Coal yard inventory scheduling system based on real-time data

Abstract

The invention discloses a coal yard inventory scheduling system based on real-time data, which comprises a sensing layer, an edge calculation layer, a digital twinborn layer, a decision-making layer and an execution layer, and is characterized in that the sensing layer is used for acquiring basic data; the edge calculation layer can collect and clean data of the sensing layer and can extract feature vectors of environment and equipment states; a three-dimensional density field can be constructed; the digital twinning layer receives data obtained by the edge calculation layer and constructs a three-dimensional dynamic digital twinning engine; after the decision-making layer receives the order information, a control instruction is generated after calculation; and after receiving the instruction issued by the decision-making layer, the execution layer performs operation according to the instruction. According to the invention, through multi-source real-time data perception, edge computing node rapid processing and digital twinborn simulation, accurate dynamic scheduling and management of coal yard inventory are realized; the digital twinborn layer is set for simulation rehearsal and mechanical analysis, the collapse risk is warned in advance, and the accident rate is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of coal yard scheduling, and in particular to a coal yard inventory scheduling system based on real-time data. Background Art

[0002] Traditional coal yard scheduling relies primarily on manual inspections and static data, which presents significant challenges: Manual equipment measurement requires long cycles, such as measuring coal pile volume, which can take hours. Inventory cannot be updated in real time, resulting in scheduling decisions relying on outdated information, which can easily lead to supply and demand imbalances. A lack of dynamic structural monitoring makes it difficult to detect potential hazards such as coal pile collapse and excessive slopes. Vehicle routing relies heavily on experience, which can easily lead to idle equipment, route conflicts, and difficulty responding to sudden environmental changes, such as extreme weather. While existing technologies have introduced some automated equipment, these isolated systems and weak data processing capabilities prevent collaborative optimization across the entire supply chain. Summary of the Invention

[0003] To this end, the present invention provides a coal yard inventory scheduling system based on real-time data to solve the problems in the prior art.

[0004] In order to achieve the above object, the present invention provides the following technical solutions:

[0005] A coal yard inventory scheduling system based on real-time data, including a perception layer, an edge computing layer, a digital twin layer, a decision layer, and an execution layer.

[0006] The perception layer includes a high-density LiDAR matrix, an inspection system with a drone equipped with an infrared thermal imager, dynamic weighing on a scale, an environmental perception network, and equipment status monitoring sensors.

[0007] The edge computing layer uses several sensors with computing and transmission capabilities in the perception layer as edge computing nodes. The edge computing nodes can collect data from the perception layer, clean the collected data, and extract feature vectors of the environment and device status. The edge computing nodes perform point cloud processing on the point cloud data of the perception layer and construct a three-dimensional density field.

[0008] The digital twin layer receives data from the edge computing layer, builds a three-dimensional dynamic digital twin engine, and integrates multispectral data to map the calorific value distribution in the digital twin engine;

[0009] After receiving the order information, the decision-making layer runs a hybrid optimization algorithm based on the feature vector preprocessed by the edge computing node, performs a simulation preview based on the virtual coal yard with physical characteristics built on the digital twin layer, and then generates equipment control instructions and sends them to the execution layer.

[0010] The execution layer, after receiving instructions from the decision-making layer, performs operations according to the instructions.

[0011] Furthermore: the deployment spacing of the laser radar is calculated according to an algorithm, and the deployment spacing algorithm is: d=2R·tan(θ / 2n), where R is the radius of the coal pile, θ is the radar field of view, n is the overlap rate, n≥30%, and the laser radar is installed on a tower.

[0012] Further: The process of point cloud processing is as follows:

[0013] (1) Background model initialization: When there is no interference from dynamic objects, collect multiple frames of static scene point clouds to build an initial background model;

[0014] (2) Dynamic background update: Periodically compare the newly collected point cloud with the background model and dynamically update the background to avoid false detection of static structure changes as dynamic objects;

[0015] (3) Downsampling of the original point cloud: voxel grid filtering or random sampling is performed on the high-density lidar point cloud to reduce the data volume and improve the efficiency of subsequent processing; key feature points are retained, including the area where the curvature of the coal pile surface changes;

[0016] (4) Dynamic area detection: Compare the current frame point cloud with the background model, extract the dynamic area through the distance threshold method or deep learning segmentation model; mark the dynamic point cloud clusters to distinguish the natural deformation of the coal pile surface from the real dynamic objects;

[0017] (5) Dynamic object tracking: Multi-target tracking of detected dynamic objects, prediction of motion trajectories; output of real-time position, speed and path of dynamic objects;

[0018] (6) Multi-view point cloud registration: align point clouds collected at multiple time periods or by multiple devices, and eliminate coordinate offsets through ICP algorithm or feature matching;

[0019] (7) Voxel processing: convert the registered point cloud into a voxel grid to fill the three-dimensional spatial structure of the coal pile; optimize the voxel density distribution and reduce voids;

[0020] (8) Density field reconstruction: Based on voxelized data, the local point cloud density field is calculated, and the mass of the coal pile is estimated by combining the density field with the volume data.

