A substation cable laying management method and system

Abstract

The application discloses a power transformation station cable laying management method and device, and relates to the technical field of power transformation stations, which comprises the following steps: constructing a three-dimensional model of a power transformation station cable channel, generating a plurality of initial cable paths satisfying space obstacle avoidance constraints by using an improved A algorithm with a heat field penalty term and a filling rate penalty term; taking the initial cable path as an initial population, iteratively optimizing by a multi-objective genetic algorithm, and obtaining an optimized target laying path; based on the target laying path and environmental temperature parameters, calling a pre-constructed physical information fusion neural network model, predicting the stress state of each key node along the path, and segmentally identifying the risk level of the target laying path to generate a stress identification path containing high-risk segment position information; and correcting and guiding according to the stress identification path combined with laying deviation until the cable laying is completed. The three-dimensional model is set to simulate the cable laying, and the reliability and processing efficiency of the cable laying are improved.

Description

Technical Field

[0001] This invention relates to the field of substation technology, specifically to a method and apparatus for managing cable laying in substations. Background Technology

[0002] In power systems, substations serve as crucial nodes for power conversion and distribution, and the rationality of their cable laying directly impacts the safe and stable operation of the power system. However, traditional cable laying methods rely on manual experience and lack precise design simulation tools, leading to problems such as unreasonable laying path planning, uneven load distribution on cable trays, and high maintenance difficulty. With the continuous expansion of substation scale and the surge in the number of cables, these problems have become increasingly prominent, seriously affecting the overall efficiency and reliability of the power system. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and device for cable laying management in substations. By combining a digital model of the cable laying process in substations with augmented reality, the efficiency and accuracy of cable laying can be improved.

[0004] This invention provides a method for managing cable laying in substations, the method comprising: A three-dimensional model of the substation cable tunnel is constructed, and an improved A model is adopted by introducing thermal field penalty terms and fill rate penalty terms. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints; Using the initial cable path as the initial population, a multi-objective genetic algorithm is used to iteratively optimize the cable length and number of components, the load distribution balance of the cable tray, and the heat source avoidance distance as optimization objectives to obtain the optimized target laying path. Based on the target laying path and ambient temperature parameters, a pre-built physical information fusion neural network model is invoked to predict the stress state of each key node along the path, and the target laying path is segmented by risk level according to the predicted stress state to generate a stress-marked path containing the location information of high-risk sections. The stress identification path is overlaid and displayed on the augmented reality terminal to obtain the spatial position of the cable construction end in real time. When the actual path deviates from the target path by more than a threshold, a correction guide is generated. When the construction end enters the high-risk section, a stress-sensitive construction prompt is generated to constrain the laying operation until the cable laying is completed.

[0005] Furthermore, the construction of the three-dimensional model of the substation cable channel adopts an improved A model that introduces thermal field penalty terms and fill rate penalty terms. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints, including: Collect design drawings and on-site measured coordinate data of substation building structure, cable trays, and electrical equipment to obtain the spatial location and dimensional parameters of each component of the substation. Based on the collected data, a substation building structure model, a cable tray model, and an electrical equipment model were constructed sequentially. This was then improved using A... The algorithm constructs the initial cable path in the substation building structure model.

[0006] Furthermore, the improved A The algorithm constructs the initial cable path in the substation building structure model, including: Determine the coordinates of the starting and ending points of the cable laying, rasterize the three-dimensional model of the substation cable channel, divide it into grid units, clarify the passability of the grid, and mark the areas where obstacles are located as impassable grids. The thermal field penalty coefficient is set according to the distance between the cable and the heat source equipment, and the thermal field penalty coefficient is set to have a negative linear relationship with the distance. Set a threshold for cable filling density in the cable tray. When the filling rate exceeds the threshold, set a filling rate penalty coefficient that increases linearly with the filling rate. By combining the thermal field penalty coefficient, the fill rate penalty coefficient, and the path length, a heuristic function is constructed. Starting from the starting point, adjacent feasible grids are searched one by one according to the principle of lowest cost until the destination is reached. Adjusting the weights of the heuristic function yields multiple conflict-free initial paths that satisfy obstacle avoidance and laying specifications.

[0007] Furthermore, using the initial cable path as the initial population, a multi-objective genetic algorithm is used to iteratively optimize the cable length and component quantity, cable tray load distribution balance, and heat source avoidance distance as optimization objectives, resulting in the optimized target laying path, which includes: Each initial cable path is encoded and converted into a computable chromosome. Each segment of the chromosome corresponds to the coordinates of the path nodes, the cable laying direction, and the cable tray allocation information, forming the initial population. Four optimization indicators were set: cable length, number of components, load distribution balance of cable tray, and heat source avoidance distance. The fitness of each chromosome was calculated by combining the four optimization indicators. Chromosomes with high fitness are retained to enter the next generation of the population. Two high-quality chromosomes are exchanged using a single-point crossover method. After the crossover, the paths are checked for conflict, and chromosomes with spatial conflicts are removed. After each iteration, the average fitness of the population is calculated. When the average fitness of multiple consecutive iterations tends to stabilize, the iteration is stopped, and the path corresponding to the chromosome with the highest fitness is output as the optimized target laying path.

[0008] Furthermore, the specific steps for weight adjustment and iterative optimization in the multi-objective genetic algorithm are as follows: Based on actual construction needs, the initial weighting of each optimization objective is set as follows: cable length and number of components account for 30%-40%, cable tray load distribution balance accounts for 25%-35%, and heat source avoidance distance accounts for 25%-35%. During the iteration process of the genetic algorithm, after every 10 generations, the fitness of the current population is reviewed, the optimization effect of each optimization objective is analyzed, and it is determined whether the weight allocation is in line with the actual engineering requirements. Based on the review results, if the optimization effect of a certain optimization objective does not meet expectations, the weight ratio of that objective should be adjusted appropriately, and the weights of other objectives should be adjusted accordingly to ensure that the total weight is 100%. After adjusting the weights, the iteration should continue.

