Mineral resource estimation system based on multi-source heterogeneous big data fusion

Abstract

The invention belongs to the technical field of mineral resource management, and discloses a mineral resource estimation system based on multi-source heterogeneous big data fusion. Comprising a data acquisition module used for acquiring mineral feature data; the data analysis module is used for analyzing the mineral characteristic data and evaluating the mining mode; the period estimation module is used for estimating a mining period based on the mining mode; the data collection module is used for collecting mineral influence data; the mining optimization module is used for optimizing a mining scheme according to the mineral product influence data and the mining period; according to the method, the limitation of a traditional static estimation method is overcome, the dynamic management and optimization capability of the mineral resources is enhanced, the economic benefit and the environmental influence can be balanced by optimizing the mining scheme, the economic benefit and the environmental protection level of the mineral resources are improved, and thus the scientificity and sustainability of mineral development are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mineral resource management, and more specifically, to a mineral resource estimation system based on multi-source heterogeneous big data fusion. Background Art

[0002] As an important strategic resource, the scientific development and sustainable management of mineral resources are the key to national economic and environmental development. However, traditional mineral resource estimation methods mostly focus on static reserve calculations, usually relying on geological exploration, drilling and sampling, and mainly focusing on static data such as the existing reserves and grade of the ore body. Although these traditional methods can provide a preliminary assessment of mineral resources, due to their inherent limitations, they are difficult to meet the needs of comprehensive evaluation of the mineral mining process, and have significant deficiencies in dynamic management, resource optimization and sustainable development, and cannot effectively support the optimization of mining plans and the prediction of ore mining cycles.

[0003] With the advancement of remote sensing technology, geophysical exploration technology and geographic information systems, multi-source heterogeneous data are gradually applied to mineral resource estimation. These data come from different channels and technical means, covering different dimensions of mineral resources, and can provide more comprehensive and accurate information for accurately reflecting the mining potential of ore bodies, evaluating mining benefits and their impact on the environment. The fusion of multi-source heterogeneous data can make up for the shortcomings of traditional static estimation methods and provide more possibilities for accurately reflecting mining potential and environmental impact. Therefore, how to evaluate the mining cycle of ore bodies and optimize mining plans based on the fusion of multi-source heterogeneous data has become a key challenge in mineral resource estimation.

[0004] In view of this, the present invention proposes a mineral resource estimation system based on multi-source heterogeneous big data fusion to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a mineral resource estimation system based on multi-source heterogeneous big data fusion, comprising:

[0006] Data acquisition module, used to acquire mineral characteristic data;

[0007] Data analysis module, used to analyze mineral characteristic data and evaluate mining methods;

[0008] The cycle estimation module estimates the mining cycle based on the mining method;

[0009] Data collection module, used to collect mineral impact data;

[0010] The mining optimization module optimizes the mining plan based on mineral impact data and mining cycle.

[0011] Furthermore, the mineral characteristic data includes reserve data, grade data, and distribution data;

[0012] The method for evaluating the mining method includes:

[0013] Set different digital tags for different mining methods and mark them as method tags; input the mineral characteristic data into the trained method evaluation model, output the corresponding method tags, and obtain the corresponding mining methods according to the method tags; the training process of the method evaluation model includes:

[0014] The method evaluation model adopts a random forest regression model. Pre-collect a sets of historical data, where the historical data includes mineral characteristic data and the corresponding method tags; sample b tree subsets from the a sets of historical data, each tree subset includes c sets of historical data, and train each decision tree according to the b tree subsets; where both a and b are integers greater than 1, and b < a, a = bc;

[0015] Initialize the model parameters, including the number of decision trees b, the maximum depth d of each decision tree, and the minimum number of samples u for internal nodes;

[0016] Recursively construct the first decision tree using one tree subset, and then sequentially train b - 1 decision trees using the remaining b - 1 tree subsets; the generation process of each decision tree is independent and the same; finally, construct a random forest containing b decision trees;

[0017] Calculate the loss function value on the b decision trees; input the mineral characteristic data into each decision tree, record the decision output, where the decision output is the method tag output by each decision tree and is marked as the prediction tag; calculate the mean square error between the average value of all prediction tags and the true value as the loss function value, where the true value is the method tag corresponding to the mineral characteristic data input into the decision tree; reduce the loss function value through parameter tuning, and select the model hyperparameters with the smallest loss function value as the model hyperparameters of the method evaluation model.

