Lithium ore exploration method and system

Abstract

The invention discloses a lithium ore exploration method and system, and belongs to the technical field of new energy lithium ore exploration. The system comprises a multi-source data fusion module, a machine learning prediction module, a three-dimensional dynamic modeling module and a real-time verification feedback module. By integrating remote sensing, geophysics, geochemistry and geologic structure data and combining a deep reinforcement learning algorithm and a three-dimensional dynamic modeling technology, efficient delineation of a lithium ore metallogenic favorable target area is achieved. When the system is applied to practical application of typical lithium metallogenic zones, the exploration accuracy is greatly improved, the exploration cost is remarkably reduced, the exploration period is remarkably shortened, and technical support is provided for large-scale development of hard rock type lithium ores. The method can be widely applied to the exploration of granite pegmatite type and salt lake type lithium ores, can drive the lithium resource detection amount to increase in the future, and assists the resource safety guarantee of a new energy industry chain.

Description

TECHNICAL FIELD

[0001] The present application relates to a lithium ore exploration method and system, belonging to the technical field of new energy lithium ore exploration. BACKGROUND

[0002] With the rapid development of new energy vehicles and energy storage systems, the demand for energy metals will continue to grow. New energy battery metal materials mainly include lithium, cobalt, nickel, manganese, chromium, etc. These metals play a key role in the production, storage and transmission of energy. Among them, lithium, as the lightest metal element, provides high energy density and is the basic element of positive electrode materials (such as lithium iron phosphate and ternary materials). Lithium metal, with its irreplaceable physical and chemical properties, has become the "white oil" of the new energy revolution, driving the popularization of electric vehicles, energy structure transformation and the restructuring of the international industrial pattern. In the future, with the maturation of solid-state battery technology and the improvement of recycling systems, the strategic value of lithium resources will continue to grow.

[0003] Hard rock lithium ore (lithium mica, lithium mica, etc.) and salt lake brine together constitute the basis for global lithium resource development, among which hard rock lithium extraction technology has high maturity and is suitable for rapid industrial production. Hard rock lithium extraction is currently an important pillar of lithium metal production (about 50%-60% of global production), and data in 2025 shows that Australian lithium mica mine is still the main source of global lithium raw materials, and its hard rock lithium extraction process (such as sulfuric acid roasting method) can quickly obtain high-purity lithium salt, supporting about 50% of battery-grade lithium compound production26. Lithium mica extraction projects in Sichuan, Jiangxi and other places in China have also improved production capacity through technological transformation. Lithium ore exploration is the core prerequisite for the sustainable development of hard rock lithium extraction industry. Current lithium ore exploration technology has improved the accuracy rate from 60% to 85% through innovation of mineralization theory (such as "five-story + basement" model), iteration of detection technology (30% improvement in electromagnetic method accuracy) and intelligent upgrading (40% improvement in AI target area screening efficiency). The breakthrough of China's lithium resource quantity ranking second in the world (adding more than 10 million tons), verifies the actual effect of technological development. However, in the future, it is still necessary to continue to deepen the integration of intelligent exploration and green technology, further improve the accuracy of exploration, and consolidate the technological advantage in the global lithium resource competition.

[0004] Traditional lithium ore exploration relies on a single data source (such as geophysical or geochemical data), lacks multi-source data collaborative analysis capability, resulting in low identification rate of concealed ore bodies (<60%) and high exploration cost. Existing models lack adaptability to complex mineralization background, making it difficult to accurately predict the distribution of deep lithium ore bodies. Artificial interpretation is inefficient and difficult to meet the needs of large-scale exploration. An intelligent analysis system integrating multi-dimensional data is needed to achieve rapid and high-precision lithium ore target prediction. Based on this, the present application is proposed to provide a method for improving the accuracy of lithium ore exploration. SUMMARY

[0005] (I) Technical problems to be solved

[0006] The technical problem to be solved by the present application is to solve the problem that traditional lithium ore exploration relies on a single data source, lacks multi-source data collaborative analysis capability, and results in low identification rate of concealed ore bodies and high exploration cost.