[0009] Furthermore, the step of predicting the stress state of each key node along the path by calling a pre-built physical information fusion neural network model based on the target laying path and ambient temperature parameters includes: The pre-built physical information fusion neural network model is retrieved, which uses historical cable laying path parameters, ambient temperature data, and measured stress data of key nodes as training samples. The training samples are normalized to eliminate the influence of data units. The training set and validation set are divided according to a preset ratio. The validation set is used to verify and adjust the model parameters to ensure that the model prediction error is controlled within a preset range. The node coordinates, directional parameters, and ambient temperature parameters of the target laying path are input into the input layer of the neural network model. The hidden layer of the model extracts features from the input data through activation functions, optimizes the network weights by incorporating the physical constraints of cable laying, and outputs the stress prediction values ​​of each key node along the target laying path from the output layer.

[0010] Furthermore, the step of segmenting the target laying path according to the predicted stress state and generating a stress-marked path containing the location information of high-risk sections includes: Based on the cable material characteristics and laying specifications, three levels of standards are set: stress safety threshold, warning threshold, and danger threshold, and the stress range corresponding to each level is clearly defined. The predicted stress values ​​of each key node are compared one by one with the stress range of the three set standards. Based on the comparison results, each key node is divided into a safe node, a warning node, and a high-risk node. Based on the risk level of key nodes, the target laying path is divided into safe section, warning section and high-risk section to ensure that the risk level of nodes within the same section of the path is consistent. A segmented coloring method is used to visually identify paths with different risk levels. High-risk sections are labeled with specific node coordinates, stress prediction values, and risk causes, ultimately generating stress-labeled paths containing complete risk information.

[0011] Furthermore, the step of overlaying and displaying the stress-identified path on the augmented reality terminal, obtaining the spatial position of the cable installation end in real time, generating correction guidance when the actual path deviates from the target path by more than a threshold, and generating stress-sensitive construction prompts to constrain the laying operation when the installation end enters the high-risk section, until the cable laying is completed, includes: The system employs a combination of visual positioning and inertial navigation, using the augmented reality terminal's camera to capture images of the scene while simultaneously acquiring motion data from the terminal's inertial sensors. The feature points of the on-site images are analyzed, and the three-dimensional coordinates of the cable construction end are calculated in real time by combining the motion data of the inertial sensor. The preset path deviation threshold range is 5cm~10cm. The real-time calculated three-dimensional coordinates of the construction end are compared with the coordinates of the target laying path to calculate the deviation value. If the deviation exceeds the preset threshold, the augmented reality terminal will provide correction guidance until the coordinates of the construction end return to the target path range.

[0012] Furthermore, the step of overlaying and displaying the stress-identified path on the augmented reality terminal, obtaining the spatial position of the cable construction end in real time, generating correction guidance when the actual path deviates from the target path by more than a threshold, and generating stress-sensitive construction prompts to constrain the laying operation when the construction end enters the high-risk section, until the cable laying is completed, also includes: The system monitors the spatial location of the construction site in real time. When the construction site is detected to have entered a high-risk section of the stress identification path, the stress-sensitive construction prompt mechanism is triggered, and stress feedback data of the construction site is collected in real time. The terminal compares the real-time collected stress feedback data with the predicted value. If the feedback stress exceeds the danger threshold, the terminal immediately generates an emergency shutdown prompt and pushes a hidden danger investigation guide. Cable laying operations can only be resumed after the hidden danger investigation is completed and the stress is restored to a safe range.

[0013] The present invention also provides a substation cable laying management device, the management device comprising: Path planning module: Used to construct 3D models of substation cable channels, employing an improved A / B model that incorporates thermal field and fill rate penalties. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints; Path optimization module: Used as the initial population of the initial cable path, the module iteratively optimizes the cable length and number of components, the load distribution balance of the cable tray, and the heat source avoidance distance using a multi-objective genetic algorithm to obtain the optimized target laying path. Path identification module: Based on the target laying path and ambient temperature parameters, it calls a pre-built physical information fusion neural network model to predict the stress state of each key node along the path, and identifies the target laying path by risk level segmentation according to the predicted stress state, generating a stress identification path containing high-risk segment location information. Path adjustment module: used to overlay and display the stress identification path on the augmented reality terminal, obtain the spatial position of the cable construction end in real time, generate correction guidance when the actual path deviates from the target path by more than a threshold, and generate stress-sensitive construction prompts to constrain the laying operation when the construction end enters the high-risk section until the cable laying is completed.

[0014] This invention provides a method and apparatus for cable laying management in substations. By combining a digital model of the cable laying process in substations with augmented reality, the efficiency and accuracy of cable laying can be improved. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the substation laying management method in an embodiment of the present invention; Figure 2 This is a schematic diagram of the three-dimensional model construction state of the substation cable channel in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the substation cable laying analysis status in an embodiment of the present invention; Figure 4 This is a schematic diagram of the optimized cable laying in a substation according to an embodiment of the present invention; Figure 5 This is a schematic diagram of a substation cable laying management device in an embodiment of the present invention. Detailed Implementation

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

[0018] Please refer to Figure 1 As for Figure 4 This invention proposes a substation cable laying management method, which includes: S1 constructs a three-dimensional model of the substation cable channel, using an improved A model that incorporates thermal field penalty terms and fill rate penalty terms. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints; Step S1 includes: Collect design drawings and on-site measured coordinate data of substation building structure, cable trays, and electrical equipment to obtain the spatial location and dimensional parameters of each component of the substation. Data acquisition is a fundamental step in obtaining information about the internal physical environment of a substation, aiming to provide comprehensive and accurate raw data for the subsequent construction of a 3D model. This can be achieved through manual on-site measurement using precision measuring equipment such as laser rangefinders and total stations to measure the building structures, cable trays, and electrical equipment within the substation one by one, recording their 3D coordinates, dimensions, and other parameters. Furthermore, 3D laser scanning technology can be used to quickly acquire point cloud data of the substation's interior using scanning equipment. Point cloud processing software can then be used to extract the geometric information and spatial location of each component. By combining existing BIM (Building Information Modeling) data or CAD design drawings, design parameters for structures, cable trays, and equipment can be extracted and verified against on-site measured data. Design drawings provide the theoretical dimensions and layout information of components, while on-site measured coordinate data reflects the actual installation location of components and potential deviations. Combining both ensures the comprehensiveness and accuracy of the data. Clearly defining the spatial location and dimensional parameters of each component is the direct goal of data acquisition, providing a geometric foundation for the accurate construction of the 3D model.

