Coal mine water disaster risk early warning method and system based on big data

Abstract

The invention discloses a coal mine water disaster risk early warning method and system based on big data, and relates to the technical field of mining safety monitoring, and the method comprises the steps: carrying out the entity relation extraction and fusion processing of coal mine multi-source historical geological data, and constructing a coal mine hydrogeological map; performing semantic analysis and structured recombination on the coal mine multi-source real-time data, and outputting a coal mine monitoring data set; injecting the coal mine monitoring data set as a dynamic updating element into the coal mine hydrogeological map to form a coal mine hydrological space-time map; identifying a water disaster risk conduction path in the coal mine hydrological space-time map by adopting a path migration algorithm, and outputting a water inrush coefficient path set; and carrying out risk level mapping and early warning information synthesis on the water inrush coefficient path set, and integrating and outputting a coal mine water disaster risk early warning report. According to the method, dynamic map updating and risk path mining are combined, and the coal mine water disaster risk is spanned from static assessment to dynamic deduction.

Description

Technical Field

[0001] This invention relates to the field of mining safety monitoring technology, and in particular to a method and system for early warning of coal mine water hazards based on big data. Background Technology

[0002] In recent years, coal mine water hazard early warning methods have gradually evolved from single-point monitoring to intelligent systems that integrate multi-source information. In the field of mining safety monitoring, real-time monitoring networks are commonly constructed using various sensors such as groundwater pressure, water level, water temperature, and stress. Statistical correlations between monitoring data and water inrush signs are established based on time-series data analysis, machine learning classification, and regression models. Simultaneously, 3D geological modeling and geographic information systems are widely used for visualizing static geological structures and aquifer distribution. Combined with hydrogeological numerical simulation methods, static assessments of water inrush risks in specific scenarios are conducted, forming a mainstream technical system of "monitoring and perception - model prediction - static evaluation."

[0003] However, existing methods have limitations in modeling the dynamic evolution mechanism of risks and the interpretability of early warnings. The lack of a unified semantic association and spatiotemporal alignment framework among multi-source heterogeneous data makes it difficult to construct a computable knowledge system that integrates prior geological knowledge and dynamic monitoring information. Early warning models often rely on data-driven black-box mapping, failing to quantitatively extrapolate and explain the mechanism of water hazard risks within the dynamic evolution path of "geological structure-mining disturbance-seepage stress coupling." This results in delayed early warning results, unclear risk transmission paths, and difficulty in supporting precise prevention and control decisions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a coal mine water hazard risk early warning method based on big data to solve the limitations in modeling the dynamic evolution mechanism of risk and the interpretability of early warning.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Firstly, this invention provides a coal mine water hazard risk early warning method based on big data, comprising: acquiring multi-source historical geological data of the coal mine; extracting and fusing entity relationships from the multi-source historical geological data of the coal mine; and constructing a coal mine hydrogeological map; wherein the multi-source historical geological data of the coal mine includes geological exploration reports, hydrogeological reports, coal mine geological spatial data, and historical water hazard accident analysis reports; collecting multi-source real-time data of the coal mine; performing semantic parsing and structured reorganization on the multi-source real-time data of the coal mine; outputting a real-time data stream of the coal mine; and using a spatiotemporal indexing algorithm to link the real-time data stream of the coal mine with the coal mine hydrogeological map. Coordinate association and semantic binding are performed to output a coal mine monitoring data set. This coal mine monitoring data set is then injected into the coal mine hydrogeological map as a dynamically updated element, outputting an enhanced hydrological map. The enhanced hydrological map undergoes topological reconstruction and relational evolution to form a coal mine hydrological spatiotemporal map. A path-walking algorithm is used to identify water hazard risk transmission paths in the coal mine hydrological spatiotemporal map. Dynamic risk propagation simulation and quantitative assessment of water inrush coefficients are performed on these paths, outputting a set of water inrush coefficient paths. Finally, risk level mapping and early warning information synthesis are performed on the water inrush coefficient path set, and a coal mine water hazard risk early warning report is integrated and output.

[0007] As a preferred embodiment of the big data-based coal mine water hazard risk early warning method of the present invention, the specific steps for extracting and fusing entity relationships from multi-source historical geological data of coal mines to construct a coal mine hydrogeological map are as follows: Preliminary geological entities and preliminary geological relationships are extracted from multi-source historical geological data of coal mines using deep learning extraction methods, forming a set of preliminary entity-relationship pairs. Perform semantic alignment and conflict resolution on the initial entity relation set, and output the geological entity set and the geological relation set; Based on the geological relationship set, the geological entity set is subjected to graph structure construction and relationship embedding to form a coal mine hydrogeological map.

[0008] As a preferred embodiment of the coal mine water hazard risk early warning method based on big data described in this invention, the specific steps of performing semantic parsing and structured reorganization on multi-source real-time coal mine data to output a real-time coal mine data stream are as follows: Identify geological entities and monitoring parameters in multi-source real-time data of coal mines, and semantically associate and contextually bind geological entities and monitoring parameters to form a set of semantic data of coal mines. Perform time-series alignment and streaming encapsulation transformation on the semantic data set of coal mines to output real-time data streams of coal mines.

[0009] As a preferred embodiment of the big data-based coal mine water hazard risk early warning method of the present invention, the step of using a spatiotemporal indexing algorithm to perform coordinate association and semantic binding between the real-time coal mine data stream and the coal mine hydrogeological map, and outputting a coal mine monitoring data set, specifically includes the following steps: A spatiotemporal encoder is used to extract the timestamp and three-dimensional coordinates of each data point from the real-time data stream of the coal mine, and a spatiotemporal coal mine data stream is generated. The spatiotemporal coal mine data stream is mapped and associated in the coal mine hydrogeological map using a coordinate matching method to generate coordinate mapping relationships. The spatiotemporal coal mine data stream and the coal mine hydrogeological map are linked and verified according to the coordinate mapping relationship, and a coal mine monitoring data set is output.

[0010] As a preferred embodiment of the big data-based coal mine water hazard risk early warning method of the present invention, the specific steps of integrating coal mine monitoring datasets into dynamically updated elements and injecting them into the coal mine hydrogeological map to output an enhanced hydrological map are as follows: The coal mine monitoring data set is discretized into spatiotemporal data segments according to timestamps and spatial coordinates, and the dynamic attributes and related events in each spatiotemporal data segment are extracted and integrated to generate a spatiotemporal data segment set. A subgraph isomorphic matching mechanism is used to perform pattern matching and structure mapping between the spatiotemporal data fragment set and the local substructure in the coal mine hydrogeological map, generating a mapping relationship table. Based on the mapping table, the associated events in the spatiotemporal data fragment set are inserted as new nodes and the dynamic attributes are inserted as node attributes into the coal mine hydrogeological map to generate a real-time intermediate state map. Perform graph structure conflict detection and attribute logic conflict detection on the real-time intermediate state graph, resolve the conflict results, and generate a resolved state graph; By reorganizing the topology of the state map through dynamic community discovery, the internal connectivity of coal mine water hazard risk areas is strengthened, and an enhanced hydrological map is output.