[0007] (II) Technical solutions

[0008] To solve the above technical problems, the present application provides a lithium ore exploration system, comprising a multi-source data fusion module, a machine learning prediction module, a three-dimensional dynamic modeling module, and a real-time verification feedback module.

[0009] The multi-source data fusion module integrates remote sensing spectral data, geophysical anomaly data, and geochemical element combination anomalies through a weighting algorithm.

[0010] The machine learning prediction module uses a deep reinforcement learning algorithm to dynamically adjust the data fusion weights in different geological scenarios and outputs a mineralization probability heat map, with an accuracy error of less than 8%.

[0011] The three-dimensional dynamic modeling module generates a three-dimensional ore body model based on an implicit modeling engine and supports real-time drilling data-driven dynamic updating of ore body boundaries, with a model resolution of 1m.

[0012] The real-time verification feedback module includes a drilling data analysis unit that collects real-time core Li2O grade and ore body spatial coordinates, a model correction unit, and a target area priority ranking unit that dynamically adjusts the exploration target area grade based on the verification results, wherein the model correction unit updates the prediction model weights within 24 hours through a backpropagation algorithm.

[0013] Further, the multi-source data fusion module comprises:

[0014] A remote sensing data processing submodule for analyzing lithium mineralization alteration features in hyperspectral remote sensing images;

[0015] A geophysical data correlation submodule that analyzes gravity, magnetic, and electromagnetic data superimposed on geological structures using a spatial interpolation algorithm;

[0016] A geochemical anomaly extraction submodule that filters Li, Rb, Cs, and other key element combination anomalies based on principal component analysis.

[0017] Further, the remote sensing data processing submodule uses a convolutional neural network to extract spectral features and combines an attention mechanism to enhance the ability to identify mineralization alteration areas.

[0018] Further, the machine learning prediction module uses a deep reinforcement learning algorithm, which specifically includes:

[0019] a dynamic weight distribution unit that automatically adjusts the multi-source data fusion weights according to the geological background;

[0020] a probability thermal map generation unit that outputs a lithium ore mineralization probability thermal map with an accuracy error of less than 8%;

[0021] an adaptive learning unit that iteratively optimizes model parameters in combination with historical exploration data and real-time drilling results.

[0022] Further, the dynamic weight distribution unit optimizes the weight strategy through a Q-learning algorithm, and preferentially allocates geophysical data weights in the fault zone area and preferentially allocates remote sensing data weights in the shallow overburden area.

[0023] Further, the three-dimensional dynamic modeling module comprises:

[0024] an implicit modeling engine that generates a three-dimensional ore body model based on geostatistics;

[0025] a real-time data driving unit that dynamically updates the model boundary and grade distribution through drilling data;

[0026] a visual interactive interface that supports three-dimensional model rotation, cross-section cutting, and automatic resource estimation.

[0027] Further, the implicit modeling engine supports multi-scale modeling, including a mine area level of 10km×10km with a resolution of 100m and an ore body level of 1km×1km with a resolution of 1m, and integrates a GIS platform to realize spatial data interaction.

[0028] Further, the drilling data analysis unit is linked with an automated drilling machine to upload core Li2O grade data to a cloud server in real time, triggering a model correction cycle of less than 24 hours.

[0029] The present application also provides a method for exploration using the above lithium ore exploration system, comprising the following steps:

[0030] Step S1: Collect hyperspectral remote sensing data with a spectral resolution of less than or equal to 5nm by a drone, and extract lithium mineralization alteration features;

[0031] Step S2: Spatially register the geophysical data and remote sensing data to generate a multi-dimensional fusion feature matrix;

[0032] Step S3: Input the fusion data into a deep reinforcement learning algorithm model to output a mineralization probability thermal map and automatically delineate target areas, wherein the probability threshold of Class A target areas is greater than 85%;

[0033] Step S4: Deploy automated drilling equipment according to the priority of the target area and upload Li2O grade data to a cloud server in real time.