[0019] Collect design drawings and on-site measured coordinate data of substation building structure, cable trays, and electrical equipment to obtain the spatial location and dimensional parameters of each component of the substation; Min-Max normalization (linear normalization) can be used to scale the data to a specific interval of [0,1] or [-1,1]; Z-score normalization (standardization) can also be used to convert the data into a distribution with a mean of 0 and a standard deviation of 1, thereby ensuring data consistency.

[0020] Based on the collected data, a substation building structure model, a cable tray model, and an electrical equipment model were constructed sequentially. This was then improved using A... The algorithm constructs the initial cable path in the substation building structure model.

[0021] When laying new cables for a substation, the engineers first obtained the original CAD design drawings of the substation. Simultaneously, the on-site team used a high-precision laser scanner to perform a 3D scan of the cable channel, cable tray structure, and surrounding electrical equipment, generating point cloud data. This was supplemented by a total station to accurately measure the coordinates and dimensions of key components. Next, the collected point cloud data was imported into specialized software to remove noise points and outliers from the scanning process.

[0022] Meanwhile, the geometric data and point cloud data extracted from the CAD drawings are subjected to Min-Max normalization processing, which scales all values ​​to the [0,1] range to eliminate dimensional differences.

[0023] Secondly, using the BIM software platform, based on the preprocessed data, the building structure model of the substation is first constructed, and then the cable tray model is refined. For each cable tray section, its cross-sectional dimensions (e.g., width 600mm, height 100mm), load-bearing limit (e.g., 50kg / m) and laying area are recorded.

[0024] Finally, the completed 3D model is exported to an augmented reality terminal. Technicians on-site use the terminal to overlay the virtual model onto the actual scene for visual comparison and measurement verification. If any deviation is found between the model and reality (for example, a section of cable tray is 5cm lower in the model than in reality), it is immediately corrected in the BIM software until the geometric accuracy deviation between the model and the actual scene is controlled within ±2mm, thus forming a high-precision, high-reliability 3D model of the cable duct.

[0025] Specifically, the improved A The algorithm constructs the initial cable path in the substation building structure model, including: Determine the coordinates of the starting and ending points of the cable laying, rasterize the three-dimensional model of the substation cable channel, divide it into grid units, clarify the passability of the grid, and mark the areas where obstacles are located as impassable grids. A thermal field penalty term is introduced, which converts the distance between the cable path and the heat source equipment into a penalty coefficient. The closer the cable path is to the heat source, the larger the penalty coefficient. The purpose of introducing the thermal field penalty term is to consider the impact of the heat source on the safe operation of the cable during the path planning process, and to avoid the cable path being too close to the heat-generating equipment, thereby reducing the risk of cable overheating.

[0026] A fill rate penalty term is introduced, and a threshold for cable fill density in the cable tray is set. When the fill rate exceeds the preset fill density threshold, the penalty coefficient increases linearly with the increase of the fill rate. The purpose of introducing the fill rate penalty term is to optimize the load distribution of the cable tray, avoid local cable trays from being overloaded due to excessive number of cables, and ensure the safety of the cable tray structure and good heat dissipation of the cables.

[0027] Construct a heuristic function that integrates the thermal field penalty term, the fill rate penalty term, and the path length cost. Using the starting point as the initial search node, search for adjacent passable grids step by step according to the cost value calculated by the heuristic function until the destination is reached. Adjust the weights of the heuristic function to generate multiple initial cable paths that are free from spatial conflicts and meet spatial obstacle avoidance constraints and basic laying requirements.

[0028] In practical applications, the starting and ending points of cable laying can first be selected in the 3D model of the substation cable channel through the user interface. The starting point can be set to the outgoing end of a certain distribution cabinet, and the ending point can be the incoming end of a certain transformer. Subsequently, the system performs voxelization processing on the 3D model to generate a grid space composed of cubic voxels with a side length of 10 centimeters.

[0029] Furthermore, each voxel is marked as "passable" or "inaccessible", with all voxels that intersect with building walls, existing equipment, and other cable tray solid models being marked as "inaccessible".

[0030] Regarding penalty settings, major heat-generating equipment such as transformers and high-voltage switchgear can be pre-identified as heat sources. The closer the cable is to the heat source, the greater the penalty. For the fill rate penalty, the fill rate threshold of the cable tray can be set to 70%. When a cable tray segment is selected, and the current fill rate of that segment plus the cross-sectional area of ​​the new cable exceeds 70%, the fill rate penalty coefficient can be set in a linear increment to ensure accurate adjustment of the laid cable.

[0031] Furthermore, the heuristic function can be: ; in, It is the actual path length from the starting point to the current grid cell n. It is the estimated Manhattan distance from the current grid cell n to the endpoint. The weight of the warm-up penalty term, The weight of the fill rate penalty term, The thermal field penalty coefficient for the current grid n. This is the fill rate penalty coefficient for the current raster n.

[0032] Furthermore, to generate multiple initial paths, the above search process can be repeated three times, adjusting each time. and The relative size of the values ​​determines three initial cable paths that satisfy spatial obstacle avoidance and are optimized for heat source avoidance and cable tray load.

[0033] Through A The algorithm incorporates a thermal field penalty term, enabling the path planning process to proactively identify and avoid heat-generating equipment within the substation. This significantly reduces the risk of accelerated aging and insulation degradation of cables due to prolonged operation in high-temperature environments, thereby improving cable reliability and lifespan. Simultaneously, a fill rate penalty term is introduced, taking into account the current load conditions when selecting cable trays. This prevents overloading of a single tray by excessive cable loads, effectively promoting a balanced load distribution within the trays, ensuring structural safety, and providing ample heat dissipation space for the cables, further enhancing cable operating efficiency and safety.

[0034] S2: Using the initial cable path as the initial population, a multi-objective genetic algorithm is used to iteratively optimize the cable length and number of components, the load distribution balance of the cable tray, and the heat source avoidance distance as optimization objectives to obtain the optimized target laying path. Specifically, step S2 includes: Each initial cable path is encoded and converted into a computable chromosome. Each segment of the chromosome corresponds to the coordinates of the path nodes, the cable laying direction, and the cable tray allocation information, forming the initial population. Four optimization indicators were set: cable length, number of components, load distribution balance of cable tray, and heat source avoidance distance. The fitness of each chromosome was calculated by combining the four optimization indicators. Chromosomes with high fitness are retained to enter the next generation of the population. Two high-quality chromosomes are exchanged using a single-point crossover method. After the crossover, the paths are checked for conflict, and chromosomes with spatial conflicts are removed. After each iteration, the average fitness of the population is calculated. When the average fitness of multiple consecutive iterations tends to stabilize, the iteration is stopped, and the path corresponding to the chromosome with the highest fitness is output as the optimized target laying path.

