A coal mining-oriented energy consumption optimization and carbon emission management method and system

Abstract

The application provides a coal mining-oriented energy consumption optimization and carbon emission management method and system, applied to the technical field of coal mining, through obtaining multi-source heterogeneous operation reference information, extracting equipment displacement characteristics and judging production space layout changes, if changes occur, redefine the equipment space topology relationship and operation physical causality logic, then perform state consistency checking on the multi-source heterogeneous operation reference information, identify missing items in the real-time sensor data stream and perform logical filling, build a virtual situation model reflecting the current operation state of the mine, finally receive artificial dispatching instructions and perform simulation calculation, output energy consumption and carbon emission influence estimation parameters, which can effectively deal with the problems of data missing and inconsistency under complex dynamic geological environment, and has the beneficial effect of simultaneously ensuring production safety, operation efficiency and carbon emission optimization.

Description

Technical Field

[0001] This application relates to the field of coal mining technology, and in particular to a method and system for energy consumption optimization and carbon emission management in coal mining. Background Technology

[0002] In modern coal mining, production scheduling centers rely on underground sensors and fiber optic communication networks to acquire real-time operational data from systems such as mining, transportation, and ventilation. This data is then used for efficiency optimization and safety assurance. However, the dynamic and unstable geological environment of coal mines, coupled with mining activities and mine pressures causing localized compression, shearing, and stretching of the rock mass, results in network equipment within the tunnels continuously experiencing random and intermittent mechanical stress. This stress triggers transient faults such as micro-bending of fiber optic cables and momentary malfunctions in sensor connections, leading to signal attenuation or data interruption. These faults are intermittent and difficult to detect. Simple retransmission and verification mechanisms cannot handle such complex data contamination, resulting in gaps, delays, packet loss, and timestamp asynchronization in the data stream, leading to severe data inconsistency. Therefore, purely automated scheduling systems cannot accurately assess equipment load, calculate material flow, or predict equipment status.

[0003] However, if automation is abandoned and information is verified with underground personnel using walkie-talkies, telephones, or other manual methods, instructions are issued entirely based on experience. While this mode can maintain production operations, it introduces decision-making delays, the risk of human error, and sluggish response. More importantly, the precision scheduling model, originally designed for dual optimization of energy consumption and carbon emissions, becomes completely unusable due to the lack of reliable real-time data.

[0004] Therefore, existing technologies cannot maintain the reliable operation of automated scheduling systems or avoid the inefficiency and high energy consumption of manual scheduling when real-time data streams are severely polluted due to intermittent and distributed sensor and communication failures caused by dynamic geological stress in underground coal mines. Thus, it is difficult to simultaneously ensure the synergistic goals of production safety, operational efficiency and carbon emission optimization.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] In view of the shortcomings of the prior art, this application provides a method and system for energy consumption optimization and carbon emission management in coal mining. By constructing an accurate virtual situation model of the mine, and on this basis, the energy consumption and carbon emission impact of manual dispatching instructions can be predicted. This can effectively address the problem of data missing and inconsistency in complex dynamic geological environments, and has the beneficial effect of simultaneously ensuring the synergistic goals of production safety, operational efficiency and carbon emission optimization.

[0007] Firstly, a method for energy consumption optimization and carbon emission management in coal mining, the method comprising the following steps: S1: Obtain multi-source heterogeneous operation reference information for the target mine. The multi-source heterogeneous operation reference information includes manual description information, equipment operation characterization information, real-time sensor data stream, and preset geological and production plan context information. S2: Extract equipment displacement features from the equipment operation characterization information, and combine them with the manual description information to determine whether the production space layout of the target mine has changed; S3: If a change occurs, the node coordinates and connection edge attributes in the equipment space topology are redefined according to the equipment displacement characteristics, and the material flow constraints in the physical causal logic are remapped based on the updated equipment space topology. S4: Using the updated device space topology and operational physical causal logic, perform state consistency verification on the multi-source heterogeneous operational reference information; S5: Identify missing items in the real-time sensor data stream, and combine the results of the state consistency verification with the geological and production plan context information, run the updated operational physical causal logic to logically fill in the missing items, and construct a virtual situation model reflecting the current operating status of the mine. S6: Receive manual dispatch instructions, introduce the manual dispatch instructions as input variables into the virtual situation model, and perform simulation calculations in combination with preset energy consumption and carbon emission calculation rules, and output the energy consumption and carbon emission impact prediction parameters corresponding to the manual dispatch instructions.

[0008] Furthermore, step S1 includes: S11: Receive human voice reports in real time, convert the voice reports into text information, extract key status words and reporting timestamps, and simultaneously receive device status, environmental parameter data and their geographical location tags entered by the mobile terminal. Store the above information in a structured manner as the human description information. S12: Acquire real-time video stream of key work area, identify pixel displacement features and shape change features of target equipment in video stream, determine equipment operation, material flow status and environmental anomalies, and convert the identification results into structured event information with confidence as the equipment operation characterization information; S13: From the production planning database and geological information system, read the preset shift production targets, equipment maintenance periods, mining progress plans, coal seam thickness, and fault locations according to time intervals and spatial working face ranges, as the preset geological and production planning context information; S14: Real-time acquisition of the device's power, current, vibration, temperature, and rotation speed time-series physical quantities, recording the device identifier, timestamp, time-series physical quantity type, and value as the real-time sensing data stream.

[0009] Furthermore, step S2 includes: S21: Read the pixel coordinate trajectory of the device in continuous video frames from the device operation characterization information, and use the displacement vector between the start and end points of the pixel coordinate trajectory as the displacement feature of the device. S22: Perform natural language parsing on the manually described information, identify the keywords such as equipment relocation, mining face advancement or roadway connection contained therein, and extract the equipment name, change time and location description associated with the keywords to generate text change information; S23: Perform a spatiotemporal correlation comparison between the displacement feature and the text change information. If the device identifiers of the two are consistent, the change time windows overlap, and the actual spatial location indicated by the displacement feature matches the location description in the text change information, then it is comprehensively determined that the production space layout has changed.

[0010] Furthermore, step S3 includes: S31: Superimpose the actual spatial displacement vector in the device displacement feature onto the original node coordinates of the device in the device spatial topology to generate updated node coordinates; S32: Based on the updated node coordinates, recalculate the spatial distance and relative azimuth angle between the device and its adjacent devices. If the spatial distance exceeds the preset connection threshold, disconnect the original connection edge. If the new node coordinates enter the connection range of another device, add a new connection edge. If the distance changes but is still within the threshold, update the connection edge attributes. The connection edge attributes include at least the material conveying distance, transmission delay time, material flow limit, start-stop timing dependency, and energy consumption coefficient. S33: For each connection edge in the updated device space topology, a directed causal edge is generated in the running physical causal logic. The direction of the directed causal edge is consistent with the material flow direction of the connection edge, and the constraint parameters of the directed causal edge inherit the connection edge attributes. S34: Write all directed causal edges into the running physical causal logic, replace all existing causal edges related to the device, and form a material flow constraint relationship that perfectly matches the current spatial topology.

[0011] Furthermore, step S4 includes: S41: Map each piece of data in the multi-source heterogeneous operation reference information to the corresponding node in the updated device space topology according to its associated device identifier; S42: For each node, a preset consistency inference rule base is invoked to perform logical constraint matching on the multi-source heterogeneous operation reference information mapped to that node. If the multi-source heterogeneous operation reference information points to the same operation state, then that operation state is taken as the verified state of the device and assigned a first confidence score. If there is a conflict, then the information source with the highest credibility is selected as the main basis according to the preset arbitration strategy and assigned a second confidence score, while recording the conflict type. The second confidence score is less than the first confidence score. S43: Based on the direction of the directed causal edge in the physical causal logic, perform causal consistency verification on equipment with material flow relationship: if the upstream equipment is stopped after verification and the downstream equipment is running after verification, it is determined to be a violation of causal logic, and the status of the downstream equipment is forcibly corrected to the pending confirmation status, and the corresponding first confidence score or second confidence score is reduced. S44: Output the verified state of each node and its corresponding first confidence score or second confidence score as the result of the state consistency verification.

