A photovoltaic power station distributed construction management method and system based on load analysis

Abstract

The application discloses a photovoltaic power station distributed construction management method and system based on load analysis, and the method comprises the following steps: outputting a dynamic load intensity thermal map according to the terrain elevation data, meteorological historical data and component mechanical parameters of a photovoltaic power station planning area; generating a wind pressure-resistant optimized construction zoning scheme based on the dynamic load intensity thermal map; outputting a conflict-free construction resource allocation matrix according to the construction zoning scheme; dynamically correcting construction parameters by using a self-adaptive particle swarm optimization algorithm based on the construction resource allocation matrix, and generating a deformation-resistant construction instruction set; driving a construction machine to execute an operation according to the deformation-resistant construction instruction set, and iteratively updating a global load distribution atlas by using a reinforcement learning model, so that a closed-loop construction management link is formed. By using the embodiment of the application, the balance between construction precision and efficiency under a complex environment can be realized by fusing multi-source data, optimizing zoning planning, intelligently scheduling equipment and closed-loop quality control.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of construction management, and particularly relates to a photovoltaic power station distributed construction management method and system based on load analysis. BACKGROUND

[0002] With the continuous expansion and complication of the construction scale of photovoltaic power stations, the traditional construction management method faces many challenges in dealing with complex terrain, dynamic weather conditions and distributed construction requirements. The existing technology mainly relies on static design drawings and manual experience for construction planning, which is difficult to respond to dynamic load changes such as sudden changes in wind speed and terrain undulations in real time, resulting in frequent construction quality problems such as support deformation and component inclination deviation. Especially in the distributed construction scene, time and space conflicts are easy to occur when multiple devices work together, resulting in low construction efficiency and resource waste. Although some research attempts to introduce finite element analysis to optimize the construction scheme, it lacks the dynamic fusion capability of real-time load data, and the adjustment of construction parameters lags behind the actual working condition changes. SUMMARY

[0003] The purpose of the present application is to provide a photovoltaic power station distributed construction management method and system based on load analysis to solve the problems in the prior art, which can balance the construction precision and efficiency in complex environments by fusing multi-source data, optimizing zoning planning, intelligently scheduling equipment and closed-loop quality control.

[0004] One embodiment of the present application provides a photovoltaic power station distributed construction management method based on load analysis, the method comprising:

[0005] According to the terrain elevation data, meteorological historical data and component mechanical parameters of the photovoltaic power station planning area, a multi-modal fusion deep neural network is used to fuse satellite remote sensing images and ground sensor time series data, and a dynamic load intensity heat map is output;

[0006] Based on the dynamic load intensity heat map, a gradient-constrained spectral clustering algorithm is used to divide the construction sub-area, and the clustering center is dynamically adjusted according to the load extreme point and the terrain mutation feature, and an anti-wind pressure optimized construction zoning scheme is generated;

[0007] According to the construction zoning scheme, combined with the real-time state data of the construction equipment, a space-time conflict resolution algorithm is used to generate an equipment scheduling path, and the algorithm outputs a conflict-free construction resource allocation matrix through dynamic time window constraints and load balancing objective functions;

[0008] Based on the construction resource allocation matrix, a digital twin model is used to simulate the construction process, real-time collection of support deformation data and component inclination data is performed, and an adaptive particle swarm optimization algorithm is used to dynamically correct the construction parameters, and an anti-deformation construction instruction set is generated;

[0009] According to the anti-deformation construction instruction set, the construction machinery is driven to perform operations, and the energy efficiency feedback data is analyzed synchronously through the edge computing node, the global load distribution atlas is iteratively updated by using the reinforcement learning model, and a closed-loop construction management link is formed.

[0010] Optionally, according to the terrain elevation data of the photovoltaic power station planning area, the meteorological historical data and the component mechanical parameters, a dynamic load intensity thermal map is output by a multimodal fusion deep neural network, which fuses satellite remote sensing images and ground sensor time series data, including:

[0011] According to the satellite remote sensing image and the terrain elevation data, a terrain-component mapping grid is generated by a geometric registration algorithm, and a spatial grid is output after eliminating cloud interference;

[0012] The spatial grid and the real-time wind speed, temperature and humidity data collected by the ground sensor are input into a spatio-temporal interpolation model, and a three-dimensional dynamic meteorological field is generated in combination with the meteorological historical data;

[0013] The three-dimensional dynamic meteorological field and the component mechanical parameters are input into a multimodal fusion network, terrain texture features are extracted by residual convolution, meteorological time series patterns are extracted by bidirectional LSTM, and an initial load prediction map is output after fusion;

[0014] The initial load prediction map is subjected to spatio-temporal noise filtering guided by optical flow, the intensity distribution is optimized according to the load correlation of adjacent grids, and a dynamically updated load intensity thermal map is generated.

[0015] Optionally, based on the dynamic load intensity thermal map, a gradient-constrained spectral clustering algorithm is used to divide the construction sub-area, the clustering center is dynamically adjusted according to the load extreme point and the terrain mutation feature, and an anti-wind pressure optimized construction zoning scheme is generated, including:

[0016] According to the dynamic load thermal map, a gradient field is calculated, and a region with a load change rate exceeding a threshold value is extracted as a gradient mutation boundary;

[0017] Based on the gradient mutation boundary, a region similarity matrix is constructed, and an adaptive Gaussian kernel function is used to quantify the load correlation strength of adjacent regions;

[0018] The region similarity matrix is input into a spectral clustering algorithm, the initial clustering center position is constrained by the terrain elevation mutation point, and preliminary construction sub-area division is performed;

[0019] The clustering center is dynamically adjusted according to real-time wind speed prediction data, so that the high load area is offset to the low wind pressure direction, and the construction sub-area boundary is optimized;

[0020] The optimized sub-area is subjected to finite element wind pressure simulation verification, and a construction zoning scheme meeting the critical wind pressure threshold is output.

[0021] Optionally, the construction resource allocation matrix is mapped to the digital twin model to drive the virtual construction machinery to perform work according to the scheduling path.

[0022] The construction zoning scheme is converted into a task node graph with time constraints, and the nodes of the task node graph include device type requirements, operation time length, and safety load threshold.

[0023] According to the device GPS positioning data and the energy consumption log, the device capability-position matrix is constructed, and the space-time cost of the device moving to each node is calculated.

[0024] Based on the task node graph and the device capability-position matrix, the dynamic time window is used to detect device path conflicts, and the initial path sequence is generated through priority scheduling.

[0025] Taking load balancing as the objective function, the mixed integer programming solver is used to optimize the initial path sequence, and the device-task allocation matrix and conflict-free scheduling path are output.

[0026] Optionally, based on the construction resource allocation matrix, the construction process is simulated through a digital twin model, real-time support frame deformation data and component inclination data are collected, an adaptive particle swarm optimization algorithm is used to dynamically correct construction parameters, and an anti-deformation construction instruction set is generated.

[0027] The construction resource allocation matrix is mapped to the digital twin model to drive the virtual construction machinery to perform work according to the scheduling path.

[0028] Real-time support frame deformation sensor data and component inclination data are collected to generate a millimeter-level precision deformation-inclination joint distribution graph.

[0029] According to the deformation-inclination joint distribution graph, define the deformation peak, inclination uniformity and progress deviation as the multi-objective fitness function, and input the adaptive particle swarm optimization algorithm.

[0030] According to the real-time deformation data, the search direction of the particle swarm is dynamically adjusted, and the optimal mechanical control parameters are solved under the constraints of construction specifications.

[0031] The optimal mechanical control parameters are encoded into an anti-deformation construction instruction set, and the integrity and traceability of the instructions are ensured through blockchain signature.

[0032] Optionally, according to the anti-deformation construction instruction set, the construction machinery is driven to perform operations, and at the same time, the energy efficiency feedback data is analyzed through the edge computing node, the global load distribution graph is iteratively updated using the reinforcement learning model, and a closed-loop construction management link is formed.

[0033] The anti-deformation construction instruction set is parsed into mechanical control signals to drive the hydraulic system and the mechanical arm to perform high-precision construction operation.

[0034] The edge computing node is used to collect installation error and mechanical energy consumption data in real time to generate energy efficiency-quality evaluation indexes.

[0035] The hierarchical reinforcement learning model is trained based on the energy efficiency-quality evaluation indexes, the upper layer optimizes load prediction parameters, the lower layer adjusts mechanical motion trajectories, the global load distribution atlas is iteratively updated, and a closed-loop construction management link is formed.

[0036] Another embodiment of the present application provides a photovoltaic power station distributed construction management system based on load analysis, which comprises:

[0037] The fusion module is used to fuse satellite remote sensing images and ground sensor time series data by a multimodal fusion deep neural network according to topographic elevation data, meteorological historical data and component mechanical parameters of the photovoltaic power station planning area, and output a dynamic load intensity thermal map.

[0038] The division module is used to divide construction sub-areas by using a gradient-constrained spectral clustering algorithm based on the dynamic load intensity thermal map, dynamically adjust the clustering center according to the load extreme point and the topographic mutation feature, and generate an anti-wind pressure optimized construction zoning scheme.

[0039] The generation module is used to generate an equipment scheduling path by using a space-time conflict resolution algorithm according to the construction zoning scheme and in combination with real-time state data of construction equipment, and output a conflict-free construction resource allocation matrix by using a dynamic time window constraint and a load balancing objective function.

[0040] The correction module is used to simulate a construction process by a digital twin model based on the construction resource allocation matrix, collect support deformation data and component inclination data in real time, dynamically correct construction parameters by using an adaptive particle swarm optimization algorithm, and generate an anti-deformation construction instruction set.

[0041] The update module is used to drive a construction machine to perform an operation according to the anti-deformation construction instruction set, analyze energy efficiency feedback data synchronously by an edge computing node, iteratively update a global load distribution atlas by using a reinforcement learning model, and form a closed-loop construction management link.

[0042] Another embodiment of the present application provides a storage medium having a computer program stored therein, wherein the computer program is configured to execute the method described in any one of the above embodiments when running.

[0043] Yet another embodiment of the present application provides an electronic device comprising a memory having a computer program stored therein and a processor configured to execute the computer program to perform the method described in any of the above embodiments.

[0044] Compared with the prior art, the photovoltaic power station distributed construction management method based on load analysis provided by the present application can output a dynamic load intensity thermal map according to the topographic elevation data of the photovoltaic power station planning area, meteorological historical data and component mechanical parameters; generate a wind pressure-resistant optimized construction zoning scheme based on the dynamic load intensity thermal map; output a conflict-free construction resource allocation matrix according to the construction zoning scheme; dynamically correct construction parameters based on the construction resource allocation matrix by using a self-adaptive particle swarm optimization algorithm to generate a deformation-resistant construction instruction set; drive the construction machinery to perform operations according to the deformation-resistant construction instruction set, and simultaneously analyze energy efficiency feedback data through an edge computing node, iteratively update a global load distribution atlas by using a reinforcement learning model, and form a closed-loop construction management link, so that the construction precision and efficiency in a complex environment can be balanced by fusing multi-source data, optimizing zoning planning, intelligently scheduling equipment and closed-loop quality control. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The hardware structure block diagram of the computer terminal of the photovoltaic power station distributed construction management method based on load analysis provided by the embodiment of the present application is shown in the figure.

[0046] Figure 2 The flowchart of the photovoltaic power station distributed construction management method based on load analysis provided by the embodiment of the present application is shown in the figure.

[0047] Figure 3 The structure diagram of the photovoltaic power station distributed construction management system based on load analysis provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0048] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be explained as a limitation of the present application.

[0049] The embodiment of the present application first provides a photovoltaic power station distributed construction management method based on load analysis, which can be applied to an electronic device such as a computer terminal, specifically a general computer and the like.

[0050] The computer terminal based on load analysis provided by the embodiment of the present application is described in detail below. Figure 1 The hardware structure block diagram of the computer terminal of the photovoltaic power station distributed construction management method based on load analysis provided by the embodiment of the present application is shown in the figure. Figure 1As shown, the computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory can include a non-volatile storage medium and an internal memory.

