Knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data

Abstract

The invention relates to a knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data. The method comprises the steps that geological data information is acquired and preprocessed, and a geological information database is formed; geologic entities, attributes of the geologic entities and mutual relations of the geologic entities in the geologic information database are recognized through the natural language processing technology, and a geologic knowledge graph is constructed; training and optimizing the prediction model to obtain a metallogenic prediction model; and obtaining a prediction result of the metallogenic potential area by using the metallogenic prediction model, and generating a metallogenic analysis report. According to the method, the geological data quality is improved through data preprocessing, the knowledge graph is constructed to integrate geological knowledge, the deep learning model is utilized to accurately extract and predict the metallogenic characteristics, and finally the metallogenic analysis report is generated, so that the prospecting efficiency is improved, the cost is reduced, and the decision scientificity is enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of geological data processing and analysis, and particularly relates to a method for constructing a knowledge graph and intelligent prospecting prediction based on multi-source heterogeneous geological data. Background Art

[0002] In the field of geological prospecting, traditional prospecting methods mainly rely on the experience of geologists and geological data. Often, these geological data are in various formats, with high-dimensional and complex features, making prospecting time-consuming and laborious, error-prone, and difficult to comprehensively and accurately grasp the internal relationships of geological information. In addition, the process of ore-forming potential assessment and report generation is cumbersome, inefficient, and the report content is not comprehensive and accurate enough, making it difficult to provide sufficient basis for exploration decisions, thus affecting the accuracy and efficiency of prospecting prediction. Therefore, it is of great significance to propose an intelligent prospecting method that can utilize multi-source heterogeneous data. Summary of the Invention

[0003] Based on this, in view of the above technical problems, it is necessary to provide a method for constructing a knowledge graph and intelligent prospecting prediction based on multi-source heterogeneous geological data that can utilize multi-source heterogeneous data for prospecting.

[0004] In a first aspect, the present application provides a method for constructing a knowledge graph and intelligent prospecting prediction based on multi-source heterogeneous geological data, including:

[0005] Obtain geological data information, preprocess the geological data information to form a geological information database;

[0006] Use natural language processing technology to identify geological entities, attributes of geological entities, and the mutual relationships between geological entities and their attributes in the geological information database, and construct a geological knowledge graph;

[0007] Use the geological knowledge graph to train and optimize a prediction model to obtain an ore-forming prediction model;

[0008] Obtain geological information data of the area to be predicted, use the ore-forming prediction model to obtain the prediction result of the ore-forming potential area, and generate an ore-forming analysis report.

[0009] Further, using natural language processing technology to identify geological entities, attributes of geological entities, and their mutual relationships in the geological information database to construct a geological knowledge graph includes:

[0010] Use natural language processing technology to identify geological entities from the geological information database to form a mineral information library, where the geological entities include: ore deposits, ore bodies, rocks, minerals, structures, strata, and geological events;

[0011] According to the chemical composition, physical properties, genetic types, occurrence states, deposit scales and uses of each mineral in the mineral information database, classify each mineral into the corresponding category among metal ores, non-metal ores and energy ores, and classify the ore types corresponding to each mineral to form mineral type information;

[0012] Based on the knowledge engineering method in the field of geology, define the relationship types between geological entities in the mineral information database, including genetic relationships, spatial relationships, and temporal relationships;

[0013] Obtain ore-forming characteristic information according to the relationship types between geological entities;

[0014] Associate the minerals in the mineral type information with the ore-forming characteristic information to form ore-forming related samples;

[0015] Calculate the correlation between the corresponding minerals and the ore-forming characteristic information in the ore-forming related samples, and summarize to obtain ore-forming information;

[0016] Take the mineral type information and related geological entities as the first-layer nodes, take the ore-forming characteristics and association relationships as the second-layer nodes, and use the ore-forming characteristics and association relationships as the edges connecting the first-layer nodes to construct a geological knowledge graph.

[0017] Furthermore, calculate the correlation between the corresponding minerals and the ore-forming characteristic information in the ore-forming related samples, and summarize to obtain ore-forming information, including:

[0018] Calculate the correlation between the corresponding minerals and the ore-forming characteristic data in the ore-forming related samples. The calculation formula is as follows:

[0019]

[0020] Among them, P(Y=1|X) represents the probability that the sample belongs to the strongly ore-forming related category under the condition of given X, e represents the natural constant, N represents the total number of samples, αi represents the Lagrange multiplier for optimizing the dual variable, wi represents the weight of the i-th sample, yi represents the label of the i-th sample, K(Zx,Zxi)=((Zx) t (Zxi)+c) d represents the kernel function, where W represents the feature weight matrix for adjusting the weight of each ore-forming characteristic, c represents the constant term for adjusting the offset of the kernel function, x and xi represent the feature vectors of the mineral, including chemical composition, physical properties, genetic types, occurrence states, deposit scales and uses, and d represents the complexity of the polynomial;

[0021] Judge the ore-forming related samples with P(Y=1|X) greater than or equal to the set threshold as strongly ore-forming related samples, and summarize the strongly ore-forming related samples to form ore-forming information.

[0022] Furthermore, the geological data information includes:

[0023] Basic geological data, chemical composition analysis data, physical exploration data, remote sensing image data, and knowledge text data;

[0024] Basic geological data includes geological maps, mineral resource distribution maps, and geological structure maps in the geological database;

[0025] Chemical composition analysis data includes chemical composition analysis data of rock, soil, and water body samples in the chemical database;

[0026] Physical exploration data includes gravity, magnetic, and electrical physical exploration data;

[0027] Remote sensing image data includes satellite remote sensing images and aerial photography images;

[0028] Knowledge text data includes geological literature, investigation reports, and geological knowledge bases.