[0011] As a preferred embodiment of the big data-based coal mine water hazard risk early warning method of the present invention, the specific steps of performing topological reconstruction and relation evolution of the enhanced hydrological map to form a coal mine hydrological spatiotemporal map are as follows: To enhance each node and relationship in the hydrological map, a timestamp attribute is added, and the graph structure is reorganized in chronological order according to the timestamp attribute to generate a reconstructed hydrological map; Dynamic propagation and aggregation operations of relationships between nodes are performed on the reconstructed hydrological map to form a spatiotemporal hydrological map of the coal mine.

[0012] As a preferred embodiment of the big data-based coal mine water hazard risk early warning method of the present invention, the specific steps of identifying the water hazard risk transmission path in the coal mine hydrological spatiotemporal map using a path walking algorithm are as follows: In the spatiotemporal map of coal mine hydrology, starting from the predefined water hazard risk source, a multi-source concurrent path walk is performed to output the initial risk transmission path set; The initial risk transmission path set is subjected to temporal reachability verification and redundancy elimination, and the water hazard risk transmission path is output.

[0013] As a preferred embodiment of the coal mine water hazard risk early warning method based on big data described in this invention, the specific steps of performing dynamic risk propagation simulation and quantitative evaluation of the water hazard risk transmission path, and outputting a set of water inrush coefficient paths, are as follows: The time-series water pressure and stress data of each water hazard risk transmission path are extracted from the real-time data stream of the coal mine. Based on the time-series water pressure and stress data, the dynamic propagation simulation of the water pressure and stress transmission relationship of each water hazard risk transmission path is carried out, and the simulated water hazard risk path set is output. Water hazard risk features are extracted from the set of simulated water hazard risk paths, and the inrush coefficient is calculated according to the water hazard risk features to generate a set of inrush coefficient risk paths. The risk path set based on the water inrush coefficient is sorted in reverse order and the risk is screened to output the water inrush coefficient path set.

[0014] As a preferred embodiment of the big data-based coal mine water hazard risk early warning method of the present invention, the specific steps of mapping risk levels and synthesizing early warning information for the set of water inrush coefficient paths, and integrating and outputting a coal mine water hazard risk early warning report are as follows: Based on the preset risk threshold of the water inrush coefficient, the risk level of the water inrush coefficient path set is divided, and a risk-labeled path set is generated. Multi-dimensional risk attributes are synthesized from the risk-labeled path set to output early warning information item data; Integrate the data of early warning information items, extract risk summaries and format reports, and output a coal mine water hazard risk early warning report.

[0015] Secondly, this invention provides a coal mine water hazard risk early warning system based on big data, including: A mapping module is used to acquire multi-source historical geological data of coal mines, extract and fuse entity relationships from the multi-source historical geological data of coal mines, and construct hydrogeological maps of coal mines. The multi-source historical geological data of coal mines includes geological exploration reports, hydrogeological reports, coal mine geological spatial data, and historical water hazard accident analysis reports. The monitoring data module is used to collect real-time data from multiple sources in the coal mine, perform semantic parsing and structured reorganization on the real-time data from multiple sources in the coal mine, output the real-time data stream of the coal mine, and use a spatiotemporal indexing algorithm to coordinate and semantically bind the real-time data stream of the coal mine with the hydrogeological map of the coal mine, outputting a set of coal mine monitoring data. The spatiotemporal mapping module is used to integrate coal mine monitoring datasets into the coal mine hydrogeological map as dynamically updated elements, output an enhanced hydrological map, and perform topological reconstruction and relation evolution on the enhanced hydrological map to form a coal mine hydrological spatiotemporal map. The water inrush coefficient module is used to identify water hazard risk transmission paths in the spatiotemporal map of coal mine hydrology using a path walking algorithm, perform dynamic risk propagation simulation and water inrush coefficient quantitative evaluation of water hazard risk transmission paths, and output a set of water inrush coefficient paths. The early warning report module is used to map the risk level of the water inrush coefficient path set and synthesize early warning information, and integrate and output a coal mine water hazard risk early warning report.

[0016] The beneficial effects of this invention are as follows: By combining dynamic map updates with risk path mining, a leap from static assessment to dynamic extrapolation of coal mine water hazard risk has been achieved. By integrating coal mine monitoring datasets as dynamically updated elements into the coal mine hydrogeological map and performing topological reconstruction, deep fusion and dynamic representation of multi-source heterogeneous data within a unified semantic framework are realized. A path-walking algorithm is employed on the coal mine hydrological spatiotemporal map to identify water hazard risk transmission paths, enabling the automatic discovery and explicit expression of hidden and complex risk channels. This allows early warning systems to not only output risk levels but also clearly reveal the spatial location, transmission path, and evolution mechanism of the risk, thus providing decision support with precise positioning and causal explanation capabilities. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a coal mine water hazard risk early warning method based on big data.

[0019] Figure 2 This is a schematic diagram of a coal mine water hazard risk early warning system based on big data.

[0020] Figure 3 This is a flowchart for outputting hydrogeological maps of coal mines.

[0021] Figure 4 A flowchart for outputting enhanced hydrographic maps. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a coal mine water hazard risk early warning method based on big data, including the following steps: S1. Obtain multi-source historical geological data of coal mines, extract and fuse entity relationships from the multi-source historical geological data of coal mines, and construct hydrogeological maps of coal mines.

[0026] We acquire multi-source historical geological data of coal mines, and use deep learning extraction methods to extract preliminary geological entities and preliminary geological relationships from the multi-source historical geological data of coal mines, forming a set of preliminary entity-relationship pairs.

[0027] Specifically, this involves acquiring multi-source historical geological data for coal mines, including geological exploration reports, hydrogeological reports, coal mine geological spatial data, and historical water hazard accident analysis reports. Geological exploration reports are obtained through paper or electronic document archives formed during the early stages of mine construction; hydrogeological reports are obtained through technical documents formed from years of hydrological observation and special research in the mining area; coal mine geological spatial data are obtained through digital layer files stored by the mine's geological survey department; and historical water hazard accident analysis reports are obtained through accident investigation and summary materials filed with the coal mine safety supervision department.