[0051] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions which, when executed, can cause the processor to perform any one of the load analysis-based photovoltaic power station distributed construction management methods.

[0052] The processor is used to provide computing and control capabilities to support the operation of the entire computer device.

[0053] The internal memory provides an environment for the execution of the computer program in the non-volatile storage medium, which, when executed by the processor, can cause the processor to perform any one of the load analysis-based photovoltaic power station distributed construction management methods.

[0054] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that, Figure 1 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0055] It should be understood that the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0056] Referring to Figure 2 The embodiments of the present application provide a load analysis-based photovoltaic power station distributed construction management method, which can include the following steps:

[0057] S201, according to the topographic elevation data of the photovoltaic power station planning area, the meteorological historical data and the component mechanical parameters, through a multi-modal fusion deep neural network, satellite remote sensing images and ground sensor time series data are fused to output a dynamic load intensity thermal map;

[0058] Specifically, a terrain-component mapping grid can be generated by a geometric registration algorithm based on satellite remote sensing images and terrain elevation data, and the spatial grid is output after eliminating cloud interference;

[0059] Satellite remote sensing images typically contain spectral information and terrain profiles of the ground, but may result in missing data in some areas due to cloud cover. Terrain elevation data is collected by LiDAR or InSAR, providing accurate altitude information. Geometric registration algorithms are used to align satellite images with terrain data in the spatial coordinate system, ensuring pixel-level matching of the two.

[0060] Data preprocessing and registration process:

[0061] Satellite image cloud removal:

[0062] A multi-temporal satellite image fusion technique is used, such as selecting cloud-free image segments for three consecutive days, and generating a complete ground cover map through sliding window mosaic. If a certain area is continuously covered by clouds (such as mountainous areas), historical images of the same period are called to fill in the gaps. For example, using 10-meter resolution images of the Sentinel-2 satellite, the time window is from January 1 to January 3, 2023, and the sliding window size is 512x512 pixels.

[0063] Terrain elevation data alignment:

[0064] Terrain data is usually stored in point cloud form and needs to be converted to raster format (such as GeoTIFF) with the same resolution as satellite images (10 meters). Through geometric registration algorithms (such as affine transformation), the latitude and longitude coordinates of satellite images are aligned with the UTM coordinate system (Universal Transverse Mercator) of terrain data. For example, the registration error is controlled within 0.5 pixels to ensure that the altitude and ground features are accurately matched.

[0065] Terrain-component mapping grid generation:

[0066] The standard size of photovoltaic modules (such as 2 meters x 1 meter) is superimposed on the terrain grid to generate a terrain-component mapping grid. Each grid cell corresponds to a module installation location, recording the center point coordinates (X, Y, Z), where Z is the altitude. For example, a 1000x1000 meter planning area is divided into 500x500 grid cells (each cell is 2 meters x 2 meters).

[0067] Cloud interference elimination technique:

[0068] ‌Cloud Detection Algorithm‌:

[0069] Based on the Short-Wave Infrared (SWIR) band and the Visible band ratio of satellite imagery (e.g., B11 / B04), set a threshold (e.g., ratio > 0.8) to mark cloud pixels. For example, if a certain area has a 30% cloud coverage, it is marked as a region that needs repair.

[0070] ‌Data Filling Strategy‌:

[0071] For cloud-covered areas, use Nearest Neighbor Interpolation or historical data to fill in the missing data. For example, if a grid cell is missing data due to cloud cover, take the average elevation value of the neighboring 8 cells (e.g., 352.6 meters) to fill in the missing data.

[0072] ‌Spatial Grid Output and Verification‌:

[0073] ‌Grid Format‌: The output is a GeoJSON file containing the latitude, longitude, elevation, slope, aspect, and cloud repair flag for each grid cell. For example, Grid ID_001 has the following attributes: {latitude 39.9°, longitude 116.5°, elevation 352 meters, slope 5°, aspect southeast, cloud repair: no}.

[0074] ‌Accuracy Verification‌: Randomly select 5% of the grid cells and compare them with field measurement data (such as differential GPS measurement) to ensure that the elevation error is ≤0.3 meters and the latitude and longitude offset is ≤0.0001 degrees.

[0075] Input the spatial grid and real-time wind speed, temperature, and humidity data collected by ground sensors into the spatiotemporal interpolation model, and combine it with historical weather data to generate a three-dimensional dynamic weather field;

[0076] ‌Ground Sensor‌: Deploy ground sensors in the photovoltaic power station area to transmit real-time data via a wireless sensor network (such as LoRaWAN).‌ Spatiotemporal Interpolation Model‌: Combine the spatial grid location and sensor data to predict weather parameters in areas where sensors have not been deployed.

[0077] ‌Sensor Data Collection and Fusion‌:

[0078] ‌Sensor Types and Deployment‌:

[0079] Anemometer sensors are installed on the top of the support, and DHT22 temperature and humidity sensors are placed 1.5 meters above the ground. Deploy them at a density of one node per 100 meters. For example, deploy 121 sensors (11x11 grid) in a 1000x1000 meter area.

[0080] ‌Real-Time Data Preprocessing‌:

[0081] Remove outliers (e.g., wind speed > 30 m / s or humidity > 100%) and smooth temporal fluctuations using Kalman Filter. For example, if a sensor's instantaneous wind speed jumps to 25 m / s, it is corrected to 18 m / s after filtering.

[0082] ‌Temporal and spatial interpolation algorithm execution‌:

[0083] ‌Kriging Interpolation‌:

[0084] Based on the Semivariogram, quantify spatial correlation and predict wind speed and temperature and humidity at unsampled points. For example, set the parameters of the Spherical Model: Range is 200 meters, Nugget is 0.1, and Sill is 1.5.

[0085] ‌Temporal dimension expansion‌:

[0086] Introduce Time Decay Factor to integrate historical meteorological data (e.g., hourly records for the past 7 days) with real-time data. For example, the current time interpolation weight is 0.8, and the data weight 1 hour ago is 0.2.

[0087] ‌Three-dimensional dynamic meteorological field construction‌:

[0088] ‌Vertical stratification modeling‌:

[0089] Layered by height (e.g., 0 meters, 10 meters, 50 meters), predict wind speed gradients at different altitudes. For example, ground wind speed is 5 m / s, 10 meters height is 6 m / s, and 50 meters height is 8 m / s.

[0090] ‌Dynamic update mechanism‌:

[0091] Re-execute interpolation every 5 minutes to generate a time series of three-dimensional meteorological fields. The output format is a NetCDF file containing timestamp, latitude, longitude, height, wind speed, temperature, and humidity fields.

[0092] Input the three-dimensional dynamic meteorological field and component mechanical parameters into the multi-modal fusion network, extract terrain texture features through residual convolution, extract meteorological time series patterns through bidirectional LSTM, and output the initial load prediction map after fusion;

[0093] Mechanical parameters of the assembly include the stiffness of the photovoltaic panel, the wind load coefficient, and the yield strength of the support material. The multi-modal fusion network consists of a residual convolutional neural network (ResNet) and a bidirectional long short-term memory network (Bi-LSTM), which process spatial and temporal features, respectively.

[0094] Residual convolutional network (ResNet) design:

[0095] Input data format: Terrain grid data (elevation, slope) is converted into a grayscale image (512x512 pixels), and meteorological field data (wind speed, temperature and humidity) is used as a multi-channel image (4 channels).

[0096] Residual block structure: Each residual block contains two 3x3 convolution layers and a skip connection. For example, the first residual block takes a 4-channel image as input and outputs a 64-channel feature map.

[0097] Terrain feature extraction: Multi-layer convolution is used to capture terrain texture (such as steep slopes and gullies), outputting a 512-dimensional feature vector.

[0098] Bidirectional LSTM (Bi-LSTM) temporal modeling:

[0099] Input time series data: Time series of three-dimensional meteorological fields (such as the past 1 hour, every 5 minutes) are expanded into time series vectors by grid cells. For example, the wind speed sequence of a certain grid is [5.0, 5.2, 5.5,..., 6.0] m / s.

[0100] Bidirectional processing: Forward LSTM scans time series data from past to future, and backward LSTM scans from future to past, capturing meteorological trends (such as rising wind speed). Output a 256-dimensional time series feature vector.

[0101] Multi-modal feature fusion and load prediction:

[0102] Feature concatenation: Concatenate the 512-dimensional terrain features of ResNet and the 256-dimensional meteorological features of Bi-LSTM into a 768-dimensional joint feature vector.

[0103] Fully connected layer prediction: Use three fully connected layers (768→256→64→1) to regress the load intensity of each grid (unit: kN / m 2 ). For example, the predicted load of a certain grid is 1.2kN / m 2 .

[0104] ‌Initial Load Prediction Map Generation‌: Render the full grid prediction results as a heat map, with color mapping (Colormap) from blue (low load) to red (high load).

[0105] ‌Temporal-Spatial Noise Filtering with Optical Flow Guidance‌: Optimize the intensity distribution based on the load correlation of adjacent grids, generate a dynamically updated load intensity heat map.

[0106] ‌Optical Flow Guided Filtering‌: Utilize the load change direction (Optical Flow Vector) of adjacent frames (temporal) and grids (spatial) to guide the filtering process, eliminate prediction noise and enhance spatial continuity.

[0107] ‌Optical Flow Estimation‌:

[0108] ‌Optical Flow Algorithm Selection‌: Use Lucas-Kanade method, calculate the optical flow vector (direction and magnitude) based on the temporal changes of load prediction map. For example, the load of a certain grid changes from 1.2kN / m 2 to 1.5kN / m 2 in 5 minutes, the optical flow vector points northeast, with a magnitude of 0.3kN / m 2 .

[0109] ‌Optical Flow Field Generation‌: Output as a vector field image, each pixel represents the direction and size of load change.

[0110] ‌Temporal-Spatial Noise Filtering Execution‌:

[0111] ‌Spatial Filtering‌: Use adaptive Gaussian filter, adjust the filter kernel size according to the optical flow magnitude. For example, use small kernel (3x3) in areas with large optical flow magnitude (load changes dramatically), use large kernel (7x7) in smooth areas.

[0112] ‌Temporal Filtering‌: Perform moving average on the load prediction map of the last 5 frames (25 minutes), suppress random noise. For example, the load of a certain grid in consecutive frames is [1.2, 1.3, 1.25, 1.28, 1.3], after averaging, it is corrected to 1.27kN / m 2 .

[0113] ‌Load Intensity Heat Map Optimization‌:

[0114] ‌Edge-Preserving Enhancement‌: Use bilateral filter to sharpen the load boundaries of terrain abrupt changes (such as the junction of steep slope and flat ground), while smoothing the flat areas.

[0115] ‌Dynamic Update Mechanism: New data is received every 5 minutes, re-executes the filtering and updates the heat map. The output is a Web Map Service (WMS) layer, supporting real-time visualization and API calls.

[0116] ‌Key Technology Examples and Parameter Descriptions:

[0117] ‌LiDAR: Laser Radar, used for high-precision topographic mapping, with an elevation data error of ≤0.1 meters.

[0118] ‌UTM Coordinate System: Universal Transverse Mercator Projection, with zone numbers such as Zone 50N, suitable for mid-latitude regions.

[0119] ‌LoRaWAN: Long Range Low Power Wireless Network, with a transmission distance of up to 10 kilometers, suitable for field sensor deployment.

[0120] ‌ResNet Residual Block: Convolutional module with skip connections, alleviating the problem of gradient vanishing in deep networks.

[0121] ‌Bi-LSTM: Bidirectional Long Short-Term Memory Network, capable of capturing the forward and backward dependencies of time series data.