[0035] Iterative optimization using a multi-objective genetic algorithm can be implemented as follows: In the population initialization phase, each initial cable path can be encoded as a real-valued chromosome. For example, a path can be represented as a sequence of nodes, each node represented by its three-dimensional coordinates, with an integer appended after each node representing the cable tray ID used in that path segment. The initial population size can be set to 100 chromosomes. During fitness calculation, weights can be assigned to four optimization metrics: cable length, number of components, cable tray load distribution balance, and heat source avoidance distance. For example, cable length can have a weight of 0.4, number of components 0.2, cable tray load distribution balance 0.2, and heat source avoidance distance 0.2. Cable length can be obtained by calculating the sum of Euclidean distances between all nodes on the path; the number of components can be estimated by counting the number of changes in cable tray IDs and the number of bends on the path; cable tray load distribution balance can be measured by calculating the standard deviation of the fill rate of each cable tray, with a smaller standard deviation indicating better balance; and heat source avoidance distance can be obtained by calculating the minimum distance from each point on the path to the nearest heat source.

[0036] The final fitness value is obtained through a linear combination of these weighted indices. In genetic operations, selection can be performed using roulette wheel, calculating the probability of selection based on the fitness value of each chromosome, and then randomly selecting one. Crossover can be performed using single-point crossover, randomly selecting a position in the chromosome's coding sequence as the crossover point, exchanging the segments after the crossover point of two parent chromosomes, and generating two offspring chromosomes.

[0037] After the intersection, rapid collision detection can be performed using a 3D model of the substation. This involves constructing bounding boxes for path segments and intersecting them with obstacle bounding boxes. If a conflict exists, the offspring chromosome is marked as invalid or directly removed. The mutation operation randomly selects the coordinates of a path node on the chromosome and randomly generates a new coordinate value within a small range (e.g., ±0.5 meters) around its current value. Spatial obstacle avoidance constraints are then verified to ensure the mutated path remains feasible. The iteration termination condition can be set to the maximum number of iterations (e.g., 500 generations) or when the average fitness improvement of the population is less than 0.001 over 50 consecutive generations. When either condition is met, the chromosome with the highest fitness is selected from the final population, decoded into the actual cable path, and used as the optimized target laying path.

[0038] Specifically, the steps for weight adjustment and iterative optimization in a multi-objective genetic algorithm are as follows: Based on actual construction needs, the initial weighting of each optimization objective is set as follows: cable length and number of components account for 30%-40%, cable tray load distribution balance accounts for 25%-35%, and heat source avoidance distance accounts for 25%-35%. During the iteration process of the genetic algorithm, after every 10 generations, the fitness of the current population is reviewed, the optimization effect of each optimization objective is analyzed, and it is determined whether the weight allocation is in line with the actual engineering requirements. Based on the review results, if the optimization effect of a certain optimization objective does not meet expectations, the weight ratio of that objective should be adjusted appropriately, and the weights of other objectives should be adjusted accordingly to ensure that the total weight is 100%. After adjusting the weights, the iteration should continue so that the output results can meet the optimization requirements.

[0039] S3: Based on the target laying path and ambient temperature parameters, call the pre-built physical information fusion neural network model to predict the stress state of each key node along the path, and mark the target laying path by risk level according to the predicted stress state to generate a stress-marked path containing the location information of high-risk sections. The step of predicting the stress state of each key node along the path by calling a pre-built physical information fusion neural network model based on the target laying path and ambient temperature parameters includes: A pre-built physical information fusion neural network model is retrieved. This model uses historical cable laying path parameters, ambient temperature data, and measured stress data at key nodes as training samples. The physical information fusion neural network model is a model that combines physical principles (such as mechanics and thermodynamics) with the data-driven capabilities of neural networks. This model not only learns patterns from data but also improves its generalization ability and prediction accuracy by embedding physical equations or constraints.

[0040] The training samples are normalized to eliminate the influence of data units. The training set and validation set are divided according to a preset ratio. The validation set is used to verify and adjust the model parameters to ensure that the model prediction error is controlled within a preset range. Furthermore, normalization aims to transform data with different dimensions and numerical ranges to a unified scale, preventing certain features from dominating model training due to excessively large values, thereby accelerating model convergence and improving training stability.

[0041] The node coordinates, directional parameters, and ambient temperature parameters of the target laying path are input into the input layer of the neural network model. The hidden layer of the model extracts features from the input data through activation functions, optimizes the network weights by incorporating the physical constraints of cable laying, and outputs the stress prediction values ​​of each key node along the target laying path from the output layer.

[0042] Specifically, predicting the stress state of key nodes using a physical information fusion neural network model can be achieved as follows: During the model preparation phase, a physical information fusion neural network model built using the TensorFlow or PyTorch framework can be invoked. In its design, the loss function of this model includes not only the prediction error term but also additional physical constraint terms based on the mechanical and thermodynamic equations of cable materials. The physical relationships between cable bending stress and curvature, and tensile stress and elongation can be introduced as regularization terms.

[0043] The training samples can be derived from the engineering archives of all cables laid in the substation over the past five years, including path parameters such as cable type, laying length, bending radius, support point location, and cable tray type. The ambient temperature data comes from the historical temperature data of each area recorded by the substation environmental monitoring system; the measured stress data of key nodes is obtained by pre-embedding fiber Bragg grating (FBG) sensors or resistance strain gauges in typical laying scenarios (such as sharp bends, wall penetrations, and cable tray entrances) to monitor and record the stress changes of the cable under different working conditions over a long period of time.