[0028] Furthermore, text cleaning is performed on geological exploration reports, hydrogeological reports, and historical water hazard accident analysis reports. This includes removing headers and footers, standardizing professional terminology, segmenting the text, and labeling chapter types. Coordinate system transformation and layer element parsing are performed on the coal mine geological spatial data to extract vector elements such as fault lines, stratigraphic boundaries, borehole locations, and roadway centerlines. The cleaned full-text text of the geological exploration reports, hydrogeological reports, and historical water hazard accident analysis reports, along with the attribute description text corresponding to the parsed coal mine geological spatial data, are used as input to a deep learning algorithm. The deep learning extraction method scans the input text character by character to identify faults, aquifers, aquitards, goafs, collapse columns, coal seams, old goaf water bodies, boreholes, and roadways as preliminary geological entities. Simultaneously, it identifies faults cutting through aquifers, aquifers located above coal seams, aquitards isolating old goaf water, goafs accumulating old goaf water, collapse columns connecting upper and lower aquifers, boreholes exposing Ordovician limestone, and roadways adjacent to faults as preliminary geological relationships. Each identified preliminary geological entity is combined with its associated preliminary geological relationship to form a set of preliminary entity relationship pairs.

[0029] Semantic alignment and conflict resolution are performed on the initial entity relation set, and the geological entity set and geological relation set are output.

[0030] Specifically, the terminology of preliminary geological entities in the set is standardized for the preliminary entity relationships. Different descriptions of the same object are unified into a standardized full name. For example, the Ordovician limestone aquifer, O2 limestone aquifer, and Ordovician limestone aquifer are unified into the Ordovician limestone aquifer, and the F12 fault and Fault No. 12 are unified into the F12 fault. Conflict resolution is performed. In cases where multiple contradictory preliminary geological relationships exist for the same pair of geological entities, the priority of the data source is used for adjudication. Hydrogeological reports have higher priority than geological exploration reports, and geological exploration reports have higher priority than historical flood accident analysis reports. Relationship content from high-priority sources is retained. For numerical attribute conflicts from the same source, such as the thickness of an aquitard being simultaneously labeled as 30 meters and 15 meters, interval merging is used to record it as 15 to 30 meters, with an added data uncertainty marker. After semantic alignment and conflict resolution, duplicate content is removed, and a set of geological entities and geological relationships that are semantically consistent, logically conflict-free, and free of redundancy is output.

[0031] Based on the geological relationship set, the geological entity set is subjected to graph structure construction and relationship embedding to form a coal mine hydrogeological map.

[0032] Specifically, graph structure construction and relation embedding are performed on the geological entity set according to the geological relation set. Each geological entity in the geological entity set is used as a graph node, and each geological relation in the geological relation set is used as a directed edge connecting two corresponding geological entities to construct the initial graph structure. The TransR relation embedding method is employed (TransR is a knowledge graph representation learning method whose core idea is to map entities and relations to different vector spaces: each entity is represented by a vector in the entity space, and each relation is also represented by a vector in an independent relation space. A projection matrix is ​​used to project the relevant entity vectors from the entity space to the relation space of the corresponding relation; in the relation space, the head entity vector is mapped through the relation...). After the system vector is translated, it should be as close as possible to the tail entity vector to model the semantic structure where the head entity's combination relationship equals the tail entity's semantic structure, effectively capturing the complex relationships between entities under different relationships. A vector representation in entity space is generated for each geological entity, and a vector representation in relation space is generated for each geological relationship. The vectors of the starting and ending geological entities are mapped from entity space to relation space of the current geological relationship through the projection matrix corresponding to the geological relationship. The semantic rationality of each structure composed of geological entities and geological relationships is evaluated in relation space. Based on the semantic rationality, each edge in the graph is assigned a corresponding weight, forming a coal mine hydrogeological map where nodes have semantic vectors, edges have semantic weights, and the overall structure reflects the static hydrogeological knowledge of the mining area.

[0033] Furthermore, the coal mine hydrogeological atlas is a semantic knowledge representation organized in a graph structure. Graph nodes represent geological entities such as faults, aquifers, aquitards, goafs, coal seams, boreholes, and roadways. Directed edges represent geological relationships such as faults cutting aquifers, aquifers being located above coal seams, and aquitards isolating old goaf water. Each graph node and directed edge carries a semantic vector and weight, reflecting entity attributes and relationship credibility. The overall structure integrates static hydrogeological knowledge from geological exploration reports, hydrogeological reports, coal mine geological spatial data, and historical water hazard accident analysis reports to express the distribution of underground aquifers, aquitard conditions, structural features, and the interrelationships of various elements in the coal mine hydrogeological atlas.

[0034] S2. Collect real-time data from multiple sources in the coal mine, perform semantic parsing and structured reorganization on the real-time data from multiple sources in the coal mine, output the real-time data stream of the coal mine, and use a spatiotemporal indexing algorithm to coordinate and semantically bind the real-time data stream of the coal mine with the hydrogeological map of the coal mine, and output the coal mine monitoring data set.

[0035] Collect real-time data from multiple sources in coal mines, identify geological entities and monitoring parameters in the real-time data, and semantically associate and contextually bind the geological entities and monitoring parameters to form a semantic data set for coal mines.

[0036] Specifically, the system collects multi-source real-time data from the coal mine. This data includes monitoring signals from underground water pressure sensors, water level gauges, water temperature probes, water inflow metering devices, microseismic monitoring arrays, stress and strain gauges, and geoelectric field monitoring instruments, as well as rainfall data provided by surface meteorological stations and information on the location and progress of mining faces pushed by the production dispatch center. Each piece of multi-source real-time data is analyzed to identify associated geological entities, including faults, Ordovician limestone aquifers, working faces, main transport roadways, and boundaries of old goaf areas. Simultaneously, the corresponding monitoring parameters are identified, including water pressure values, water level values, water temperature values, water inflow, microseismic event energy, and stress and strain. The data includes values, geoelectric field strength, daily rainfall, and working face advance distance. Semantic associations are established between each identified geological entity and monitoring parameter in the same data record. For example, faults are linked to the energy of microseismic events reported by nearby microseismic monitoring arrays, and Ordovician limestone aquifers are linked to water pressure sensor readings. Descriptive context is added to each pair of geological entities and monitoring parameters, including equipment number, spatial proximity, and mining activity status, based on contextual information such as equipment installation location and mining progress. After processing all data records, all combinations of geological entities and monitoring parameters with contextual descriptions are aggregated to form a coal mine semantic data set.

[0037] Perform time-series alignment and streaming encapsulation transformation on the semantic data set of coal mines to output real-time data streams of coal mines.

[0038] Specifically, each geological entity and monitoring parameter combination record in the coal mine semantic data set is timestamped using a unified time benchmark, and records from different monitoring devices are merged into the same time slice. All records within each time slice are sorted and structured according to geological entity type and monitoring parameter category to form batch units with a clear time order. Continuous time slices are encapsulated according to streaming data format specifications, and streaming sequence numbers and time slice start and end identifiers are added to generate a continuously outputting, structurally consistent, and time-ordered data sequence, i.e., the real-time data stream of the coal mine.

[0039] A spatiotemporal encoder is used to extract the timestamp and three-dimensional coordinates of each data point from the real-time coal mine data stream to generate a spatiotemporal coal mine data stream.