[0122] This method first integrates multi-dimensional data sources, including terrain elevation, weather records, and photovoltaic component mechanical properties, and realizes the spatio-temporal alignment and feature extraction of satellite remote sensing images and ground sensor data through a deep neural network. The multi-modal fusion network uses residual convolutional structures to capture terrain spatial features, combined with bidirectional LSTM to handle weather temporal changes, ultimately generating intensity heat maps reflecting dynamic load distribution such as wind pressure and snow load, breaking through the limitations of traditional static load analysis and achieving fine-grained prediction of photovoltaic power station global dynamic load. The heat map provides a data basis for subsequent construction zoning, ensuring that construction planning matches the real environment load and reducing the risk of structural failure from the source.

[0123] S202, based on the dynamic load intensity heat map, a gradient-constrained spectral clustering algorithm is used to divide the construction sub-area, and the clustering center is dynamically adjusted according to the load extreme point and terrain mutation characteristics, to generate an anti-wind pressure optimized construction zoning scheme;

[0124] Specifically, the gradient field can be calculated based on the dynamic load heat map, and the area where the load change rate exceeds the threshold is extracted as the gradient mutation boundary;

[0125] ‌Gradient Field: By calculating the rate of change of dynamic load intensity in space, it identifies areas with sharp fluctuations in load.‌ Gradient Mutation Boundary: The boundary between adjacent grid cells where the difference in load intensity exceeds a pre-set threshold, representing the potential basis for construction zoning.

[0126] Gradient field calculation method:

[0127] Sobel operator application: Apply Sobel edge detection operator (a convolution-based image gradient calculation method) to calculate the gradient magnitude and direction of each grid cell on the three-dimensional dynamic load thermograph. For example, the horizontal direction convolution kernel is [-1, 0, 1; -2, 0, 2; -1, 0, 1], and the vertical direction is [-1, -2, -1; 0, 0, 0; 1, 2, 1], respectively calculating the X / Y direction gradient components.

[0128] Gradient magnitude formula: Gradient magnitude = √(Gx 2 + Gy 2 ), where Gx is the X direction gradient and Gy is the Y direction gradient. For example, a certain grid Gx = 0.8kN / m 2 ·m, Gy = 1.2kN / m 2 ·m, then the gradient magnitude ≈ 1.44kN / m 2 ·m (unit: load change per meter length).

[0129] Gradient direction calculation: Gradient direction angle θ = arctan(Gy / Gx), used to identify the load change trend (such as southeast, northwest).

[0130] Gradient mutation boundary extraction:

[0131] Threshold setting: Based on historical data statistics, set the gradient magnitude threshold to be 2 times the standard deviation of the normal load change rate (such as a certain area with normal gradient range 0-0.5kN / m 2 ·m, threshold set to 0.5×2 = 1.0kN / m 2 ·m).

[0132] Boundary marking: Scan the gradient field and mark all grid cells with gradient magnitude ≥ 1.0kN / m 2 ·m, merge adjacent cells to form continuous boundary lines. For example, a certain area boundary consists of 3 discontinuous line segments, total length 120 meters.

[0133] Noise filtering: Remove isolated small area mutation regions (such as single grid cell or length <5 meters line), and retain the main mutation boundary. For example, filter out the isolated line segment with length 3 meters, and retain the boundary with length ≥10 meters.

[0134] Verification and visualization:

[0135] Manual sampling verification: Randomly select 5% of the boundary line segments, compare with the actual terrain mutation points (such as steep slopes, ridge lines), and ensure that the boundary matching degree with the actual terrain is ≥90%.

[0136] ‌Heat map overlay display‌: Overlay the gradient mutation boundary with the load heat map on the GIS platform, intuitively showing the spatial distribution of high load gradient areas.

[0137] Based on the gradient mutation boundary, a regional similarity matrix is constructed, and an adaptive Gaussian kernel function is used to quantify the correlation strength of adjacent regions;

[0138] ‌The regional similarity matrix (Similarity Matrix) is used to describe the correlation strength between different construction sub-regions, and is the input basis of the spectral clustering algorithm. The adaptive Gaussian kernel function dynamically adjusts the kernel function parameters according to the gradient boundary, enhancing the similarity weight of local regions.

[0139] ‌Region division and labeling‌:

[0140] ‌Region definition‌: Divide the photovoltaic power station planning area into N adjacent sub-regions (such as N=50), and each sub-region is surrounded by a gradient mutation boundary. For example, region A contains 100 grid cells, and region B contains 85 cells.

[0141] ‌Region attribute assignment‌: Record the average load intensity, area, terrain complexity (slope standard deviation) and other attributes of each sub-region. For example, region A has an average load of 1.5kN / m 2 , area 2000 square meters, slope standard deviation 8°.

[0142] ‌Adaptive Gaussian kernel function design‌:

[0143] ‌Kernel function formula‌: Similarity S(i,j)=exp(-d 2 (i,j) / (σ_iσ_j)), where d(i,j) is the Euclidean distance between regions i and j (comprehensive load intensity, slope difference), and σ_i is the adaptive bandwidth parameter of region i.

[0144] ‌Bandwidth parameter σ adjustment‌: For regions adjacent to the gradient mutation boundary, reduce σ to enhance local similarity; for flat areas, increase σ to expand the correlation range. For example, the boundary region σ=50 meters, and the flat region σ=200 meters.

[0145] ‌Similarity matrix generation‌: Calculate the similarity scores of all region pairs to generate an N×N matrix. For example, the similarity between regions A and B is 0.7, A and C is 0.3, and B and C is 0.5.

[0146] ‌Matrix optimization and sparsification‌:

[0147] Sparse connection strategy: Only keep edges with similarity score ≥ 0.5 (region association), reduce computational complexity. For example, select 300 valid connections from the original 50x50=2500 elements.

[0148] Symmetry correction: Force the similarity matrix to be symmetric (S(i,j)=S(j,i)), avoid clustering bias caused by one-way association.

[0149] Input the region similarity matrix into the spectral clustering algorithm, and constrain the initial clustering center position by the terrain elevation mutation point, and perform preliminary construction sub-region division;

[0150] Spectral clustering (Spectral Clustering) is a clustering method based on graph theory, which realizes data division through eigenvalue decomposition of Laplacian matrix. Terrain elevation mutation points (such as cliffs and gullies) as prior knowledge to constrain the clustering center position, ensure that the construction partition matches the terrain risk area.

[0151] Laplacian matrix construction:

[0152] Degree matrix (Degree Matrix): Diagonal matrix D, diagonal element D(i,i)=Σ_jS(i,j), represents the total similarity of region i. For example, D(A,A)=0.7+0.3+...+0.2=4.5 for region A.

[0153] Laplacian matrix calculation: Standardize the Laplacian matrix L=I-D^(-1 / 2)SD^(-1 / 2), where I is the identity matrix. For example, the dimension of matrix L is 50x50.

[0154] Eigenvalue decomposition and clustering initialization:

[0155] Feature vector extraction: Perform eigenvalue decomposition on matrix L, and select the top k smallest eigenvalues corresponding to the eigenvectors (k is the preset number of clusters, such as k=5). For example, the eigenvector dimension is 50x5.

[0156] Terrain-constrained clustering center: From the eigenvector space, preferentially select regions close to the elevation mutation point as the initial clustering center. For example, select 3 as the initial center in the region where the terrain mutation point (slope>15°) is located.

[0157] k-means clustering execution:

[0158] Feature vector dimension reduction: Arrange the top k eigenvectors as an Nxk matrix by row as input data. For example, 50 regions x 5 eigenvectors.

[0159] k-means iterative optimization: Randomly initialize cluster centers (or specify terrain-constrained centers), iteratively update center positions until convergence. For example, after 10 iterations, the cluster centers move less than 0.001, stop the calculation.

[0160] Preliminary construction sub-area division: Output the cluster label of each area, and merge the areas with the same label into construction sub-areas. For example, label 1 corresponds to sub-areas A, C, E, and label 2 corresponds to B, D.

[0161] According to the real-time wind speed prediction data, dynamically adjust the cluster center to make the high load area deviate to the low wind pressure direction, and optimize the construction sub-area boundary;

[0162] Real-time wind speed prediction: The future 1-hour wind speed and direction data are provided by a meteorological model (such as WRF, Weather Research and Forecasting Model). Low wind pressure direction: areas with lower wind speed or wind direction that forms a shielding effect with the terrain, which can reduce the construction risk.

[0163] Wind speed-load coupling analysis:

[0164] Wind pressure calculation formula: Wind pressure P = 0.5 × ρ × v 2 × C_p, where ρ is the air density (1.225 kg / m 3 ), v is the wind speed, and C_p is the wind pressure coefficient (related to the inclination of the component). For example, when the wind speed is 10 m / s, P = 0.5 × 1.225 × 10 2 × 1.0 ≈ 61.25 N / m 2 .

[0165] Load-wind pressure superposition: Superimpose dynamic load intensity and wind pressure to generate a comprehensive stress distribution map. For example, a certain area originally has a load of 1.2 kN / m 2 , superimposed with wind pressure of 0.06 kN / m 2 , total load of 1.26 kN / m 2 .

[0166] Cluster center dynamic adjustment strategy:

[0167] Wind pressure risk assessment: Mark areas with comprehensive stress exceeding the safety threshold (such as 1.5 kN / m 2 ) as high-risk areas, and adjust the cluster center to move away from the area.

[0168] Offset direction selection: According to the wind direction prediction (such as the dominant wind direction in the next 1 hour is northwest), adjust the cluster center of the high-risk area to the southeast (leeward side). For example, the original center coordinates (X = 100, Y = 200) are adjusted to (X = 105, Y = 205).

[0169] ‌Offset Calculation‌: Offset Δ = α × v 2 where α is the adjustment coefficient (e.g., α = 0.001 m -1 s 2 ), and v is the predicted wind speed. For example, when v = 12 m / s, Δ = 0.001 × 12 2 ≈ 0.144 meters.

[0170] ‌Sub-region Boundary Optimization‌:

[0171] ‌Boundary Redivision‌: Re-execute k-means clustering based on new cluster centers, update construction sub-region boundaries. For example, the boundary of sub-region A is contracted by 10 meters north, forming a new block A'.

[0172] ‌Continuity Constraint‌: Ensure that the adjusted sub-region boundaries are continuous, avoiding fragmented segmentation. For example, merge isolated blocks with an area <500 square meters to adjacent sub-regions.

[0173] Perform finite element wind pressure simulation verification on the optimized sub-regions, output a construction zoning scheme that meets the critical wind pressure threshold.

[0174] ‌Finite Element Analysis (FEA)‌: Discretize the structure model to simulate stress distribution under wind pressure load, verify the wind pressure resistance performance of the construction zoning.‌Critical wind pressure threshold‌ is determined by photovoltaic component mechanical parameters (such as support yield strength).

[0175] ‌Finite Element Model Construction‌:

[0176] ‌Geometry Modeling‌: Based on the 3D terrain data (elevation, slope) of the construction sub-region, establish a support-component assembly model in ANSYS or COMSOL. For example, a sub-region model contains 500 support nodes and 2000 triangular mesh elements.

[0177] ‌Material Property Definition‌: Set the support material to Q235 steel (elastic modulus 210 GPa, Poisson's ratio 0.3), and the component glass panel (elastic modulus 70 GPa, Poisson's ratio 0.23).

[0178] ‌Boundary Condition Loading‌: Fix the bottom of the support (full constraint), and apply wind pressure load on the component surface (according to the wind pressure distribution map calculated in step 4).

[0179] ‌Wind Pressure Simulation and Stress Analysis‌:

[0180] ‌Statics solution: Calculate stress distribution of the scaffold under various wind pressure loads, identify the maximum equivalent stress (Von Mises Stress). For example, the maximum stress in a certain area is 180 MPa, which is lower than the yield strength of Q235 steel, 235 MPa.

[0181] ‌Critical threshold verification: If the maximum stress of all sub-regions is ≤ critical threshold (e.g. 200 MPa), the zoning scheme is considered qualified; otherwise, return to step 4 to adjust the cluster center again.