[0044] In the sample preprocessing stage, the collected training samples are subjected to Min-Max normalization, scaling all numerical features (such as length, temperature, and stress values) to the [0,1] interval. Subsequently, the dataset is randomly divided into an 80% training set and a 20% validation set. During model training, the mean squared error (MSE) of the model is evaluated using the validation set every 10 epochs, and hyperparameters such as the learning rate and batch size are adjusted based on the validation set performance until the validation set MSE does not significantly decrease for 20 consecutive epochs, ensuring the model maintains a low prediction error even on unseen data. In the parameter input stage, detailed information about a target laying path, optimized using a multi-objective genetic algorithm, is input into the model.

[0045] In the feature extraction and prediction stage, the hidden layers of the neural network model can adopt a multi-layer fully connected network structure and use ReLU as the activation function to efficiently extract nonlinear features from the input parameters.

[0046] Finally, the model's output layer (e.g., a linear layer) outputs a predicted stress value (e.g., tensile stress, bending stress, shear stress) every meter or every key geometric feature point (e.g., bends, support points) along the target laying path. During the result verification phase, the model's predicted stress values ​​are compared with measured stress data from similar paths in historical laying cases. For example, the mean absolute percentage error (MAPE) between the predicted and measured values ​​is calculated. If the MAPE exceeds 5%, the prediction is considered inaccurate. In this case, the input parameters can be adjusted; for example, a fine-tuning of the ambient temperature parameter by ±2°C, or a small correction to the bending radius of a node in the target laying path, followed by a re-prediction. This process is iterated until the MAPE drops below 5%, resulting in highly accurate stress predictions and providing reliable data for subsequent risk assessment.

[0047] Specifically, the step of segmenting the target laying path according to risk level based on the predicted stress state and generating a stress-marked path containing the location information of high-risk sections includes: Based on the cable material characteristics and laying specifications, three levels of standards are set: stress safety threshold, warning threshold, and danger threshold, and the stress range corresponding to each level is clearly defined. Based on the stress comparison results, the risk level of key nodes is classified, and the entire target laying path is divided into several continuous segments to ensure that all nodes in the same segment have the same risk level.

[0048] The predicted stress values ​​of each key node are compared one by one with the stress ranges of the three set standards. Based on the comparison results, each key node is divided into safe nodes, warning nodes, and high-risk nodes. If multiple consecutive nodes on the path are determined to be safe nodes, then these nodes constitute a safe segment. If consecutive nodes are determined to be warning nodes, then a warning segment is formed. If consecutive nodes are determined to be high-risk nodes, then a high-risk segment is formed. This segmentation process helps to simplify the complex path risk information, making it easier for construction personnel to understand and manage.

[0049] Based on the risk level of key nodes, the target laying path is divided into safe section, warning section and high-risk section to ensure that the risk level of nodes within the same section of the path is consistent. A segmented coloring method is used to visually identify paths with different risk levels. High-risk sections are labeled with specific node coordinates, stress prediction values, and risk causes, ultimately generating stress-labeled paths containing complete risk information.

[0050] Furthermore, the target laying path, segmented by risk level, is visually presented through color coding, allowing construction personnel to easily identify different risk areas. For example, safe sections can be colored green, warning sections yellow, and high-risk sections red. In addition, high-risk sections require further detailed information, such as specific node coordinates, predicted stress values, and potential causes of the high risk. The final stress-marked path is a visually intuitive representation containing complete risk information.

[0051] Furthermore, preset material colors can be assigned to path segments with different risk levels in 3D modeling software or augmented reality (AR) applications. For high-risk segments, specific information can be marked by displaying text labels, pop-up information boxes, or arrow indicators next to or above the path. Another approach is to use Geographic Information System (GIS) or CAD software to associate path data with risk level data, generate a layer with color coding and detailed annotations, and then export this layer in a format that can be loaded and displayed on augmented reality terminals.

[0052] Specifically, at the substation cable laying site, construction workers wear augmented reality (AR) glasses. First, the AR glasses capture real-time images of the surrounding environment through their built-in camera. Combined with inertial sensor data, they utilize visual-inertial odometry (VIO) technology for simultaneous localization and mapping (SLAM). Simultaneously, they identify several QR codes pre-attached to the substation cable trays as reference points, precisely aligning the AR glasses' coordinate system with the pre-imported 3D cable channel model's coordinate system, eliminating discrepancies between the virtual model and the real environment. After calibration, the AR glasses' display overlays the optimized stress marker paths as semi-transparent virtual lines above the actual cable trays. Safe sections are displayed in green, warning sections in yellow, and high-risk sections in red with a slight flickering effect to highlight their risk. When construction workers focus their gaze on a red high-risk section, an information box automatically pops up next to that section of the path on the augmented reality glasses interface. This box displays the cable's "Model: YJV-22-10kV," "Recommended Laying Angle: 30°," and "Current Stress Level: High Risk." Workers can zoom in or out of the path view by making a "pinch" gesture in the air and rotate the view by a "swipe" gesture, allowing them to examine details of the high-risk section from different angles. For example, they can observe stress concentration at cable bends, thus gaining a more comprehensive understanding of the causes of the risk and key avoidance strategies.

[0053] S4: Overlay the stress identification path onto the augmented reality terminal to obtain the spatial position of the cable construction end in real time. When the actual path deviates from the target path by more than a threshold, generate a correction guide. When the construction end enters a high-risk section, generate a stress-sensitive construction prompt to constrain the laying operation until the cable laying is completed.

[0054] Specifically, step S4 includes: The system employs a combination of visual positioning and inertial navigation, using the augmented reality terminal's camera to capture images of the scene while simultaneously acquiring motion data from the terminal's inertial sensors. The feature points of the on-site images are analyzed, and the three-dimensional coordinates of the cable construction end are calculated in real time by combining the motion data of the inertial sensor. The preset path deviation threshold range is 5cm~10cm. The real-time calculated three-dimensional coordinates of the construction end are compared with the coordinates of the target laying path to calculate the deviation value. If the deviation exceeds the preset threshold, the augmented reality terminal will provide correction guidance until the coordinates of the construction end return to the target path range.

[0055] In the location acquisition step, a combination of visual positioning and inertial navigation is used. The augmented reality terminal's camera captures images of the scene, while simultaneously acquiring motion data from the terminal's inertial sensors. Visual positioning refers to determining the relative or absolute position of an object or device by analyzing the image information captured by the camera.