[0012] Furthermore, in step S5, identifying missing items in the real-time sensor data stream, and combining the results of the state consistency check with the geological and production planning context information, and running the updated operational physical causal logic to logically fill in the missing items includes the following steps: S51: Based on the spatial topology of the equipment, identify the upstream and downstream related equipment that have a material flow relationship with the faulty equipment that generated the missing item; S52: Based on the physical causal logic of the operation and the result of the state consistency verification, obtain the real-time start-stop status of the upstream and downstream related equipment, and combine the geological and production planning context information to deduce the theoretical operating conditions of the faulty equipment during the missing period. S53: Obtain regional aggregated power load information, extract the total power change characteristics from the aggregated power load information, and verify whether the total power change characteristics match the rated power of the equipment under the theoretical operating conditions; S54: If a match is found, the missing item is logically filled in according to the preset data template corresponding to the theoretical operating condition.

[0013] Furthermore, in step S5, constructing a virtual situational model reflecting the current operating status of the mine includes the following steps: S55: The logical filling result of the missing item is timestamped with the non-missing item in the real-time sensing data stream to form a complete operating state sequence of the faulty device on a continuous time axis. S56: Based on the equipment spatial topology, map the complete operating state sequence to the corresponding node of the faulty equipment in the topology graph, and synchronously update the material flow and load status of upstream and downstream related equipment that have a material flow relationship with the faulty equipment; S57: Extract features and vectorize the artificial description information, equipment operation characterization information, aggregated power load information, and preset geological and production plan context information according to a unified spatiotemporal benchmark, and bind them to the corresponding nodes in the equipment spatial topology relationship; S58: Iteratively aggregate the status information of all bound nodes to generate a comprehensive digital image reflecting the current operating status of the target mine, which serves as the virtual situation model.

[0014] Furthermore, step S52 includes: S521: From the result of the state consistency verification, read the verified state of each device in the upstream device set and downstream device set that are connected to the faulty device by a directed causal edge, and its corresponding first confidence score or second confidence score. The verified state with the first confidence score or second confidence score higher than the preset adoption threshold is taken as the real-time start / stop state of the device. S522: Extract constraints from the geological and production planning context information; S523: Using the real-time start-stop state as the observation variable and the constraint conditions as the reasoning boundary, and utilizing the causal transmission rules defined in the operational physical causal logic, perform forward causal propagation and reverse demand reasoning along the directed causal edge to calculate the set of possible operating conditions of the faulty equipment under all constraint conditions. S524: Select the operating condition with the highest confidence level from the set of possible operating conditions as the theoretical operating condition.

[0015] Furthermore, step S6 includes: S61: Parse the manual dispatch instruction and extract the target device identifier, action type, and action execution time window from the instruction; S62: The action type and time window are used as external input variables and injected into the state variables of the corresponding target device node in the virtual situation model, overriding the original running state of the node within the time window; S63: Based on the injected virtual situation model, the operating condition evolution of all devices within the time window is simulated sequentially according to the preset simulation step size. The operating condition evolution follows the material flow constraints in the physical causal logic of the operation and the connection edge attributes in the spatial topology of the devices. S64: During the simulation, according to the preset energy consumption calculation rules, the theoretical energy consumption value of each device node in each simulation step is accumulated. S65: Calculate the theoretical carbon emissions for each device node by multiplying the theoretical energy consumption value by a preset carbon emission factor, and sum them up to obtain the total carbon emissions. S66: Compare the total energy consumption and total carbon emissions before and after executing the manual scheduling instruction, calculate the difference, and output the energy consumption and carbon emission impact prediction parameters corresponding to the manual scheduling instruction.

[0016] Secondly, an energy consumption optimization and carbon emission management system for coal mining, the system being used to implement the steps of any of the above methods, the system comprising: Acquisition Module: Acquires multi-source heterogeneous operation reference information for the target mine; the multi-source heterogeneous operation reference information includes manual description information, equipment operation characterization information, real-time sensor data stream, and preset geological and production plan context information; Judgment module: Extracts equipment displacement features from the equipment operation characterization information and combines them with the manual description information to determine whether the production space layout of the target mine has changed; Remapping module: If a change occurs, the node coordinates and connection edge attributes in the equipment space topology are redefined according to the equipment displacement characteristics, and the material flow constraints in the running physical causal logic are remapped based on the updated equipment space topology. Verification module: Using the updated device space topology and operational physical causal logic, the state consistency of the multi-source heterogeneous operational reference information is verified; Construction module: Identifies missing items in the real-time sensor data stream, and combines the results of the state consistency verification with the geological and production plan context information, runs the updated operational physical causal logic to logically fill in the missing items, and constructs a virtual situation model reflecting the current operating status of the mine; Parameter estimation module: Receives manual scheduling instructions, introduces the manual scheduling instructions as input variables into the virtual situation model, and performs simulation calculations in combination with preset energy consumption and carbon emission calculation rules, and outputs the energy consumption and carbon emission impact estimation parameters corresponding to the manual scheduling instructions.

[0017] Beneficial Effects: The energy consumption optimization and carbon emission management method and system proposed in this application for coal mining acquires multi-source heterogeneous operation reference information, extracts equipment displacement characteristics and judges changes in production space layout. If changes occur, the spatial topology relationship of equipment and the causal logic of operation are redefined. Then, the state consistency of the multi-source heterogeneous operation reference information is checked, missing items in the real-time sensor data stream are identified and logically filled, and a virtual situation model reflecting the current operating state of the mine is constructed. Finally, manual scheduling instructions are received and simulation calculations are performed to output the predicted parameters of energy consumption and carbon emission impact. This method can effectively deal with the problem of data missingness and inconsistency in complex dynamic geological environments, construct an accurate virtual situation model of the mine, and on this basis, realize the prediction of energy consumption and carbon emission impact of manual scheduling instructions. It has the beneficial effect of simultaneously ensuring the synergistic goals of production safety, operational efficiency and carbon emission optimization. Attached Figure Description

[0018] Figure 1 This is a flowchart of an energy consumption optimization and carbon emission management method for coal mining proposed in this application.

[0019] Figure 2 This is a structural diagram of an energy consumption optimization and carbon emission management system for coal mining proposed in this application.

[0020] Labeling Explanation: 201. Acquisition Module; 202. Judgment Module; 203. Remapping Module; 204. Verification Module; 205. Construction Module; 206. Parameter Prediction Module. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. The components of the embodiments of this application described and marked in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] Faced with the problems of severe pollution of real-time operating data, failure of purely automated optimization models, and low efficiency and high energy consumption of purely manual scheduling and decision-making modes, how to design a method that combines manual and automated optimization models to achieve effective optimization and control of energy consumption and carbon emissions has become a major challenge in existing technologies.

[0024] Therefore, please refer to Figure 1 This application proposes a method for energy consumption optimization and carbon emission management in coal mining, the method comprising the following steps: S1: Obtain multi-source heterogeneous operation reference information for the target mine. The multi-source heterogeneous operation reference information includes human description information, equipment operation characterization information, real-time sensor data stream, and preset geological and production plan context information. S2: Extract equipment displacement features from equipment operation characterization information and combine them with manual description information to determine whether the production space layout of the target mine has changed; S3: If a change occurs, redefine the node coordinates and connection edge attributes in the equipment space topology based on the equipment displacement characteristics, and remap the material flow constraints in the physical causal logic based on the updated equipment space topology. S4: Use the updated device space topology and operational physical causal logic to perform state consistency verification on multi-source heterogeneous operational reference information; S5: Identify missing items in the real-time sensor data stream, and combine the results of the state consistency verification with geological and production plan context information to run the updated operational physical causal logic to logically fill in the missing items and construct a virtual situation model that reflects the current operating status of the mine. S6: Receive manual dispatch instructions, introduce the manual dispatch instructions as input variables into the virtual situation model, and perform simulation calculations in combination with preset energy consumption and carbon emission calculation rules, and output the energy consumption and carbon emission impact prediction parameters corresponding to the manual dispatch instructions.

[0025] The core working principle of this application lies in abandoning the reliance on uninterrupted, delay-free, and precisely synchronized raw real-time data streams, and instead constructing a fault-tolerant comprehensive data processing and state reconstruction module. This module actively aggregates and integrates information from multiple sources, with different structures and asynchronous types, specifically including: human voice reports, video stream analysis results from key areas, regionally aggregated power data, and basic switching status indicators of equipment. For this fragmented and potentially contradictory information, the module does not discard it directly due to timestamp misalignment or intermittent data gaps. Instead, it applies a set of logical association and inference rules, combined with state consistency verification operations, to discover and resolve conflicts between information.

[0026] Building on this foundation, the module further incorporates pre-defined geological condition data, production operation plans, and other contextual information, as well as causal constraint logic based on the physical world, to logically deduce and fill in key missing data.