[0182] ‌Simulation result optimization iteration:

[0183] ‌High stress area marking: Mark sub-regions with maximum stress exceeding the threshold, generate a risk distribution map. For example, sub-region B has a maximum stress of 210 MPa, which needs to be re-divided.

[0184] ‌Zoning scheme update: According to the simulation results, further shift the cluster center of high-risk areas, repeat steps 4-5 until all areas pass the verification.

[0185] ‌Key technology example and parameter description:

[0186] ‌Sobel operator: Edge detection convolution kernel, used to calculate image gradient.

[0187] ‌k-means clustering: Distance-based unsupervised clustering algorithm, iteratively optimize center position.

[0188] ‌FEA (Finite Element Analysis): Numerical simulation method for structural mechanics analysis.

[0189] ‌Q235 steel: A carbon structural steel with a yield strength of 235 MPa.

[0190] ‌Von Mises stress: An equivalent strength indicator that integrates multi-directional stress.

[0191] Through the thermal map gradient field to identify the load mutation boundary, combined with the terrain elevation change characteristics, using the improved spectral clustering algorithm to dynamically adjust the construction sub-region division. The algorithm automatically classifies high load areas and steep terrain areas, and optimizes the cluster center position according to real-time wind speed prediction, so that the zoning boundary conforms to the wind pressure distribution law, realizes the wind pressure adaptive optimization of construction zoning, and avoids the mechanical arrangement in high-risk areas. The dynamic adjustment mechanism improves the response ability of the zoning scheme to sudden weather changes, ensuring the balance between construction safety and efficiency.

[0192] S203, according to the construction zoning scheme, combined with the real-time state data of construction equipment, using the space-time conflict resolution algorithm to generate equipment scheduling path, the algorithm through dynamic time window constraint and load balance objective function, output conflict-free construction resource allocation matrix;

[0193] Specifically, the construction zoning scheme can be converted into a task node graph with time constraints, the nodes of the task node graph containing equipment type requirements, operation time, and safety load threshold values;

[0194] The task node graph with time constraints is to abstract the construction zoning scheme of the photovoltaic power station into a network structure with time attributes, each node representing a construction task, and the edges representing the dependency relationship or movement path between tasks. The equipment type requirement refers to the type of machinery required to complete the task (such as a crane, forklift, pile driver); the operation time refers to the estimated time required for the task; and the safety load threshold value refers to the maximum mechanical load allowed in the region (unit: kN / m 2 ), preventing overloading of construction equipment from causing ground collapse or damage to the support.

[0195] Node attribute definition and mapping:

[0196] Construction zoning mapping: Each construction sub-region (such as sub-region A, B) corresponds to one or more task nodes. For example, sub-region A needs to complete two tasks of support installation and component laying, creating nodes A1 (support installation) and A2 (component laying) respectively.

[0197] Equipment type requirement: specify equipment according to process requirements. For example, support installation node A1 requires a crane (Crane) and a pile driver (Pile Driver), and component laying node A2 requires a forklift (Forklift) and a drone (Drone).

[0198] Operation time calculation: based on historical construction data and area estimation. For example, the operation time of support installation node A1 = area (500m 2 ) ÷ crane efficiency (50m 2 / h) = 10 hours.

[0199] Safety load threshold value setting: determined according to the results of finite element simulation (step 3). For example, the foundation bearing capacity of sub-region A is 15kN / m 2 , and the safety threshold is set to 12kN / m 2 (reserve 20% safety margin).

[0200] Time constraint modeling:

[0201] Task dependency relationship: define the sequence between tasks. For example, support installation node A1 must be completed before component laying node A2.

[0202] ‌Time Window Constraint‌: Set the task executable time period combined with weather forecast data (e.g., no aerial work in rainy days). For example, the time window of crane operation node A1 is 8:00-18:00 daily (avoiding the night strong wind period).

[0203] ‌Task Node Graph Generation and Verification‌:

[0204] ‌Graph Structure Generation‌: Use a graph database (e.g., Neo4j) to store the relationship between nodes and edges. For example, node A1 is connected to A2, and the edge attribute is "must be started after A1 is completed."

[0205] ‌Logical Verification‌: Check if there are circular dependencies or time window conflicts. For example, if the start time of node A2 is earlier than the end time of A1, trigger an alarm and adjust the time window.

[0206] According to the device GPS positioning data and energy consumption log, construct the device capability-position matrix, and calculate the space-time cost of the device moving to each node;

[0207] ‌Device Capability-Position Matrix‌ is a two-dimensional table, with rows representing devices and columns representing task nodes, and element values representing the comprehensive cost of the device moving to the node and executing the task. The space-time cost includes moving time, energy consumption cost, and device capability matching degree.

[0208] ‌Device Capability Parameter Definition‌:

[0209] ‌Mechanical Capability Dimensions‌: Load capacity (Max Load, unit: tons), moving speed (Speed, unit: km / h), energy efficiency (Energy Efficiency, unit: kWh / km). For example, the load capacity of the crane is 20 tons, the moving speed is 5 km / h, and the energy efficiency is 3 kWh / km.

[0210] ‌Device State Monitoring‌: Obtain real-time location (latitude and longitude coordinates) through GPS and record historical energy consumption data through energy consumption sensors. For example, the crane is currently located at (116.3°E, 39.9°N), and the average energy consumption in the past week is 25 kWh / h.

[0211] ‌Space-Time Cost Calculation Model‌:

[0212] ‌Moving Time Calculation‌: Estimate based on the distance between the current location of the device and the target node (calculate the spherical distance using the Haversine formula) and the moving speed. For example, the crane is 2 kilometers away from node A1, and the moving time = 2 km ÷ 5 km / h = 0.4 hours (24 minutes).

[0213] ‌Energy cost calculation: Mobile energy cost = distance × energy efficiency + operation energy cost × operation duration. For example, the energy cost of the crane moving to A1 = 2 km × 3 kWh / km + operation energy cost 10 kWh / h × 10 hours = 6 + 100 = 106 kWh.

[0214] ‌Capability matching score: If the device capability is lower than the task requirement (e.g., insufficient load capacity), the matching score = 0; otherwise, the matching score = 1. For example, node A1 requires a load capacity ≥ 15 tons, and the crane has a capacity of 20 tons, so the matching score = 1.

[0215] ‌Device-node matrix construction:

[0216] ‌Matrix filling: Each device calculates the space-time cost and matching score for each node, generating matrix elements. For example, the space-time cost of the crane for node A1 = moving time 0.4 hours + energy cost 106 kWh, and the matching score = 1.

[0217] ‌Matrix optimization: Remove device-node pairs with matching score = 0 (e.g., forklift cannot perform hoisting tasks) to reduce computational complexity.

[0218] Based on the task node graph and device capability-location matrix, dynamic time window is used to detect device path conflicts, and an initial path sequence is generated through priority scheduling;

[0219] ‌Dynamic Time Window (DTW) is the time period allocated to each device to perform tasks, which is updated in real time as the task progresses. Path conflict refers to multiple devices entering the same area within the same time window, causing spatial collision or load exceeding limit.

[0220] ‌Dynamic Time Window Definition and Update:

[0221] ‌Initial Time Window Assignment: Based on the time constraints of the task node graph, each device is assigned an initial task time window. For example, the crane's time window for task A1 is Day1 8:00-18:00.

[0222] ‌Real-time Adjustment Mechanism: If a device is delayed due to failure, its subsequent task time window is automatically postponed. For example, if the crane is delayed for 2 hours at A1, the subsequent task time window is postponed to Day1 10:00-20:00.

[0223] ‌Path Conflict Detection Algorithm:

[0224] ‌Spatial Conflict Detection: Through device path superposition analysis, identify devices that may enter the same geographic grid (e.g., 10m × 10m) within the same time period. For example, the crane and forklift both enter grid (X=100, Y=200) at Day1 14:00-14:30.

[0225] Load conflict detection: The total load of equipment in the same area within the same time period is counted, and if it exceeds the safety threshold, a conflict is marked. For example, the total load of a crane (load capacity 20 tons) and a forklift (load capacity 5 tons) in grid A is 25 tons, which exceeds the safety threshold of 20 tons.

[0226] Priority scheduling strategy:

[0227] Prioritization rules: Prioritize tasks based on urgency, energy efficiency, or contract terms (e.g., VIP customer priority). For example, bracket installation has higher priority than component laying, and cranes have higher priority than forklifts.

[0228] Conflict resolution actions:

[0229] Delayed execution: Low-priority devices wait for high-priority devices to complete their tasks before entering. For example, a forklift waits for a crane to leave grid A before entering.

[0230] Route replanning: Replanning detour routes for low-priority equipment. For example, a forklift is rerouted from its original route to detour via the north-facing road, increasing the travel distance by 50 meters but avoiding conflicts.

[0231] Initial path sequence generation:

[0232] Sequence format: The task execution order and time arrangement for each device. For example, crane path sequence: A1 (Day 1 8:00-18:00) → B1 (Day 2 8:00-12:00); forklift path sequence: A2 (Day 1 18:30-22:30) → B2 (Day 2 13:00-17:00).

[0233] Visualization output: Overlay device paths and time windows on the GIS platform, distinguish device types with different colors, and intuitively display the initial scheduling plan.

[0234] Using load balancing as the objective function, a mixed-integer programming solver is used to optimize the initial path sequence, outputting a device-task allocation matrix and conflict-free scheduling paths.

[0235] Mixed Integer Programming (MIP) is a mathematical optimization method that handles both continuous variables (such as time) and discrete variables (such as device-task allocation). Load balancing refers to distributing the load (task volume and energy consumption) of devices as evenly as possible, avoiding overloading some devices while leaving others idle.

[0236] Objective function definition:

[0237] Load balancing objective: Minimize the maximum load (Max Load) or load variance (LoadVariance) across all devices. For example, Objective Function = Minimize Max(Device1 Load, Device2 Load,..., DeviceN Load).

[0238] Load calculation method: Load = Σ(Task Duration × Device Energy Consumption Rate + Movement Energy Consumption). For example, Crane Load = 10 hours × 10 kWh / h + Movement Energy Consumption 6 kWh = 106 kWh.

[0239] Constraint setting:

[0240] Time window constraint: Devices must arrive and perform tasks within the task time window. For example, Crane arrival time at Node A1 ≥ 8:00, departure time ≤ 18:00.

[0241] Resource uniqueness: The same task can only be performed by one device. For example, Node A1 can only be assigned to the crane or backup crane, not repeated.

[0242] Safety load constraint: The total load of devices in the same area ≤ safety threshold. For example, Total load of devices within Grid A ≤ 20 tons.

[0243] Solver configuration and execution:

[0244] Solver selection: Commercial solvers such as Gurobi or open-source tools COIN-OR CBC are used to support MIP model solving.

[0245] Model input: Encode the objective function, constraints, and initial path sequence into a mathematical model. For example, define variable x_ij to represent whether device i performs task j (0 / 1 variable), and variable t_i to represent the task start time of device i (continuous variable).

[0246] Solving process: Iteratively search for feasible solutions and gradually approach the optimal solution. For example, the solver finds the optimal solution within 30 minutes, reducing the crane load from 106 kWh to 98 kWh, the forklift load from 80 kWh to 85 kWh, and the overall load variance by 40%.

[0247] Output results and applications:

[0248] Device-task allocation matrix: Display the task list assigned to each device in table form. For example, Crane assigns tasks A1, B1, C1; Forklift assigns tasks A2, B2, C2.

[0249] ‌Conflict-free scheduling path‌: Optimized device movement path and time window, ensuring no spatial or load conflicts. For example, the crane and forklift in grid A work at different times, Day1 8:00-12:00 and Day1 14:00-18:00.

[0250] ‌Construction instruction issuance‌: Distribute the matrix and path to the automatic driving system of construction machinery (such as AGV navigation module), start the automated construction process.

[0251] ‌Key technology examples and parameter explanations‌:

[0252] ‌Haversine formula‌: A formula for calculating the distance between two points on a sphere, input is latitude and longitude coordinates, output is kilometers.