[0040] Specifically, the spatiotemporal indexing algorithm is a joint indexing mechanism that integrates the time dimension and three-dimensional spatial coordinates. It is used to achieve accurate alignment between streaming monitoring data and static map nodes. The execution process includes three stages: spatiotemporal information extraction, spatial location mapping, and semantic consistency verification. A spatiotemporal encoder is used to parse the embedded timestamp field of each data point in the coal mine real-time data stream, and the corresponding three-dimensional coordinates are retrieved from the unified measurement coordinate system of the coal mine based on the geological entities associated with the data point and the installation location of the monitoring equipment. When the data point directly contains coordinate information, it is directly extracted and used. When the data point is only associated with the roadway name, working face number, or geological entity name, the three-dimensional coordinates are obtained by matching the roadway centerline, working face design location, or geological entity spatial range in the coal mine geological spatial data. The extracted timestamp and three-dimensional coordinates are appended to the data point to form a complete data record containing timestamp, three-dimensional coordinates, geological entity, monitoring parameters, and contextual description. All complete data records are reassembled in the original order of the coal mine real-time data stream to generate a spatiotemporal coal mine data stream.

[0041] The spatiotemporal coal mine data stream is mapped and associated in the coal mine hydrogeological map using a coordinate matching method to generate coordinate mapping relationships.

[0042] Specifically, for each data point in the spatiotemporal coal mine data stream, using three-dimensional coordinates as the query basis, a coordinate matching method is executed in the node spatial index of the coal mine hydrogeological map to calculate the Euclidean distance between the three-dimensional coordinates and the spatial location represented by each geological entity node in the coal mine hydrogeological map, and the geological entity node with the smallest distance is selected as the matching target; a correspondence is established between the current data point and the successfully matched geological entity node, and recorded as a pairing entry between the data point and the geological entity node; after traversing all data points in the spatiotemporal coal mine data stream, all pairing entries are summarized to generate a coordinate mapping relationship.

[0043] The spatiotemporal coal mine data stream and the coal mine hydrogeological map are linked and verified according to the coordinate mapping relationship, and a coal mine monitoring data set is output.

[0044] Specifically, based on the pairing entries between each data point and geological entity node in the coordinate mapping relationship, the monitoring parameters of the corresponding data points in the spatiotemporal coal mine data stream are attempted to be linked to the corresponding geological entity nodes in the coal mine hydrogeological atlas. Attribute link verification refers to checking whether the monitoring parameter type belongs to the monitoring attribute category that the current geological entity node is allowed to associate with. For example, water pressure, water level, water temperature, and water inflow are only allowed to be linked to aquifer, old working water body, or aquitard nodes; microseismic event energy is only allowed to be linked to fault or collapse column nodes; working face advance distance is only allowed to be linked to working face nodes; and daily rainfall is only allowed as a global environmental parameter without being bound to a geological entity. If the monitoring parameter type is incompatible with the semantic role of the geological entity node or there is a spatial logic conflict in the context description (such as a water pressure sensor in the roadway being bound to an aquifer far from the roadway), the verification is judged as a failure and the current data point is discarded. If the monitoring parameter type matches and the context logic is consistent, the verification is judged as successful. All verified data points and their binding information with geological entity nodes are retained to form a coal mine monitoring data set.

[0045] S3. The coal mine monitoring dataset is integrated into the coal mine hydrogeological map as a dynamically updated element, outputting an enhanced hydrological map. The enhanced hydrological map is then topologically reconstructed and its relationships evolved to form a spatiotemporal hydrological map of the coal mine.

[0046] The coal mine monitoring data set is discretized into spatiotemporal data segments based on timestamps and spatial coordinates. Dynamic attributes and related events within each spatiotemporal data segment are extracted and integrated to generate a spatiotemporal data segment set.

[0047] Specifically, all data points in the coal mine monitoring dataset are grouped within a fixed time window. Data points within the same time window whose Euclidean distance between their three-dimensional coordinates is less than a set neighborhood radius (the set neighborhood radius is determined by analyzing the impact range of typical coal mine water hazard events, specifically based on the measured spatial diffusion distance of the water inrush point's impact area in historical water hazard accident analysis reports, combined with the current mine roadway layout density and aquifer permeability characteristics, typically ranging from 10 to 50 meters) are grouped into the same spatiotemporal data segment. For each spatiotemporal data segment, the monitoring data contained therein are extracted. The parameter values ​​are dynamic attributes, including water pressure, water level, water temperature, water inflow, microseismic event energy, stress and strain, geoelectric field strength, and working face advance distance. Simultaneously, abnormal phenomena indicated by multiple dynamic attributes are identified as associated events, including sudden increases in water pressure accompanied by rising microseismic frequency, abnormal increases in water temperature ahead of the working face, and rapid rise in water level in old working areas after rainfall. Each spatiotemporal data segment and its corresponding dynamic attributes and associated events are organized into a structured record. After traversing all time windows, all structured records are collected and integrated to generate a spatiotemporal data segment set.

[0048] A subgraph isomorphic matching mechanism is used to perform pattern matching and structure mapping between the spatiotemporal data fragment set and the local substructure in the coal mine hydrogeological map, generating a mapping relationship table.

[0049] Specifically, for each spatiotemporal data segment in the spatiotemporal data segment set, the geological entities and their interrelationships involved in the associated events are extracted to form a subgraph to be matched. All local substructures composed of the same type of geological entities and the same type of geological relationships are traversed in the coal mine hydrogeological atlas. The subgraph isomorphic matching mechanism determines whether there is structural equivalence between the subgraph to be matched and the local substructure by comparing node labels (such as faults, aquifers, and impermeable layers) and edge labels (such as faults cutting aquifers and impermeable layers isolating old workings). When there is a one-to-one correspondence between nodes and edges with completely identical labels, the match is considered successful, and the spatial location correspondence and structural mapping path between each geological entity in the current spatiotemporal data segment and the corresponding geological entity node in the coal mine hydrogeological atlas are recorded. If no local substructure meeting the conditions is found, the current spatiotemporal data segment is marked as having no matching item. After completing the matching process for all spatiotemporal data segments, all successfully matched mapping records are summarized to generate a mapping relationship table.

[0050] Furthermore, the subgraph isomorphism matching mechanism is a graph theory method used to determine whether a graph (called the query graph) is completely identical to a subgraph in another graph (called the target graph) in terms of structure and node / edge labels. Specifically, the subgraph isomorphism matching mechanism finds a one-to-one mapping from nodes in the query graph to some nodes in the target graph, such that each edge in the query graph also has an edge of the same type at the corresponding position in the target graph, and the semantic labels of all nodes and edges (such as fault, aquifer, and fault cutting aquifer) are strictly matched, thereby confirming that the subgraph to be matched is structurally equivalent to the local substructure.