[0034] More specifically, in step S3, when the target area verification result deviates from the prediction by more than 15%, the model retraining process is triggered, and the update cycle is less than or equal to 48 hours.

[0035] (III) Beneficial effects

[0036] The above technical solutions of the present application have the following advantages:

[0037] The present application provides a multi-source information fusion technology based on geological, geophysical, geochemical and remote sensing data, combined with machine learning algorithm and three-dimensional geological modeling, a lithium mine intelligent exploration system and method, for efficiently delineating lithium ore favorable target areas and improving exploration accuracy. The present application can be widely used in granite pegmatite type and salt lake type lithium mine exploration, and can drive the growth of proven lithium resources in the future, and help to ensure the safety of new energy industry chain resources.

[0038] In addition to the technical problems solved by the present application, the technical features of the technical solutions described above, and the advantages brought by these technical features, other technical features of the present application and the advantages brought by these technical features will be further described with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0040] Figure 1 The present application is a lithium mine exploration system schematic diagram. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0042] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0043] Example 1

[0044] This embodiment provides a lithium ore exploration system, such as Figure 1 As shown, it includes a multi-source data fusion module, a machine learning prediction module, a 3D dynamic modeling module, and a real-time verification feedback module. The multi-source data fusion module integrates remote sensing spectral data, geophysical anomaly data, and geochemical element combination anomalies using a weighted algorithm. The machine learning prediction module employs a deep reinforcement learning (DRL) algorithm to dynamically adjust the data fusion weights under different geological scenarios and outputs a mineralization probability heatmap with an accuracy error of less than 8%. The 3D dynamic modeling module generates a 3D orebody model based on an implicit modeling engine and supports real-time drilling data-driven dynamic updates of the orebody boundary, with a model resolution of up to 1 meter. The real-time verification feedback module includes a drilling data parsing unit, a model correction unit, and a target area priority ranking unit. The model correction unit updates the prediction model weights every 24 hours using a backpropagation algorithm. The drilling data parsing unit is linked with the automated drilling rig, uploading core Li2O grade data to the cloud server in real time, triggering a model correction cycle of ≤24 hours.

[0045] The core modules are described in further detail below:

[0046] 1. Multi-source data fusion module

[0047] Data input and preprocessing:

[0048] Integrating hyperspectral remote sensing (e.g., GF-5, spectral resolution 5nm), geophysical (CSAMT, gravity gradient), geochemical (Li elemental analysis of soil / rock samples), and geological structural (fault, fold) data, dimensional differences are eliminated through spatial registration (error <0.1 pixels) and standardization (Z-score normalization).

[0049] Feature fusion technology:

[0050] A weighted fusion algorithm is used, for example:

[0051] Remote sensing data weights (0.4): Distribution of altered minerals extracted based on CNN;

[0052] Geophysical data weight (0.3): based on resistivity anomaly correlation with gravity gradient;

[0053] Geochemical data weight (0.2): based on Li element anomaly intensity;

[0054] Tectonic data weight (0.1): based on fracture zone and spatial coupling degree of mineralization.

[0055] Output result: generate a multi-dimensional feature matrix with a spatial resolution of 10m x 10m for machine learning model input.

[0056] 2. Machine learning prediction module (model)

[0057] Model architecture:

[0058] Build a deep reinforcement learning (DRL) network, including:

[0059] State space: multi-dimensional feature matrix (remote sensing + geophysical + geochemical + tectonic);

[0060] Action space: mineralization probability prediction (0-100%);

[0061] Reward function: based on Li2O grade error of validation drill holes to adjust weight inversely.

[0062] Training process:

[0063] Input historical data (such as 12 known ore spots in a certain area), optimize decision path through Monte Carlo tree search (MCTS), and test set accuracy ≥92% after model convergence.

[0064] Dynamic optimization:

[0065] When the real-time drilling data and prediction deviation >15%, trigger model retraining, update cycle ≤48 hours.