[0253] ‌Neo4j‌: A graph database for storing node and edge relationship data.

[0254] ‌Gurobi‌: A commercial mathematical optimization solver that supports mixed integer programming, linear programming, etc.

[0255] ‌AGV (Automated Guided Vehicle)‌: An automated guided vehicle that travels along a predetermined path through a navigation system (such as laser or magnetic strip).

[0256] Convert the construction partition into a task node graph with time and space constraints, and build a movement cost model based on device GPS positioning and energy consumption data. The algorithm detects device path conflicts through dynamic time window, optimizes the scheduling sequence with load balancing as the goal, ensures that more resources are allocated to high-load areas while avoiding spatial and temporal overlap of device operations, solves the resource competition problem of large-scale distributed construction, and maximizes device utilization through mathematical optimization. Load balancing strategy prolongs the life of high-load devices and reduces the probability of construction interruptions by reducing spatial and temporal conflicts.

[0257] S204, based on the construction resource allocation matrix, simulate the construction process through the digital twin model, real-time collect support deformation data and component inclination data, adopt adaptive particle swarm optimization algorithm to dynamically correct construction parameters, generate anti-deformation construction instruction set;

[0258] Specifically, the construction resource allocation matrix can be mapped to the digital twin model to drive virtual construction machinery to perform work according to the scheduling path;

[0259] ‌Digital Twin Model‌ is a virtual mirror of physical construction process, which realizes accurate simulation through real-time data synchronization.‌ Construction resource allocation matrix‌ contains device-task allocation relationship and scheduling path, which needs to be converted into mechanical action instruction in virtual environment to drive digital twin model to simulate construction process.

[0260] ‌Construction and initialization of digital twin model:

[0261] ‌Three-dimensional scene modeling: Use Unreal Engine or Unity 3D engine to build a high-precision virtual construction site based on the BIM (Building Information Modeling) model of the construction area. For example, the geometric size and material properties (such as elastic modulus, density) of the photovoltaic support are consistent with the physical entity, and the support model precision reaches millimeter level (±1mm).

[0262] ‌Mechanical model import: Convert the CAD model of construction machinery such as cranes and forklifts to FBX format and add kinematic constraints (such as limiting the rotation angle of the boom to 0°~270°). For example, the hydraulic telescopic boom model of the crane contains 6 degrees of freedom (DOF) to simulate the motion trajectory of the real machine.

[0263] ‌Resource allocation matrix mapping: Analyze the device-task relationship in the construction resource allocation matrix and convert task nodes (such as "Node A1 needs crane operation from 8:00-10:00") into time series instructions. For example, the crane task "A1→B1" in the matrix is mapped to the virtual crane moving from coordinates (X=100, Y=200) to (X=150, Y=300).

[0264] ‌Virtual construction machinery driving mechanism:

[0265] ‌Path trajectory generation: Based on the GPS coordinates (latitude and longitude) of the scheduling path, generate the moving path of the virtual machinery through coordinate conversion algorithms (such as UTM to local coordinate system). For example, the moving path of the crane is composed of a series of waypoints (Waypoint) with a spacing of 0.5 meters, and the path smoothing process uses Bézier Curve.

[0266] ‌Action instruction analysis: Decompose operation instructions (such as "install support") into mechanical arm movement, hydraulic pressure regulation, and other bottom-level control signals. For example, the "install support" instruction corresponds to a crane boom lowering speed of 0.2m / s and a gripper clamping force of 500N.

[0267] ‌Real-time synchronization mechanism: Synchronize sensor data (such as motor speed) of physical devices to digital twin model through OPC UA (Open Platform Communications Unified Architecture) protocol to ensure consistency between virtual and real states. For example, the crane hydraulic pressure sensor data is updated to the virtual model at a frequency of 100Hz.

[0268] ‌Simulation verification and exception handling:

[0269] ‌Collision Detection‌: Enable the physical engine (e.g. NVIDIA PhysX) in the virtual environment to detect collisions between the machinery and the support, terrain. For example, if the boom is within <0.1 meters of the support, trigger an alarm and mark the path risk point.

[0270] ‌Timing Deviation Correction‌: Compare the virtual progress (e.g. Task A1 takes 9.5 hours) with the actual progress (10 hours), and automatically adjust the subsequent task time window. For example, after the virtual model detects the delay, the start time of Task B1 is postponed from 10:00 to 10:30.

[0271] ‌Visual Monitoring‌: Display the machinery load, energy consumption, etc. in the form of a heat map in the digital twin interface, support multi-view (top view, side view, first person) observation of construction details.

[0272] Real-time collection of support deformation sensor data and component inclination data to generate millimeter-level precision deformation-inclination joint distribution map;

[0273] ‌Support Deformation Sensor‌: Typically a strain gauge (Strain Gauge) or fiber Bragg grating sensor (FBG), deployed at key stress points of the support;‌ Component Inclination Data‌: Collected by a gyroscope (Gyroscope) or inclinometer (Inclinometer), used to monitor the installation angle of photovoltaic panels.‌ Deformation-Inclination Joint Distribution Map‌: Integrates both types of data, reflecting the coupling relationship between structural deformation and component posture during construction.

[0274] ‌Sensor Deployment and Data Collection‌:

[0275] ‌Deformation Sensor Arrangement‌: Install strain gauges at key locations such as support columns and beams, with a sensor every 2 meters. For example, a 20-meter support is equipped with 11 strain gauges, with a sampling frequency of 50 Hz and a measurement accuracy of ±0.01 mm.

[0276] ‌Inclination Sensor Installation‌: Install MEMS (Micro-Electro-Mechanical Systems) inclination sensors on the four corners of photovoltaic components to measure X / Y axis inclination angles. For example, a component inclination measurement range of ±30° with a resolution of 0.01°.

[0277] ‌Wireless Transmission Network‌: Use Zigbee or LoRa wireless networks to aggregate sensor data to edge gateways. For example, each gateway covers a radius of 500 meters, supports simultaneous connection of 200 sensor nodes, and data transmission delay is <100ms.

[0278] ‌Data Preprocessing and Fusion‌:

[0279] Noise Filtering: Apply Kalman Filter to raw deformation data to remove vibration disturbances (e.g., transient noise caused by mechanical operations). For example, a certain strain gauge's raw data fluctuates within ±0.05mm, while the filtered data stabilizes at ±0.01mm.

[0280] Temporal-Spatial Alignment: Unify the timestamps of deformation data and tilt data to the GPS clock and align them according to a spatial grid (e.g., 1m x 1m). For example, the deformation data (0.12mm) and tilt data (5.2°) of grid G001 are associated with the same time point (2023-10-01 10:00:00).

[0281] Outlier Repair: Repair the measurement values of failed nodes based on neighboring sensor data. For example, when a certain tilt sensor fails, take the average of the data from the four neighboring sensors (e.g., 5.0°, 5.1°, 5.3°, 5.2° → repaired value: 5.15°).

[0282] Joint Distribution Map Generation and Visualization:

[0283] Interpolation Algorithm Application: Use Kriging interpolation to convert discrete sensor data into continuous distribution maps. For example, after interpolation, the deformation data generates a 0.1m x 0.1m resolution deformation field, with color mapping from blue (0mm) to red (2mm).

[0284] Tilt Vector Superimposition: Indicate the direction and size of the tilt angle on the deformation heat map with arrows. For example, in a certain area with 0.5mm deformation, the tilt arrow points northwest, and the length represents a 5° tilt angle.

[0285] Dynamic Update Mechanism: Refresh the distribution map every 5 seconds, supporting historical data backtracking (e.g., deformation trend animation for the past 1 hour). The output format is a WebGL layer, which can be rendered in real-time on the browser side.

[0286] Based on the deformation-tilt joint distribution map, define peak deformation, tilt uniformity, and schedule deviation as multi-objective fitness functions, input adaptive particle swarm optimization algorithm;

[0287] Multi-Objective Fitness Function Quantifies Multiple Dimensions of Construction Quality: Peak Deformation needs to be minimized to avoid structural damage; Tilt Uniformity needs to be maximized to ensure component efficiency; Schedule Deviation needs to be minimized to maintain the construction period. Adaptive Particle Swarm Optimization (APSO) finds a balanced solution among multiple objectives by dynamically adjusting search parameters.

[0288] Fitness function definition:

[0289] Deformation peak calculation: Take the maximum deformation value among all grid cells. For example, the maximum deformation in the current distribution is 1.8 mm, which corresponds to the target value f1=1.8.

[0290] Tilt uniformity measure: Calculate the standard deviation of all component tilt angles (Standard Deviation), the smaller the standard deviation, the higher the uniformity. For example, the tilt data is [5.0°, 5.1°, 4.9°, 5.2°], the standard deviation ≈ 0.12°, the target value f2=1 / 0.12≈8.33 (maximize).

[0291] Progress deviation evaluation: Compare the actual progress with the planned progress, calculate the absolute value of the deviation in hours. For example, a task planned to be completed in 10 hours actually took 10.5 hours, the deviation f3=0.5.

[0292] Multi-objective weighted integration: Set weights according to engineering priorities (such as deformation weight w1=0.5, uniformity w2=0.3, progress w3=0.2), fitness function F= w1f1 + w2(1 / f2) + w3*f3, need to minimize F.

[0293] Adaptive particle swarm optimization algorithm configuration:

[0294] Particle coding design: Each particle represents a set of mechanical control parameters (such as boom speed, clamping force). For example, particle=[speed 0.2m / s, clamping force 500N, hydraulic pressure 20MPa].

[0295] Inertia weight adjustment: Dynamically reduce the inertia weight (Inertia Weight) according to the number of iterations, enhance global search in the early stage, and focus on local optimization in the later stage. For example, the initial weight w=0.9, reduce 0.02 per generation, the minimum is 0.4.

[0296] Learning factor optimization: Individual learning factor (C1) and social learning factor (C2) are adaptively adjusted according to particle distribution density. For example, if the particle aggregation degree is high, increase C1 (focus on individual experience), otherwise increase C2 (focus on group experience).

[0297] Algorithm execution and convergence control:

[0298] Population initialization: Randomly generate 50 particles, parameter range meets mechanical operation specifications (such as speed 0.1~0.5m / s).

[0299] Iteration update rule: Calculate the fitness of each particle in each generation, update the individual optimal (Pbest) and global optimal (Gbest). For example, if the fitness of a particle decreases from 8.5 to 7.2, update its Pbest.

[0300] Termination condition setting: Stop when the global optimal fitness changes by less than 1% for 10 consecutive generations or reaches the maximum number of iterations (e.g., 100 generations). Output the optimal parameter combination (e.g., speed 0.25 m / s, clamping force 480 N).

[0301] Adjust the search direction of the particle swarm dynamically according to real-time deformation data, and solve the optimal mechanical control parameters under the constraints of construction specifications;

[0302] Real-time deformation data is continuously input into the optimization algorithm by edge computing nodes, dynamically correcting the search direction to avoid local optimization. Construction specification constraints include mechanical operation safety range (such as maximum lifting weight limit) and engineering standards (such as support installation precision requirements), which need to be processed through constraint processing technology to ensure the effectiveness of the solution.

[0303] Real-time data-driven search direction adjustment:

[0304] Deformation gradient analysis: Calculate the gradient field of the current deformation distribution map, identify high deformation growth areas (such as gradient > 0.1 mm / m). The optimization algorithm preferentially reduces the mechanical load parameters in these areas. For example, if the deformation gradient of area A is 0.15 mm / m, reduce the crane operating speed in this area.

[0305] Dynamic weight distribution: Adjust the fitness function weight according to the severity of deformation. For example, when the deformation peak value of a certain area exceeds 1.5 mm, increase the deformation weight w1 from 0.5 to 0.7, and strengthen the optimization of this target.

[0306] Real-time update of particle swarm parameters: Adjust the particle velocity update formula according to the deformation trend prediction (such as ARIMA model). For example, if the predicted deformation will continue to increase, increase the search step of the particle in the direction of low load parameters.