[0051] Based on the mapping table, the associated events in the spatiotemporal data fragment set are inserted as new nodes and the dynamic attributes are inserted as node attributes into the coal mine hydrogeological map to generate a real-time intermediate state map.

[0052] Specifically, each record in the mapping table is traversed. Based on the correspondence between the spatiotemporal data fragments indicated in the records and the geological entity nodes in the coal mine hydrogeological map, the associated events in the spatiotemporal data fragments are created as new event nodes in the coal mine hydrogeological map. The event node name includes the event type and spatial location identifier. At the same time, the dynamic attributes in the current spatiotemporal data fragments are attached to the geological entity nodes specified in the mapping table according to the attribute type, as new attributes or attribute update values ​​for the current geological entity nodes. For example, the water pressure value is written into the water pressure attribute field of the Ordovician limestone aquifer node. For spatiotemporal data fragments marked as having no matching items in the mapping table, temporary context nodes are added in the coal mine hydrogeological map with the three-dimensional coordinates as the center, and the associated events and dynamic attributes are bound to the temporary context nodes. After completing all insertion operations, a map version integrating real-time monitoring information is generated, namely the real-time intermediate state map.

[0053] The system performs graph structure conflict detection and attribute logic conflict detection on the real-time intermediate state graph, resolves the conflicts based on the detection results, and generates a resolved state graph.

[0054] Specifically, graph structure conflict detection involves traversing all newly added edges and event nodes in the real-time intermediate state graph to check for connections that violate common sense in coal mine hydrogeology. For example, aquitard nodes may be directly connected to both strong aquifers and old workings without faults or collapse columns acting as water channels, or microseismic event nodes may be connected to non-tectonic geological entities such as coal seams. Attribute logic conflict detection verifies the physical logic between current dynamic attributes for each geological entity node. For graph structure conflicts, unreasonable newly added edges or event nodes are deleted based on the principle of prioritizing static knowledge in the coal mine hydrogeological graph. For attribute logic conflicts, reconciliation is achieved by retaining high-confidence source data or using time-decay weighted fusion. After all conflict handling is completed, a logically consistent and structurally sound resolved state graph is generated. This resolved state graph is a real-time intermediate state graph after conflict correction, where nodes and edges retain effective real-time monitoring information while conforming to the basic hydrogeological laws of the mining area.

[0055] By reorganizing the topology of the state map through dynamic community discovery, the internal connectivity of coal mine water hazard risk areas is strengthened, and an enhanced hydrological map is output.

[0056] Specifically, the dynamic community discovery method is a graph analysis method based on the connection density and attribute similarity between nodes. It is used to automatically identify closely related sub-graph clusters (i.e., communities) in a graph structure that evolves over time. Based on edge weights, node semantic labels, and dynamic attribute change trends, nodes with high cohesion (such as multiple water pressure anomalies, microseismic events, and water-conducting faults being adjacent to each other) are aggregated into a community. The dynamic community discovery method is applied to the resolution state map to identify communities with potential water hazard risks. Virtual connections are added between geological entity nodes within the community or the weights of existing edges are increased to enhance structural cohesion. The original connections of non-risk communities are kept unchanged. The reconstructed map highlights the clustering characteristics of risk areas while preserving the original geological relationships, and outputs an enhanced hydrological map.

[0057] To enhance each node and relationship in the hydrological map, a timestamp attribute is added, and the graph structure is reorganized according to the time sequence based on the timestamp attribute to generate a reconstructed hydrological map.

[0058] Specifically, for each geological entity node and each geological relationship in the enhanced hydrological map, a corresponding timestamp attribute is added. Static geological entity nodes use the timestamp from when the coal mine hydrogeological map was constructed, while dynamic event nodes and monitoring attribute nodes use the start time of the time window of their respective spatiotemporal data segments. Based on the timestamp attributes of all nodes and relationships, the enhanced hydrological map is divided into continuous time slices in chronological order. Each time slice contains nodes that exist or have been updated at the current time and their connections. Cross-time associations of nodes are preserved between adjacent time slices, and the states of the same geological entity in different time slices are connected by time link edges, forming a reconstructed hydrological map with time as the main order, maintaining the original topology within each time slice, and reflecting the state evolution between slices.

[0059] Dynamic propagation and aggregation operations of relationships between nodes are performed on the reconstructed hydrological map to form a spatiotemporal hydrological map of the coal mine.

[0060] Specifically, starting from water source geological entities such as aquifers and old workings, dynamic attributes such as water pressure and temperature values ​​are propagated along the time-increasing direction according to the edge type and weight to adjacent geological entities such as aquitards, faults, and goafs. The propagation process considers time interval decay and spatial barrier effects. The multi-source dynamic attributes received by the same target geological entity node at different time slices are weighted and aggregated, with the weight determined by the semantic credibility and time freshness of the edge. Tectonic activity indicators such as microseismic event energy and stress-strain values ​​are temporally correlated along geological relationships such as faults and collapse columns to form an event transmission chain with causal logic. After completing the full map traversal, each geological entity node carries a spatiotemporal state that integrates multi-source real-time information, forming a coal mine hydrological spatiotemporal map that reflects the evolution of hydrological conditions in the mining area over time.

[0061] S4. The path walking algorithm is used to identify the water hazard risk transmission path in the hydrological spatiotemporal map of the coal mine, and the dynamic risk propagation simulation and water inrush coefficient quantitative evaluation of the water hazard risk transmission path are carried out to output the water inrush coefficient path set.

[0062] In the spatiotemporal map of coal mine hydrology, starting from the predefined water hazard risk source, a multi-source concurrent path walk is performed to output the initial risk transmission path set.

[0063] Specifically, the predefined water hazard risk sources include Ordovician limestone aquifers, old working water bodies, strongly water-rich faults, and areas with active rainfall infiltration. A path-walking algorithm (a traversal method in a graph structure that starts from a specified initial node and expands along adjacent nodes based on edge direction, weight, and time constraints, recording the traversal path; in the spatiotemporal map of coal mine hydrology, the path-walking algorithm is used to simulate the potential transmission process of water hazard risk from water sources to protected target areas) is employed, with each water hazard risk source as the starting node in the spatiotemporal map of coal mine hydrology. Multiple path traversal processes are initiated simultaneously on the graph. Each path traversal process visits adjacent geological entity nodes in sequence along the time-increasing direction, based on the direction and weight of the edges in the graph. These nodes include aquitards, coal seams, goafs, roadways, and collapse columns. During the traversal process, the sequence of geological entity nodes traversed, the geological relationship type of the connecting edges, and the corresponding timestamps are recorded. When a path reaches a protected target area such as a working face or main transport roadway, the current path extension is terminated, and all paths generated by various water hazard risk sources are aggregated to output an initial risk transmission path set.