[0021] Further: The processing method of the edge computing node is as follows:

[0022] (1) The deployed laser radar scans the coal pile surface in real time to generate high-precision three-dimensional point cloud data; it simultaneously accesses environmental sensor data to correct point cloud acquisition errors;

[0023] (2) Receive and clean point cloud data in real time, remove noise points, standardize data format, and adapt to subsequent processing modules;

[0024] (3) Temporarily store the original point cloud and pre-processing results to support efficient throughput of sudden large-scale scanning data;

[0025] (4) Point cloud registration: align the point clouds scanned from multiple perspectives to eliminate coordinate offsets caused by equipment movement or environmental jitter; use the ICP algorithm for registration to build a complete three-dimensional model of the coal pile; and dynamically update the registration results.

[0026] (5) Based on the coal yard ground point cloud, a reference plane is fitted as the reference surface for volume calculation, and the surface of the registered point cloud is reconstructed to generate a closed coal pile surface and calculate the volume of the coal pile;

[0027] (6) Detect and identify structural risks of coal piles;

[0028] (7) The real-time calculation module monitors the volume change rate and issues an alarm for sudden volume changes;

[0029] (8) Identify foreign matter or classify coal quality in coal piles based on a lightweight AI model deployed on edge computing nodes;

[0030] (9) Push the volume calculation results to the coal yard management system; visualize and output the three-dimensional coal pile model, mark the key parameters, and synchronize the volume data to the cloud.

[0031] Further: The steps for building the digital twin engine are as follows:

[0032] (1) Point cloud data preprocessing and enhancement:

[0033] Receive point cloud data processed by edge computing nodes; synchronize access to coal yard environmental parameters and operation logs; segment and remove residual interference, retaining the main point cloud of the coal pile; enhance details in high-density areas and smooth low-density noise based on density field information; and use pre-trained models to distinguish semantic categories of coal piles, ground, and loading and unloading equipment.

[0034] (2) 3D model reconstruction and structure optimization:

[0035] 1) Voxelization: converting point clouds into high-resolution voxel grids, retaining density field data as voxel attributes; dynamically adjusting voxel resolution to balance computational efficiency and model accuracy;

[0036] 2) Surface reconstruction: using the MarchingCubes algorithm to generate a triangular mesh model of the coal pile surface from the voxel grid;

[0037] 3) Hole repair: Filling in missing areas of the model due to occlusion based on density field interpolation;

[0038] 4) Construct a hierarchical model, with a coarse-grained model for global visualization and a fine-grained model to support local mechanical analysis;

[0039] (3) Physical property binding:

[0040] Density binding, mapping the voxel density field to coal pile mass distribution properties;

[0041] Calorific value texture mapping, assigning calorific value parameters to different areas according to coal classification, supporting combustion efficiency simulation;

[0042] Humidity parameter binding, associated with environmental sensor data, marking the humidity gradient on the surface of the coal pile.

[0043] Furthermore: the digital twin engine supports multi-dimensional data overlay display; and provides a section analysis tool to view the density and moisture distribution inside the coal pile in real time.

[0044] Furthermore: the decision-making process of the decision-making layer is implemented as follows:

[0045] (1) Environmental perception

[0046] 1) Real-time data collection and equipment status monitoring: sensors are used to obtain real-time operating parameters of loaders and transport vehicles, including GPS location, oil pressure, engine speed, load status, and remaining fuel. Current job order data is synchronized from the enterprise ERP system or scheduling platform, including coal type, demand, priority, delivery time, and customer location. Weather stations and on-site sensors are integrated to collect environmental data and assess its impact on operational efficiency.

[0047] 2) Multi-source data integration: equipment status, order requirements, and environmental parameters are uniformly connected to edge computing nodes, timestamp alignment and data format standardization are performed to build a global perception data pool;

[0048] (2) Status code

[0049] 1) Spatial feature extraction: Convert the 3D point cloud data from the coal yard LiDAR scan into a voxel grid to preserve the spatial characteristics of the coal pile. Key areas, including loading and unloading areas, roads, and coal storage areas, are identified based on a semantic segmentation model and encoded into a spatial topology map.