[0056] Furthermore, feature-point-based visual odometry estimates camera motion by matching feature points across consecutive image frames. Inertial navigation, on the other hand, uses inertial sensors (such as accelerometers and gyroscopes) to measure the linear acceleration and angular velocity of a device, and calculates the device's position, velocity, and attitude through integration. Inertial navigation systems based on Kalman filtering or extended Kalman filtering fuse accelerometer and gyroscope data to provide more accurate attitude and position estimates, improving the robustness and accuracy of position acquisition.

[0057] Specifically, in the coordinate calculation step, image recognition is used to analyze feature points in the field image, extracting discriminative feature points from the acquired field image. These feature points can be corner points, edge points, or texture features. Combining inertial sensor data, the three-dimensional coordinates of the cable construction end are calculated in real time by integrating the field image feature points and inertial sensor data.

[0058] In the deviation judgment step, a preset path deviation threshold of 5-10cm is used. The real-time calculated 3D coordinates of the construction end are compared with the coordinates of the target laying path to calculate the deviation value. The preset path deviation threshold is a standard used to judge whether the construction end deviates from the target laying path, and its value ranges from 5-10cm, which can be adjusted according to the actual construction accuracy requirements and cable characteristics. The calculation of the deviation value refers to quantifying the distance between the current position of the construction end and the target laying path. This is done by projecting the 3D coordinates of the construction end onto the target laying path and calculating the distance between the projection point and the construction end.

[0059] During the deviation guidance step, if the deviation value exceeds a preset threshold, the augmented reality terminal immediately generates deviation guidance through both voice prompts and on-screen pop-ups, clearly indicating the direction of deviation, the distance of deviation, and the adjustment method, until the coordinates of the construction end return to the target laying path range. Voice prompts refer to playing preset voice information through the augmented reality terminal's speaker to convey the deviation situation and correction suggestions to the construction personnel.

[0060] Specifically, step S4 also includes: The system monitors the spatial location of the construction site in real time. When the construction site is detected to have entered a high-risk section of the stress identification path, the stress-sensitive construction prompt mechanism is triggered, and stress feedback data from the construction site is collected in real time and transmitted to the system. The terminal compares the real-time collected stress feedback data with the predicted value. If the feedback stress exceeds the danger threshold, the terminal immediately generates an emergency shutdown prompt and pushes a hidden danger investigation guide. Cable laying operations can only be resumed after the hidden danger investigation is completed and the stress is restored to a safe range.

[0061] At substation cable laying sites, construction workers wear augmented reality (AR) glasses to perform their work. When a worker holds cable laying tools and begins laying cables in high-risk sections (e.g., near high-power transformers or cable tray bends, areas marked as high-risk in the path identification module), the AR glasses' built-in visual positioning system, combined with an inertial measurement unit (IMU), calculates the spatial position of the worker and their laying tools in real time.

[0062] The system compares this real-time location with the three-dimensional coordinate range of the high-risk section in the pre-loaded stress identification path. Once the AR glasses detect that the construction worker has entered a high-risk section—for example, if the system determines that the worker's laying tool has entered an area 50 cm from the transformer cooling duct—the AR glasses interface will immediately display a visual prompt accompanied by a voice announcement: "Attention! You have entered a high-risk section. Please reduce your laying speed. Do not exceed 200 Newtons in traction. The real-time stress monitoring frequency has been increased to 10 times per second." Simultaneously, the AR glasses interface will overlay the predicted stress value for the area (e.g., the estimated maximum stress is 150 MPa), the cause of the risk (e.g., "thermal stress concentration, excessively small bending radius"), and specific avoidance points (e.g., "Ensure the cable bending radius is not less than 30 cm"). Throughout the laying process in the high-risk section, the fiber optic strain sensor integrated on the laying tool continuously collects stress data of the cable during traction and bending. This data is transmitted in real-time to the backend management system via the AR glasses' wireless communication module (e.g., Wi-Fi).

[0063] The backend system continuously compares these real-time stress feedback data with the predicted stress value of the high-risk section and the preset danger threshold (e.g., 250 MPa). If the construction worker mishandles the installation process, causing the cable stress to momentarily exceed the danger threshold, such as reaching 260 MPa, the AR glasses will immediately emit a piercing alarm sound and display a red "Emergency Stop!" message on the screen, while simultaneously vibrating to alert the construction worker. The system will also automatically push a hazard inspection guide, instructing the construction worker to check for cable jams, abnormal laying tools, or cable damage. Only after the construction worker has completed the inspection according to the guide, and the system confirms that the cable stress has returned to a safe range (e.g., below 180 MPa), will the system lift the stoppage and allow the construction worker to continue the laying operation.

[0064] Specifically, by constructing a high-precision 3D model of the substation cable channel, a precise spatial basis is provided for subsequent route planning, which contrasts sharply with the limitations of traditional 2D drawings. Furthermore, an improved A model is adopted, incorporating thermal field penalty terms and fill rate penalty terms. The algorithm generates the initial cable path, incorporating heat source avoidance and cable tray load balancing considerations from the early stages of path planning. This makes the quality of the initial path far superior to traditional paths that only consider spatial obstacle avoidance.

[0065] Example 2: Please refer to Figure 5This invention also proposes a substation cable laying management device, which achieves full digital management of cable laying through modular design, effectively replacing traditional manual experience-based methods. The core innovation of this embodiment lies in the systematic integration of the path planning module 10, path optimization module 20, path identification module 30, and path adjustment module 40, thereby realizing intelligent management throughout the entire process of cable laying planning, optimization, risk assessment, and construction guidance, achieving the effect of improving laying efficiency and reliability.

[0066] Specifically, the path planning module 10 constructs a three-dimensional model of the substation cable channel and adopts an improved A model by introducing thermal field penalty terms and fill rate penalty terms. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints. A thermal field penalty term quantifies the distance relationship between the cable path and heat source equipment, assigning higher costs to path segments closer to the heat source to guide the algorithm in planning cable paths away from the heat source. A fill rate penalty term quantifies the fill density of the cable tray after cable laying. When the fill rate of the cable tray exceeds a preset threshold, the cost of the path through that cable tray is increased to ensure a balanced load distribution. For example, during algorithm execution, if a path segment is close to heat source equipment such as a transformer, the thermal field penalty term will significantly increase the cost of that path segment, prompting the algorithm to automatically select an alternative path away from the heat source. Simultaneously, if the current fill rate of a cable tray is close to its design limit, the fill rate penalty term will dynamically increase the cost of the path through that cable tray, guiding the algorithm to allocate some cables to cable trays with lower loads.