[0064] Furthermore, water hazard risk sources are predefined based on static hydrogeological knowledge from historical water hazard accident analysis reports, hydrogeological reports, and coal mine hydrogeological maps. These sources include Ordovician limestone aquifers, old goaf water bodies, highly water-rich faults, and areas with active rainfall infiltration. Among these, the Ordovician limestone aquifers are identified as the main water source due to their high water pressure and strong recharge capacity. Old goaf water bodies are determined based on the historical water accumulation range and water level monitoring data of the goaf areas. Highly water-rich faults are determined through fault conductivity evaluation and microseismic activity frequency. Areas with active rainfall infiltration are delineated by combining the distribution of surface subsidence areas, overburden thickness, and the historical rainfall-water inrush response relationship. All water hazard risk sources are set based on actual geological conditions, historical disaster records, and hydrogeological monitoring evidence to ensure coverage of water inrush hazard sources.

[0065] The initial risk transmission path set is subjected to temporal reachability verification and redundancy elimination, and the water hazard risk transmission path is output.

[0066] Specifically, for each path in the initial risk transmission path set, the timestamps corresponding to the geological entity node sequence are checked to ensure that the time order is not strictly decreasing. If there is a case where the timestamp of the later node is earlier than the timestamp of the earlier node, the current path is determined not to meet the temporal accessibility requirement and is removed. For paths that pass the temporal verification, a structural comparison is performed. If any two paths have the same sequence of geological entity node types, the same sequence of geological relationship edges, and cover the same time interval, they are considered redundant paths, and only one of them is retained. At the same time, paths that do not reach the working face or the main transport tunnel and other protected target areas are removed. After completing the temporal accessibility verification and redundancy elimination processing, the remaining paths are output as valid results to form the water hazard risk transmission path.

[0067] The time-series water pressure and stress data of each water hazard risk transmission path are extracted from the real-time data stream of the coal mine. Based on the time-series water pressure and stress data, the dynamic propagation simulation of the water pressure and stress transmission relationship of each water hazard risk transmission path is carried out, and the simulated water hazard risk path set is output.

[0068] Specifically, for each water hazard risk transmission path, based on the spatial location and timestamp of each geological entity node in the path, the water pressure and stress-strain values ​​reported by the corresponding monitoring equipment within the matching time window are retrieved from the real-time data stream of the coal mine, forming a time-series water pressure sequence and a time-series stress sequence specific to the current path. The time-series water pressure sequence and time-series stress sequence are then input into the dynamic propagation simulation process of water pressure and stress transmission relationship according to the path node sequence. During the simulation, geological attribute parameters such as the thickness of the aquitard, the fault water conductivity, and the permeability of the coal and rock mass are combined to statistically analyze the attenuation or cumulative effect of water pressure along the path and the stress superposition response segment by segment. After dynamic propagation simulation, each path generates an enhanced path containing the risk evolution process. All enhanced paths are aggregated and output as a set of simulated water hazard risk paths.

[0069] Water hazard risk features are extracted from the set of simulated water hazard risk paths, and the inrush coefficient is calculated according to the water hazard risk features to generate a set of inrush coefficient risk paths.

[0070] Specifically, for each path in the simulated flood risk path set, five flood risk characteristics are extracted from the terminal node and the entire path: peak water pressure at the end of the path (the last effective water pressure value of the path), water pressure rise rate (the maximum slope obtained by the time difference of continuous water pressure values ​​in the path is used as the water pressure rise rate), cumulative stress increment (the sum of stress and strain values ​​of each node in the path), safe thickness ratio of aquitard (the ratio of actual aquitard thickness to critical safe thickness), and number of times the path crosses a water-conducting structure (the number of times the path passes through faults or collapse columns). Each flood risk characteristic (such as peak water pressure, water pressure rise rate, cumulative stress increment, safe thickness ratio of aquitard, and number of times the path crosses a water-conducting structure) is normalized and mapped to the [0,1] interval to eliminate dimensional differences. The normalized characteristic values ​​are then substituted into the weighted line. The formula combines the weights of historical water inrush cases in the mining area, which are determined through statistical regression (the weights include peak water pressure weight, water pressure rise rate weight, cumulative stress increment weight, safe thickness ratio of aquitard layer weight, and number of times water-conducting structures cross, totaling five weights. The weight determination is based on the historical water inrush case dataset of the mining area. Through multiple linear regression, with whether water inrush occurred as the dependent variable and the five normalized features as independent variables, the contribution coefficients of each feature to the water inrush result are fitted, and the final weight values ​​are determined after optimization through cross-validation and residual analysis. The calibration process ensures that the weights reflect the relative influence of each factor in actual water inrush events). The dimensionless water inrush coefficient of each path is calculated, and each path is bound to the corresponding water inrush coefficient to form a path water inrush coefficient pair. All path water inrush coefficient pairs are integrated to generate a set of risk paths for water inrush coefficients.

[0071] The formula for calculating the water inrush coefficient is as follows: ; in, Indicates the inrush coefficient. Indicates the peak water pressure weight. This represents the peak water pressure at the end of the path after normalization. Indicates the weight of the rate of increase in water pressure. Indicates the cumulative stress increment weight. This indicates the weighting of the safe thickness of the waterproof layer. Indicates the weight of the number of times the water-conducting structure crosses. This represents the normalized rate of increase in water pressure. This represents the normalized cumulative stress increment. This represents the normalized safe thickness ratio of the waterproof layer. This indicates the number of times the path crosses a water-conducting structure after normalization.

[0072] Furthermore, the water inrush coefficient is a dimensionless comprehensive index used to quantify the degree of water hazard risk in coal mines. It reflects the likelihood of water inrush from a water-bearing body into the mining space under corresponding geological and mining conditions. The value is calculated by weighted integration of multiple factors such as water pressure intensity, water pressure dynamic change rate, stress disturbance accumulation, safety margin of aquitard, and the degree of development of water-conducting structures. The larger the water inrush coefficient, the higher the water hazard risk, indicating the existence of water inrush danger and the need to take early warning or prevention measures.

[0073] The risk path set based on the water inrush coefficient is sorted in reverse order and the risk is screened to output the water inrush coefficient path set.

[0074] Specifically, reverse sorting and risk screening refers to sorting all paths in the risk path set based on their corresponding water inrush coefficient values ​​from largest to smallest. For example, if path A has a water inrush coefficient of 0.82, path B has 0.75, and path C has 0.63, the sorting result is path A, path B, and path C. A risk screening threshold of 0.70 is set. This threshold is determined based on the statistical distribution of water inrush coefficients before water inrush events in historical water inrush cases in the mining area. It is calculated by analyzing the minimum water inrush coefficient of paths that have experienced water inrushes and combining it with a safety margin (usually the 95th percentile or the optimal cutoff point of the ROC curve) to ensure that the screened paths have a high probability of water inrush. Only paths with a water inrush coefficient value greater than or equal to 0.70 are retained; therefore, paths A and B are retained, and path C is removed. After sorting and screening, the remaining paths constitute the water inrush coefficient path set.