[0066] 3. Three-dimensional dynamic modeling module

[0067] Modeling technology:

[0068] Use implicit modeling algorithm (such as RBF neural network) to integrate surface geological mapping, drilling data and geophysical inversion results to generate three-dimensional ore body model, supporting:

[0069] Ore body boundary automatic delineation: based on Li2O grade threshold (such as 0.5%) to divide mineralization zone;

[0070] Resource estimation: calculate resource quantity through block model, error <5%.

[0071] Visualization function: support any profile cutting in three-dimensional space, ore body volume rendering and exploration path planning, export standard industrial format (such as Leapfrog, Surpac).

[0072] 4. Real-time verification feedback module

[0073] Data closed loop mechanism:

[0074] After drilling data is uploaded to the system in real time, the following processes are triggered:

[0075] Data verification: eliminate outliers (such as Li2O grade > 5%);

[0076] Model correction: update DRL network weights through back propagation, and preferentially optimize high deviation areas (such as target areas with prediction error > 20%);

[0077] Target area reevaluation: dynamically adjust the target area grade according to new data (for example, the original B-class target area is upgraded to A-class after verification).

[0078] Feedback efficiency: from data entry to model update completion, the time consumption is ≤2 hours, and the target area correction result is pushed to the exploration team mobile terminal.

[0079] The lithium ore exploration system of the embodiment has the following advantages:

[0080] Full-process automation: realize unmanned operation from data acquisition to target delineation, and improve exploration efficiency by more than 50%; high-precision dynamic modeling: three-dimensional model spatial resolution reaches 1m level, supporting accurate positioning of deep (>1500m) concealed ore bodies; cross-scene compatibility: suitable for hard rock type (pegmatite), salt lake type (brine) and clay type lithium ore exploration.

[0081] The system described in the embodiment supports salt lake type lithium ore exploration, which predicts the brine lithium enrichment area by fusing underground water chemical data (Li + Concentration, pH value) and remote sensing thermal infrared data, and the prediction accuracy is ≥90%.

[0082] The system described in the embodiment is integrated with a GIS platform, supports online visual display of exploration results and automatic estimation of resource quantity, and the estimation error is <5%.

[0083] Embodiment 2

[0084] The embodiment is further optimized and refined on the basis of embodiment 1, and specifically, the multi-source data fusion module comprises:

[0085] Remote sensing data processing submodule, used for analyzing lithium mineralization alteration characteristics (such as lithium mica and eucolite spectral response) in hyperspectral remote sensing images;

[0086] The geophysical data correlation submodule superimposes and analyzes gravity, magnetic, and electromagnetic data with geological structures through a spatial interpolation algorithm.

[0087] The geochemical anomaly extraction submodule screens Li, Rb, Cs, and other key element combination anomalies based on principal component analysis (PCA).

[0088] More specifically, in this embodiment, the remote sensing data processing submodule uses a convolutional neural network (CNN) to extract spectral features and combines an attention mechanism (Attention) to enhance the ability to identify mineralized alteration areas.

[0089] Embodiment 3

[0090] This embodiment is a further optimization and refinement of the machine learning prediction module based on Embodiment 2, specifically,

[0091] The machine learning prediction module uses a deep reinforcement learning (DRL) algorithm, specifically including:

[0092] The dynamic weight allocation unit automatically adjusts the weight of multi-source data fusion according to the geological background (such as fault zones and rock mass contact zones);

[0093] The probability heat map generation unit outputs a lithium ore mineralization probability heat map with an accuracy error of <8%;

[0094] The adaptive learning unit iteratively optimizes model parameters in combination with historical exploration data and real-time drilling results.

[0095] More specifically, in this embodiment, the dynamic weight allocation unit optimizes the weight strategy through a Q-learning algorithm, preferentially allocating geophysical data weight (accounting for >60%) in fault zones and preferentially allocating remote sensing data weight (accounting for >70%) in shallow overburden zones.