[0029] Furthermore, the prediction model is constructed based on the following method:

[0030] Select deep learning algorithms such as convolutional neural networks, graph neural networks, recurrent neural networks and their variants to construct the prediction model. Among them, convolutional neural networks are used to process remotely sensed images and geospatial correlation geological data in the geophysical field, graph neural networks are used to process graph structure data of geological knowledge graphs, and recurrent neural networks and their variants are used to analyze time series data of the temporal evolution of geological events;

[0031] Obtain ore-forming instance data in the geological knowledge graph and train the prediction model, including evaluating the performance of the prediction model using cross-validation techniques:

[0032] Randomly divide the ore-forming instance data into a training set, a cross-validation set, and a test set in a 7:3 ratio. Among them, the training set and the cross-validation set account for 70%, and the test set accounts for 30%. The training set is used to train the prediction model, the cross-validation set is used to optimize and tune the parameters of the prediction model, and the test set is used to evaluate the performance of the prediction model. The performance indicators include accuracy, recall, and F1 score. The calculation formulas are as follows:

[0033]

[0034] Among them, Precision represents the proportion of samples that are actually positive among the samples predicted as positive by the model, Recall represents the proportion of samples that are actually positive and are correctly predicted as positive by the model, TP represents the number of samples correctly predicted as positive by the model, FP represents the number of samples wrongly predicted as positive by the model, and FN represents the number of samples wrongly predicted as negative by the model;

[0035] Calculate the F1 score of each prediction model, select the prediction model with the best F1 score, and obtain the ore-forming prediction model.

[0036] Further, the ore-forming prediction model is trained and optimized using ore-forming examples in the geological knowledge graph to obtain an ore-forming prediction model, including:

[0037] Generate hyperparameter combinations using random search and grid search methods;

[0038] Evaluate the performance of the ore-forming prediction model based on different hyperparameter combinations using cross-validation techniques, and select the ore-forming prediction model with the optimal performance.

[0039] Further, obtain the geological information data of the area to be predicted, use the ore-forming prediction model to obtain the prediction result of the ore-forming potential area, and generate an ore-forming analysis report, including:

[0040] Use Internet of Things technology to deploy sensors in the area to be predicted to obtain real-time geological data;

[0041] Preprocess the real-time geological data to obtain standard data to be measured;

[0042] Based on the standard data to be measured, use the ore-forming prediction model to generate an ore analysis report. The ore-forming analysis report includes the type and distribution of mineral resources, ore-forming conditions and characteristics, resource quantity and potential assessment, spatial law scale of potential distribution, exploration risk assessment and suggestions, and exploration path and plan.

[0043] Further, obtain the geological information data of the area to be predicted, use the ore-forming prediction model to obtain the prediction result of the ore-forming potential area, and generate an ore-forming analysis report, including:

[0044] Obtain real-time geological data, combine the standard data to be measured with historical geological data information, and update the geological database and the ore-forming prediction model in real time;

[0045] Based on the real-time updated geological data, use the ore-forming prediction model to obtain the prediction result of the change trend of the ore-forming potential area. Among them, the prediction result includes: potential increase or decrease information of mineral resources, improvement or deterioration information of ore-forming conditions, and expansion or contraction information of the ore-forming potential area;

[0046] Optimize the exploration path and plan according to the prediction result of the change trend of the ore-forming potential area.

[0047] In a second aspect, the present application also provides a knowledge graph construction and intelligent prospecting prediction device based on multi-source heterogeneous geological data, including:

[0048] A data acquisition and preprocessing module, which is used to obtain geological data information and preprocess the geological data information to form a geological information database;

[0049] A knowledge graph construction module, which is used to utilize natural language processing technology to obtain geological entity information, attribute information of geological entities, relationship information between geological entities, and relationship information between geological entities and geological entity attributes in a geological information database, and construct a geological knowledge graph;

[0050] An intelligent prediction model construction module, which is used to utilize the geological knowledge graph to train and optimize a prediction model to obtain an ore-forming prediction model;

[0051] An ore-forming prediction and analysis report generation module, which is used to obtain geological information data of an area to be predicted, utilize the ore-forming prediction model to obtain a prediction result of an ore-forming potential area, and generate an ore-forming analysis report.

[0052] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data are implemented.

[0053] The knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data provided by this application forms a database by acquiring and preprocessing geological data, constructs a geological knowledge graph using natural language processing technology, then trains an ore-forming prediction model based on this graph, and finally realizes the accurate prediction of the ore-forming potential of the target area and the generation of a report. This process effectively solves problems such as the complexity of geological data, low prospecting efficiency, and inaccurate ore-forming assessment, and significantly improves the efficiency and accuracy of geological prospecting. Description of the Drawings

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0055] Figure 1 It is a flowchart of a knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data of the present invention;

[0056] Figure 2 It is a structural schematic diagram of a knowledge graph construction and intelligent prospecting prediction device based on multi-source heterogeneous geological data of the present invention. Detailed Embodiments

[0057] To make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] As Figure 1 shown, a method for constructing a knowledge graph and intelligent ore prospecting prediction based on multi-source heterogeneous geological data is provided. In this embodiment, the method is exemplified by being applied to a terminal. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] S101. Obtain geological data information, preprocess the geological data information, and form a geological information database.

[0060] Specifically, the preprocessing of the data information may include: data cleaning: removing noise, missing values, and error values in the data. Data conversion: converting unstructured data into structured data. Data standardization: standardizing data from different sources to ensure the consistency and comparability of the data. Data fusion: fusing multi-source data to form a comprehensive geological data set. The geological data information includes geological investigation reports, exploration data, remote sensing data, laboratory analysis data, covering information in multiple aspects such as strata, rocks, minerals, structures, geochemistry, and geophysics.

[0061] S102. Use natural language processing technology to obtain geological entity information, attribute information of geological entities, relationship information between geological entities, and relationship information between geological entities and geological entity attributes in the geological information database, and construct a geological knowledge graph.