[0027] Through the above processing, this module generates a virtual situational awareness model reflecting the overall current operating status of the mine. Although this virtual situational awareness model is not built based on millisecond-level precise data, its comprehensive information coverage and reliable state estimation are sufficient to support operational monitoring and decision-making. Subsequently, this virtual situational awareness model is presented to the dispatcher through a visual interface, along with a scenario simulation-based auxiliary decision-making tool. The dispatcher can input the dispatching instructions to be executed based on their own experience, and the module will then quickly simulate the multidimensional impact range and degree that the instructions may produce based on the generated virtual situational awareness model and preset energy consumption and carbon emission calculation rules.

[0028] This mechanism transforms the dispatcher's experience-based decision-making from blind trial-and-error into an iterative optimization process with quantifiable feedback. Ultimately, this method converts automated optimization capabilities, which were previously rendered ineffective due to data quality issues, into intelligent assistance for human decision-making, thereby enabling operation and control based on assisted optimization under traditional manual management conditions.

[0029] Specifically, acquiring multi-source heterogeneous operational reference information for the target mine aims to break away from the reliance on single, easily interfered traditional sensor data streams. By establishing a diversified information collection channel, a rich and resilient data foundation can be provided for the subsequent construction of a robust mine operation status model.

[0030] Furthermore, step S1 includes: S11: Receive real-time human voice reports, convert the voice reports into text information, extract key status words and report timestamps, and simultaneously receive device status, environmental parameter data and their geographical location tags entered by the mobile terminal. Store the above information in a structured manner as human description information. S12: Acquire real-time video streams of key work areas, identify pixel displacement and shape change features of target equipment in the video stream, determine equipment operation actions, material flow status and environmental anomalies, and convert the identification results into structured event information with confidence as equipment operation characterization information; S13: From the production planning database and geological information system, read the preset shift output targets, equipment maintenance periods, mining progress plans, coal seam thickness, and fault locations according to time intervals and spatial working face ranges, as preset geological and production planning context information; S14: Real-time acquisition of the device's power, current, vibration, temperature, and rotation speed time-series physical quantities, recording the device identifier, timestamp, time-series physical quantity type, and value as the real-time sensing data stream.

[0031] At the control console in the dispatch center, a communication system integrating an automatic speech recognition engine is deployed. This engine is specifically optimized for the background noise and industry terminology of the coal mine working environment. When the shift leader at the underground working face reports via digital walkie-talkie, for example, "Mining machine A has encountered a hard rock layer, making cutting difficult," this voice report is captured in real time. The speech recognition engine converts it into text: "Mining machine A, hard rock layer, cutting difficult." Subsequently, the natural language processing module analyzes the text, extracting key status terms, such as the equipment identifier "mining machine A," the event description "cutting difficult," and associating it with the current report timestamp.

[0032] Meanwhile, underground inspection personnel can use a handheld industrial-grade explosion-proof tablet computer to manually input equipment status information such as "B belt conveyor is operating normally and without deviation" or "D tunnel ventilation seems insufficient" through a pre-set form application. The application automatically records the timestamp of the input and the geographic location tag obtained through the built-in GPS module or base station positioning.

[0033] All of this information, obtained through voice and manual input, is eventually formatted uniformly, for example, stored as a JSON object containing fields such as information source, time, location, and key content, and then stored in an unstructured information database as manually described information.

[0034] Industrial-grade explosion-proof network cameras are typically installed at key locations such as coal faces and transfer points in main transport roadways. These cameras transmit high-definition video streams in real time to a video analytics server cluster on the surface via an internal fiber optic ring network. The server runs a deep learning-based visual analysis algorithm. For example, a target detection and segmentation model based on the YOLOv8 architecture can be used. This model, trained on a large amount of underground video data, can accurately identify targets such as coal mining machines, scraper conveyors, hydraulic supports, and personnel.

[0035] For coal mining machines, the algorithm continuously tracks the pixel area of ​​its cutting drum and determines whether it is in a cutting state by analyzing the continuous movement trajectory and shape changes of this area. For scraper conveyors, the algorithm determines the material flow status, i.e., whether there is coal flow or no load, by detecting whether there are large clusters of moving pixels on its surface.

[0036] In the event of environmental anomalies, the algorithm detects smoke by analyzing whether there are abnormal, diffuse pixel areas in the image, or detects water accumulation by recognizing dynamic changes in reflective areas on the ground.

[0037] All these identification results, such as A coal mining machine is cutting coal and B conveyor belt coal flow is normal, will be converted into a structured event record, which includes event type, target equipment, timestamp, and a confidence score indicating the accuracy of identification, such as 0.95.

[0038] The specific process of transforming the recognition results is as follows: the detection results of each frame or keyframe output by the video analysis algorithm are structurally encapsulated according to a preset event template. Each structured event information includes at least the following fields: a unique equipment identifier, used to associate the corresponding node in the spatial topology; event start and end timestamps, accurate to the millisecond level; action type, including coal mining machine cutting / stopping, conveyor with coal flow / no load, hydraulic support lifting, etc.; confidence score, directly using the Softmax probability value output by the deep learning model for the current recognition result; and environmental feature parameters, such as the proportion of coal flow coverage area, dust concentration level, light intensity, and other auxiliary information. The final structured event information serves as equipment operation characterization information.

[0039] The implementation method for pre-defined geological and production planning context information is as follows: Pre-defined shift production targets, equipment maintenance periods, mining progress plans, coal seam thickness, and fault locations are retrieved from the production planning database and geological information, categorized by time interval and spatial working face range. A central data warehouse is established and connected to the mine's production management and geological information via a data interface. Data synchronization tasks are automatically executed before the start of each shift, or according to a pre-defined time cycle, such as every hour. This task sends a Structured Query Language (SCL) command to the production planning database. For example, the command `SELECT target_tonnage, maintenance_schedule FROM production_plan WHERE shift_date = CURRENT_DATE AND workface_id = 'A'`. Here, `SELECT` specifies that data should be retrieved from the database; `target_tonnage` represents the target production output data field to be retrieved; `maintenance_schedule` represents the equipment maintenance plan data field to be retrieved; `FROM production_plan` specifies that the data comes from a database table named `production_plan`; `WHERE` sets the query conditions; `shift_date = CURRENT_DATE` is one of the filtering conditions, indicating that records with the shift date being the current date should be selected; and `workface_id = 'A'` is another filtering condition, indicating that records with the workface identifier 'A' should be selected. The overall meaning of this query language command is: Retrieve the target production output and equipment maintenance plan for workface A as of the current date from the production planning database.

[0040] A query is sent to the geological information system, and based on the spatial coordinates of the current mining face, data such as the coal seam thickness distribution map and the contour map of the known fault locations are retrieved. This relatively static or slowly changing information is loaded into memory as a background knowledge base, providing important prior constraints and references for subsequent logical reasoning and data filling.

[0041] The implementation method for real-time sensor data streams is as follows: Intrinsically safe power sensors, current transformers, vibration sensors, and temperature sensors are installed on key equipment such as coal mining machines, scraper conveyors, main ventilation fans, and belt conveyors. These sensors continuously collect data at a sampling frequency of not less than 1Hz and transmit it to the surface data acquisition and monitoring system through an underground ring network. Each data record is formatted as {equipment ID, timestamp, data type (e.g., power / kW), value}. This real-time sensor data stream, together with the aforementioned manually described information, equipment operation characterization information, and geological and production plan context information, constitutes multi-source heterogeneous operational reference information, providing a basic data source for subsequent state consistency verification and logic filling.

[0042] In step S2, the coal mine production process is a dynamic evolution; the advancement of the mining face and the extension of the roadways will lead to changes in the location and interrelationships of equipment. Therefore, this step aims to capture this dynamic change in physical space in real time, ensuring that subsequent analysis and modeling are based on the latest and most accurate physical layout. Furthermore, it is understandable that if the final judgment in step S2 indicates no change, the current equipment space topology and operational physical causal logic can remain unchanged, and step S4 and subsequent steps can be executed directly based on the current equipment space topology and operational physical causal logic.