[0307] Construction specification constraint processing technology:

[0308] Hard constraint processing: Directly exclude solutions that violate safety specifications. For example, if the crane speed parameter of a certain particle is 0.6 m / s (exceeding the maximum allowed value of 0.5 m / s), it is marked as an invalid solution and regenerated.

[0309] Soft constraint penalty: Apply a penalty term to slightly illegal solutions. For example, when the clamping force is less than 450 N, the fitness function increases the penalty term P=10*(450-actual value), forcing the algorithm to move away from the illegal area.

[0310] Feasible Solution Repair: Automatically corrects some violated parameters. For example, if the hydraulic pressure parameter is 25 MPa (exceeding the upper limit of 22 MPa), it will be corrected to 22 MPa and the fitness will be recalculated.

[0311] Multi-Objective Pareto Front Generation:

[0312] Non-dominated Sorting: Sort particles by objective function values, preferentially retaining particles that are not dominated by other solutions. For example, the first layer contains three solutions with fitness F=7.2, F=7.5, and F=7.8, none of which dominate each other.

[0313] Crowding Distance Calculation: Evaluate the distribution density of solutions in the target space to preserve diversity. For example, a solution with a crowding distance of 0.15 indicates that there are fewer surrounding solutions, so it is preferentially retained.

[0314] Elitism Strategy: Retain the top 10% of optimal solutions in each generation to avoid losing high-quality solutions. For example, among 50 particles, 5 elite solutions are directly passed to the next generation.

[0315] Encode optimal mechanical control parameters as anti-deformation construction instruction sets, and ensure instruction integrity and traceability through blockchain signatures.

[0316] Anti-deformation construction instruction sets are the output of optimization algorithms and need to be converted into executable instruction sequences for machines. Blockchain signatures use hash algorithms (such as SHA-256) and digital certificates to ensure that instructions cannot be tampered with during transmission and storage, and operation records can be traced.

[0317] Instruction Set Encoding and Formatting:

[0318] Instruction Structure Design: Encapsulate control parameters, timestamps, and device IDs in JSON format. For example:

[0319] {

[0320] "device_id": "Crane-001",

[0321] "timestamp": "2023-10-01T10:00:00Z",

[0322] "commands": [

[0323] {"action": "move", "target": {"x": 100, "y": 200}, "speed":0.25},

[0324] {"action": "clamp", "force": 480, "duration": 60} ]

[0326] }

[0327] Command Priority Flag: Sets the execution priority (level 0-9) based on the urgency of the task. For example, the bracket installation command has a priority of 7, and the component debugging command has a priority of 3.

[0328] Timing synchronization mechanism: The clocks of all devices are synchronized via NTP (Network Time Protocol) to ensure that instructions are executed according to the planned timing. For example, the timestamp error in the instruction is controlled within ±1ms.

[0329] Blockchain signature and notarization process:

[0330] Hash value calculation: Perform a SHA-256 hash operation on the instruction set JSON string to generate a unique digest. For example, the instruction set hash value is "a1b2c3...f6e5d4".

[0331] Digital signature generation: The hash value is encrypted using the device's private key to generate a digital signature. For example, the private key is a 2048-bit RSA key, and the signature length is 256 bytes.

[0332] Blockchain on-chain: The instruction set, hash value, signature, and timestamp are packaged into a transaction and written to the distributed ledger of the Hyperledger Fabric blockchain. For example, the transaction ID is "TX-20231001-1000" and the block height is #12345.

[0333] Command issuance and execution verification:

[0334] Edge node verification: After receiving the instruction, the device uses the public key to verify the validity of the signature and recalculates the hash value for consistency. For example, the public key is pre-installed in the device's security chip (HSM).

[0335] Abnormal Command Interception: If the hash value does not match or the signature is invalid, an alarm is triggered and the command is discarded. For example, if a tampered command is detected, it is automatically reported to the monitoring center.

[0336] ‌Execution Record Backtracking: After completing a task, the device signs the execution results (such as actual time consumption, final deformation value) and returns them to the blockchain, forming a complete traceability chain. For example, the result record contains "Crane-001 completed task A1 at 2023-10-01T10:30:00Z with deformation 0.8mm".

[0337] ‌Key Technology Examples and Parameter Descriptions:

[0338] ‌UTM Coordinate System (Universal Transverse Mercator): A global zoned projection coordinate system suitable for local area high-precision positioning.

[0339] ‌OPC UA Protocol: An industrial automation communication protocol that supports cross-platform device data interaction.

[0340] ‌MEMS Sensor (Micro-Electro-Mechanical Systems): A micro-electromechanical system used for high-precision tilt measurement.

[0341] ‌SHA-256: Secure Hash Algorithm, generating a 256-bit (32-byte) unique data digest.

[0342] ‌Hyperledger Fabric: An enterprise-level blockchain framework that supports permission management and privacy protection.

[0343] The digital twin model maps the physical construction state in real time, and constructs a deformation-tilt joint distribution graph through millimeter-level deformation monitoring data. The adaptive particle swarm optimization algorithm takes the deformation peak value and tilt uniformity as the optimization objective, dynamically adjusts the mechanical control parameters under the constraints of construction specifications, generates anti-deformation operation instructions, and realizes online optimization and risk pre-control of the construction process, suppressing the structure deformation within the safety threshold. The combination of digital twin and optimization algorithm significantly improves the construction accuracy under complex terrain.

[0344] S205, according to the anti-deformation construction instruction set, drive the construction machinery to execute the operation, synchronize through the edge computing node to analyze the energy efficiency feedback data, use the reinforcement learning model to iteratively update the global load distribution atlas, form a closed-loop construction management link.

[0345] Specifically, the anti-deformation construction instruction set can be parsed into mechanical control signals to drive the hydraulic system and robotic arm to perform high-precision construction operations;

[0346] The anti-deformation construction instruction set is a JSON format instruction file verified by blockchain signature, containing device action parameters, target coordinates, and execution timing. The parsing process involves instruction decoding, coordinate conversion, and signal adaptation, ultimately generating pulse signals (PWM) or digital signals (Digital I / O) executable by the underlying mechanical control system.

[0347] ‌Instruction parsing and decoding‌:

[0348] ‌Instruction structure parsing‌:

[0349] The instruction set is encapsulated in JSON format, with device ID, action type, target parameters, etc. defined as key-value pairs. For example:

[0350] {

[0351] "device_id": "HydraulicArm-001",

[0352] "timestamp": "2023-10-01T10:00:00Z",

[0353] "actions": [

[0354] {"type": "move_xy", "x": 150.25, "y": 200.75, "speed": 0.3},

[0355] {"type": "clamp", "force": 520, "duration": 5} ]

[0357] }。

[0358] Each action field needs to be mapped to the corresponding control instruction. For example, move_xy corresponds to the two-dimensional plane movement of the mechanical arm, and clamp corresponds to the clamping action of the gripper.

[0359] ‌Blockchain signature verification‌:

[0360] Use the device's pre-set public key (Public Key) to verify the consistency of the instruction hash value (Hash Value) and digital signature (Digital Signature). For example, the hash algorithm uses SHA-256 (Secure Hash Algorithm 256-bit), and the signature algorithm uses ECDSA (Elliptic Curve Digital Signature Algorithm). If the verification fails, the instruction will be discarded and an alarm will be triggered.

[0361] Control Signal Generation and Adaptation:

[0362] Coordinate Conversion:

[0363] Convert GPS coordinates (e.g., WGS-84 latitude and longitude) to local engineering coordinate systems (e.g., UTM, Universal Transverse Mercator). For example, convert the target point (116.3°E, 39.9°N) to UTM Zone 50N coordinate system (X=324567, Y=4321098).

[0364] Signal Type Adaptation:

[0365] Hydraulic System Control: Control hydraulic flow through the opening of proportional valves, generating 4-20mA analog signals. For example, a target pressure of 20MPa corresponds to a current signal of 16mA.

[0366] Mechanical Arm Motion: Use pulse width modulation (PWM) signals to drive servo motors. For example, a rotation speed of 0.3m / s corresponds to a PWM duty cycle of 45%.

[0367] Real-Time Feedback Closed Loop:

[0368] Monitor the mechanical arm position in real time through Hall sensors and feed back to the PLC (Programmable Logic Controller). For example, when the actual position deviates from the target position by more than 2mm, trigger the PID (Proportional-Integral-Derivative) control algorithm to adjust the motor speed.

[0369] High Precision Execution and Safety Assurance:

[0370] Multi-Axis Synchronous Control:

[0371] The X / Y / Z axis motion of the mechanical arm is driven synchronously by the motion controller, ensuring smooth trajectories. For example, use the EtherCAT (Ethernet for Control Automation Technology) bus protocol to achieve multi-axis linkage, with a synchronization period of ≤1ms.

[0372] Overload Protection Mechanism:

[0373] Real-time monitoring of clamping force by pressure transducer. If the clamping force exceeds the safety threshold (e.g. 600N), the pressure is immediately released and the action is interrupted. For example, the clamping force threshold is dynamically adjusted according to the mechanical parameters of the assembly to prevent fracturing the photovoltaic panel.

[0374] ‌Precision calibration‌:

[0375] Periodic calibration of the mechanical arm positioning accuracy using a laser tracker. For example, the repeatability accuracy after calibration is ±0.1mm, meeting the photovoltaic panel installation requirements.

[0376] Real-time collection of assembly installation error and mechanical energy consumption data by edge computing nodes to generate efficiency-quality evaluation metrics.

[0377] ‌Edge computing nodes are embedded devices (such as NVIDIA Jetson TX2) deployed on the construction site, responsible for real-time processing of sensor data and generation of evaluation metrics. The efficiency-quality evaluation metrics (EQM) quantitatively evaluate the energy consumption and installation quality of the construction process, providing training data for reinforcement learning models.

[0378] ‌Data collection and preprocessing‌:

[0379] ‌Installation error measurement‌:

[0380] ‌Visual measurement system: Industrial cameras (such as Basler ace 2) are used to capture the installation position of the assembly, and OpenCV (Open Source Computer Vision Library) image processing algorithms are used to calculate the deviation between the actual coordinates and the design coordinates. For example, image resolution 2448x2048, measurement error ±0.5mm.

[0381] ‌Laser range finder: Measures the spacing error between assemblies using the TOF (Time-of-Flight) principle. For example, the design spacing between adjacent assemblies is 1.5m, and when the actual measurement is 1.48m, it is marked as a -0.02m deviation.

[0382] ‌Mechanical energy consumption statistics‌:

[0383] ‌Electric energy metering module: Integrates ADE7953 chip to collect real-time device power consumption data (such as voltage, current, power factor). For example, the instantaneous power during crane operation is 15kW, and the cumulative energy consumption = 15kW x 2h = 30kWh.

[0384] ‌Hydraulic Energy Consumption Calculation: Calculate the hydraulic system energy consumption based on Flow Meter and Pressure Sensor data. For example, Flow 10 L / min x Pressure 20 MPa x Time 1 h → Energy Consumption = (10 x 20 x 60) / 600 = 20 kWh (assuming system efficiency 60%).

[0385] ‌Energy Efficiency-Quality Index Calculation:

[0386] ‌Installation Quality Score (QS):

[0387] Define three sub-indexes for weighted calculation:

[0388] ‌Position Error: Root Mean Square Error (RMSE) of all component coordinate deviations. For example, RMSE = 0.8 mm, weight 40%.

[0389] ‌Tilt Consistency: Standard Deviation of component tilt angle. For example, Standard Deviation = 0.15°, weight 30%.

[0390] ‌Structural Integrity: Peak Deformation of support. For example, Peak Deformation = 1.2 mm, weight 30%.

[0391] Final QS = 0.4 x (1 - RMSE / 2 mm) + 0.3 x (1 - StdDev / 0.5°) + 0.3 x (1 - PeakDeform / 2 mm), numerical range 0-1 (1 is optimal).