[0096] Embodiment 4

[0097] This embodiment is a further optimization and refinement of the three-dimensional dynamic modeling module based on Embodiment 3, specifically,

[0098] The three-dimensional dynamic modeling module includes:

[0099] The implicit modeling engine generates a three-dimensional ore body model based on geostatistics (Kriging interpolation);

[0100] The real-time data driving unit dynamically updates the model boundary and grade distribution through drilling data;

[0101] The visual interactive interface supports three-dimensional model rotation, cross-section cutting, and automatic resource estimation.

[0102] More specifically, in this embodiment, the implicit modeling engine supports multi-scale modeling, including mine area level (10km x 10km, resolution 100m) and ore body level (1km x 1km, resolution 1m), and integrates a GIS platform to realize spatial data interaction.

[0103] Embodiment 5

[0104] This embodiment is a further optimization and refinement of the real-time verification feedback module based on embodiment 4, specifically,

[0105] The real-time verification feedback module comprises:

[0106] The drilling data analysis unit collects the core Li2O grade and ore body spatial coordinates in real time;

[0107] The model correction unit updates the prediction model weights through the back propagation algorithm;

[0108] The target area priority sorting unit dynamically adjusts the exploration target area level (A / B / C type) according to the verification results.

[0109] The system of any of the above embodiments supports salt lake type lithium mine exploration, and predicts the lithium enrichment area of brine by integrating groundwater chemical data (Li + concentration, pH value) and remote sensing thermal infrared data.

[0110] Embodiment 6

[0111] This embodiment provides a method for exploration using the lithium mine exploration system described in embodiment 1, which comprises the following steps:

[0112] Step S1: Collect hyperspectral remote sensing data (spectral resolution ≤5nm) by unmanned aerial vehicle, and extract lithium mineralization alteration characteristics;

[0113] Step S2: Spatially register geophysical data (CSAMT, gravity gradient) and remote sensing data to generate a multi-dimensional fusion feature matrix;

[0114] Step S3: Input the fusion data into the DRL model, output the metallogenic probability heat map and automatically delineate the target area, and the probability threshold of A type target area is >85%;

[0115] Step S4: Deploy automated drilling equipment according to the target area priority, and upload Li2O grade data to the cloud server in real time.

[0116] In step S3, when the target area verification result deviates from the prediction by more than 15%, the model retraining process is triggered, and the update cycle is ≤48 hours.

[0117] This embodiment covers the whole process of data collection, fusion, prediction and verification.

[0118] Embodiment 7

[0119] The embodiment provides a method for realizing real-time verification feedback based on the system in Embodiment 1, and the method comprises the following steps:

[0120] Step T1: analyzing core Li2O grade data uploaded by an automatic drilling machine, and removing outliers (such as outliers with a grade of >5%);

[0121] Step T2: updating DRL model parameters through a back propagation algorithm, and preferentially correcting high deviation areas (target areas with a prediction error of >20%);

[0122] Step T3: dynamically adjusting target area grades, upgrading B-type target areas that meet the standard of resource quantity after verification to A-type, and pushing to an exploration terminal.

[0123] In step T1, drilling data is uploaded to a cloud server in real time in an electronic form, and the data submission date is the time when the data enters a specified electronic system.

[0124] The embodiment designs an independent method for a real-time verification feedback link, and emphasizes data cleaning, model updating and dynamic target area management.

[0125] The application designs a lithium ore prospecting system and method of "multi-source data fusion-machine learning prediction-three-dimensional dynamic modeling". Multi-source data deep fusion: for the first time, spatial correlation analysis of hyperspectral remote sensing and geophysical data is introduced to solve the problem of concealed ore body identification. Adaptive weight distribution: based on the dynamic weight adjustment technology of DRL, the prediction error of the model in a complex geological area (such as a fault zone) is reduced to within 8%. Three-dimensional real-time modeling: realize the dynamic interaction of exploration data and geological model, and support intelligent planning of drilling path. Compared with the traditional method, the exploration cycle of the application is shortened by about 40%, and the success rate of target area verification is increased from 58% to more than 86%.

[0126] The specific embodiments of the application are described in detail above in combination with the drawings, but the application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the application.