[0050] 2) Time series feature modeling: extract the time series data of the device within the past hour and construct a time series matrix; analyze the time series data through long-short-term memory networks to capture the trend of device status changes;

[0051] 3) Multimodal feature fusion: combining spatial voxel features, temporal encoding results, and environmental parameters into a unified high-dimensional feature vector as the input state representation of the mathematical optimization model;

[0052] (3) Hybrid optimization solution

[0053] 1) Reinforcement learning strategy generation: calling a pre-trained reinforcement learning model to generate a set of candidate scheduling strategies based on the high-dimensional feature vector of the current state;

[0054] 2) Mathematical programming solution: Convert the scheduling problem into a mathematical optimization model. The objective function is to minimize total cost, and the constraints include equipment capacity, delivery deadline, and safety distance. Use the solver to solve the mathematical optimization model and output the optimal scheduling solution.

[0055] 3) Digital twin simulation verification: Load candidate strategies into the digital twin engine, simulate the entire execution process, and evaluate the actual effects; adjust strategy weights based on simulation results to select the most robust and feasible solutions;

[0056] (4) Instruction issuance

[0057] 1) Decompose the optimized scheduling plan into a sequence of device executable instructions;

[0058] 2) Send instructions to the target device; receive the device execution status in real time and dynamically trigger the exception handling mechanism.

[0059] Going a step further: after the job is completed, the execution layer will feed back the information to the enterprise ERP / MES system.

[0060] This invention offers the following advantages: Through multi-source real-time data sensing, rapid edge computing node processing, and digital twin simulation, it enables precise dynamic scheduling and management of coal yard inventory. Furthermore, by implementing digital twin simulation previews and mechanical analysis, it provides early warning of landslide risks, reducing accident rates.

[0061] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] To more intuitively illustrate the prior art and the present application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be considered as limiting conditions for implementing the present application. For example, based on the technical concepts disclosed in this application and the exemplary drawings, those skilled in the art are capable of easily making routine adjustments or further optimizations to the addition / reduction / attribution division of certain units (components), the specific shapes, positional relationships, connection methods, and dimensional ratios.

[0063] Figure 1 A system block diagram of a coal yard inventory scheduling system based on real-time data provided in one embodiment of the present application.

[0064] Figure 2A schematic diagram of a data interaction architecture between coal yard inventory scheduling systems based on real-time data provided by the present invention.

[0065] Figure 3 This is a flowchart of point cloud processing in a coal yard inventory scheduling system based on real-time data provided by the present invention.

[0066] Figure 4 This is a structural diagram of an edge computing node in a coal yard inventory scheduling system based on real-time data provided by the present invention.

[0067] Figure 5 This is a flowchart of the construction of a digital twin engine in a coal yard inventory scheduling system based on real-time data provided by the present invention.

[0068] Figure 6 This is a scheduling decision flow chart of a coal yard inventory scheduling system based on real-time data provided by the present invention. DETAILED DESCRIPTION

[0069] The following specific embodiments illustrate the implementation of the present invention. People familiar with this technology can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. It should be understood that these embodiments are only to further illustrate the present invention and cannot be understood as limiting the scope of protection of the present invention. Technical engineers in this field can make some non-essential improvements and adjustments to the present invention based on the content of the above invention; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0070] See also Figure 1-Figure 2 A coal yard inventory scheduling system based on real-time data includes a perception layer, an edge computing layer, a digital twin layer, a decision layer, and an execution layer.

[0071] The perception layer is a multimodal data perception layer, including a high-density LiDAR matrix (5G+LiDAR), a drone inspection system equipped with infrared thermal imagers, dynamic weighing based on RFID binding, an environmental perception network (temperature, humidity, dust, and water seepage detection), and equipment status monitoring sensors.

[0072] The lidar array uses Velodyne VLS-128 lidars (40° vertical viewing angle, 2.4 million points per second). The lidar spacing is calculated using an algorithm: d = 2R·tan(θ / 2n), where R is the radius of the coal pile, θ is the radar field of view, and n is an overlap ratio ≥ 30%. The lidars are mounted on a tower. The lidar array's scanning strategy automatically adjusts the scanning frequency (1-5Hz) based on the coal pile height. Interference from moving vehicles is automatically filtered using the RANSAC algorithm (point cloud preprocessing). The coal pile outline is then calculated using a spatial calibration algorithm. Rain and fog compensation and dust filtering are performed to achieve interference mitigation. Rain and fog compensation utilizes a 1550nm wavelength laser penetration solution, and dust filtering utilizes an adaptive threshold algorithm based on particle swarm optimization.