[0075] S5. Map risk levels and synthesize early warning information for the set of water inrush coefficient paths, and output a coal mine water hazard risk early warning report.

[0076] Based on a preset risk threshold for the water inrush coefficient, the risk level of the water inrush coefficient path set is classified, and a risk-labeled path set is generated.

[0077] Specifically, the risk level of the water inrush coefficient path set is classified based on a preset water inrush coefficient risk threshold. The water inrush coefficient risk threshold is divided into three intervals: a water inrush coefficient less than 0.60 corresponds to a low risk level, a water inrush coefficient greater than or equal to 0.60 and less than 0.75 corresponds to a medium risk level, and a water inrush coefficient greater than or equal to 0.75 corresponds to a high risk level. The water inrush coefficient risk threshold is determined based on the statistical distribution of water inrush coefficients before water inrush events in historical water inrush cases in the mining area and safety control requirements. For each path in the water inrush coefficient path set, the water inrush coefficient value is read, and the risk threshold interval to which the water inrush coefficient value belongs is determined. Based on the determination result, the current path is assigned a corresponding risk level label, namely low risk, medium risk, or high risk. Each path is combined with the corresponding risk level label to form a structured record. After all paths are processed, all structured records are collected to generate a risk-labeled path set.

[0078] Multi-dimensional risk attributes are synthesized from the risk-labeled path set to output early warning information item data.

[0079] Specifically, for each path in the risk-marked path set, multi-dimensional risk attributes are extracted. These attributes include the type of water hazard risk source at the path's starting point, the name of the protected target area at the path's ending point, water-conducting structures such as faults or collapse columns traversed by the path, the water inrush coefficient value, the risk level label, the number of geological entity nodes included in the path, the monitoring equipment numbers involved, and the timestamp of the most recent dynamic attribute anomaly. These multi-dimensional risk attributes are then structured according to unified early warning information fields, including the risk source, affected area, water inrush coefficient, risk level, key structures, monitoring points, and anomaly time. For paths with medium or high risk levels, response measures matching the risk characteristics are entered in the treatment suggestion field, such as increasing the frequency of water pressure monitoring, verifying the integrity of the aquitard, or suspending operations at nearby workfaces. Each path, after attribute integration, forms a complete early warning information entry. After all paths are processed, early warning information entry data, composed of all early warning information entries, is output.

[0080] Integrate the data of early warning information items, extract risk summaries and format reports, and output a coal mine water hazard risk early warning report.

[0081] Specifically, the warning information items are sorted from highest to lowest risk level, with high-risk items placed at the beginning of the main report, medium-risk items in the second position, and low-risk items excluded from the main report. Common characteristics are analyzed from the high- and medium-risk warning information items, including frequently occurring water hazard risk source types, concentrated protected target areas, repeatedly involved water-conducting structures, and concentrated intervals of water inrush coefficients. Based on this, three categories of risk summary content are extracted: main risk areas, key disaster-causing factors, and key prevention and control targets. Following the standard format of coal mine water hazard risk warning reports, the report structure is constructed sequentially, including a cover page, risk summary, detailed risk path list, spatial distribution description, summary of disposal suggestions, and appendices. The risk summary content is filled into the risk summary section, and the sorted warning information items are filled into the detailed risk path list section, with the spatial distribution description describing the location relationship of each risk path's roadway or mining area. After completing the content filling and format verification, the coal mine water hazard risk warning report is output.

[0082] This embodiment also provides a coal mine water hazard risk early warning system based on big data, including: A mapping module is used to acquire multi-source historical geological data of coal mines, extract and fuse entity relationships from the multi-source historical geological data of coal mines, and construct hydrogeological maps of coal mines. The multi-source historical geological data of coal mines includes geological exploration reports, hydrogeological reports, coal mine geological spatial data, and historical water hazard accident analysis reports. The monitoring data module is used to collect real-time data from multiple sources in the coal mine, perform semantic parsing and structured reorganization on the real-time data from multiple sources in the coal mine, output the real-time data stream of the coal mine, and use a spatiotemporal indexing algorithm to coordinate and semantically bind the real-time data stream of the coal mine with the hydrogeological map of the coal mine, outputting a set of coal mine monitoring data. The spatiotemporal mapping module is used to integrate coal mine monitoring datasets into the coal mine hydrogeological map as dynamically updated elements, output an enhanced hydrological map, and perform topological reconstruction and relation evolution on the enhanced hydrological map to form a coal mine hydrological spatiotemporal map. The water inrush coefficient module is used to identify water hazard risk transmission paths in the spatiotemporal map of coal mine hydrology using a path walking algorithm, perform dynamic risk propagation simulation and water inrush coefficient quantitative evaluation of water hazard risk transmission paths, and output a set of water inrush coefficient paths. The early warning report module is used to map the risk level of the water inrush coefficient path set and synthesize early warning information, and integrate and output a coal mine water hazard risk early warning report.

[0083] In summary, this invention achieves a leap from static assessment to dynamic projection of coal mine water hazard risk by combining dynamic map updates with risk path mining. By integrating coal mine monitoring datasets as dynamically updated elements into the coal mine hydrogeological map and performing topological reconstruction, deep fusion and dynamic representation of multi-source heterogeneous data within a unified semantic framework are achieved. Furthermore, a path-walking algorithm is employed on the coal mine hydrogeological spatiotemporal map to identify water hazard risk transmission paths, enabling the automatic discovery and explicit expression of hidden and complex risk channels. This allows early warning systems to not only output risk levels but also clearly reveal the spatial location, transmission path, and evolution mechanism of the risk, thus providing decision support with precise positioning and causal explanation capabilities.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for early warning of coal mine water hazard risks based on big data, characterized in that: include, The process involves acquiring multi-source historical geological data of coal mines, extracting and fusing entity relationships from the multi-source historical geological data, and constructing a hydrogeological atlas of the coal mines. The multi-source historical geological data of the coal mines includes geological exploration reports, hydrogeological reports, coal mine geological spatial data, and historical water hazard accident analysis reports. Collect real-time data from multiple sources in the coal mine, perform semantic parsing and structured reorganization on the real-time data from multiple sources in the coal mine, output the real-time data stream of the coal mine, and use a spatiotemporal indexing algorithm to coordinate and semantically bind the real-time data stream of the coal mine with the hydrogeological map of the coal mine, outputting a set of coal mine monitoring data; The coal mine monitoring dataset is integrated into the coal mine hydrogeological map as a dynamically updated element, outputting an enhanced hydrological map. The enhanced hydrological map is then subjected to topological reconstruction and relational evolution to form a spatiotemporal hydrological map of the coal mine. The path walking algorithm is used to identify the water hazard risk transmission path in the spatiotemporal map of coal mine hydrology. Dynamic risk propagation simulation and water inrush coefficient quantitative evaluation of the water hazard risk transmission path are carried out, and the water inrush coefficient path set is output. Risk level mapping and early warning information synthesis are performed on the set of water inrush coefficient paths, and a coal mine water hazard risk early warning report is output.

2. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The specific steps for extracting and fusing entity relationships from multi-source historical geological data of coal mines to construct a hydrogeological map of the coal mine are as follows: Preliminary geological entities and preliminary geological relationships are extracted from multi-source historical geological data of coal mines using deep learning extraction methods, forming a set of preliminary entity-relationship pairs. Perform semantic alignment and conflict resolution on the initial entity relation set, and output the geological entity set and the geological relation set; Based on the geological relationship set, the geological entity set is subjected to graph structure construction and relationship embedding to form a coal mine hydrogeological map.

3. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The specific steps for performing semantic parsing and structured reorganization on multi-source real-time data from coal mines to output a real-time data stream from the coal mine are as follows: Identify geological entities and monitoring parameters in multi-source real-time data of coal mines, and semantically associate and contextually bind geological entities and monitoring parameters to form a set of semantic data of coal mines. Perform time-series alignment and streaming encapsulation transformation on the semantic data set of coal mines to output real-time data streams of coal mines.

4. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The process involves using a spatiotemporal indexing algorithm to correlate and semantically bind real-time coal mine data streams with coal mine hydrogeological maps, outputting a coal mine monitoring data set. The specific steps are as follows: A spatiotemporal encoder is used to extract the timestamp and three-dimensional coordinates of each data point from the real-time data stream of the coal mine, and a spatiotemporal coal mine data stream is generated. The spatiotemporal coal mine data stream is mapped and associated in the coal mine hydrogeological map using a coordinate matching method to generate coordinate mapping relationships. The spatiotemporal coal mine data stream and the coal mine hydrogeological map are linked and verified according to the coordinate mapping relationship, and a coal mine monitoring data set is output.

5. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The specific steps for integrating coal mine monitoring data into dynamically updated elements and injecting them into the coal mine hydrogeological map to output an enhanced hydrogeological map are as follows: The coal mine monitoring data set is discretized into spatiotemporal data segments according to timestamps and spatial coordinates, and the dynamic attributes and related events in each spatiotemporal data segment are extracted and integrated to generate a spatiotemporal data segment set. A subgraph isomorphic matching mechanism is used to perform pattern matching and structure mapping between the spatiotemporal data fragment set and the local substructure in the coal mine hydrogeological map, generating a mapping relationship table. Based on the mapping table, the associated events in the spatiotemporal data fragment set are inserted as new nodes and the dynamic attributes are inserted as node attributes into the coal mine hydrogeological map to generate a real-time intermediate state map. Perform graph structure conflict detection and attribute logic conflict detection on the real-time intermediate state graph, resolve the conflict results, and generate a resolved state graph; By reorganizing the topology of the state map through dynamic community discovery, the internal connectivity of coal mine water hazard risk areas is strengthened, and an enhanced hydrological map is output.

6. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The specific steps for performing topological reconstruction and relational evolution on the enhanced hydrological map to form a spatiotemporal hydrological map of the coal mine are as follows: To enhance each node and relationship in the hydrological map, a timestamp attribute is added, and the graph structure is reorganized in chronological order according to the timestamp attribute to generate a reconstructed hydrological map; Dynamic propagation and aggregation operations of relationships between nodes are performed on the reconstructed hydrological map to form a spatiotemporal hydrological map of the coal mine.

7. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The specific steps for identifying the water hazard risk transmission path in the spatiotemporal map of coal mine hydrology using the path walking algorithm are as follows: In the spatiotemporal map of coal mine hydrology, starting from the predefined water hazard risk source, a multi-source concurrent path walk is performed to output the initial risk transmission path set; The initial risk transmission path set is subjected to temporal reachability verification and redundancy elimination, and the water hazard risk transmission path is output.

8. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The specific steps for performing dynamic risk propagation simulation and quantitative assessment of the water inrush risk transmission path, and outputting a set of water inrush coefficient paths, are as follows: The time-series water pressure and stress data of each water hazard risk transmission path are extracted from the real-time data stream of the coal mine. Based on the time-series water pressure and stress data, the dynamic propagation simulation of the water pressure and stress transmission relationship of each water hazard risk transmission path is carried out, and the simulated water hazard risk path set is output. Water hazard risk features are extracted from the set of simulated water hazard risk paths, and the inrush coefficient is calculated according to the water hazard risk features to generate a set of inrush coefficient risk paths. The risk path set based on the water inrush coefficient is sorted in reverse order and the risk is screened to output the water inrush coefficient path set.

9. The coal mine water hazard risk early warning method based on big data as described in claim 1, characterized in that: The specific steps for mapping risk levels and synthesizing early warning information from the set of water inrush coefficient paths to output a coal mine water hazard risk early warning report are as follows: Based on the preset risk threshold of the water inrush coefficient, the risk level of the water inrush coefficient path set is divided, and a risk-labeled path set is generated. Multi-dimensional risk attributes are synthesized from the risk-labeled path set to output early warning information item data; Integrate the data of early warning information items, extract risk summaries and format reports, and output a coal mine water hazard risk early warning report.

10. A coal mine water hazard risk early warning system based on big data, based on the coal mine water hazard risk early warning method based on big data as described in any one of claims 1 to 9, characterized in that: include, A mapping module is used to acquire multi-source historical geological data of coal mines, extract and fuse entity relationships from the multi-source historical geological data of coal mines, and construct hydrogeological maps of coal mines. The multi-source historical geological data of coal mines includes geological exploration reports, hydrogeological reports, coal mine geological spatial data, and historical water hazard accident analysis reports. The monitoring data module is used to collect real-time data from multiple sources in the coal mine, perform semantic parsing and structured reorganization on the real-time data from multiple sources in the coal mine, output the real-time data stream of the coal mine, and use a spatiotemporal indexing algorithm to coordinate and semantically bind the real-time data stream of the coal mine with the hydrogeological map of the coal mine, outputting a set of coal mine monitoring data. The spatiotemporal mapping module is used to integrate coal mine monitoring datasets into the coal mine hydrogeological map as dynamically updated elements, output an enhanced hydrological map, and perform topological reconstruction and relation evolution on the enhanced hydrological map to form a coal mine hydrological spatiotemporal map. The water inrush coefficient module is used to identify water hazard risk transmission paths in the spatiotemporal map of coal mine hydrology using a path walking algorithm, perform dynamic risk propagation simulation and water inrush coefficient quantitative evaluation of water hazard risk transmission paths, and output a set of water inrush coefficient paths. The early warning report module is used to map the risk level of the water inrush coefficient path set and synthesize early warning information, and integrate and output a coal mine water hazard risk early warning report.

Images