[0043] Furthermore, step S2 includes: S21: Read the pixel coordinate trajectory of the device in continuous video frames from the device operation characterization information, and use the displacement vector between the start and end points of the pixel coordinate trajectory as the displacement feature of the device. S22: Perform natural language parsing on manually described information, identify keywords such as equipment relocation, mining face advancement, or roadway connection, and extract the equipment name, change time, and location description associated with the keywords to generate text change information; S23: Compare the displacement features with the text change information in a spatiotemporal correlation. If the equipment identification of the two is consistent, the change time window overlaps, and the actual spatial location indicated by the displacement features matches the location description in the text change information, then it is determined that the production space layout has changed.

[0044] The video analysis module, after identifying a specific device, such as a hydraulic support, continuously locks onto the target in consecutive video frames. In the first frame, it records the centroid pixel coordinates of the hydraulic support image, for example, (x1, y1). After a period of time, such as 30 seconds later, in another frame, its centroid pixel coordinates are recorded again as (x2, y2). By calculating the pixel displacement vector (x2-x1, y2-y1) between these two coordinate points and performing coordinate system transformation based on the camera's intrinsic and extrinsic parameters, the spatial displacement vector of the hydraulic support in the real-world coordinate system can be obtained. This displacement vector objectively reflects the physical movement of the device.

[0045] Meanwhile, when a voice-to-text report is received, such as "Working face E has completed this round of coal cutting and is now beginning to move the support forward, expected to be completed in half an hour," the natural language processing module first performs keyword matching to identify the action of moving the support forward, which indicates a change in spatial layout. Subsequently, using named entity recognition technology, it extracts the main equipment support involved in the action, as well as the associated location Working face E and the time information of half an hour. This information is then integrated into a structured text change information record.

[0046] Finally, the displacement features are compared spatiotemporally with the text change information, a process of cross-validation using multi-source information. Specifically, the displacement features of the hydraulic supports analyzed from the video are matched with the change information extracted from the text. First, it is confirmed whether the equipment identification is consistent and whether the hydraulic supports in the video belong to the support group of working face A. Second, it is confirmed whether the time windows overlap; whether the time when the support movement was detected in the video falls within the time period described in the manual report. Finally, it is confirmed whether the spatial location is consistent; whether the direction and distance of support movement obtained from the video analysis are consistent with the normal procedures of the mining face advancement. Only when all three conditions are met simultaneously is it finally confirmed that the production space layout has changed. This comprehensive judgment method effectively avoids incorrect judgments caused by errors from a single information source, such as misidentification of other moving objects in the video or human error.

[0047] As a specific implementation method, during the advancement of the fully mechanized mining face, video acquisition nodes continuously capture the motion images of the hydraulic supports. The processing flow first uses feature point matching technology to locate the structural markers on the side of the supports in a continuous video sequence. By calculating the pixel displacement vectors of feature points between adjacent frames and combining them with the scale of the fixed reference point of the working face, the actual physical distance the supports move towards the coal face is calculated. Simultaneously, the voice processing unit receives a report from the mining team leader, which includes a description of completing the current round of coal cutting and starting the support relocation. Natural language processing logic identifies keywords such as support relocation and working face advancement, and associates them with the current time window. The judgment logic compares the displacement direction and amount of the ten consecutive supports obtained from video analysis with the construction plan in the report. If the total displacement shown in the video reaches a cycle step, and the report text confirms the operation, then a change in the production space layout is confirmed. This method uses the physical facts of visual observation to verify the logical intent of the manual report, ensuring the reliability of spatial coordinate updates.

[0048] After confirming changes in the production space layout, it is necessary to update the equipment space topology and operational physical causal logic synchronously to ensure consistency between the digital model and physical reality.

[0049] Further, step S3 includes: S31: Superimpose the actual spatial displacement vector in the equipment displacement feature onto the original node coordinates of the equipment in the equipment spatial topology to generate updated node coordinates; S32: Based on the updated node coordinates, recalculate the spatial distance and relative azimuth angle between the device and its adjacent devices. If the spatial distance exceeds the preset connection threshold, disconnect the original connection edge. If the new node coordinates enter the connection range of another device, add a new connection edge. If the distance changes but is still within the threshold, update the connection edge attributes. The connection edge attributes include at least the material conveying distance, transmission delay time, material flow limit, start-stop timing dependency, and energy consumption coefficient. S33: For each connection edge in the updated device space topology, a directed causal edge is generated in the running physical causal logic. The direction of the directed causal edge is consistent with the material flow direction of the connection edge, and the constraint parameters of the directed causal edge inherit the connection edge attributes. S34: Write all directed causal edges into the running physical causal logic, replace all existing causal edges related to the device, and form a material flow constraint relationship that perfectly matches the current spatial topology.

[0050] In the digital twin map of the mine, each key piece of equipment is represented by a node with three-dimensional coordinates. When the actual spatial displacement vector of a piece of equipment, such as a coal mining machine, is obtained from video analysis, this vector is added to the current three-dimensional coordinates of the coal mining machine node to calculate its new position on the map, i.e., the updated node coordinates.

[0051] Based on the new node locations, the connection relationships between devices are dynamically adjusted. If the spatial distance between the updated coal mining machine node and the downstream scraper conveyor node exceeds a preset physical connection threshold, such as 5 meters, the direct material transfer relationship between them is considered broken, and the edge connecting these two nodes is deleted in the topology graph. Conversely, if the coal mining machine moves to a new location and enters the effective connection range of another transfer device, such as transfer device B, a new connection edge is added between the coal mining machine node and transfer device B node. If the distance between devices changes but remains within the connection threshold, for example, if the length of the scraper conveyor increases due to the advancement of the working face, the existence of the connection edge is not changed, but its attributes are updated.

[0052] The connection edge attributes include at least the material conveying distance, which directly affects the calculation of conveying energy consumption; the transmission delay time, which is the time required for material to be conveyed from one side to the other; the upper limit of material flow rate, which is determined by the physical structure size of the equipment and the maximum design speed, and may be affected by changes in the connection angle; the start-stop timing dependency, which clarifies the logical order in which downstream equipment must start before upstream equipment in order to prevent coal piling; and the energy consumption coefficient related to the new distance, which is the electrical energy consumed by conveying a unit mass of material per unit distance.

[0053] The physical causal logic is a rule base describing the interactions between devices. For example, a connection edge from the coal mining machine to the scraper conveyor represents the physical flow path of coal. In the causal logic, this would correspond to generating a directed causal edge from the state of the coal mining machine to the state of the scraper conveyor, indicating that the operating state of the coal mining machine is the cause of whether the scraper conveyor is under load. The constraint parameters of this causal edge are completely inherited from the properties of the physical connection edge.

[0054] The mapping mechanism transforms physical connections into logical constraints. Each connection edge is abstracted as a directed dependency path at the logical layer. The starting point of a directed causal edge represents the cause device, and the ending point represents the result device. If there is a physical material flow from the scraper conveyor to the transfer machine, a logical causal edge is generated. The inheritance of constraint parameters means that the upper limit of material flow rate is transformed into a capacity limit in causal logic, and the transmission delay is transformed into the lag time of state change.

[0055] This mapping ensures that the logical inference process follows the laws of energy and mass conservation, so that even when sensor data is missing, the operating load of intermediate equipment can still be calculated based on upstream output and downstream reception.

[0056] Finally, all newly generated directed causal edges are written into the running physical causal logic, replacing all existing causal edges related to the mobile device, forming material flow constraints that perfectly match the current spatial topology. This is an atomic update operation, ensuring that after the update, the entire causal logic library accurately reflects the latest physical layout and material flow paths of the mine, providing a solid and correct foundation for subsequent state consistency checks and logical reasoning.

[0057] Next, the updated equipment space topology and operational causal logic are used to verify the consistency of the multi-source heterogeneous operational reference information. Due to the diverse sources of information, conflicts or contradictions may exist between different pieces of information. This step uses a unified logical framework that matches the current physical reality to cross-validate and filter all information, eliminating errors and integrating consensus, thereby improving the accuracy of judging the true state of the mine.

[0058] Furthermore, step S4 includes: S41: Map each piece of data in the multi-source heterogeneous operation reference information to the corresponding node in the updated device space topology according to its associated device identifier; S42: For each node, call the preset consistency inference rule base to perform logical constraint matching on the multi-source heterogeneous operation reference information mapped to that node. If the multi-source heterogeneous operation reference information points to the same operation state, then the operation state is taken as the verified state of the device and assigned a first confidence score. If there is a conflict, then select the information source with the highest credibility as the main basis according to the preset arbitration strategy and assign a second confidence score, while recording the conflict type. The second confidence score is less than the first confidence score. S43: Based on the direction of the directed causal edge in the physical causal logic, perform causal consistency verification on equipment with material flow relationship: if the upstream equipment is stopped after verification and the downstream equipment is running after verification, it is determined to be a violation of causal logic, and the status of the downstream equipment is forcibly corrected to the pending confirmation status, and the corresponding first confidence score or second confidence score is reduced. S44: Output the verified state of each node and its corresponding first confidence score or second confidence score as the result of the state consistency verification.