[0073] The drone inspection system first flies along a pre-set path, ensuring full coverage of the target area while maintaining a safe distance and a stable attitude. The ground station simultaneously monitors thermal imaging, marking abnormal areas (such as localized high-temperature spots). The system also stores raw infrared data, visible light images, GPS coordinates, and environmental parameters.

[0074] The processing flow for the data collected during flight is as follows:

[0075] a. Non-uniformity correction (NUC): eliminates the pixel response differences of the infrared detector itself;

[0076] b. Temperature calibration: Based on the blackbody radiation reference value, the original signal is converted into a temperature value;

[0077] c. Humidity compensation: Based on the humidity data recorded during inspection, correct the attenuation effect of humidity on thermal radiation;

[0078] d. Atmospheric transmittance correction: Adjust the radiation value of distant targets based on atmospheric parameters (such as CO2 and water vapor content);

[0079] e. Radiation calibration: convert the corrected data into standard radiation units;

[0080] f. Pseudo-color mapping: Convert temperature distribution into intuitive color images to highlight abnormal areas.

[0081] Equipment status monitoring sensors can obtain real-time operating parameters of loaders and transport vehicles, including GPS location, oil pressure, engine speed, load status, remaining fuel, etc.

[0082] The edge computing layer establishes node sensors with edge computing capabilities among several sensors in the perception layer. The edge computing nodes can collect sensor data, clean the collected data, and extract environmental feature vectors.

[0083] The edge computing layer can process the original point cloud data collected by the perception layer and construct a three-dimensional density field.

[0084] See Figure 3 , the process of point cloud processing is as follows:

[0085] (1) Background model initialization: When there is no interference from dynamic objects (such as vehicles and people), collect multiple frames of static scene point clouds and construct an initial background model (such as based on statistical mean or Gaussian mixture model).

[0086] (2) Dynamic background update: Periodically compare the newly collected point cloud with the background model, and dynamically update the background (such as the shape change of the coal pile due to loading and unloading) to avoid misdetecting static structure changes as dynamic objects; use a progressive update algorithm to balance the sensitivity to environmental changes and computational overhead.

[0087] (3) Downsampling of the original point cloud: Perform voxel grid filtering or random sampling on the high-density lidar point cloud to reduce the data volume (for example, from a million-level point cloud to a hundred thousand-level point cloud) and improve the efficiency of subsequent processing; retain key feature points (such as the area where the curvature of the coal pile surface changes) to avoid over-simplification that affects accuracy.

[0088] (4) Dynamic area detection: Compare the current frame point cloud with the background model, and extract dynamic areas (such as moving loading and unloading vehicles and inspection robots) through the distance threshold method or deep learning segmentation model; mark dynamic point cloud clusters to distinguish between natural deformation of the coal pile surface and real dynamic objects.

[0089] (5) Dynamic object tracking: Perform multi-target tracking of detected dynamic objects (e.g., based on Kalman filtering or Hungarian algorithm) and predict motion trajectories; output the real-time position, speed, and path of dynamic objects (e.g., the route of a vehicle in a coal yard) to support obstacle avoidance and job scheduling.

[0090] (6) Multi-view point cloud registration: align point clouds collected at multiple time periods or by multiple devices (such as fixed radar and vehicle-mounted radar data), and eliminate coordinate offsets through the ICP algorithm or feature matching (FPFH descriptor); to address the problem of sparse surface features of coal piles, introduce semantic-assisted registration (such as using the outline of loading and unloading equipment as a matching reference).

[0091] (7) Voxelization: Convert the registered point cloud into a voxel grid to fill the three-dimensional spatial structure of the coal pile; optimize the voxel density distribution and reduce voids (such as missing point clouds caused by dust occlusion).

[0092] (8) Density field reconstruction: Based on voxelized data, the local point cloud density field is calculated to reflect the compactness of different areas of the coal pile (such as the density difference between loose coal powder on the surface and the compacted coal pile inside); the density field is combined with volume data to estimate the mass of the coal pile (the density parameters of the coal type need to be preset).

[0093] See Figure 4 , the processing flow of the edge computing node is as follows:

[0094] Data access, collecting sensor data, that is, collecting the deployed laser radar (LiDAR) to scan the coal pile surface in real time to generate high-precision three-dimensional point cloud data; synchronously access environmental sensor data (such as temperature, humidity, and wind speed) to correct point cloud collection errors.