[0018] Furthermore, the method for recursively constructing the first decision tree using one tree subset includes:

[0019] Divide the c sets of historical data in one tree subset into a training set α and a validation set β, where α p is the p-th set of historical data in the training set α, p ∈ [1, α′], α′ < c, and α′ is the number of historical data in the training set α; use the feature t h to represent one data in the mineral characteristic data, and calculate the information gain rate Z(α, t h ) of the h-th feature t h ;

[0020] For each feature t hThe corresponding information gain rate Z(α,t h ) is calculated and the maximum information gain rate Z(α, t h ) corresponds to the feature t h As an internal node, and according to the maximum information gain rate Z(α, t h ) corresponds to the feature t h Divide the training set α into decision subsets, each with the same number of historical data;

[0021] For each maximum information gain rate Z(α, t h ) is divided into decision subsets, and the algorithm recursive process is repeated until the number of historical data in the decision subset is less than or equal to the minimum number of samples u of internal nodes, or the number of recursions is greater than or equal to the maximum depth d of each decision tree, and the recursion ends; the trained decision tree model is evaluated using the validation set β, and finally the construction of the first decision tree model is completed.

[0022] Furthermore, Z(α, t h ) is: In the formula, h∈[1, k], k is the number of data in the mineral characteristic data, αg is the training set α according to the characteristic t h The decision subsets divided, g∈[1, k′], k′ is an integer greater than 1, H(αg) is the information entropy of ag, H(α) is the information entropy of the training set α, where the calculation methods of H(α) and H(αg) include:

[0023]

[0024]

[0025] In the formula, αg p Indicates the mode label corresponding to the pth group of historical data in αg.

[0026] Furthermore, the method for estimating the mining cycle includes:

[0027] According to the reserve data, distribution data and mining method, the mining rate range is estimated; the attenuation factor q is preset; the mining rate change expression is: V(t) = V min +(V max -V min )×e -qt ; In the formula, V(t) is the mining rate corresponding to the mining time t, V min is the minimum mining rate in the mining rate range, V max is the maximum mining rate in the mining rate range, e is a natural constant, and t is the mining time; the expression of reserve data is: In the formula, S is the reserve data, and T is the mining cycle. The numerical solution method is used to solve the value of the mining cycle from the expression of the reserve data.

[0028] Further, the step of estimating the production rate range includes:

[0029] Step A1: Set the search interval [m,n] and precision ε;

[0030] Step A2: Determine the search sequence F(x);

[0031] Step A3: Determine the partition coefficient F(r) from the search sequence;

[0032] Step A4: Divide the search interval using a partition coefficient to obtain two points ψ1 and ψ2;

[0033] Step A5: Calculate the feasibility corresponding to ψ1 and ψ2 respectively;

[0034] Step A6: Update the search interval;

[0035] Step A7: Calculate the width υ of the search interval;

[0036] Step A8: If the width υ is less than or equal to the precision ε, the search interval is used as the mining rate range; if the width υ is greater than the precision ε, return to step A3.

[0037] Furthermore, in step A2, the search sequence

[0038] In step A3, the method for determining the partition coefficient F(r) is: calculating the initial partition coefficient δ′, Subtract the initial partition coefficient from each value in the search sequence to obtain the value difference; sort each value difference from small to large, obtain the value in the search sequence corresponding to the first value difference, and use it as the partition coefficient F(r);

[0039] In the step A4, Where, F(r-2) is the value of the first two digits of the search sequence before the partition coefficient F(r), and F(r-1) is the value of the first digit of the search sequence before the partition coefficient F(r);

[0040] In step A5, the feasibility is obtained by using the reserve data, distribution data, mode labels and numerical values ​​corresponding to the points as analysis data, inputting the analysis data into the trained rate estimation model, and predicting the corresponding feasibility;

[0041] In step A6, the method for updating the search interval is: marking the feasibility corresponding to ψ1 as the first feasibility The feasibility corresponding to ψ2 is marked as the second feasibility like The search interval is updated to Right now like The search interval is updated to Right now

[0042] In step A7, the width υ is obtained by subtracting the minimum value from the maximum value of the search interval.