[0067] Furthermore, the path optimization module 20 uses the initial cable path as the initial population and employs a multi-objective genetic algorithm to iteratively optimize the cable length and component quantity, the load distribution balance of the cable tray, and the heat source avoidance distance, obtaining the optimized target laying path. The algorithm encodes each initial path as a chromosome and simulates natural selection, crossover, and mutation operations during the iteration process to gradually optimize the population. For example, in the crossover operation, some node sequences of two better paths are exchanged to generate a sub-path that combines the advantages of both; in the mutation operation, the key node positions of a path are randomly adjusted to explore a better solution space. Through multiple iterations, the algorithm finally converges to a set of comprehensively optimal paths that minimize the cable length, the number of components, the load distribution balance of the cable tray, and the heat source avoidance distance.

[0068] Based on this, the path identification module 30, using the target laying path and ambient temperature parameters, calls a pre-built physical information fusion neural network model to predict the stress state at each key node along the path. This model, during its training phase, incorporates physical constraints such as cable material properties, bending radius, elevation changes, and ambient temperature, enabling accurate stress distribution prediction. According to the predicted stress state, the path identification module 30 segments the target laying path by risk level, generating stress-identified paths that include information on the locations of high-risk segments. For example, based on the yield strength of the cable insulation material, a stress threshold is set. Path segments with predicted stress values ​​below the safety threshold are identified as safe segments, those between the safety and warning thresholds are identified as warning segments, and those exceeding the warning threshold are identified as high-risk segments. The start and end coordinates of the high-risk segments are precisely recorded.

[0069] Finally, the path adjustment module 40 overlays the stress-marked path onto the augmented reality terminal, acquiring the spatial position of the cable installation end in real time. When the actual path deviates from the target path by more than a threshold, a correction guide is generated, such as dynamically displaying an arrow on the augmented reality terminal screen indicating the direction and distance of deviation. When the installation end enters a high-risk section, a stress-sensitive construction prompt is generated, such as displaying specific operational constraints like "High-risk area, please lay slowly and avoid excessive bending," to standardize the laying operation. For example, during construction, the augmented reality terminal's built-in visual positioning system tracks the three-dimensional coordinates of the cable installation end in real time. If the installation end is detected to deviate from the target path by more than 5 centimeters, a correction guide is immediately triggered. If the installation end enters a pre-marked high-risk section, the laying speed is forcibly limited and the operator is prompted to adjust the operating force until the cable laying is completed.

[0070] Through the above technical solutions, this application achieves end-to-end digital management from path planning, optimization, risk assessment to construction guidance. Compared to traditional methods that rely solely on two-dimensional drawings and manual experience, this application avoids heat source and cable tray overload risks in the early stages of path planning through thermal field penalty terms and filling rate penalty terms. It systematically balances multiple optimization objectives through a multi-objective genetic algorithm, achieves quantitative early warning of stress risks through a physical information fusion neural network model, and finally provides real-time and accurate construction constraints with the help of augmented reality terminals. This significantly improves the planning accuracy and construction safety of cable laying, effectively reduces the difficulty of later maintenance, and improves the work efficiency and laying quality of cable laying.

[0071] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0072] Furthermore, the above provides a detailed description of the substation cable laying management method and device provided in the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for managing cable laying in substations, characterized in that, The management method includes: A three-dimensional model of the substation cable tunnel is constructed, and an improved A model is adopted by introducing thermal field penalty terms and fill rate penalty terms. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints; Using the initial cable path as the initial population, a multi-objective genetic algorithm is used to iteratively optimize the cable length and number of components, the load distribution balance of the cable tray, and the heat source avoidance distance as optimization objectives to obtain the optimized target laying path. Based on the target laying path and ambient temperature parameters, a pre-built physical information fusion neural network model is invoked to predict the stress state of each key node along the path, and the target laying path is segmented by risk level according to the predicted stress state to generate a stress-marked path containing the location information of high-risk sections. The stress identification path is overlaid and displayed on the augmented reality terminal to obtain the spatial position of the cable construction end in real time. When the actual path deviates from the target path by more than a threshold, a correction guide is generated. When the construction end enters the high-risk section, a stress-sensitive construction prompt is generated to constrain the laying operation until the cable laying is completed.

2. The substation cable laying management method according to claim 1, characterized in that, The construction of the three-dimensional model of the substation cable channel adopts an improved A model that introduces thermal field penalty terms and fill rate penalty terms. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints, including: Collect design drawings and on-site measured coordinate data of substation building structure, cable trays, and electrical equipment to obtain the spatial location and dimensional parameters of each component of the substation. Based on the collected data, a substation building structure model, a cable tray model, and an electrical equipment model were constructed sequentially. This was then improved using A... The algorithm constructs the initial cable path in the substation building structure model.

3. The substation cable laying management method as described in claim 2, characterized in that, The improved A The algorithm constructs the initial cable path in the substation building structure model, including: Determine the coordinates of the starting and ending points of the cable laying, rasterize the three-dimensional model of the substation cable channel, divide it into grid units, clarify the passability of the grid, and mark the areas where obstacles are located as impassable grids. The thermal field penalty coefficient is set according to the distance between the cable and the heat source equipment, and the thermal field penalty coefficient is set to have a negative linear relationship with the distance. Set a threshold for cable filling density in the cable tray. When the filling rate exceeds the threshold, set a filling rate penalty coefficient that increases linearly with the filling rate. By combining the thermal field penalty coefficient, the fill rate penalty coefficient, and the path length, a heuristic function is constructed. Starting from the starting point, adjacent feasible grids are searched one by one according to the principle of lowest cost until the destination is reached. Adjusting the weights of the heuristic function yields multiple conflict-free initial paths that satisfy obstacle avoidance and laying specifications.