[0095] The stream processing engine receives and cleans point cloud data in real time, removes noise points (such as flying birds and temporary equipment interference), standardizes the data format (such as converting to PCD or PLY format), and adapts to subsequent processing modules.

[0096] Data caching temporarily stores raw point clouds and pre-processing results, supporting efficient throughput of sudden large-scale scanning data.

[0097] Point cloud registration aligns point clouds scanned from multiple perspectives (such as multiple scans of a coal pile by a drone), eliminating coordinate offsets caused by equipment movement or environmental jitter. It uses the ICP algorithm or NDT (normal distribution transform) for registration to construct a complete 3D model of the coal pile. The registration results are dynamically updated to adapt to real-time changes in the coal pile shape caused by loading and unloading operations.

[0098] Volume calculation is based on the coal yard ground point cloud. A datum plane is fitted as the reference surface for volume calculation. The registered point cloud is then reconstructed (such as triangulated meshing or voxelization) to generate a closed coal pile surface. The coal pile volume (unit: cubic meter) is calculated using spatial integration or slice projection methods, with an error within ±2%. Recalculation is triggered hourly or on demand to support real-time inventory statistics.

[0099] Anomaly detection: Detect and identify structural risks such as coal pile collapse and excessive slope (e.g., using curvature analysis or comparison with historical models).

[0100] Real-time calculation: The real-time calculation module monitors the rate of volume change and issues alarms for sudden volume changes, such as those triggered by abnormal loading and unloading (such as theft or dumping accidents).

[0101] Model inference involves deploying lightweight AI models (such as PointNet++) to identify foreign objects (such as mixed debris) in coal piles or classify coal quality (such as lump coal and pulverized coal). This lightweight AI model is accelerated by TensorRT, achieving inference response times within seconds at the edge.

[0102] Output the volume calculation results to the coal yard management system to generate daily / weekly inventory reports (including volume, density estimation, and calorific value analysis). Visualize and output a 3D coal pile model, annotating key parameters (maximum height, bottom area, and volume change trends).

[0103] Closed-loop control: Abnormal events (such as landslides) trigger audible and visual alarms, automatically dispatching inspection drones for retesting. Volume data is synchronized to the cloud for long-term trend analysis and supply chain optimization.

[0104] The digital twin layer receives data from the edge computing layer, builds a three-dimensional dynamic digital twin engine (digital twin model), and integrates multispectral data to map the calorific value distribution in the digital twin engine.

[0105] See Figure 5 , the steps to build the digital twin engine are as follows:

[0106] (1) Point cloud data preprocessing and enhancement

[0107] Laser point cloud input: Receives point cloud data processed by the edge computing node (registration, dynamic object removal, and density field reconstruction have been completed); synchronously accesses coal yard environmental parameters (temperature, humidity, wind speed) and operation logs (loading and unloading time, coal type).

[0108] Noise filtering and optimization: RANSAC algorithm: Segment and remove residual interference (such as temporary equipment and flying birds), retaining the main point cloud of the coal pile.

[0109] Density field guided filtering: Based on density field information, it enhances the details of high-density areas and smoothes low-density noise (such as loose coal powder on the surface).

[0110] Semantic segmentation: Use pre-trained models to distinguish semantic categories such as coal piles, ground, and loading and unloading equipment, providing labels for subsequent physical attribute binding.

[0111] (2) 3D model reconstruction and structure optimization

[0112] Voxelization: Convert the point cloud into a high-resolution voxel grid (e.g., 0.05 m³), ​​retaining density field data as voxel attributes; dynamically adjust voxel resolution to balance computational efficiency and model accuracy (e.g., using fine voxels in areas with frequent loading and unloading).

[0113] Surface reconstruction: The MarchingCubes algorithm is used to generate a triangular mesh model of the coal pile surface from the voxel grid to ensure smooth transitions and closed surfaces.

[0114] Hole repair: Fill in missing areas of the model caused by occlusion (such as dust and equipment) based on density field interpolation.

[0115] Multi-scale modeling: Build hierarchical models with coarse-grained models for global visualization and fine-grained models to support local mechanical analysis.