[0062] Specifically, using natural language processing technology, such as text mining, information extraction, semantic analysis and other methods, geological entities are identified from the text data in the geological information database, such as specific concepts or objects such as ore deposits, ore bodies, rock types, minerals, geological structures, and stratigraphic units, and various attributes of the geological entities are obtained, such as physical properties of rocks such as color, hardness, and density, chemical composition and crystal structure characteristics of minerals, and information such as the scale, grade, and shape of ore deposits. Determine the associations between geological entities and between geological entities and their attributes. For example, certain ore bodies may be related to specific magmatic activities or sedimentary environments. Store the obtained geological entities, attributes, and relationships in a graph database to form a geological knowledge graph.

[0063] S103. Use the geological knowledge graph to train and optimize a prediction model to obtain an ore-forming prediction model.

[0064] Specifically, the features in the geological knowledge graph are used as training data to train the prediction model. During the process, labels can be used to clarify the learning objectives of the model. For example, the attributes and relationships of geological entities can be used as input features, and whether there is mineralization can be used as a label. The model can learn whether the attributes and relationships of these geological entities are mineralized. During the optimization process, by continuously adjusting the model parameters and optimizing the model structure, the fitting ability of the model to the mineralization law is improved.

[0065] S104. Obtain the geological information data of the area to be predicted, use the mineralization prediction model to obtain the prediction result of the mineralization potential area, and generate a mineralization analysis report.

[0066] Specifically, when it is necessary to evaluate the mineralization potential of a specific area, obtain the geological information data of the area and perform necessary preprocessing to make it meet the input requirements of the mineralization prediction model. Use the trained mineralization prediction model to analyze the data to obtain the prediction result of the mineralization potential area, and generate a mineralization analysis report according to the prediction result.

[0067] A method for constructing a knowledge graph and intelligent prospecting prediction based on multi-source heterogeneous geological data provided by this application preprocesses the obtained geological data information to form a geological information database, which solves the problems of inconsistent geological data formats and complexity, enables the originally diverse, high-dimensional and complex geological data to be uniformly stored and managed, reduces data complexity, and facilitates subsequent analysis. Using natural language processing technology to identify geological entities and their mutual relationships in the geological information database to construct a geological knowledge graph improves the efficiency of geologists in obtaining and using geological information, makes the internal connections of geological information clearer, and improves the accuracy and efficiency of prospecting prediction. Training and optimizing the mineralization prediction model based on the geological knowledge graph realizes fast and accurate mineralization potential assessment and generation of mineralization analysis reports, solves the problems of low efficiency and inaccurate content in mineralization potential assessment and report generation in traditional methods, provides more comprehensive and powerful support for exploration decision-making, and thus improves the practicality of geological prospecting as a whole.

[0068] In one embodiment, using natural language processing technology to identify geological entities, the attributes of geological entities and their mutual relationships in the geological information database to construct a geological knowledge graph includes:

[0069] S201. Using natural language processing technology, identify geological entities from the geological information database to form a mineral information library, where the geological entities include: ore deposits, ore bodies, rocks, minerals, structures, strata and geological events.

[0070] Specifically, natural language processing technology is used to store the identified information on mineral deposits, ore bodies, rocks, minerals, structures, strata and geological events in a special database to form a mineral information library, providing a data basis for subsequent analysis and processing.

[0071] S202, classify each mineral into the corresponding categories of metal mines, non-metal mines and energy mines according to the chemical composition, physical properties, genesis type, occurrence state, deposit scale and use of each mineral in the mineral information database, and classify the corresponding mineral species of each mineral to form mineral type information.

[0072] Specifically, key features of minerals, such as chemical composition, physical properties, genesis type, etc., are obtained from the mineral information database, and then classification rules are formulated based on these features. For example, metal ores are mainly composed of metal elements and are used to extract metal elements or compounds; non-metallic ores use their physical or chemical properties, but are not mainly used to extract metals; energy ores are mainly used to provide energy. Through these rules, minerals are classified and their corresponding mineral species are recorded, ultimately forming mineral type information.

[0073] S203, based on the knowledge engineering method in the field of geology, define the relationship types between geological entities in the mineral information database, including causal relationship, spatial relationship, and temporal relationship.

[0074] Among them, knowledge engineering is a technology that converts the knowledge and experience of human experts into knowledge that can be processed and applied by computers. Through the knowledge engineering method, various types of relationships between geological entities can be defined, so that computers can understand and infer the internal connections of geological phenomena. Causal relationship: refers to the causal relationship between geological entities, that is, the formation of one geological entity is caused by another geological entity or geological event. For example, the formation of an ore body may be related to geological events such as magma intrusion, sedimentation or metamorphism. Spatial relationship: refers to the spatial location and distribution characteristics of geological entities, such as the contact relationship between the ore body and the surrounding rock, the occurrence of the rock, the distribution of the structure, etc. For example, ore bodies are usually distributed along tectonic fracture zones. By analyzing spatial relationships, the extension direction and possible occurrence location of the ore body can be predicted. Temporal relationship: refers to the occurrence sequence and evolution process of geological entities in the geological history period, such as the deposition sequence of strata, the stage of tectonic movement, the time series of mineralization events, etc. For example, by analyzing the temporal relationship, it can be determined in which geological period the ore body was formed, thereby inferring the changes in the mineralization environment and mineralization conditions.

[0075] S204, obtaining mineralization characteristic information according to the relationship type between geological entities.

[0076] Among them, by analyzing the genetic, spatial, and temporal relationships between geological entities, metallogenic characteristic information is constructed, including elements such as the shape, occurrence, ore type, mineral assemblage, and wall rock alteration of the ore body. The metallogenic characteristic information reflects the formation conditions and characteristics of minerals.

[0077] S205, associate the minerals in the mineral type information with the metallogenic characteristic information to form metallogenic-related samples.

[0078] Specifically, metallogenic-related samples refer to a data set with clear ore type information and metallogenic characteristic information. Each sample represents a specific mineral instance and contains information on aspects such as the classification of minerals, metallogenic characteristics, and geological background.

[0079] S206, calculate the correlation between the corresponding minerals and the metallogenic characteristic information in the metallogenic-related samples, and summarize to obtain metallogenic information.