[0043] Furthermore, the training process of the rate estimation model includes:

[0044] Collect y groups of analysis data in advance, set corresponding feasibility for each of the y groups of analysis data, where y is an integer greater than 1, and convert the analysis data and the corresponding feasibility into a corresponding set of feature vectors;

[0045] Each group of feature vectors is used as the input of the rate estimation model. The rate estimation model takes a group of prediction feasibility corresponding to each group of analysis data as output, and takes the actual feasibility corresponding to each group of analysis data as the prediction target, where the actual feasibility is the pre-set feasibility corresponding to the analysis data; minimizing the sum of prediction errors of all analysis data is used as the training target; the rate estimation model is trained until the sum of prediction errors converges and the training is stopped; the rate estimation model is a deep neural network model.

[0046] Furthermore, the mineral impact data includes historical market data and environmental data; the historical market data includes historical supply and demand data and historical price data; the historical supply and demand data is the production and consumption of mineral resources at historical moments; the historical price data is the transaction price of mineral resources at historical moments;

[0047] Input historical supply and demand data into the supply and demand forecasting model to forecast future supply and demand data, including the production and consumption of mineral resources within the mining cycle; input historical price data into the price forecasting model to forecast future price data, including the transaction price of mineral resources within the mining cycle; both the supply and demand forecasting model and the price forecasting model are LSTM models in the recurrent neural network model; optimize the mining plan within the mining cycle based on future supply and demand data and future price data, including mining rate and mining method;

[0048] The steps to optimize mining plans include:

[0049] Step B1: construct N mining sets, set different digital labels for different mining sets, and mark them as set labels, and construct a mining matrix based on all set labels;

[0050] Step B2: preset the constraint table capacity and the maximum number of iterations, and determine the evaluation function;

[0051] Step B3: Randomly select a set label from the mining matrix as the current solution, add it to the constraint table, and take the current solution as the optimal solution;

[0052] Step B4: According to the mining matrix, obtain the set labels adjacent to the current solution and use them as adjacent solutions;

[0053] Step B5: Calculate the evaluation value of each adjacent solution;

[0054] Step B6: Determine whether each adjacent solution exists in the constraint table; if so, mark the corresponding adjacent solution as a constraint solution; if not, mark the corresponding adjacent solution as a candidate solution;

[0055] Step B7: Mark the candidate solution with the largest evaluation value as the optimal candidate solution. If the evaluation value of the constraint solution is greater than the evaluation value of the optimal candidate solution, update the optimal candidate solution to the constraint solution. If the evaluation value of the constraint solution is less than or equal to the evaluation value of the optimal candidate solution, delete the constraint solution.

[0056] Step B8: Update the current solution to the optimal candidate solution, and determine whether to update the optimal solution; add all candidate solutions to the constraint table;

[0057] Step B9: Determine whether the number of iterations is equal to the maximum number of iterations; if so, the iteration is terminated and the mining set corresponding to the set label corresponding to the optimal solution is obtained; if not, the number of iterations is increased by one and the process returns to step B4.