[0116] (3) Physical property binding and simulation empowerment

[0117] Physical property mapping includes:

[0118] Density binding, mapping the voxel density field to coal pile mass distribution properties (unit: kg / m³);

[0119] Calorific value texture mapping, which assigns calorific value parameters (unit: MJ / kg) to different regions according to coal classification (such as lignite, anthracite), supporting combustion efficiency simulation;

[0120] Humidity parameter binding, associated with environmental sensor data, marking the humidity gradient of the coal pile surface;

[0121] Mechanical simulation and stability analysis include:

[0122] Finite element analysis (FEA) simulates stress distribution in coal piles under loads (such as rainfall penetration and mechanical compaction) to predict areas at risk of collapse;

[0123] Slope safety warning: combining the coal pile geometry model and friction coefficient to calculate the critical slope and mark the exceeding area;

[0124] Dynamic update mechanism:

[0125] Real-time synchronization of edge node processing results (such as coal pile deformation caused by new loading and unloading) triggers local model reconstruction and attribute updates.

[0126] The interactive 3D twin interface supports the overlay display of multi-dimensional data (such as calorific value distribution thermograms and stress cloud maps). It also provides cross-section analysis tools for real-time viewing of the density and moisture distribution within the coal pile.

[0127] Based on volume and density data, the mass of the coal pile is automatically calculated and an inventory report is generated. The finite element analysis results trigger a landslide warning and are pushed to the inspector's handheld terminal. The coal scheduling strategy is optimized by combining calorific value distribution and boiler parameters.

[0128] The lightweight twin model is retained at the edge to support real-time alerts and local decision-making; high-precision models and historical data are synchronized to the cloud for long-term trend analysis (such as coal quality degradation prediction).

[0129] See Figure 6 ,After receiving the order information, the decision layer runs the ,hybrid optimization algorithm based on the relevant feature vectors pre-processed by the ,edge computing nodes (including real-time collected equipment status data, ,environmental data, etc.), and conducts a simulation preview of the ,virtual coal yard with physical characteristics constructed based on the ,digital twin layer, and then generates equipment control instructions and ,sends them to the execution layer.

[0130] The decision-making process of the decision-making layer is implemented as follows:

[0131] (1) Environmental perception

[0132] 1) Real-time data collection: Equipment status monitoring uses sensors to obtain real-time operating parameters of loaders and transport vehicles, including GPS location, oil pressure, engine speed, load status, remaining fuel, etc.; synchronize current job order data from the enterprise ERP system or scheduling platform, including coal type, demand, priority, delivery time, and customer location; integrate weather stations with on-site sensors to collect environmental data such as wind speed, temperature, humidity, and visibility, and evaluate their impact on operational efficiency (for example, strong winds may limit drone inspections).

[0133] 2) Multi-source data integration: Device status, order requirements, and environmental parameters are uniformly connected to edge computing nodes, timestamp alignment and data format standardization are performed to build a global perception data pool.

[0134] (2) Status code

[0135] 1) Spatial feature extraction: The 3D point cloud data from the coal yard LiDAR scan is converted into a voxel grid with a resolution of 128×128×128, preserving spatial features such as the volume and density distribution of the coal pile. Key areas such as loading and unloading areas, roads, and coal storage areas are identified based on a semantic segmentation model and encoded into a spatial topology map.

[0136] 2) Time series feature modeling: Extract time series data such as the equipment's operating trajectory, operation frequency, and fault records within the past hour to construct a time series matrix. Analyze the time series data using a long short-term memory (LSTM) network to capture trends in equipment status changes (such as fuel consumption decay patterns and load fluctuation cycles).

[0137] 3) Multimodal feature fusion: Combine spatial voxel features, temporal encoding results, and environmental parameters into a unified high-dimensional feature vector as the input state representation of the optimization model.

[0138] (3) Hybrid optimization solution

[0139] 1) Reinforcement learning strategy generation: Calling pre-trained reinforcement learning models (such as PPO and DQN) to generate a set of candidate scheduling strategies based on current state features, such as vehicle path planning and loader task allocation priority.

[0140] 2) Mathematical programming solution: The scheduling problem is transformed into a mathematical optimization model with the objective function of minimizing total cost (fuel consumption, time delay, etc.). Constraints include equipment capacity, delivery deadlines, and safe spacing. Solvers such as CPLEX and Gurobi are used to efficiently solve the model and output the optimal scheduling solution (such as vehicle-coal pile matching and loading and unloading time windows).

[0141] 3) Digital twin simulation verification: Load candidate strategies into the digital twin engine, simulate the entire execution process (such as vehicle movement and coal loading and unloading), and evaluate the actual results (such as task completion rate and resource utilization). Adjust the strategy weights based on the simulation results to select the most robust feasible solution.

[0142] (4) Instruction issuance

[0143] 1) Control instruction generation: Decompose the optimized scheduling plan into a sequence of executable instructions for the equipment, such as the loader's operating path coordinates, the transport vehicle's navigation route, and the loading and unloading robot's motion parameters.