[0080] Among them, statistical methods or machine learning algorithms are used to calculate the correlation between mineral properties and metallogenic characteristics, which can quantify the influence degree of different properties on the metallogenic potential. For example, calculate the correlation between the iron content of minerals and the scale of the ore body, or the correlation between the hardness of minerals and the wall rock alteration characteristics. Through these calculations, it can be determined which mineral properties have a significant impact on the metallogenic characteristics. Samples with a correlation greater than or equal to the set threshold are judged as strongly metallogenic-related samples, and these strongly metallogenic-related samples are summarized to form metallogenic information.

[0081] S207, use the mineral type information and related geological entities as the first-layer nodes, and the metallogenic characteristics and association relationships as the second-layer nodes, and use the metallogenic characteristics and association relationships as the edges connecting the first-layer nodes to construct a geological knowledge graph.

[0082] Among them, by defining geological entities as the first-layer nodes, metallogenic characteristics and association relationships as the second-layer nodes, and connecting them through association relationships. For example, a certain ore body node can be connected to a geological event node describing its origin or a tectonic node describing its spatial location, and a graph database is used to store these nodes and relationships to form a geological knowledge graph. In this way, geological entities and their attributes and relationships can be intuitively represented, providing support for geological research and resource exploration.

[0083] In one embodiment, calculating the correlation between the corresponding minerals and the metallogenic characteristic information in the metallogenic-related samples and summarizing to obtain metallogenic information includes:

[0084] S301, calculate the correlation between the corresponding minerals and the metallogenic characteristic data in the metallogenic-related samples, and the calculation formula is as follows:

[0085]

[0086] Among them, P(Y = 1|X) represents the probability that a sample belongs to the strongly metallogenic-related category given X, e represents the natural constant, N represents the total number of samples, αi represents the Lagrange multiplier used to optimize the dual variable, wi represents the weight of the i-th sample, yi represents the label of the i-th sample, and K(Zx, Zxi) = ((Zx) t (Zxi)+c) d represents the kernel function, where Z represents the feature weight matrix used to adjust the weight of each metallogenic feature, c represents the constant term used to adjust the offset of the kernel function, x and xi represent the feature vectors of minerals, including chemical composition, physical properties, genetic type, occurrence state, deposit scale, and use, and d represents the complexity of the polynomial.

[0087] Specifically, this formula calculates the similarity between samples through the kernel function K(Zx, Zxi), and combines the Lagrange multiplier αi and the sample label yi to optimize the model, and finally calculates the probability that a sample belongs to the strongly metallogenic-related category, which can effectively evaluate the correlation between mineral characteristics and metallogenic potential and provide a scientific basis for metallogenic prediction.

[0088] S302. Judge the metallogenic-related samples with P(Y = 1|X) greater than or equal to the set threshold as strongly metallogenic-related samples, and summarize the strongly metallogenic-related samples to form metallogenic information.

[0089] Among them, the threshold can be adjusted according to the needs of actual applications. For example, if you want to increase the metallogenic correlation of samples, you can increase the threshold; if you want to reduce the metallogenic correlation of samples, you can lower the threshold. Summarize all strongly metallogenic-related samples to provide support for metallogenic prediction.

[0090] In one embodiment, the geological data information includes:

[0091] Basic geological data, chemical composition analysis data, physical exploration data, remote sensing image data, and knowledge text data;

[0092] The basic geological data includes geological maps, mineral resource distribution maps, and geological structure maps in the geological database;

[0093] The chemical composition analysis data includes the chemical composition analysis data of rock, soil, and water samples in the chemical database;

[0094] The physical exploration data includes gravity, magnetic, and electrical physical exploration data;

[0095] The remote sensing image data includes satellite remote sensing images and aerial photography images;

[0096] The knowledge text data includes geological literature, investigation reports, and geological knowledge bases.

[0097] Among them, geological maps record the distribution and interrelationships of geological units such as strata, rock masses, and structures. Geological maps contain information such as the age, lithology, thickness, and attitude of strata. Mineral resource distribution maps show the geographical locations, types, and scales of mineral resources. Geological structure maps depict the shapes and distributions of geological structures, such as faults, folds, and joints. Chemical composition analysis of rock samples analyzes the contents of major elements such as Si, Al, Fe, etc. and trace elements such as Cu, Zn, Au, etc. in rocks. Gravity exploration data measure the variations in the Earth's gravity field, which can be used to identify geological bodies with large underground density differences, such as magmatic rock bodies and ore bodies. Chemical composition analysis of soil samples analyzes the chemical components in soils, which can be used to identify soil geochemical anomalies and indicate potential mineralized areas. Chemical composition analysis of water samples analyzes the chemical components in water bodies, which can be used to identify hydrochemical anomalies and indicate the enrichment of ore-forming elements in groundwater. Magnetic exploration data measure the variations in the Earth's magnetic field, which can be used to identify the distributions of underground magnetic minerals, such as magnetite and ilmenite. Electrical method exploration data measure the electrical properties of underground rocks, such as resistivity and polarization rate, which can be used to identify the distributions of ore bodies and groundwater. Satellite remote sensing images, such as multi-spectral and thermal infrared images provided by satellites, can be used to identify surface vegetation, soil, rock types, and geological structures. Aerial photography images obtain high-resolution surface images through aerial photography, which can be used to detail the topographic features and detailed characteristics of geological bodies on the surface. Geological literature, including academic papers, monographs, conference papers, etc., records the latest research results and theories in geological research. Investigation reports, including geological investigation reports, mineral resource assessment reports, etc., provide detailed data and analysis results of field investigations. Geological knowledge bases integrate professional knowledge and experience in the geological field, including ore-forming theories, geological models, exploration methods, etc. These data can help identify ore-forming parent rocks and enrichment areas of ore-forming elements. The referenced geological data information is relatively diverse. Through comprehensive analysis and processing, it can provide comprehensive and accurate data support for geological prospecting.