[0059] In one specific implementation, for example, an event from video analytics indicating that coal mining machine A is cutting coal would be associated with the node representing coal mining machine A in the topology graph; a manual report of conveyor belt blockage B would be associated with the B conveyor belt node. This step categorizes all the scattered information, ensuring that each device node has all the relevant information.

[0060] A predefined consistency inference rule base defines the logical relationships between different information sources. For example, a rule could be: if the video analysis result is "running," the device PLC reports a "running" status, and the associated power sensor reading is greater than the no-load threshold, then these information points to the same running state, i.e., normal load operation. In this case, the verified state of the node is determined to be normal load operation, and a high first confidence score, such as 0.95, is assigned. If an information conflict occurs, for example, if the video analysis result is "stopped," but the PLC reports a "running" status, then an arbitration strategy is triggered.

[0061] The arbitration strategy can be pre-set as follows: for the basic judgment of whether something is running, the PLC has higher priority. Therefore, the PLC information is selected as the primary basis, and the verified status is judged as running, but it may be idle or faulty. Due to the conflict, a lower second confidence score is assigned, such as 0.6, and the conflict type is recorded as "video and PLC status mismatch".

[0062] The arbitration strategy is used to handle conflicting scenarios where multiple sources of information point to different outcomes. The execution process pre-determines the reliability levels of the information sources, typically following the principle that industrial control signals are superior to video analysis signals, and video analysis signals are superior to verbal reports. When the video display device is running but the control signal indicates shutdown, the control signal is selected as the verified state based on priority. The calculation of the second confidence score considers the quantity and degree of deviation of conflicting information. If multiple low-priority information sources point to a state opposite to that of a high-priority source, the initial confidence score is deducted proportionally while selecting the high-priority state; the resulting value is the second confidence score. This mechanism quantifies the degree of doubt in the current judgment, providing a risk reference for subsequent missing value filling.

[0063] Causal consistency verification is performed on equipment with material flow relationships. This is a system-level verification based on the verification of individual equipment status. For example, in causal logic, there exists a directed edge from coal mining machine A to scraper conveyor B. If, in the previous step, the verified status of coal mining machine A was "stopped," while the verified status of scraper conveyor B was "loaded," this clearly violates physical causal logic because if there is no material coming in from upstream, there cannot be a load downstream. In this case, it is determined to be a violation of causal logic, and the status of the downstream equipment, scraper conveyor B, is forcibly corrected to an "unconfirmed" state, with its corresponding confidence score significantly reduced.

[0064] After the two layers of verification described above, each device node in the topology graph will have a current most reliable operating state and a confidence score that quantifies that reliability. This result set provides high-quality input for subsequent missing data imputation.

[0065] After the state consistency verification is completed, the missing items in the real-time sensor data stream can be logically filled in, and the final virtual situation model can be constructed.

[0066] It should be noted that there is a one-way dependency between the state consistency verification result output in step S4 and the logical filling in step S5: the consistency verification result provides a reliable upstream and downstream observation basis and confidence constraint for filling, but the data generated by logical filling will not correct the completed verification result in reverse (that is, the post-verification state of each node and its confidence score remain unchanged after entering step S5), thereby ensuring the determinism and convergence of the processing flow.

[0067] Furthermore, in step S5, missing items in the real-time sensor data stream are identified, and combined with the results of the state consistency check and the geological and production planning context information, the updated operational physical causal logic is run to logically fill in the missing items, including the following steps: S51: Based on the spatial topology of equipment, identify the upstream and downstream related equipment that have a material flow relationship with the faulty equipment that generated the missing item; S52: Based on the results of the physical causal logic and state consistency verification, obtain the real-time start-stop status of upstream and downstream related equipment, and combine the geological and production planning context information to deduce the theoretical operating conditions of the faulty equipment during the missing period. S53: Obtain regional aggregated power load information, extract the total power change characteristics from the aggregated power load information, and verify whether the total power change characteristics match the rated power of the equipment under theoretical operating conditions; S54: If a match is found, the missing items will be logically filled in according to the preset data template corresponding to the theoretical operating conditions.

[0068] In one specific implementation, during the logic filling stage, firstly, assuming that the power sensor data of coal mining machine A is interrupted, by querying the topology map, its direct downstream equipment can be quickly located as scraper conveyor B.

[0069] Then, from the state consistency verification results, the verified state of scraper conveyor B was found to be high-load operation, with a first confidence score of 0.95. Simultaneously, from the geological and production planning context, it was learned that mining machine A is currently located in a known hard rock area, and the current shift's production task has not yet been completed. Based on this information, the following deductions were made: high-load operation of downstream equipment implies a continuous material supply from upstream, therefore mining machine A must be in operation; combined with its geological information of being located in a hard rock area, it can be further inferred that its operating load must be very high. Therefore, the theoretical operating condition of mining machine A during the data-missing period was deduced as high-power cutting.

[0070] Next, the total power data of the power supply circuit of the working face where the A coal mining machine is located was retrieved. The analysis revealed that during the period when the data was missing, the total power curve remained at a high level, and its value was approximately equal to the operating power of the B scraper conveyor plus the rated power of the A coal mining machine under high-power cutting conditions.

[0071] The variation characteristics of the total power closely match the theoretically derived operating conditions. The verification process first extracts the total power curve from the main power supply circuit of the substation in the mining area where the faulty equipment is located. The aggregated power load information originates from the main power supply circuit of the substation in the mining area where the faulty equipment is located.

[0072] The calculation logic identifies step-like jumps or continuous fluctuations in the curve. The magnitude of these jumps is related to the startup or shutdown power of a specific device.

[0073] The expected power contribution value is obtained by multiplying the rated power of the equipment under theoretical operating conditions by the derived load rate.

[0074] The expected power contribution is superimposed onto the load baseline of other known operating equipment in the area to obtain the predicted total load.

[0075] If the deviation between the predicted total load and the actual measured aggregate power is within a preset error range, such as ±5%, the match is considered successful. This method eliminates isolated misjudgments that may occur in logical deduction through closed-loop verification of the total energy.

[0076] Once the match is confirmed, a data generation template preset for high-power cutoff conditions is used, for example, based on 90% of the rated power, with an additional random fluctuation that conforms to the characteristics of the equipment, to generate power data points for the missing time period, which are then filled into the database and marked as logical fill data.

[0077] In addition, if a mismatch occurs, logical filling is abandoned, the missing item is marked as unreliable missing and an alarm is triggered. At the same time, historical data from the same period or a conservative estimate of the device's rated power is used as temporary filling, and an extremely low confidence score of less than 0.3 is assigned.

[0078] Furthermore, in step S5, constructing a virtual situational model reflecting the current operating status of the mine includes the following steps: S55: Align the logical filling results of missing items with the non-missing items in the real-time sensing data stream using timestamps to form a complete sequence of the faulty device's operating status on a continuous time axis. S56: Based on the equipment spatial topology, map the complete operating status sequence to the corresponding node of the faulty equipment in the topology diagram, and synchronously update the material flow and load status of upstream and downstream related equipment that have a material flow relationship with the faulty equipment. S57: Extract and vectorize the features of manually described information, equipment operation characterization information, aggregated power load information, and preset geological and production plan context information according to a unified spatiotemporal benchmark, and bind them to the corresponding nodes in the equipment spatial topology. S58: Iteratively aggregate the status information of all bound nodes to generate a comprehensive digital image reflecting the current operating status of the target mine, which serves as a virtual situational model.

[0079] In the stage of constructing the virtual situation model, firstly, the power data of the A coal mining machine after being filled in is spliced ​​together with the real data collected after the sensor returned to normal in chronological order to form a complete power state time series without interruption.

[0080] Then, the complete power sequence of coal mining machine A is assigned to its node in the topology graph. Based on this power sequence, its coal production can be estimated, and this coal production will be used as input to update the material flow rate and load status of the downstream scraper conveyor node B.

[0081] Subsequently, all other information from various sources, such as text reports from humans, events from video analysis, and regional total power data, are processed to extract key features and convert them into numerical vector forms, which are then bound to the corresponding devices or regional nodes in the topology graph.