[0392] ‌Efficiency Score (ES):

[0393] Calculate based on Energy per Unit Area and Completion Rate:

[0394] ‌Energy Intensity: Total Energy Consumption (kWh) / Installed Area (m 2 ). For example, 30 kWh / 50 m 2 = 0.6 kWh / m 2 .

[0395] ‌Completion Rate: Number of tasks completed on time / Total number of tasks. For example, 8 out of 10 tasks completed on time → 80%.

[0396] Final ES = 0.7 x (1 - EnergyIntensity / 1 kWh / m 2 ) + 0.3 x CompletionRate, value range 0-1.

[0397] ‌Overall Evaluation Metric (EQM)‌:

[0398] EQM = α x QS + β x ES, where α, β are weights (e.g., α = 0.6, β = 0.4), reflecting the quality-first strategy.

[0399] ‌Edge Computing and Data Upload‌:

[0400] ‌Local Real-Time Computing‌:

[0401] Edge nodes are equipped with TensorRT (NVIDIA's deep learning inference framework) to accelerate index calculation. For example, QS calculation delay < 50ms, meeting real-time requirements.

[0402] ‌Data Compression and Encryption‌:

[0403] Use Zstandard (ZSTD) algorithm to compress data and encrypt through TLS (Transport Layer Security) protocol to upload to the cloud. For example, the original data volume is 1MB, compressed to 200KB, reducing the upload bandwidth occupancy by 80%.

[0404] ‌Caching and Offline Transmission‌:

[0405] In weak network environment, data is temporarily stored in local SSD (Solid State Drive), and automatically transmitted after network recovery. For example, cache capacity 512GB, supporting offline storage of 72 hours of data.

[0406] Based on the energy efficiency-quality evaluation index, a hierarchical reinforcement learning model is trained. The upper layer optimizes the load prediction parameters, the lower layer adjusts the mechanical motion trajectory, and the global load distribution map is iteratively updated to form a closed-loop construction management link.

[0407] ‌Hierarchical Reinforcement Learning Model‌ is divided into Upper-Level Policy and Lower-Level Policy. The upper layer is responsible for global optimization (such as load prediction model parameter optimization), and the lower layer is responsible for local control (such as mechanical arm motion trajectory adjustment), through collaborative iteration to realize closed-loop optimization.

[0408] ‌Hierarchical Model Architecture Design‌:

[0409] Upper-Level:

[0410] State Space: Contains global load distribution map, historical EQM indicators, and weather forecast data. For example, the state vector consists of 1000 load grid values (1m x 1m per grid), past 24-hour QS / ES sequences, and future 6-hour wind speed predictions.

[0411] Action Space: Adjusts hyperparameters of the multi-modal fusion network (deep neural network in Step 1), such as learning rate and kernel count. For example, Action Space = [Learning Rate: 0.001~0.01, Kernel Count: 16~64].

[0412] Reward Function: Based on the improvement rate of the EQM indicator. For example, if the current EQM is 0.8 and the optimized EQM is 0.85, the reward value is +5.

[0413] Lower-Level:

[0414] State Space: Real-time mechanical position and orientation, sensor readings (deformation, inclination), and current task progress. For example, State Vector = [X Coordinate, Y Coordinate, Clamping Force, Deformation Value, Remaining Task Number].

[0415] Action Space: Adjusts mechanical motion parameters such as speed, acceleration, and path smoothing. For example, Action Space = [Speed: 0.1~0.5 m / s, Acceleration: 0.05~0.2 m / s 2 ].

[0416] Reward Function: Combines local EQM (e.g., single-task quality score) with energy efficiency. For example, if Task A1 has a QS of 0.9 and an energy consumption of 0.5 kWh / m 2 , the reward value is +3.

[0417] Model Training and Iterative Optimization:

[0418] Upper-Level Training Process:

[0419] Uses the PPO (Proximal Policy Optimization) algorithm to update the policy every 24 hours:

[0420] ‌Data Sampling: Load the past 7 days of global load data and EQM records from the cloud.

[0421] ‌Policy Update: Calculate the Advantage Function to update network parameters, constrain the policy change range (KL Divergence ≤ 0.05).

[0422] ‌Parameter Delivery: Synchronize the optimized hyperparameters to the multi-modal fusion network to generate a new load prediction model.

[0423] ‌Lower Layer Training Process:

[0424] Use DQN (Deep Q-Network) algorithm, update policy every 1 hour:

[0425] ‌Experience Replay: Store mechanical execution records (state, action, reward), randomly extract batches (Batch) to train the network.

[0426] ‌Target Network: Delay the update of the target Q network (Update Interval = 1000 steps) to stabilize the training process.

[0427] ‌Action Exploration: Add ε-greedy policy (ε = 0.1 → 10% probability of random action) to avoid local optimization.

[0428] ‌Global Load Distribution Map Update:

[0429] ‌Data Fusion and Interpolation:

[0430] According to the real-time construction data (deformation, inclination) and the optimized prediction model, use Kriging Interpolation to update the load map. For example, the measured deformation in a certain area is 1.0mm, the predicted value is 0.9mm → correction factor 0.95, updated value = 0.95 × 0.9 + 0.05 × 1.0 = 0.905mm.

[0431] ‌Map Visualization and Verification:

[0432] Display the updated load distribution in the form of a heat map on the digital twin platform and compare it with the finite element simulation results. For example, the maximum load deviation ≤ 5% is considered as passing the verification.

[0433] ‌Closed Loop Link Triggering Condition:

[0434] When the EQM index improves by less than 1% or the load prediction error exceeds 10% for three consecutive iterations, trigger manual intervention procedures (such as adjusting reinforcement learning model structure or resetting training data).

[0435] ‌Key technology examples and parameter explanations‌:

[0436] ‌CAN bus (Controller Area Network): A communication protocol for vehicles and industrial machinery that supports high-reliability data transmission.

[0437] ‌EtherCAT: A real-time industrial Ethernet protocol used for synchronous communication in multi-axis motion control systems.

[0438] ‌PID control (Proportional-Integral-Derivative): A closed-loop control algorithm that adjusts system output through proportional, integral, and derivative components.

[0439] ‌PPO (Proximal Policy Optimization): A reinforcement learning algorithm based on policy gradient that improves stability by constraining the magnitude of policy updates.

[0440] ‌KL divergence (Kullback-Leibler Divergence): A measure of the difference between two probability distributions used to constrain the magnitude of policy updates.

[0441] Construction machinery performs high-precision operations according to the instruction set, and edge nodes analyze installation errors and energy consumption data in real time. The hierarchical reinforcement learning model achieves dynamic updating of load distribution maps and continuous improvement of construction strategies through dual iteration of upper-layer optimization of load prediction parameters and lower-layer adjustment of mechanical trajectories, forming a complete closed loop of "perception-decision-execution-optimization" and enabling construction management to have self-learning capabilities. Through continuous data feedback and model iteration, the quality and disaster resistance of power plant construction are continuously improved.

[0442] As can be seen, based on the topographic elevation data of the photovoltaic power station planning area, historical meteorological data, and component mechanical parameters, a dynamic load intensity heat map is output; based on the dynamic load intensity heat map, an anti-wind pressure optimized construction zoning scheme is generated; according to the construction zoning scheme, a conflict-free construction resource allocation matrix is output; based on the construction resource allocation matrix, an adaptive particle swarm optimization algorithm is used to dynamically correct construction parameters, generating an anti-deformation construction instruction set; according to the anti-deformation construction instruction set, construction machinery is driven to perform operations, and energy efficiency feedback data is analyzed synchronously through edge computing nodes, and a reinforcement learning model is used to iteratively update the global load distribution map, forming a closed-loop construction management link, so that through the fusion of multi-source data, optimization of zoning planning, intelligent scheduling of equipment, and closed-loop quality control, the balance of construction precision and efficiency in complex environments can be achieved.

[0443] Yet another embodiment of the present application provides a photovoltaic power station distributed construction management system based on load analysis, referring to Figure 3 , the system can include:

[0444] The fusion module 301 is configured to fuse satellite remote sensing images and ground sensor time series data through a multimodal fusion deep neural network according to terrain elevation data, meteorological historical data and component mechanical parameters of a photovoltaic power station planning area, and output a dynamic load intensity heat map.

[0445] The division module 302 is configured to divide a construction sub-area using a gradient-constrained spectral clustering algorithm based on the dynamic load intensity heat map, dynamically adjust the clustering center according to the load extreme point and the terrain mutation feature, and generate a wind pressure-resistant optimized construction zoning scheme.

[0446] The generation module 303 is configured to generate a device scheduling path using a space-time conflict resolution algorithm according to the construction zoning scheme and in combination with real-time state data of construction equipment, output a conflict-free construction resource allocation matrix through a dynamic time window constraint and a load balancing objective function.

[0447] The correction module 304 is configured to simulate a construction process through a digital twin model based on the construction resource allocation matrix, real-time collect support frame deformation data and component inclination data, dynamically correct construction parameters using an adaptive particle swarm optimization algorithm, and generate a deformation-resistant construction instruction set.

[0448] The update module 305 is configured to drive a construction machine to perform an operation according to the deformation-resistant construction instruction set, synchronously analyze energy efficiency feedback data through an edge computing node, iteratively update a global load distribution atlas using a reinforcement learning model, and form a closed-loop construction management link.

[0449] As can be seen, a dynamic load intensity heat map is output according to terrain elevation data, meteorological historical data and component mechanical parameters of a photovoltaic power station planning area, a wind pressure-resistant optimized construction zoning scheme is generated based on the dynamic load intensity heat map, a conflict-free construction resource allocation matrix is output according to the construction zoning scheme, deformation-resistant construction parameters are dynamically corrected using an adaptive particle swarm optimization algorithm based on the construction resource allocation matrix, a deformation-resistant construction instruction set is generated, a construction machine is driven to perform an operation according to the deformation-resistant construction instruction set, energy efficiency feedback data is synchronously analyzed through an edge computing node, a global load distribution atlas is iteratively updated using a reinforcement learning model, and a closed-loop construction management link is formed, so that the balance between construction precision and efficiency in a complex environment can be achieved through fusion of multi-source data, optimized zoning planning, intelligent scheduling of equipment and closed-loop quality control.

[0450] The embodiment of the present application also provides a storage medium, wherein the storage medium stores a computer program, and the computer program is arranged to execute the steps in any of the method embodiments.

[0451] Specifically, in the present embodiment, the storage medium can be arranged to store a computer program for executing the following steps:

[0452] S201, according to the terrain elevation data of the photovoltaic power station planning area, the meteorological historical data and the component mechanical parameters, through a multi-modal fusion deep neural network, satellite remote sensing images and ground sensor time series data are fused to output a dynamic load intensity heat map;

[0453] S202, based on the dynamic load intensity heat map, a gradient-constrained spectral clustering algorithm is used to divide the construction sub-area, the clustering center is dynamically adjusted according to the load extreme point and the terrain mutation feature, and an anti-wind pressure optimized construction zoning scheme is generated;

[0454] S203, according to the construction zoning scheme, combined with the real-time state data of the construction equipment, a space-time conflict resolution algorithm is used to generate an equipment scheduling path, the algorithm outputs a conflict-free construction resource allocation matrix through dynamic time window constraints and load balancing objective functions;

[0455] S204, based on the construction resource allocation matrix, a digital twin model is used to simulate the construction process, real-time collection of support deformation data and component inclination data is performed, and an adaptive particle swarm optimization algorithm is used to dynamically correct the construction parameters to generate an anti-deformation construction instruction set;

[0456] S205, according to the anti-deformation construction instruction set, the construction machinery is driven to perform operations, energy efficiency feedback data is analyzed synchronously through an edge computing node, and a reinforcement learning model is used to iteratively update a global load distribution atlas to form a closed-loop construction management link.