[0098] In one embodiment, the prediction model is constructed based on the following method:

[0099] S401, select deep learning algorithms of convolutional neural networks, graph neural networks, recurrent neural networks and their variants to construct the prediction model. Among them, convolutional neural networks are used to process remotely sensed images and geospatial correlation geological data in the geophysical field, graph neural networks are used to process graph-structured data of geological knowledge graphs, and recurrent neural networks and their variants are used to analyze time series data of the temporal evolution of geological events.

[0100] Among them, convolutional neural networks are mainly used to process remote sensing images and geophysical field data, which have spatial correlation. Through convolutional layers, activation functions, pooling layers, and fully connected layers, the features of images can be extracted to identify geological structures and abnormal areas. Graph neural networks are mainly used to process the graph-structured data of geological knowledge graphs. Through graph convolution operations, the relationships between nodes in the graph can be captured, the features of nodes can be extracted, and entity classification and link prediction can be performed. Recurrent neural networks and their variants are mainly used to analyze the time series data of geological events, capture time-dependent relationships, and predict the evolution of geological events. Using the corresponding modeling methods according to the corresponding data makes full use of the characteristics of different types of geological data and improves the accuracy and efficiency of ore-forming prediction.

[0101] S402. Obtain the ore-forming instance data in the geological knowledge graph and train the prediction model, including using cross-validation techniques to evaluate the performance of the prediction model:

[0102] Randomly divide the ore-forming instance data into a training set, a cross-validation set, and a test set according to a 7:3 ratio, where the training set and the cross-validation set account for 70%, and the test set accounts for 30%. The training set is used to train the prediction model, the cross-validation set is used to optimize and adjust the parameters of the prediction model, and the test set is used to evaluate the performance of the prediction model. The performance indicators include accuracy, recall, and F1 score, and the calculation formulas are as follows:

[0103]

[0104] Among them, Precision represents the proportion of samples that are actually positive among the samples predicted as positive by the model, Recall represents the proportion of samples that are actually positive and are correctly predicted as positive by the model, TP represents the number of samples correctly predicted as positive by the model, FP represents the number of samples wrongly predicted as positive by the model, and FN represents the number of samples wrongly predicted as negative by the model.

[0105] Specifically, the training set: is used to train the prediction model to help the model learn ore-forming characteristics and patterns. The cross-validation set: divides the training set into K subsets of equal size. Each time, K - 1 subsets are selected as the training set, and the remaining 1 subset is used as the validation set. Perform K times of training and validation, and finally take the average value of the K evaluation results as the evaluation index of the model performance, which is used to evaluate the performance of the model on unseen data, adjust the model parameters, and prevent overfitting. The test set: is used to finally evaluate the performance of the model to ensure the generalization ability of the model on new data.

[0106] S403. Calculate the F1 score of each prediction model, select the prediction model with the best F1 score, and obtain the ore-forming prediction model.

[0107] Among them, the F1-score is the harmonic mean of precision and recall, which can comprehensively consider the precision and recall of the model and is a relatively comprehensive performance evaluation index. By selecting the model with the optimal F1-score, it can be ensured that the model has a high precision when predicting the metallogenic potential area and will not miss too many positive samples, thereby improving the accuracy and reliability of metallogenic prediction.

[0108] In one embodiment, the metallogenic prediction model is trained and optimized using metallogenic examples in the geological knowledge graph, and the obtained metallogenic prediction model includes:

[0109] S501, generating hyperparameter combinations using random search and grid search methods.

[0110] Specifically, the selection of hyperparameters directly affects the complexity, learning rate, model capacity, generalization ability, and convergence speed of the model. A reasonable selection of hyperparameters can improve the performance of the model. Grid search ensures finding the global optimal solution by traversing multiple hyperparameter combinations; random search randomly samples within the predefined hyperparameter space to find a combination close to the optimal with a lower computational cost, which is suitable for high-dimensional parameter spaces. In actual applications, the method for generating hyperparameter combinations can be selected according to specific circumstances.

[0111] S502, evaluating the performance of the metallogenic prediction model based on different hyperparameter combinations using cross-validation techniques, and screening out the metallogenic prediction model with the optimal performance.

[0112] Specifically, cross-validation techniques are used to evaluate the performance of the metallogenic prediction model under different hyperparameter combinations. By dividing the dataset into multiple small parts, then using one part for testing and the rest for training, and repeating this process multiple times, the performance of the model is evaluated by synthesizing the results of multiple times. For example, according to the five-fold cross-validation theory, 4 sample sets are used for training and 1 sample set is used for validation. This process is repeated 5 times, and each sequence is used as a validation set respectively. Finally, the training set and validation set obtained from the five-fold cross-validation are merged to evaluate the performance of the model. Finally, the model with the optimal performance is screened out.

[0113] In one embodiment, geological information data of the area to be predicted is obtained, and the prediction result of the metallogenic potential area is obtained using the metallogenic prediction model, and a metallogenic analysis report is generated, including:

[0114] S601, using Internet of Things technology, deploying sensors in the area to be predicted to obtain real-time geological data.

[0115] Among them, sensors are deployed in the area to be predicted, and real-time geological data is obtained by connecting the sensors to the Internet. Sensors are generally deployed at key locations in the area to be predicted to ensure full coverage of the monitoring area. For example, they are deployed on the surface above the ore body or in potential landslide areas, in surface cracks, in areas sensitive to groundwater level changes near the ore body, and in the rock mass inside or around the ore body.

[0116] S602, pre-processing the actual geological data to obtain standard test data.

[0117] Specifically, the actual geological data is preprocessed and cleaned, including: Noise removal: The actual geological data may contain measurement errors or outliers. These noise data can be identified and removed through statistical analysis, such as standard deviation and interquartile range. Processing missing values: There may be missing values ​​in the data, which can be processed by filling or deleting records with missing values, such as deleting records with more missing values, and filling missing values ​​with mean, median or mode for data with few missing values.

[0118] S603, based on the standard test data, generate a mineral analysis report using the mineralization prediction model. The mineralization analysis report includes the type and distribution of mineral resources, mineralization conditions and characteristics, resource quantity and potential assessment, spatial regularity and scale of potential distribution, exploration risk assessment and recommendations, and exploration paths and plans.

