain the gradient score corresponding to each inspection point, and the gradient score sequence is output; wherein, each of the time gradient thresholds is determined based on the priority of each inspection point; According to the preset DTW algorithm, the gradient score sequence is calculated using DTW scoring, which is used as the time deviation score.

3. The compliance assessment method for power inspection activities as described in claim 1, characterized in that, The step of calculating a spatial similarity score based on the trajectory data and inspection standard data includes: The actual spatial coordinate point sequence in the trajectory data is aligned and denoised with the standard spatial coordinate point sequence in the inspection standard data. Based on the processed sequence of actual spatial coordinate points and the sequence of standard spatial coordinate points, the corresponding Hausdorff distance is calculated according to the preset Hausdorff distance algorithm and linearly mapped to the spatial similarity score.

4. The compliance assessment method for power inspection activities as described in claim 1, characterized in that, The step of mapping the trajectory data onto a three-dimensional mesh to generate a spacetime cube includes: A grid spatial coordinate system is constructed using the X and Y axes as geographic spatial coordinates and the Z axis as the time axis. The trajectory data is mapped to the grid space coordinate system and converted into the corresponding three-dimensional tensor to obtain the spacetime cube.

5. The compliance assessment method for power inspection activities as described in claim 4, characterized in that, The step of extracting spatiotemporal features and outputting a spatiotemporal score through a preset convolutional neural network model includes: The spatiotemporal cube is convolved to extract motion features, behavior pattern features, and path temporal pattern features; The motion features, behavior pattern features, and path temporal pattern features are concatenated and fused, and mapped to the output node through a fully connected layer. After normalization by the sigmoid function, a score is predicted, and the corresponding spatiotemporal score is output. Specifically, when the predicted task type is a classification task, the compliance probability is output as the spatiotemporal score, and when the task type is a regression task, the compliance degree score is output as the spatiotemporal score.

6. A compliance assessment system for power line inspection activities, characterized in that, include: Data acquisition module, scoring calculation module, model processing module, and result output module; The data acquisition module is used to acquire the trajectory data of the inspection personnel and the preset inspection standard data. The scoring calculation module is used to calculate the time deviation score and spatial similarity score based on the trajectory data and inspection standard data. The model processing module is used to map the trajectory data onto a three-dimensional mesh to generate a spatiotemporal cube, and extract the spatiotemporal features therein through a preset convolutional neural network model and output a spatiotemporal score; wherein, the spatiotemporal features include motion features, behavioral pattern features and path temporal pattern features; the motion features include trajectory direction change features, the behavioral pattern features include equipment inspection dwell time features, and the path temporal pattern features include equipment access sequence features; The result output module is used to weight and fuse the time deviation score, spatial similarity score, and spatiotemporal score to obtain the compliance assessment result of the inspection behavior of the inspection personnel.

7. The compliance assessment system for power inspection activities as described in claim 6, characterized in that, The scoring calculation module includes: a first sequence extraction unit, a second sequence extraction unit, a score calculation unit, and a time deviation calculation unit; The first sequence extraction unit is used to extract the arrival time and stay time of the inspection personnel at each inspection point from the trajectory data to obtain a first time sequence. The second sequence extraction unit is used to extract the standard arrival time and standard dwell time corresponding to each inspection point from the inspection standard data to obtain a second time series; The score calculation unit is used to calculate the original time deviation corresponding to each inspection point based on the first time series and the second time series, and match it with the preset time gradient thresholds to obtain the gradient score corresponding to each inspection point, and output the gradient score sequence; wherein, each of the time gradient thresholds is determined based on the priority of each inspection point; The time deviation calculation unit is used to perform DTW score calculation on the gradient score sequence according to the preset DTW algorithm, and use it as the time deviation score.

8. The compliance assessment system for power inspection activities as described in claim 6, characterized in that, The scoring calculation module further includes: a preprocessing unit and a spatial similarity calculation unit; The preprocessing unit is used to align and denoise the actual spatial coordinate point sequence in the trajectory data with the standard spatial coordinate point sequence in the inspection standard data. The spatial similarity calculation unit is used to calculate the corresponding Hausdorff distance based on the processed sequence of actual spatial coordinate points and the sequence of standard spatial coordinate points, according to a preset Hausdorff distance algorithm, and linearly map it to the spatial similarity score.

9. A compliance assessment device for power inspection activities, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements the steps of a compliance assessment method for power inspection activities according to any one of claims 1-5.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of a compliance assessment method for power inspection activities according to any one of claims 1-5.

Images