[0457] As can be seen, according to the terrain elevation data of the photovoltaic power station planning area, the meteorological historical data and the component mechanical parameters, a dynamic load intensity heat map is output; based on the dynamic load intensity heat map, an anti-wind pressure optimized construction zoning scheme is generated; according to the construction zoning scheme, a conflict-free construction resource allocation matrix is output; based on the construction resource allocation matrix, an adaptive particle swarm optimization algorithm is used to dynamically correct the construction parameters to generate an anti-deformation construction instruction set; according to the anti-deformation construction instruction set, the construction machinery is driven to perform operations, energy efficiency feedback data is analyzed synchronously through an edge computing node, and a reinforcement learning model is used to iteratively update a global load distribution atlas to form a closed-loop construction management link, so that through the fusion of multi-source data, optimized zoning planning, intelligent scheduling of equipment and closed-loop quality control, the balance between construction precision and efficiency in complex environments can be realized.

[0458] The embodiment of the present application also provides an electronic device, comprising a memory and a processor, the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any of the above method embodiments.

[0459] Specifically, the electronic device described above can further comprise a transmission device connected with the processor and an input and output device connected with the processor.

[0460] Specifically, in the embodiment, the processor can be configured to execute the following steps through the computer program:

[0461] S201, according to the terrain elevation data of the photovoltaic power station planning area, the meteorological historical data and the component mechanical parameters, through a multi-modal fusion deep neural network, satellite remote sensing images and ground sensor time series data are fused, and a dynamic load intensity thermal map is output;

[0462] S202, based on the dynamic load intensity thermal map, a gradient-constrained spectral clustering algorithm is used to divide a construction sub-area, a cluster center is dynamically adjusted according to a load extreme point and a terrain mutation feature, and an anti-wind pressure optimized construction partition scheme is generated;

[0463] S203, according to the construction partition scheme, combined with real-time state data of construction equipment, a space-time conflict resolution algorithm is used to generate an equipment scheduling path, the algorithm outputs a conflict-free construction resource allocation matrix through a dynamic time window constraint and a load balancing objective function;

[0464] S204, based on the construction resource allocation matrix, a digital twin model is used to simulate a construction process, real-time collection of support deformation data and component inclination data is performed, an adaptive particle swarm optimization algorithm is used to dynamically correct construction parameters, and an anti-deformation construction instruction set is generated;

[0465] S205, according to the anti-deformation construction instruction set, a construction machine is driven to perform an operation, energy efficiency feedback data is analyzed synchronously through an edge computing node, a global load distribution atlas is iteratively updated using a reinforcement learning model, and a closed-loop construction management link is formed.

[0466] It can be seen that, according to the terrain elevation data of the photovoltaic power station planning area, the meteorological historical data and the mechanical parameters of the components, a dynamic load intensity thermal map is output; based on the dynamic load intensity thermal map, an anti-wind pressure optimized construction partition scheme is generated; according to the construction partition scheme, a conflict-free construction resource allocation matrix is output; based on the construction resource allocation matrix, an adaptive particle swarm optimization algorithm is used to dynamically correct the construction parameters, and an anti-deformation construction instruction set is generated; according to the anti-deformation construction instruction set, the construction machinery is driven to execute the operation, and the energy efficiency feedback data is analyzed synchronously through the edge computing node, the global load distribution atlas is iteratively updated by using the reinforcement learning model, and a closed-loop construction management link is formed, so that the balance of construction precision and efficiency in a complex environment can be realized by fusing multi-source data, optimizing partition planning, intelligently scheduling equipment and closed-loop quality control.

[0467] The above describes the structure, features and effects of the present application in detail according to the embodiments shown in the drawings. The above description is only the preferred embodiments of the present application, but the present application is not limited to the embodiments shown in the drawings. Any changes or modifications made in accordance with the concept of the present application, or equivalent embodiments with equivalent changes, are still within the scope of the present application.

Claims

1. A load analysis-based photovoltaic power plant distributed construction management method, characterized by, The method comprises: According to the terrain elevation data of the photovoltaic power station planning area, the meteorological historical data and the component mechanical parameters, a dynamic load intensity thermal map is output by a multi-modal fusion deep neural network by fusing satellite remote sensing images and ground sensor time series data; wherein, according to the satellite remote sensing image and the terrain elevation data, a terrain-component mapping grid is generated by a geometric registration algorithm, and a spatial grid is output after eliminating cloud interference; the spatial grid and the real-time wind speed, temperature and humidity data collected by the ground sensor are input into a space-time interpolation model to generate a three-dimensional dynamic meteorological field in combination with the meteorological historical data; the three-dimensional dynamic meteorological field and the component mechanical parameters are input into a multi-modal fusion network, the terrain texture features are extracted by residual convolution, the meteorological time series mode is extracted by bidirectional LSTM, and the initial load prediction map is output after fusion; the initial load prediction map is subjected to space-time noise filtering guided by optical flow, the intensity distribution is optimized according to the load correlation of adjacent grids, and a dynamically updated load intensity thermal map is generated; Based on the dynamic load intensity thermal map, a gradient-constrained spectral clustering algorithm is used to divide the construction sub-area, the clustering center is dynamically adjusted according to the load extreme point and the terrain mutation feature, and an anti-wind pressure optimized construction partition scheme is generated; According to the construction partition scheme, combined with the real-time state data of the construction equipment, a device scheduling path is generated by using a space-time conflict resolution algorithm, the algorithm outputs a conflict-free construction resource allocation matrix through dynamic time window constraints and load balancing objective functions; Based on the construction resource allocation matrix, the construction process is simulated by a digital twin model, the support deformation data and the component inclination data are collected in real time, the adaptive particle swarm optimization algorithm is used to dynamically correct the construction parameters, and an anti-deformation construction instruction set is generated; According to the anti-deformation construction instruction set, the construction machinery is driven to execute the operation, the energy efficiency feedback data is analyzed synchronously through the edge computing node, the global load distribution atlas is iteratively updated by using the reinforcement learning model, and a closed-loop construction management link is formed.

2. The method of claim 1, wherein, Based on the dynamic load intensity thermal map, a gradient-constrained spectral clustering algorithm is used to divide the construction sub-area, the clustering center is dynamically adjusted according to the load extreme point and the terrain mutation feature, and an anti-wind pressure optimized construction partition scheme is generated, which comprises: Calculate the gradient field according to the dynamic load thermal map, and extract the area with a load change rate exceeding a threshold value as the gradient mutation boundary; Based on the gradient mutation boundary, a region similarity matrix is constructed, and an adaptive Gaussian kernel function is used to quantify the load correlation strength of adjacent regions; The region similarity matrix is input into the spectral clustering algorithm, the initial clustering center position is constrained by the terrain elevation mutation point, and the preliminary construction sub-area division is performed; The clustering center is dynamically adjusted according to the real-time wind speed prediction data, so that the high load area is offset to the low wind pressure direction, and the construction sub-area boundary is optimized; The optimized sub-area is subjected to finite element wind pressure simulation verification, and a construction partition scheme meeting the critical wind pressure threshold value is output.

3. The method of claim 2, wherein, According to the construction partition scheme, combined with the real-time state data of the construction equipment, a device scheduling path is generated by using a space-time conflict resolution algorithm, the algorithm outputs a conflict-free construction resource allocation matrix through dynamic time window constraints and load balancing objective functions, which comprises: The construction zoning scheme is converted into a task node graph with time constraints, and nodes of the task node graph include equipment type requirements, operation time length, and safety load threshold; A device capability-position matrix is constructed according to device GPS positioning data and energy consumption logs, and the space-time cost of moving the device to each node is calculated; Based on the task node graph and the device capability-position matrix, a dynamic time window is used to detect device path conflicts, and an initial path sequence is generated through priority scheduling; A mixed integer programming solver is used to optimize the initial path sequence with load balancing as the objective function, and a device-task allocation matrix and a conflict-free scheduling path are output.

4. The method of claim 3, wherein, Based on the construction resource allocation matrix, the construction process is simulated through a digital twin model, real-time support deformation data and component inclination data are collected, an adaptive particle swarm optimization algorithm is used to dynamically correct construction parameters, and an anti-deformation construction instruction set is generated, including: The construction resource allocation matrix is mapped to the digital twin model, and the virtual construction machinery is driven to perform work according to the scheduling path; Real-time support deformation sensor data and component inclination data are collected to generate a millimeter-level precision deformation-inclination joint distribution map; According to the deformation-inclination joint distribution map, define deformation peak, inclination uniformity and progress deviation as a multi-objective fitness function, input adaptive particle swarm optimization algorithm; According to the real-time deformation data, dynamically adjust the search direction of the particle swarm, and solve the optimal mechanical control parameters under the constraints of the construction specification; The optimal mechanical control parameters are encoded into an anti-deformation construction instruction set, and the integrity and traceability of the instructions are ensured through blockchain signature.

5. The method of claim 4, wherein, According to the anti-deformation construction instruction set, the construction machinery is driven to perform operations, and energy efficiency feedback data is analyzed through edge computing nodes, and a reinforcement learning model is used to iteratively update the global load distribution atlas, forming a closed-loop construction management link, including: The anti-deformation construction instruction set is parsed into mechanical control signals to drive the hydraulic system and mechanical arm to perform high-precision construction operations; Through the edge computing node, real-time component installation error and mechanical energy consumption data are collected to generate energy efficiency-quality evaluation indicators; Based on the energy efficiency-quality evaluation indicators, a hierarchical reinforcement learning model is trained, the upper layer optimizes the load prediction parameters, and the lower layer adjusts the mechanical motion trajectory, iteratively updates the global load distribution atlas, and forms a closed-loop construction management link.

6. A load analysis-based photovoltaic power plant distributed construction management system, characterized by, The system comprises: The fusion module is used for fusing satellite remote sensing images and ground sensor time series data by a multi-modal fusion deep neural network according to terrain elevation data of a photovoltaic power station planning area, meteorological historical data and component mechanical parameters, and outputting a dynamic load intensity thermal map; wherein, according to the satellite remote sensing images and the terrain elevation data, a terrain-component mapping grid is generated by a geometric registration algorithm, and the spatial grid is output after eliminating cloud interference; the spatial grid and real-time wind speed, temperature and humidity data collected by the ground sensor are input into a space-time interpolation model, and a three-dimensional dynamic meteorological field is generated in combination with the meteorological historical data; the three-dimensional dynamic meteorological field and the component mechanical parameters are input into the multi-modal fusion network, terrain texture features are extracted by residual convolution, meteorological time series patterns are extracted by bidirectional LSTM, and an initial load prediction map is output after fusion; the initial load prediction map is subjected to space-time noise filtering guided by optical flow, the intensity distribution is optimized according to the load correlation of adjacent grids, and a dynamically updated load intensity thermal map is generated; The division module is used for dividing a construction sub-area by using a gradient-constrained spectral clustering algorithm based on the dynamic load intensity thermal map, dynamically adjusting the clustering center according to the load extreme point and the terrain mutation feature, and generating a wind pressure-resistant optimized construction zoning scheme; The generation module is used for generating an equipment dispatching path by using a space-time conflict resolution algorithm according to the construction zoning scheme and in combination with real-time state data of construction equipment, and outputting a conflict-free construction resource allocation matrix by a dynamic time window constraint and a load balancing objective function; The correction module is used for simulating a construction process by a digital twin model based on the construction resource allocation matrix, collecting support deformation data and component inclination data in real time, dynamically correcting construction parameters by using an adaptive particle swarm optimization algorithm, and generating a deformation-resistant construction instruction set; The update module is used for driving a construction machine to perform an operation according to the deformation-resistant construction instruction set, synchronously analyzing energy efficiency feedback data by an edge computing node, and iteratively updating a global load distribution atlas by using a reinforcement learning model, so as to form a closed-loop construction management link.

7. A storage medium, characterized by The storage medium stores a computer program, wherein the computer program is configured to execute the method of any one of claims 1-5 when running.

8. An electronic device comprising a memory and a processor, characterized in that The memory stores a computer program, and the processor is configured to execute the method of any one of claims 1-5 by running the computer program.

Images