[0082] Finally, all information bound to all nodes in the topology graph, including sensor data, fill data, and feature vectors of various heterogeneous information, is comprehensively calculated and visualized. Ultimately, a dynamic, multi-dimensional, comprehensive digital image reflecting the current operating status of the entire mine is presented on the large screen in the dispatch center. This image is the final constructed virtual situational awareness model.

[0083] As a specific implementation method, when the current sensor of the main conveyor belt experiences intermittent data loss due to fiber optic cable compression, the logic filling program is immediately initiated. First, the position of the conveyor belt in the topology is located, identifying the three upstream transfer points and the downstream main coal bunker entrance. The operating status of the upstream transfer equipment after consistency verification is checked, revealing that the transfer machine is operating at full load and video confirms continuous coal flow. Coal seam hardness parameters from the geological context are extracted, and the theoretical material flow rate at the current coal production rate is calculated. Based on the operational physical causal logic, the minimum power range of the scraper conveyor required to maintain this material flow rate is calculated. The aggregated power of the main power supply circuit for the mining area is read, and after deducting the permanent loads such as ventilation and drainage, the remaining power fluctuation characteristics highly match the calculated conveyor operating conditions. The execution logic then uses this theoretical power value to fill the blank segments in the sensor data stream. The completed data sequence, along with wind speed data, is converged into a spatiotemporal alignment matrix. Through multi-dimensional vectorized aggregation, a comprehensive digital image covering the entire mine production chain is constructed, achieving a complete restoration of the current mine operating status.

[0084] In a preferred embodiment, the deduction process of the theoretical operating conditions of the faulty equipment during the missing period can be further refined. Further, step S52 includes: S521: From the result of the state consistency verification, read the verified state of each device in the upstream device set and downstream device set that are connected to the faulty device by a directed causal edge, as well as its corresponding first confidence score or second confidence score. The verified state with the first confidence score or second confidence score higher than the preset adoption threshold is taken as the real-time start / stop state of the device. S522: Extract constraints from geological and production planning context information; S523: Using real-time start-stop status as the observed variable and constraints as the reasoning boundary, and utilizing the causal propagation rules defined in the physical causal logic of operation, forward causal propagation and reverse demand reasoning are performed along the directed causal edge to calculate the set of possible operating conditions of the faulty equipment under all constraints. S524: Select the operating condition with the highest confidence level from the set of possible operating conditions as the theoretical operating condition.

[0085] The preset adoption threshold can be set according to the actual situation, for example, it can be set to 0.7. The verified state when the first confidence score or the second confidence score is higher than the preset adoption threshold is used as the real-time start and stop state of the device, which ensures that the initial observation variables used for inference are highly reliable.

[0086] Secondly, the constraints include: based on the fault location in the geological map and the current mining progress plan, if the working face is in the fault crossing stage, a constraint for reducing equipment speed will be generated; based on the comparison between the shift production target and the current completed production, if the target has not been achieved, a production constraint that prohibits unnecessary shutdowns will be generated; based on the equipment maintenance plan, if the current time is within the scheduled maintenance period of a certain piece of equipment, a constraint for mandatory shutdown of the corresponding equipment will be generated.

[0087] Then, starting from the operating status of downstream equipment, the engine infers in reverse the state that the faulty upstream equipment must be in to ensure material supply. Simultaneously, all derived possible states must not violate any safety, geological, or production plan constraints. In practice, the inference engine first reads the real-time start / stop status of all upstream and downstream related equipment from the state consistency verification results, treating these states as known observation variables. Then, the engine loads a pre-built operational physics causal logic, stored in the form of a directed graph. Nodes represent equipment, and directed edges represent causal relationships of materials or energy. Each edge is accompanied by a causal strength parameter (e.g., "the probability that the downstream equipment must operate when the upstream equipment is running is 0.95").

[0088] The reasoning process is divided into two directions: forward causal propagation starts from the upstream node and transmits the operating requirements downstream along the direction of the directed edges: if the upstream equipment is running, then the causal logic requires that the downstream equipment must be in a running state; backward demand reasoning starts from the downstream node and transmits the production requirements upstream along the opposite direction of the directed edges: if the downstream equipment needs to receive materials, then the upstream equipment must supply materials.

[0089] Using a Bayesian network, faulty devices and their upstream and downstream related devices are treated as nodes in the network. Directed causal edges defined in the physical causal logic represent the dependencies between nodes. Each directed edge is accompanied by a conditional probability table; for example, if an upstream device is running, the probability that a downstream device is running is 0.95, and the probability that it is stopped is 0.05. The real-time start / stop status of upstream and downstream devices obtained from state consistency verification is used as observational evidence input into the network. Evidence with a confidence level below 1 can be considered soft evidence (i.e., a node is in a certain state with a certain probability).

[0090] Based on these observational evidences and conditional probability tables, the Bayesian network inference engine calculates the posterior probability of the faulty device node being in each candidate operating condition (such as running, shutting down, or low speed).

[0091] The posterior probability is the joint confidence level of the working condition, which integrates information from all observed variables in the upstream and downstream processes as well as the probability constraints in the causal logic.

[0092] Finally, the candidate operating condition with the highest posterior probability is selected as the theoretical operating condition output.

[0093] Once the virtual situational awareness model is constructed, the dispatcher can use it for decision support. The specific implementation of this process is as follows: Further, step S6 includes: S61: Parse manual dispatch instructions and extract the target device identifier, action type, and action execution time window from the instructions; S62: The action type and time window are used as external input variables and injected into the state variables of the corresponding target device node in the virtual situation model, overriding the original running state of the node within the time window. S63: Based on the injected virtual situation model, the operating condition evolution of all devices within the execution time window is simulated sequentially according to the preset simulation step size. The operating condition evolution follows the material flow constraints in the physical causal logic of operation and the connection edge attributes in the spatial topology of the devices. S64: During the simulation, according to the preset energy consumption calculation rules, the theoretical energy consumption value of each device node in each simulation step is accumulated. S65: Calculate the theoretical carbon emissions for each device node by multiplying the theoretical energy consumption value by the preset carbon emission factor, and sum them up to obtain the total carbon emissions. S66: Compare the total energy consumption and total carbon emissions before and after executing the manual dispatch command, calculate the difference, and output the estimated parameters of energy consumption and carbon emission impact corresponding to the manual dispatch command.

[0094] In one specific implementation, when the dispatcher enters a command on the interactive interface, such as increasing the speed of the main ventilation fan C by 10% for 1 hour, it will be parsed into structured data, including the target equipment identifier A main ventilation fan, the action type adjustment speed, the action parameter increase of 10%, and the time window of the next 1 hour.

[0095] Then, the parsed action type and time window are used as external input variables and injected into the state variables of the corresponding target device node in the virtual situational awareness model, overwriting the original operating state of the node within the time window. After overwriting the target device state, the upstream device is traced backward and the forward propagation is carried out along the directed causal edge in the physical causal logic of the operation. If a logical contradiction occurs where the upstream device is down but the downstream device is running, the downstream device state is automatically corrected to an unconfirmed state and the confidence level is reduced, while a consistency alarm is issued to the scheduler.

[0096] If no logical contradiction occurs, find the node representing the main ventilation fan C in the digital copy of the virtual situation model, and modify its rotational speed state variable for the next hour to the current rotational speed multiplied by 1.1.

[0097] Next, based on the virtual situational model injected with new instructions, the simulation engine sequentially simulates the evolution of the operating conditions of all equipment within a preset simulation step size, such as 1 minute. The simulation process strictly adheres to the material flow constraints in the physical causal logic and the connection edge attributes in the spatial topology of the equipment. The state of the entire mine evolves step by step with the time step. In particular, for the start-stop timing dependencies recorded in the connection edge attributes, the simulation engine enforces the following rules: Enforcement rules during startup: The start / stop sequence dependency attribute of each connection edge defines the order constraints for starting and stopping two devices. For example, for a connection edge from device A to device B, if its start / stop sequence dependency attribute is "A starts first, B starts later; B stops first, A stops later", it means that when materials flow from A to B, to prevent coal accumulation or no-load operation, it must be ensured that A starts before B can start, and B stops before A can stop.