[0144] 2) Real-time delivery and execution: Instructions are delivered to target devices via Time-Sensitive Networking (TSN). Device execution status (such as location updates and task progress) is received in real time, dynamically triggering exception handling mechanisms (such as path replanning and faulty device replacement).

[0145] The execution layer, after receiving instructions from the decision-making layer, performs operations according to the instructions.

[0146] After the operation is completed, the information will be fed back to the enterprise ERP / MES system.

[0147] In this embodiment, the perception layer and the edge computing layer adopt a lightweight communication protocol; the digital twin layer and the decision layer interact through a state space interface; and the execution layer device supports plug-and-play.

[0148] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A coal yard inventory scheduling system based on real-time data, characterized in that: Including perception layer, edge computing layer, digital twin layer, decision layer and execution layer, The perception layer includes a high-density LiDAR matrix, an inspection system with a drone equipped with an infrared thermal imager, dynamic weighing on a scale, an environmental perception network, and equipment status monitoring sensors. The edge computing layer uses several sensors with computing and transmission capabilities in the perception layer as edge computing nodes. The edge computing nodes can collect data from the perception layer, clean the collected data, and extract feature vectors of the environment and device status. The edge computing nodes perform point cloud processing on the point cloud data of the perception layer and construct a three-dimensional density field. The digital twin layer receives data from the edge computing layer, builds a three-dimensional dynamic digital twin engine, and integrates multispectral data to map the calorific value distribution in the digital twin engine; After receiving the order information, the decision-making layer runs a hybrid optimization algorithm based on the feature vector preprocessed by the edge computing node, performs a simulation preview based on the virtual coal yard with physical characteristics built on the digital twin layer, and then generates equipment control instructions and sends them to the execution layer. The execution layer, after receiving instructions from the decision-making layer, performs operations according to the instructions.

2. A coal yard inventory scheduling system based on real-time data according to claim 1, characterized in that: The deployment spacing of the laser radar is calculated according to an algorithm. The deployment spacing algorithm is: d=2R·tan(θ / 2n), where R is the radius of the coal pile, θ is the radar field of view, and n is the overlap rate. n≥30%. The laser radar is installed on a tower.

3. The coal yard inventory scheduling system based on real-time data according to claim 1 is characterized in that: The process of point cloud processing is as follows: (1) Background model initialization: When there is no interference from dynamic objects, collect multiple frames of static scene point clouds to build an initial background model; (2) Dynamic background update: Periodically compare the newly collected point cloud with the background model and dynamically update the background to avoid false detection of static structure changes as dynamic objects; (3) Downsampling of the original point cloud: voxel grid filtering or random sampling is performed on the high-density lidar point cloud to reduce the data volume and improve the efficiency of subsequent processing; key feature points are retained, including the area where the curvature of the coal pile surface changes; (4) Dynamic area detection: Compare the current frame point cloud with the background model, extract the dynamic area through the distance threshold method or deep learning segmentation model; mark the dynamic point cloud clusters to distinguish the natural deformation of the coal pile surface from the real dynamic objects; (5) Dynamic object tracking: Multi-target tracking of detected dynamic objects and prediction of motion trajectories; Output the real-time position, speed and path of dynamic objects; (6) Multi-view point cloud registration: align point clouds collected at multiple times or by multiple devices, and eliminate coordinate offsets through ICP algorithm or feature matching; (7) Voxel processing: convert the registered point cloud into a voxel grid to fill the three-dimensional spatial structure of the coal pile; optimize the voxel density distribution and reduce voids; (8) Density field reconstruction: Based on voxelized data, the local point cloud density field is calculated, and the mass of the coal pile is estimated by combining the density field with the volume data.

4. The coal yard inventory scheduling system based on real-time data according to claim 1 is characterized in that: The processing method of the edge computing node is as follows: (1) The deployed laser radar scans the coal pile surface in real time to generate high-precision three-dimensional point cloud data; it simultaneously accesses environmental sensor data to correct point cloud acquisition errors; (2) Receive and clean point cloud data in real time, remove noise points, standardize data format, and adapt to subsequent processing modules; (3) Temporarily store the original point cloud and pre-processing results to support efficient throughput of sudden large-scale scanning data; (4) Point cloud registration: align the point clouds scanned from multiple perspectives to eliminate coordinate offsets caused by equipment movement or environmental jitter; use the ICP algorithm for registration to build a complete three-dimensional model of the coal pile; and dynamically update the registration results. (5) Based on the coal yard ground point cloud, a reference plane is fitted as the reference surface for volume calculation, and the surface of the registered point cloud is reconstructed to generate a closed coal pile surface and calculate the volume of the coal pile; (6) Detect and identify structural risks of coal piles; (7) The real-time calculation module monitors the volume change rate and issues an alarm for sudden volume changes; (8) Identify foreign matter or classify coal quality in coal piles based on a lightweight AI model deployed on edge computing nodes; (9) Push volume calculation results to the coal yard management system; Visually output the 3D coal pile model, annotate key parameters, and synchronize volume data to the cloud.