[0119] Specifically, based on the standard test data, through the analysis of the mineralization prediction model, comprehensive geological information is provided, including: the types and distribution of mineral resources include the types of mineral resources that may exist in the predicted area and their distribution areas, such as metal mines, non-metallic mines, and energy mines. Mineralization conditions and characteristics include geologic background, lithologic characteristics, geochemical anomalies, and mineralization characteristics, such as the shape, scale, and occurrence of ore bodies. Resource quantity and potential assessment include the reserves and potential economic value of mineral resources. The spatial regularity and scale of potential distribution are used to help determine the target area for prospecting. Exploration risk assessment and recommendations include the risks that may be encountered during the exploration process, such as geological disasters and technical difficulties, and provide corresponding recommendations and countermeasures. Exploration paths and plans, including specific exploration paths and plans, including exploration starting points, end points, alternative paths, drilling locations, sampling point layouts, and exploration methods, provide guidance for actual exploration work.

[0120] In one embodiment, geological information data of the area to be predicted is obtained, and the prediction results of the mineralization potential area are obtained using the mineralization prediction model, and a mineralization analysis report is generated, including:

[0121] S701, real-time acquisition of actual geological data, combining standard test data with historical geological data information, and real-time update of geological database and mineralization prediction model.

[0122] Specifically, the real-time geological data from sensors in the area to be measured is obtained, and the real-time updated geological data can reflect the latest geological changes. Combining the historical geological data information to update the geological database and the metallogenic prediction model enables the metallogenic prediction model to timely capture the dynamic changes in the metallogenic potential areas.

[0123] S702. Based on the real-time updated geological data, using the metallogenic prediction model, obtain the prediction results of the change trend of the metallogenic potential areas. Among them, the prediction results include: potential increase or decrease information of mineral resources, improvement or deterioration information of metallogenic conditions, and expansion or contraction information of metallogenic potential areas.

[0124] Specifically, use the metallogenic prediction model to analyze the updated data over a period of time. By comparing the change trends of the metallogenic potential areas during this period, obtain the prediction results of the change trends. These information include: potential increase or decrease information of mineral resources: Through model prediction, the potential increase or decrease of mineral resources can be evaluated. For example, the resource volume in some areas may increase or decrease due to geological activities or mining activities. Improvement or deterioration information of metallogenic conditions: The model can predict the changes in metallogenic conditions, such as the sedimentary environment of the strata, the intensity of tectonic activities, the frequency of magmatic activities, etc. Expansion or contraction information of metallogenic potential areas: Through three-dimensional geological modeling and spatial analysis methods, predict the changes in the shape, location, and scale of the metallogenic potential areas. For example, some areas may expand due to geological activities, while other areas may contract due to resource mining. Obtaining multi-faceted prediction results based on the updated geological data makes the prediction information forward-looking and comprehensive.

[0125] S703. According to the prediction results of the change trend of the metallogenic potential areas, optimize the exploration path and plan.

[0126] Among them, the optimization of the exploration path includes: adjusting the exploration path according to the change trend of the metallogenic potential areas. For example, if it is predicted that the metallogenic potential in a certain area increases, the exploration points in this area can be increased; if the metallogenic potential in a certain area decreases, the exploration points can be reduced or adjusted. The optimization of the exploration plan includes: adjusting the exploration methods and strategies according to the improvement or deterioration of the metallogenic conditions. For example, if the metallogenic conditions improve, more efficient exploration technologies can be adopted; if the metallogenic conditions deteriorate, the exploration depth or methods can be adjusted to adapt to the new geological environment.

[0127] Generally speaking, a method for constructing a knowledge graph and intelligent ore prospecting prediction based on multi-source heterogeneous geological data provided by the present application obtains and preprocesses geological data information, extracts geological entities and their attributes and relationships by using natural language processing technology, and constructs a geological knowledge graph. This knowledge graph not only covers geological entities such as ore deposits, ore bodies, rocks, and minerals, but also includes multi-dimensional relationship information such as genetic relationships, spatial relationships, and temporal relationships, providing a rich feature and knowledge basis for ore-forming prediction and solving the problems of difficult multi-source data fusion and information silos. Through deep learning algorithms, such as convolutional neural networks to process remote sensing images and geophysical field data, graph neural networks to process the graph structure data of the geological knowledge graph, and recurrent neural networks to analyze the time series data of geological events, while using appropriate ore-forming prediction models according to the corresponding data, these models are trained and optimized. The cross-validation technique is used to evaluate the performance of these trained and optimized models, ensuring the accuracy and reliability of the models and solving the problems of low prediction model accuracy and poor generalization ability. In practical applications, the Internet of Things technology is used to deploy sensors to obtain the actual geological data of the area to be predicted, combined with standard test data and historical geological data, to update the geological database and ore-forming prediction model in real time. By integrating geological data in real time, the problems of scattered data sources and difficult integration are solved. Based on the real-time updated geological data, an ore-forming analysis report is generated, including the types and distributions of mineral resources, the characteristics of ore-forming conditions, the evaluation of resource volume and potential, the spatial law scale of potential distribution, the evaluation and suggestions of exploration risks, and the exploration path and plan. At the same time, the change trend of the ore-forming potential area is predicted, including the potential increase or decrease of mineral resources, the improvement or deterioration of ore-forming conditions, and the expansion or contraction of the ore-forming potential area, so as to optimize the exploration path and plan. By adopting dynamic path optimization technology, the exploration path is adjusted according to the real-time updated geographical and environmental information, ensuring the efficiency and safety of exploration activities and solving the problem of fixed exploration path and difficulty in adjusting according to real-time data. Through the integration of multi-source heterogeneous geological data and the application of deep learning technology, the accuracy and efficiency of ore-forming prediction are significantly improved, providing a scientific, intelligent, efficient, comprehensive, and forward-looking intelligent solution for geological exploration and mineral resource development.