[0058] Furthermore, in step B1, a mining method is randomly selected from all mining methods, and a value is randomly selected from the mining rate range to construct a mining set, and a total of N mining sets are constructed, and the N mining sets are all different;

[0059] In step B2, the expression of the evaluation function is: In the formula, f is the evaluation value, jx is the economic benefit, hy is the environmental impact, All are weight coefficients;

[0060] The method for obtaining economic benefits and environmental impacts is as follows: according to future supply and demand data and future price data, the predicted supply and demand data and the predicted price data are obtained; the predicted supply and demand data is one of the future supply and demand data, the predicted price data is one of the future price data, and the predicted supply and demand data and the predicted price data correspond to the same time; the predicted supply and demand data, the predicted price data and the set labels are used as test data, the test data are input into the trained benefit prediction model, and the corresponding economic benefits are predicted; the environmental data and the set labels are used as research data, the research data are input into the trained impact prediction model, and the corresponding environmental impact is predicted; the training process of the benefit prediction model and the impact prediction model are consistent with the training process of the rate estimation model, and both are neural network models;

[0061] In step B8, the method for determining whether to update the optimal solution is as follows: if the evaluation value of the optimal candidate solution is greater than the evaluation value of the optimal solution, the optimal solution is updated to the optimal candidate solution; if the evaluation value of the optimal candidate solution is less than or equal to the evaluation value of the optimal solution, the optimal solution is not updated.

[0062] The technical effects and advantages of the mineral resource estimation system based on multi-source heterogeneous big data fusion of the present invention are as follows:

[0063] By acquiring mineral characteristic data, a preliminary assessment of the mining method is made; and based on the mining method, a dynamic and accurate estimation of the mining rate range and mining cycle of mineral resources is achieved; at the same time, multi-source heterogeneous data, including environmental data and historical market data, are integrated to achieve dynamic optimization of the mining method and mining rate within the mining cycle. This not only overcomes the limitations of traditional static estimation methods and enhances the dynamic management and optimization capabilities of mineral resources, but also achieves a balance between economic benefits and environmental impacts by optimizing mining plans, thereby improving the economic benefits and environmental protection level of mineral resources, thereby improving the scientific nature and sustainability of mineral development. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a schematic diagram of a mineral resource estimation system based on multi-source heterogeneous big data fusion according to Example 1 of the present invention;

[0065] Figure 2 This is a flow chart of the mining plan optimization method of Example 1 of the present invention. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] Example 1

[0068] See also Figure 1 As shown, the mineral resource estimation system based on multi-source heterogeneous big data fusion described in this embodiment includes a data acquisition module, a data analysis module, a cycle estimation module, a data collection module and a mining optimization module; each module is connected by wired and / or wireless means to realize data transmission between modules;

[0069] The data acquisition module is used to obtain mineral characteristic data.

[0070] Mineral characteristic data include reserve data, grade data and distribution data;

[0071] Reserve data refers to the reserves of mineral resources available for mining in the ore body, usually expressed as the volume or mass of the ore; grade data refers to the content of valuable minerals (such as metal minerals) in the ore, and is an important indicator for determining the economic value of minerals; distribution data refers to the spatial distribution of the ore body on the surface or underground, including the size, shape, depth, extension direction, etc. of the ore body, which is used to clarify the structure and characteristics of the ore body; it should be noted that the methods for obtaining mineral characteristic data are all existing technologies and will not be elaborated on here; mineral characteristic data will affect the choice of mining method, and different mining methods are applicable to different mineral characteristic data. For example, when the grade data is high and the reserve data is large, the applicable mining method is open-pit mining to reduce the mining cost per unit ore; when the grade data is low and the reserve data is small, the applicable mining method is underground mining to reduce unnecessary mining of low-grade ores.

[0072] The data analysis module is used to analyze mineral characteristic data and evaluate mining methods.

[0073] Methods for assessing mining practices include:

[0074] Different digital labels are set for different mining methods and marked as method labels. Mining methods include open-pit mining, underground mining, hydraulic mining, etc. The mineral feature data is input into the trained method evaluation model, and the corresponding method label is output. The corresponding mining method is obtained according to the method label. The specific training process of the method evaluation model includes:

[0075] The method evaluation model adopts a random forest regression model, and collects a group of historical data in advance, the historical data includes mineral characteristic data and method labels corresponding to the mineral characteristic data; the method labels corresponding to the mineral characteristic data are collected by technical personnel in the process of historical mineral resource estimation, a group of historical data are collected, under the conditions of each group of historical data, the corresponding mining methods are obtained in turn according to actual experience, and corresponding method labels are set for the a group of historical data; b tree subsets are sampled from the a group of historical data, each tree subset includes c groups of historical data, and each decision tree is trained according to the b tree subsets; wherein a and b are both integers greater than 1, and b<a, a=bc.