5. The coal yard inventory scheduling system based on real-time data according to claim 1 is characterized in that: The steps for building the digital twin engine are as follows: (1) Point cloud data preprocessing and enhancement: Receive point cloud data processed by edge computing nodes; synchronize access to coal yard environmental parameters and operation logs; segment and remove residual interference to retain the main point cloud of the coal pile; Based on density field information, it enhances the details of high-density areas and smoothes low-density noise. It also uses pre-trained models to distinguish the semantic categories of coal piles, ground, and loading and unloading equipment. (2) 3D model reconstruction and structure optimization: 1) Voxelization: converting point clouds into high-resolution voxel grids, retaining density field data as voxel attributes; dynamically adjusting voxel resolution to balance computational efficiency and model accuracy; 2) Surface reconstruction: using the MarchingCubes algorithm to generate a triangular mesh model of the coal pile surface from the voxel grid; 3) Hole repair: Filling in missing areas of the model due to occlusion based on density field interpolation; 4) Construct a hierarchical model, with a coarse-grained model for global visualization and a fine-grained model to support local mechanical analysis; (3) Physical property binding: Density binding, mapping the voxel density field to coal pile mass distribution properties; Calorific value texture mapping, assigning calorific value parameters to different areas according to coal classification, supporting combustion efficiency simulation; Humidity parameter binding, associated with environmental sensor data, marking the humidity gradient on the surface of the coal pile.

6. The coal yard inventory scheduling system based on real-time data according to claim 5 is characterized in that: The digital twin engine supports multi-dimensional data overlay display and provides a section analysis tool to view the density and moisture distribution inside the coal pile in real time.

7. The coal yard inventory scheduling system based on real-time data according to claim 1 is characterized in that: The decision-making process of the decision-making layer is implemented as follows: (1) Environmental perception 1) Real-time data collection and equipment status monitoring: sensors are used to obtain real-time operating parameters of loaders and transport vehicles, including GPS location, oil pressure, engine speed, load status, and remaining fuel. Current job order data is synchronized from the enterprise ERP system or scheduling platform, including coal type, demand, priority, delivery time, and customer location. Weather stations and on-site sensors are integrated to collect environmental data and assess its impact on operational efficiency. 2) Multi-source data integration: integrating equipment status, order requirements, and environmental parameters into edge computing nodes, aligning timestamps and standardizing data formats to build a global perception data pool; (2) Status code 1) Spatial feature extraction: Convert the 3D point cloud data from the coal yard LiDAR scan into a voxel grid to preserve the spatial characteristics of the coal pile. Key areas, including loading and unloading areas, roads, and coal storage areas, are identified based on a semantic segmentation model and encoded into a spatial topology map. 2) Time series feature modeling: extract the time series data of the device within the past hour and construct a time series matrix; analyze the time series data through long-short-term memory networks to capture the trend of device status changes; 3) Multimodal feature fusion: combining spatial voxel features, temporal encoding results, and environmental parameters into a unified high-dimensional feature vector as the input state representation of the mathematical optimization model; (3) Hybrid optimization solution 1) Reinforcement learning strategy generation: calling a pre-trained reinforcement learning model to generate a set of candidate scheduling strategies based on the high-dimensional feature vector of the current state; 2) Mathematical programming solution: Convert the scheduling problem into a mathematical optimization model. The objective function is to minimize total cost, and the constraints include equipment capacity, delivery deadline, and safety distance. Use the solver to solve the mathematical optimization model and output the optimal scheduling solution. 3) Digital twin simulation verification: Load candidate strategies into the digital twin engine, simulate the entire execution process, and evaluate the actual effects; adjust strategy weights based on simulation results to select the most robust and feasible solutions; (4) Instruction issuance 1) Decompose the optimized scheduling plan into a sequence of device executable instructions; 2) Send instructions to the target device; receive the device execution status in real time and dynamically trigger the exception handling mechanism.

8. The coal yard inventory scheduling system based on real-time data according to claim 1 is characterized in that: After the job is completed, the execution layer will feed back the information to the enterprise ERP / MES system.

Images