[0098] Stop Phase Enforcement Rules: When a device node receives a stop command, the simulation engine queries all downstream connection edges (i.e., material outflow directions) originating from that device. For each downstream connection edge, if its start / stop timing dependency requires this device to stop before the downstream device, the engine checks the current simulation status of the downstream device. If the downstream device has not yet stopped (i.e., is still running), the stop command for this device is suspended and not executed; the engine adds the stop request to the waiting queue and records the identifier of the dependent downstream device. Once the dependent downstream device has stopped, the engine automatically executes the stop for this device.

[0099] Timeout and Conflict Handling: If a start or stop command exceeds a preset timeout threshold (e.g., 30 seconds) in the waiting queue, and the state of the dependent device still does not meet the requirements, the simulation engine determines it as a timing dependency violation and handles it according to the preset strategy: ① If it is a start command, the device is forcibly started and a warning message is recorded (indicating a potential risk of temporary coal accumulation in actual production); ② If it is a stop command, the device is forcibly stopped and a warning is recorded (indicating potential energy waste due to idleness). The conflict handling strategy can be adjusted through the system configuration file.

[0100] Multi-level dependency transitivity: If the startup of device C depends on device B, and the startup of device B depends on device A, the simulation engine automatically forms a dependency chain. The engine resolves the dependencies in reverse order (starting from the top), ensuring that device A starts and stabilizes before device B can start, and finally device C can start. The stopping process is executed in forward order (starting from the bottom).

[0101] Through the above rules, the simulation engine can automatically maintain the correct start-up and shutdown sequence between devices during the time stepping process, so that the operating condition evolution of the virtual situation model is consistent with the actual operating logic of the physical site, thereby ensuring the reliability of the energy consumption and carbon emission simulation results.

[0102] During the simulation, the theoretical energy consumption value of each device node within each simulation step is accumulated according to preset energy consumption calculation rules. For each device, the energy consumption calculation rule is based on its current operating status, such as running, stopped, speed, load rate, etc., and the running time, to call a preset power-load characteristic curve function in the device node's attributes to calculate the power consumption within that simulation step. For example, for the C main ventilation fan, based on its new speed, the corresponding power is found from its characteristic curve, and then multiplied by the simulation step time to obtain the energy consumption for that step.

[0103] Subsequently, the theoretical energy consumption value is multiplied by a preset carbon emission factor to calculate the theoretical carbon emissions for each equipment node, and these are summed to obtain the total carbon emissions. This carbon emission factor is determined based on the source of electricity used by the mine, such as the average carbon emission intensity of the local power grid.

[0104] Finally, the total energy consumption and total carbon emissions before and after executing the manual dispatch command are compared, the difference is calculated, and the estimated parameters of the energy consumption and carbon emission impact corresponding to the manual dispatch command are output. Before executing the above simulation, a baseline simulation is performed first, that is, without injecting any new commands, the model is allowed to evolve naturally in its current state for 1 hour to obtain a baseline total energy consumption and total carbon emissions. Then, the simulation results after injecting the command are compared with the baseline results, and the difference is calculated. Finally, a clear prediction message is displayed on the dispatcher's screen: Executing this command is expected to reduce total energy consumption by X degrees and reduce total carbon emissions by Y tons in the next hour.

[0105] Please refer to Figure 2 This application also proposes an energy consumption optimization and carbon emission management system for coal mining. The system is used to implement the steps of any of the above methods, and the system includes: The acquisition module 201 acquires multi-source heterogeneous operational reference information for the target mine, including manual descriptions, equipment operation characterization information, real-time sensor data streams, and preset geological and production plan context information, providing a comprehensive data foundation for subsequent analysis and processing. This integration of multi-source heterogeneous information can more comprehensively reflect the actual operating status of the mine, compensating for the shortcomings of a single data source.

[0106] The judgment module 202 extracts equipment displacement features from the equipment operation characterization information and combines them with manual description information to determine whether the production space layout of the target mine has changed. By combining equipment displacement features and manual description information, the system can accurately identify changes in the mine's production space layout, which is crucial for dynamically adjusting the mine model and avoiding data inconsistencies and model failures caused by changes in spatial layout.

[0107] The remapping module 203, when the production space layout changes, redefines the node coordinates and connection edge attributes in the equipment space topology based on equipment displacement characteristics, and remaps the material flow constraints in the physical causal logic based on the updated equipment space topology. This module ensures that the mine's digital model can accurately reflect changes in the physical world in real time, providing a reliable foundation for subsequent simulation and decision-making. By redefining node coordinates and connection edge attributes and remapping material flow constraints, the system can adapt to dynamic changes in the mine's production environment, ensuring the model's effectiveness.

[0108] The verification module 204 uses the updated device space topology and operational physical causal logic to perform state consistency verification on multi-source heterogeneous operational reference information. By verifying the consistency of multi-source heterogeneous information, the system can identify and handle data conflicts and inconsistencies, improve data reliability and accuracy, and provide high-quality data input for subsequent virtual situation model construction.

[0109] Module 205 identifies missing items in the real-time sensor data stream and, combining the results of state consistency verification with geological and production planning context information, runs updated operational physical causal logic to logically fill in the missing items, constructing a virtual situational model reflecting the current operational status of the mine. This module is crucial for solving the data contamination problem. By logically filling in the missing items, the system can construct a complete and accurate virtual situational model, overcoming the data interruption problem caused by sensor failures and providing a continuous and reliable data foundation for subsequent scheduling simulations.

[0110] The parameter estimation module 206 receives manual dispatch instructions, incorporates these instructions as input variables into the virtual situation model, and performs simulation calculations based on preset energy consumption and carbon emission calculation rules. It then outputs the estimated energy consumption and carbon emission impact parameters corresponding to the manual dispatch instructions. By introducing manual dispatch instructions into the virtual situation model for simulation, the system can estimate the impact of dispatch instructions on energy consumption and carbon emissions before actual execution, thereby supporting decision-makers in selecting the optimal dispatch scheme and achieving the goals of energy consumption optimization and carbon emission management.

[0111] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for energy consumption optimization and carbon emission management in coal mining, characterized in that, The method includes the following steps: S1: Obtain multi-source heterogeneous operation reference information for the target mine. The multi-source heterogeneous operation reference information includes manual description information, equipment operation characterization information, real-time sensor data stream, and preset geological and production plan context information. S2: Extract equipment displacement features from the equipment operation characterization information, and combine them with the manual description information to determine whether the production space layout of the target mine has changed; Step S2 includes: S21: Read the pixel coordinate trajectory of the device in continuous video frames from the device operation characterization information, and use the displacement vector between the start and end points of the pixel coordinate trajectory as the displacement feature of the device. S22: Perform natural language parsing on the manually described information, identify the keywords such as equipment relocation, mining face advancement or roadway connection contained therein, and extract the equipment name, change time and location description associated with the keywords to generate text change information; S23: Perform a spatiotemporal correlation comparison between the displacement feature and the text change information. If the device identifiers of the two are consistent, the change time windows overlap, and the actual spatial position indicated by the displacement feature matches the position description in the text change information, then it is comprehensively determined that the production space layout has changed. S3: If a change occurs, the node coordinates and connection edge attributes in the equipment space topology are redefined according to the equipment displacement characteristics, and the material flow constraints in the physical causal logic are remapped based on the updated equipment space topology. S4: Using the updated device space topology and operational physical causal logic, perform state consistency verification on the multi-source heterogeneous operational reference information; S5: Identify missing items in the real-time sensor data stream, and combine the results of the state consistency verification with the geological and production plan context information, run the updated operational physical causal logic to logically fill in the missing items, and construct a virtual situation model reflecting the current operating status of the mine. S6: Receive manual dispatch instructions, introduce the manual dispatch instructions as input variables into the virtual situation model, and perform simulation calculations in combination with preset energy consumption and carbon emission calculation rules, and output the energy consumption and carbon emission impact prediction parameters corresponding to the manual dispatch instructions.

2. The method for energy consumption optimization and carbon emission management in coal mining according to claim 1, characterized in that, Step S1 includes: S11: Receive human voice reports in real time, convert the voice reports into text information, extract key status words and reporting timestamps, and simultaneously receive device status, environmental parameter data and their geographical location tags entered by the mobile terminal. Store the above information in a structured manner as the human description information. S12: Acquire real-time video stream of key work area, identify pixel displacement features and shape change features of target equipment in video stream, determine equipment operation, material flow status and environmental anomalies, and convert the identification results into structured event information with confidence as the equipment operation characterization information; S13: From the production planning database and geological information system, read the preset shift output target, equipment maintenance period, mining progress plan, coal seam thickness, fault location, as the preset geological and production planning context information according to time interval and spatial working face range; S14: Real-time acquisition of the device's power, current, vibration, temperature, and rotation speed time-series physical quantities, recording the device identifier, timestamp, time-series physical quantity type, and value as the real-time sensing data stream.