[0128] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, there is no strict order limit for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0129] Based on the same inventive concept, an embodiment of the present application also provides a knowledge graph construction and intelligent prospecting prediction device based on multi-source heterogeneous geological data for implementing the above-mentioned knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data. The implementation solutions provided by this device to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more device embodiments provided below can refer to the limitations on a knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data in the above text, and will not be repeated here.

[0130] In an exemplary embodiment, as Figure 2 shown, a knowledge graph construction and intelligent prospecting prediction device based on multi-source heterogeneous geological data is provided, including:

[0131] A data acquisition and preprocessing module 101, configured to obtain geological data information and preprocess the geological data information to form a geological information database;

[0132] A knowledge graph construction module 102, configured to use natural language processing technology to obtain geological entity information, attribute information of geological entities, relationship information between geological entities, and relationship information between geological entities and geological entity attributes in the geological information database, and construct a geological knowledge graph;

[0133] An intelligent prediction model construction module 103, configured to use the geological knowledge graph to train and optimize a prediction model to obtain a mineralization prediction model;

[0134] A mineralization prediction and analysis report generation module 104, configured to obtain geological information data of an area to be predicted, use the mineralization prediction model to obtain a prediction result of a mineralization potential area, and generate a mineralization analysis report.

[0135] In one embodiment, the data acquisition and preprocessing module 101 includes:

[0136] The acquisition unit is used to obtain basic geological data, including geological maps, mineral resource distribution maps, and geological structure maps in the geological database; chemical composition analysis data includes chemical composition analysis data of rock, soil, and water samples in the chemical database; physical exploration data includes gravity, magnetic, and electrical physical exploration data; remote sensing image data includes satellite remote sensing images and aerial photography images; knowledge text data includes geological literature, investigation reports, and geological knowledge bases.

[0137] The knowledge graph construction module 102 includes:

[0138] The metallogenic information analysis unit is used to:

[0139] Using natural language processing technology, identify geological entities from the geological information database to form a mineral information library, where geological entities include: ore deposits, ore bodies, rocks, minerals, structures, strata, and geological events;

[0140] According to the chemical composition, physical properties, genetic types, occurrence states, ore deposit scales, and uses of each mineral in the mineral information library, classify each mineral into the corresponding categories of metal ores, non-metal ores, and energy ores, and classify the ore types corresponding to each mineral to form mineral type information;

[0141] Based on the knowledge engineering method in the field of geology, define the relationship types between geological entities in the mineral information library, including genetic relationships, spatial relationships, and temporal relationships;

[0142] Obtain metallogenic characteristic information according to the relationship types between geological entities;

[0143] Associate the minerals in the mineral type information with the metallogenic characteristic information to form metallogenic-related samples;

[0144] Calculate the correlation between the corresponding minerals and the metallogenic characteristic information in the metallogenic-related samples, and summarize to obtain metallogenic information;

[0145] The correlation calculation unit is used to calculate the correlation between the corresponding minerals and the metallogenic characteristic data in the metallogenic-related samples, and judge the metallogenic-related samples with a correlation greater than or equal to the set threshold as strongly metallogenic-related samples.

[0146] The intelligent prediction model construction module 103 includes:

[0147] The model evaluation unit is used to evaluate the performance of the prediction model using cross-validation technology;

[0148] The model optimization unit is used to generate hyperparameter combinations using random search and grid search methods, evaluate the performance of the metallogenic prediction models based on different hyperparameter combinations using cross-validation technology, and screen out the metallogenic prediction model with the best performance.

[0149] The ore-forming prediction and analysis report generation module 104 includes:

[0150] A change prediction unit, configured to obtain real-time geological data, combine standard data to be measured and historical geological data information, and update the geological database and the ore-forming prediction model in real time; based on the real-time updated geological data, use the ore-forming prediction model to obtain the prediction result of the change trend of the ore-forming potential area, where the prediction result includes: potential increase or decrease information of mineral resources, improvement or deterioration information of ore-forming conditions, and expansion or contraction information of the ore-forming potential area; optimize the exploration path and plan according to the prediction result of the change trend of the ore-forming potential area.

[0151] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of a method for constructing a knowledge graph and intelligent ore prospecting prediction based on multi-source heterogeneous geological data as described above are implemented.

[0152] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0153] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can refer to the partial descriptions of the method embodiments. The device embodiments described above are only illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0154] The above embodiments only represent several implementation manners of the embodiments of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several deformations and improvements can be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A knowledge graph construction and intelligent prospecting prediction method based on multi-source heterogeneous geological data, characterized in that: The method comprises: Acquiring geological data information, and preprocessing the geological data information to form a geological information database; Using natural language processing technology to obtain geological entity information, attribute information of geological entities, relationship information between geological entities, and relationship information between geological entities and attributes of geological entities in the geological information database, and construct a geological knowledge map; Using the geological knowledge graph to train and optimize the prediction model to obtain a mineralization prediction model; Obtain geological information data of the area to be predicted, use the mineralization prediction model to obtain the prediction results of the mineralization potential area, and generate a mineralization analysis report.

2. The method according to claim 1, characterized in that The method of using natural language processing technology to identify geological entities, attributes of geological entities and their mutual relationships in the geological information database and constructing a geological knowledge map includes: Using natural language processing technology, geological entities are identified from the geological information database to form a mineral information database, wherein the geological entities include: ore deposits, ore bodies, rocks, minerals, structures, strata and geological events; According to the chemical composition, physical properties, genesis type, occurrence state, deposit scale and use of each of the minerals in the mineral information database, each of the minerals is classified into corresponding categories of metal minerals, non-metallic minerals and energy minerals, and the mineral species corresponding to each of the minerals are classified to form mineral category information; Based on the knowledge engineering method in the field of geology, define the relationship types between geological entities in the mineral information database, including genetic relationship, spatial relationship, and temporal relationship; Obtaining mineralization characteristic information according to the relationship type between the geological entities; Associating the minerals in the mineral resource type information with the metallogenic characteristic information to form a metallogenic related sample; Calculating the correlation between the corresponding minerals in the mineralization-related samples and the mineralization characteristic information, and summarizing the mineralization information; The mineral type information and related geological entities are used as the first-layer nodes, the metallogenic characteristics and association relationships are used as the second-layer nodes, and the metallogenic characteristics and association relationships are used as the edges connecting the first-layer nodes to construct a geological knowledge graph.