4. The substation cable laying management method as described in claim 1, characterized in that, Using the initial cable path as the initial population, a multi-objective genetic algorithm is used to iteratively optimize the cable length and number of components, the load distribution balance of the cable tray, and the heat source avoidance distance as optimization objectives, resulting in the optimized target laying path, which includes: Each initial cable path is encoded and converted into a computable chromosome. Each segment of the chromosome corresponds to the coordinates of the path nodes, the cable laying direction, and the cable tray allocation information, forming the initial population. Four optimization indicators were set: cable length, number of components, load distribution balance of cable tray, and heat source avoidance distance. The fitness of each chromosome was calculated by combining the four optimization indicators. Chromosomes with high fitness are retained to enter the next generation of the population. Two high-quality chromosomes are exchanged using a single-point crossover method. After the crossover, the paths are checked for conflict, and chromosomes with spatial conflicts are removed. After each iteration, the average fitness of the population is calculated. When the average fitness of multiple consecutive iterations tends to stabilize, the iteration is stopped, and the path corresponding to the chromosome with the highest fitness is output as the optimized target laying path.

5. The substation cable laying management method as described in claim 4, characterized in that, The specific steps for weight adjustment and iterative optimization in a multi-objective genetic algorithm are as follows: Based on actual construction needs, the initial weighting of each optimization objective is set as follows: cable length and number of components account for 30%-40%, cable tray load distribution balance accounts for 25%-35%, and heat source avoidance distance accounts for 25%-35%. During the iteration process of the genetic algorithm, after every 10 generations, the fitness of the current population is reviewed, the optimization effect of each optimization objective is analyzed, and it is determined whether the weight allocation is in line with the actual engineering requirements. Based on the review results, if the optimization effect of a certain optimization objective does not meet expectations, the weight ratio of that objective should be adjusted appropriately, and the weights of other objectives should be adjusted accordingly to ensure that the total weight is 100%. After adjusting the weights, the iteration should continue.

6. The substation cable laying management method as described in claim 1, characterized in that, The step of predicting the stress state of each key node along the path by calling a pre-built physical information fusion neural network model based on the target laying path and ambient temperature parameters includes: The pre-built physical information fusion neural network model is retrieved, which uses historical cable laying path parameters, ambient temperature data, and measured stress data of key nodes as training samples. The training samples are normalized to eliminate the influence of data units. The training set and validation set are divided according to a preset ratio. The validation set is used to verify and adjust the model parameters to ensure that the model prediction error is controlled within a preset range. The node coordinates, directional parameters, and ambient temperature parameters of the target laying path are input into the input layer of the neural network model. The hidden layer of the model extracts features from the input data through activation functions, optimizes the network weights by incorporating the physical constraints of cable laying, and outputs the stress prediction values ​​of each key node along the target laying path from the output layer.

7. The substation cable laying management method as described in claim 1, characterized in that, The step of segmenting the target laying path according to the predicted stress state and generating a stress-marked path containing the location information of high-risk sections includes: Based on the cable material characteristics and laying specifications, three levels of standards are set: stress safety threshold, warning threshold, and danger threshold, and the stress range corresponding to each level is clearly defined. The predicted stress values ​​of each key node are compared one by one with the stress range of the three set standards. Based on the comparison results, each key node is divided into a safe node, a warning node, and a high-risk node. Based on the risk level of key nodes, the target laying path is divided into safe section, warning section and high-risk section to ensure that the risk level of nodes within the same section of the path is consistent. A segmented coloring method is used to visually identify paths with different risk levels. High-risk sections are labeled with specific node coordinates, stress prediction values, and risk causes, ultimately generating stress-labeled paths containing complete risk information.

8. The substation cable laying management method as described in claim 1, characterized in that, The process of overlaying the stress identification path onto the augmented reality terminal, obtaining the spatial position of the cable installation end in real time, generating correction guidance when the actual path deviates from the target path by more than a threshold, and generating stress-sensitive construction prompts to constrain the laying operation when the installation end enters the high-risk section, until the cable laying is completed, includes: The system employs a combination of visual positioning and inertial navigation, using the augmented reality terminal's camera to capture images of the scene while simultaneously acquiring motion data from the terminal's inertial sensors. The feature points of the on-site images are analyzed, and the three-dimensional coordinates of the cable construction end are calculated in real time by combining the motion data of the inertial sensor. The preset path deviation threshold range is 5cm~10cm. The real-time calculated three-dimensional coordinates of the construction end are compared with the coordinates of the target laying path to calculate the deviation value. If the deviation exceeds the preset threshold, the augmented reality terminal will provide correction guidance until the coordinates of the construction end return to the target path range.

9. The substation cable laying management method as described in claim 1, characterized in that, The process of overlaying and displaying the stress identification path on the augmented reality terminal, obtaining the spatial position of the cable construction end in real time, generating correction guidance when the actual path deviates from the target path by more than a threshold, and generating stress-sensitive construction prompts to constrain the laying operation when the construction end enters the high-risk section, until the cable laying is completed, also includes: The system monitors the spatial location of the construction site in real time. When the construction site is detected to have entered a high-risk section of the stress identification path, the stress-sensitive construction prompt mechanism is triggered, and stress feedback data of the construction site is collected in real time. The terminal compares the real-time collected stress feedback data with the predicted value. If the feedback stress exceeds the danger threshold, the terminal immediately generates an emergency shutdown prompt and pushes a hidden danger investigation guide. Cable laying operations can only be resumed after the hidden danger investigation is completed and the stress is restored to a safe range.

10. A substation cable laying management device, characterized in that, The management device includes: Path planning module: Used to construct 3D models of substation cable channels, employing an improved A / B model that incorporates thermal field and fill rate penalties. The algorithm generates multiple initial cable paths that satisfy spatial obstacle avoidance constraints; Path optimization module: Used as the initial population of the initial cable path, the module iteratively optimizes the cable length and number of components, the load distribution balance of the cable tray, and the heat source avoidance distance using a multi-objective genetic algorithm to obtain the optimized target laying path. Path identification module: Based on the target laying path and ambient temperature parameters, it calls a pre-built physical information fusion neural network model to predict the stress state of each key node along the path, and identifies the target laying path by risk level segmentation according to the predicted stress state, generating a stress identification path containing high-risk segment location information. Path adjustment module: used to overlay and display the stress identification path on the augmented reality terminal, obtain the spatial position of the cable construction end in real time, generate correction guidance when the actual path deviates from the target path by more than a threshold, and generate stress-sensitive construction prompts to constrain the laying operation when the construction end enters the high-risk section until the cable laying is completed.

Images