3. The method for energy consumption optimization and carbon emission management in coal mining according to claim 1, characterized in that, Step S3 includes: S31: Superimpose the actual spatial displacement vector in the device displacement feature onto the original node coordinates of the device in the device spatial topology to generate updated node coordinates; S32: Based on the updated node coordinates, recalculate the spatial distance and relative azimuth angle between the device and its adjacent devices. If the spatial distance exceeds the preset connection threshold, disconnect the original connection edge. If the new node coordinates enter the connection range of another device, add a new connection edge. If the distance changes but is still within the threshold, update the connection edge attributes. The connection edge attributes include at least the material conveying distance, transmission delay time, material flow limit, start-stop timing dependency, and energy consumption coefficient. S33: For each connection edge in the updated device space topology, a directed causal edge is generated in the running physical causal logic. The direction of the directed causal edge is consistent with the material flow direction of the connection edge, and the constraint parameters of the directed causal edge inherit the connection edge attributes. S34: Write all directed causal edges into the running physical causal logic, replace all existing causal edges related to the device, and form a material flow constraint relationship that perfectly matches the current spatial topology.

4. The method for energy consumption optimization and carbon emission management in coal mining according to claim 1, characterized in that, Step S4 includes: S41: Map each piece of data in the multi-source heterogeneous operation reference information to the corresponding node in the updated device space topology according to its associated device identifier; S42: For each node, a preset consistency inference rule base is invoked to perform logical constraint matching on the multi-source heterogeneous operation reference information mapped to that node. If the multi-source heterogeneous operation reference information points to the same operation state, then that operation state is taken as the verified state of the device and assigned a first confidence score. If there is a conflict, then the information source with the highest credibility is selected as the main basis according to the preset arbitration strategy and assigned a second confidence score, while recording the conflict type. The second confidence score is less than the first confidence score. S43: Based on the direction of the directed causal edge in the physical causal logic, perform causal consistency verification on equipment with material flow relationship: if the upstream equipment is stopped after verification and the downstream equipment is running after verification, it is determined to be a violation of causal logic, and the status of the downstream equipment is forcibly corrected to the pending confirmation status, and the corresponding first confidence score or second confidence score is reduced. S44: Output the verified state of each node and its corresponding first confidence score or second confidence score as the result of the state consistency verification.

5. The method for energy consumption optimization and carbon emission management in coal mining according to claim 4, characterized in that, In step S5, identifying missing items in the real-time sensor data stream and combining the results of the state consistency check with the geological and production planning context information, the updated operational physical causal logic is run to logically fill in the missing items, including the following steps: S51: Based on the spatial topology of the equipment, identify the upstream and downstream related equipment that have a material flow relationship with the faulty equipment that generated the missing item; S52: Based on the physical causal logic of the operation and the result of the state consistency verification, obtain the real-time start-stop status of the upstream and downstream related equipment, and combine the geological and production planning context information to deduce the theoretical operating conditions of the faulty equipment during the missing period. S53: Obtain regional aggregated power load information, extract the total power change characteristics from the aggregated power load information, and verify whether the total power change characteristics match the rated power of the equipment under the theoretical operating conditions; S54: If a match is found, the missing item is logically filled in according to the preset data template corresponding to the theoretical operating condition.

6. The method for energy consumption optimization and carbon emission management in coal mining according to claim 5, characterized in that, In step S5, constructing a virtual situational model reflecting the current operating status of the mine includes the following steps: S55: The logical filling result of the missing item is timestamped with the non-missing item in the real-time sensing data stream to form a complete operating state sequence of the faulty device on a continuous time axis. S56: Based on the equipment spatial topology, map the complete operating state sequence to the corresponding node of the faulty equipment in the topology graph, and synchronously update the material flow and load status of upstream and downstream related equipment that have a material flow relationship with the faulty equipment; S57: Extract features and vectorize the artificial description information, equipment operation characterization information, aggregated power load information, and preset geological and production plan context information according to a unified spatiotemporal benchmark, and bind them to the corresponding nodes in the equipment spatial topology relationship; S58: Iteratively aggregate the status information of all bound nodes to generate a comprehensive digital image reflecting the current operating status of the target mine, which serves as the virtual situation model.

7. The method for energy consumption optimization and carbon emission management in coal mining according to claim 5, characterized in that, Step S52 includes: S521: From the result of the state consistency verification, read the verified state of each device in the upstream device set and downstream device set that are connected to the faulty device by a directed causal edge, and its corresponding first confidence score or second confidence score. The verified state with the first confidence score or second confidence score higher than the preset adoption threshold is taken as the real-time start / stop state of the device. S522: Extract constraints from the geological and production planning context information; S523: Using the real-time start-stop state as the observation variable and the constraint conditions as the reasoning boundary, and utilizing the causal transmission rules defined in the operational physical causal logic, perform forward causal propagation and reverse demand reasoning along the directed causal edge to calculate the set of possible operating conditions of the faulty equipment under all constraint conditions. S524: Select the operating condition with the highest confidence level from the set of possible operating conditions as the theoretical operating condition.

8. The method for energy consumption optimization and carbon emission management in coal mining according to claim 1, characterized in that, Step S6 includes: S61: Parse the manual dispatch instruction and extract the target device identifier, action type, and action execution time window from the instruction; S62: The action type and time window are used as external input variables and injected into the state variables of the corresponding target device node in the virtual situation model, overriding the original running state of the node within the time window; S63: Based on the injected virtual situation model, the operating condition evolution of all devices within the time window is simulated sequentially according to the preset simulation step size. The operating condition evolution follows the material flow constraints in the physical causal logic of the operation and the connection edge attributes in the spatial topology of the devices. S64: During the simulation, according to the preset energy consumption calculation rules, the theoretical energy consumption value of each device node in each simulation step is accumulated. S65: Calculate the theoretical carbon emissions for each device node by multiplying the theoretical energy consumption value by a preset carbon emission factor, and sum them up to obtain the total carbon emissions. S66: Compare the total energy consumption and total carbon emissions before and after executing the manual scheduling instruction, calculate the difference, and output the energy consumption and carbon emission impact prediction parameters corresponding to the manual scheduling instruction.

9. An energy consumption optimization and carbon emission management system for coal mining, characterized in that, The system is used to implement the steps of the method according to any one of claims 1-8, and the system includes: Acquisition Module: Acquires multi-source heterogeneous operation reference information for the target mine. The multi-source heterogeneous operation reference information includes manual description information, equipment operation characterization information, real-time sensor data stream, and preset geological and production plan context information. Judgment module: Extracts equipment displacement features from the equipment operation characterization information and combines them with the manual description information to determine whether the production space layout of the target mine has changed; The judgment module is also used to read the pixel coordinate trajectory of the device in continuous video frames from the device operation characterization information, and use the displacement vector between the start point and the end point of the pixel coordinate trajectory as the displacement feature of the device. Natural language parsing is performed on the manually described information to identify keywords such as equipment relocation, mining face advancement, or roadway connection, and the equipment name, change time, and location description associated with the keywords are extracted to generate text change information; The displacement feature is compared with the text change information in a spatiotemporal correlation. If the device identifiers of the two are consistent, the change time windows overlap, and the actual spatial position indicated by the displacement feature matches the position description in the text change information, then it is determined that the production space layout has changed. Remapping module: If a change occurs, the node coordinates and connection edge attributes in the equipment space topology are redefined according to the equipment displacement characteristics, and the material flow constraints in the running physical causal logic are remapped based on the updated equipment space topology. Verification module: Using the updated device space topology and operational physical causal logic, the state consistency of the multi-source heterogeneous operational reference information is verified; Construction module: Identifies missing items in the real-time sensor data stream, and combines the results of the state consistency verification with the geological and production plan context information, runs the updated operational physical causal logic to logically fill in the missing items, and constructs a virtual situation model reflecting the current operating status of the mine; Parameter estimation module: Receives manual scheduling instructions, introduces the manual scheduling instructions as input variables into the virtual situation model, and performs simulation calculations in combination with preset energy consumption and carbon emission calculation rules, and outputs the energy consumption and carbon emission impact estimation parameters corresponding to the manual scheduling instructions.

Images