3. The method according to claim 2, characterized in that The calculating the correlation between the corresponding minerals in the mineralization-related samples and the mineralization characteristic information and summarizing the mineralization information includes: The correlation between the corresponding minerals and the mineralization characteristic data in the mineralization-related samples is calculated using the following formula: Where P(Y=1|X) represents the probability that a sample belongs to a strongly related category of mineralization under the condition of given X, e represents a natural constant, N represents the total number of samples, αi represents the Lagrange multiplier used to optimize the dual variable, wi represents the weight of the i-th sample, yi represents the label of the i-th sample, K(Zx,Zxi)=((Zx) t (Zxi)+c) d represents the kernel function, where W represents the characteristic weight matrix, which is used to adjust the weight of each mineralization feature, c represents the constant term, which is used to adjust the offset of the kernel function, x and xi represent the characteristic vectors of the mineral, including chemical composition, physical properties, genetic type, occurrence state, deposit scale and use, and d represents the complexity of the polynomial; P(Y ＝ 1|X) is greater than or equal to the set threshold value. The mineralization-related samples are judged as strongly mineralization-related samples, and the strongly mineralization-related samples are summarized to form mineralization information.

4. The method according to claim 3, characterized in that The geological data information includes: Basic geological data, chemical composition analysis data, physical exploration data, remote sensing image data and knowledge text data; The basic geological data include geological maps, mineral resource distribution maps and geological structure maps in the geological database; The chemical composition analysis data include chemical composition analysis data of rock, soil and water samples in the chemical database; The physical exploration data include gravity, magnetic and electrical physical exploration data; The remote sensing image data includes satellite remote sensing images and aerial photography images; The knowledge text data includes geological literature, investigation reports, and geological knowledge base.

5. The method according to any one of claims 1 to 4, characterized in that The prediction model is constructed based on the following method: The prediction model is constructed by using deep learning algorithms of convolutional neural networks, graph neural networks, recurrent neural networks and their variants, wherein the convolutional neural network is used to process the associated geological data of remote sensing images and geophysical field data space, the graph neural network is used to process the graph structure data of geological knowledge graphs, and the recurrent neural network and its variants are used to analyze the time series data of the temporal evolution of geological events; Obtaining mineralization example data in the geological knowledge map and training the prediction model, including evaluating the prediction model performance using cross-validation techniques: The mineralization case data is randomly divided into training set, cross-validation set and test set in a ratio of 7:3, of which the training set and cross-validation set account for 70% and the test set accounts for 30%. The training set is used to train the prediction model, the cross-validation set is used to optimize the parameters of the prediction model, and the test set is used to evaluate the performance of the prediction model. The performance indicators include accuracy, recall rate and F1 score, and the calculation formula is as follows: Among them, precision represents the proportion of samples predicted by the model as positive that are actually positive, recall represents the proportion of samples actually positive that are correctly predicted by the model as positive, TP represents the number of samples correctly predicted by the model as positive, FP represents the number of samples incorrectly predicted as positive by the model, and FN represents the number of samples incorrectly predicted as negative by the model. The F1 score of each prediction model is calculated, and the prediction model with the best F1 score is selected to obtain the mineralization prediction model.

6. The method according to claim 5, characterized in that The method of training and optimizing the prediction model by using the mineralization examples in the geological knowledge map to obtain the mineralization prediction model includes: Generate hyperparameter combinations using random search and grid search methods; The cross-validation technology is used to evaluate the performance of the mineralization prediction model based on different hyperparameter combinations, and the mineralization prediction model with the best performance is screened out.

7. The method according to claim 1, characterized in that The method of obtaining geological information data of the area to be predicted, obtaining prediction results of the mineralization potential area by using the mineralization prediction model, and generating a mineralization analysis report includes: Using IoT technology, deploy sensors in the area to be predicted to obtain real-time geological data; Preprocessing the actual geological data to obtain standard data to be measured; Based on the standard test data, a mineralization prediction model is used to generate a mineralization analysis report, which includes mineral resource types and distribution, mineralization conditions and characteristics, resource quantity and potential assessment, spatial regularity and scale of potential distribution, exploration risk assessment and recommendations, and exploration paths and plans.

8. The method according to claim 7, characterized in that The method of obtaining geological information data of the area to be predicted, obtaining prediction results of the mineralization potential area by using the mineralization prediction model, and generating a mineralization analysis report includes: Acquire the actual geological data in real time, combine the standard data to be measured with historical geological data information, and update the geological database and mineralization prediction model in real time; Based on real-time updated geological data, using the mineralization prediction model, the change trend prediction results of the mineralization potential area are obtained, wherein the prediction results include: potential increase or decrease information of mineral resources, improvement or deterioration information of mineralization conditions and expansion or contraction information of mineralization potential area; According to the prediction results of the changing trend of the mineralization potential area, the exploration path and plan are optimized.

9. A knowledge graph construction and intelligent prospecting prediction device based on multi-source heterogeneous geological data, characterized in that: The device comprises: A data acquisition and preprocessing module is used to acquire geological data information and preprocess the geological data information to form a geological information database; A knowledge graph construction module is used to identify geological entities, attributes of geological entities and their mutual relationships in the geological information database using natural language processing technology to construct a geological knowledge graph; An intelligent prediction model building module is used to train and optimize the prediction model using the geological knowledge map to obtain a mineralization prediction model; The mineralization prediction and analysis report generation module is used to obtain geological information data of the area to be predicted, use the mineralization prediction model to obtain the prediction results of the mineralization potential area, and generate a mineralization analysis report.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images