Claims

1. A lithium ore exploration system characterized by: The multi-source data fusion module, the machine learning prediction module, the three-dimensional dynamic modeling module, and the real-time verification feedback module are included. The multi-source data fusion module integrates remote sensing spectral data, geophysical anomaly data and geochemical element combination anomaly through a weighting algorithm. The machine learning prediction module adopts a deep reinforcement learning algorithm, dynamically adjusts the data fusion weight in different geological scenes, and outputs a metallogenic probability heat map, with an accuracy error of less than 8%. The three-dimensional dynamic modeling module generates a three-dimensional ore body model based on an implicit modeling engine, and supports real-time drilling data-driven dynamic updating of the ore body boundary, with a model resolution of 1m. The real-time verification feedback module includes a drilling data analysis unit that collects core Li2O grade and ore body spatial coordinates in real time, a model correction unit, and a target area priority ranking unit that dynamically adjusts the exploration target area grade according to the verification results.

2. The lithium ore exploration system of claim 1, wherein, The multi-source data fusion module includes: A remote sensing data processing submodule for analyzing lithium mineralization alteration features in hyperspectral remote sensing images; A geophysical data correlation submodule that uses spatial interpolation algorithms to analyze gravity, magnetic, and electromagnetic data in relation to geological structures; An geochemical anomaly extraction submodule that filters Li, Rb, Cs, and other key element combination anomalies based on principal component analysis.

3. The lithium ore exploration system of claim 1, wherein: The remote sensing data processing submodule uses a convolutional neural network to extract spectral features and an attention mechanism to enhance the ability to identify mineralized alteration areas.

4. The lithium ore exploration system of claim 1, wherein, The machine learning prediction module uses a deep reinforcement learning algorithm, which includes: A dynamic weight allocation unit that automatically adjusts multi-source data fusion weights based on geological background; A probability heat map generation unit that outputs a lithium mineralization probability heat map with an accuracy error of less than 8%; An adaptive learning unit that iteratively optimizes model parameters based on historical exploration data and real-time drilling results.

5. The lithium ore exploration system of claim 4, wherein, The dynamic weight allocation unit uses a Q-learning algorithm to optimize weight strategies, prioritizing geophysical data weights in fault zone areas and remote sensing data weights in shallow overburden areas.

6. The lithium ore exploration system of claim 1, wherein: The three-dimensional dynamic modeling module includes: An implicit modeling engine that generates a three-dimensional ore body model based on geostatistics; A real-time data-driven unit that dynamically updates model boundaries and grade distribution based on drilling data; A visual interactive interface that supports three-dimensional model rotation, cross-section cutting, and automatic resource estimation.

7. The lithium ore exploration system of claim 6, wherein: The implicit modeling engine supports multi-scale modeling, including a 10km x 10km, 100m resolution mine area level and a 1km x 1km, 1m resolution ore body level, and integrates a GIS platform for spatial data interaction.

8. The lithium ore exploration system of claim 1, wherein: The drilling data analysis unit is linked with an automated drilling rig to upload core Li2O grade data to a cloud server in real time, triggering a model correction cycle of less than 24 hours.

9. A method of prospecting using the lithium ore prospecting system of claim 1, characterized by, The method includes the following steps: Step S1: Collect hyperspectral remote sensing data with a spectral resolution of ≤5nm using a drone, and extract lithium mineralization alteration features; Step S2: Spatially register geophysical data with remote sensing data to generate a multi-dimensional fusion feature matrix; Step S3: input the fusion data into the deep reinforcement learning algorithm model, output the metallogenic probability thermal map and automatically delineate the target area, wherein the probability threshold of class A target area is > 85%; Step S4: deploy the automatic drilling equipment according to the target area priority, and upload the Li2O grade data to the cloud server in real time.

10. A method of lithium ore exploration according to claim 9, characterised in that: In step S3, when the deviation between the target area verification result and the prediction is more than 15%, the model retraining process is triggered, and the update cycle is ≤ 48 hours.

Images