[0076] Initialization model parameters, including the number of decision trees b, the maximum depth of each decision tree d, and the minimum number of samples of internal nodes u, are all preset by those skilled in the art according to actual conditions.

[0077] Recursively construct the first decision tree using a subset of trees:

[0078] Divide the c group of historical data in a tree subset into a training set α and a validation set β, α p is the pth group of historical data in the training set α, p∈[1,α′], α′<c, α′ is the number of historical data in the training set α; use feature t h Represents a data in the mineral characteristic data, calculates the hth characteristic t h The information gain rate Z(α,t h ), Z(α,t h ) is: In the formula, h∈[1,k], k is the number of data in the mineral characteristic data, αg is the training set α according to the characteristic t h The decision subsets divided, g∈[1,k′], k′ is an integer greater than 1, H(αg) is the information entropy of αg, H(α) is the information entropy of the training set α, where the calculation methods of H(α) and H(αg) include:

[0079]

[0080]

[0081] In the formula, αg p Indicates the mode label corresponding to the p-th group of historical data in αg;

[0082] For each feature t h The corresponding information gain rate Z(α,t h ) is calculated and the maximum information gain rate Z(α,t h ) corresponds to the feature t h As an internal node, and according to the maximum information gain rate Z(α,th ) corresponds to the feature t h Divide the training set α into decision subsets so that the number of historical data in each decision subset is the same;

[0083] For each maximum information gain rate Z(α,t h ) is divided into decision subsets, and the algorithm recursive process is repeated until the number of historical data in the decision subset is less than or equal to the minimum number of samples u of internal nodes, or the number of recursions is greater than or equal to the maximum depth d of each decision tree, and the recursion ends; the trained decision tree model is evaluated using the validation set β, and finally the construction of the first decision tree model is completed.

[0084] The remaining b-1 tree subsets are used in turn to train b-1 decision trees; the generation process of each decision tree is independent and the same; finally, a random forest containing b decision trees is constructed.

[0085] Calculate the loss function value on b decision trees; input the mineral characteristic data into each decision tree, record the decision output, which is the method label output by each decision tree and marked as the prediction label; calculate the mean square error between the average value of all prediction labels and the true value as the loss function value, and the true value is the method label corresponding to the mineral characteristic data input into the decision tree; reduce the loss function value through parameter tuning, and select the model hyperparameter with the smallest loss function value as the model hyperparameter of the method evaluation model.

[0086] The cycle estimation module estimates the mining cycle based on the mining method.

[0087] Methods for estimating the mining life cycle include:

[0088] According to the reserve data, distribution data and mining method, the mining rate range is estimated; the attenuation factor q is preset, and the attenuation factor q is preset by technical personnel in this field according to the actual situation; the mining rate change expression is: V(t) = V min +(V max -V min )×e -qt ; In the formula, V(t) is the mining rate corresponding to the mining time t, V min is the minimum mining rate in the mining rate range, V max is the maximum mining rate in the mining rate range, e is a natural constant, and t is the mining time; the expression of reserve data is: In the formula, S is the reserve data, and T is the mining cycle. The numerical solution method (such as Newton's method, bisection method, etc.) is used to solve the value of the mining cycle from the expression of the reserve data.

[0089] It should be understood that in the process of mineral mining, the mining rate is usually higher in the initial stage of mining, but as mining progresses, the mining rate tends to gradually decrease due to changes in factors such as ore body conditions, mining methods, equipment use, and ore grade.

[0090] The steps to estimate the range of production rates include:

[0091] Step A1: Set the search interval [m,n] and precision ε;

[0092] Step A2: Determine the search sequence F(x);

[0093] Step A3: Determine the partition coefficient F(r) from the search sequence;

[0094] Step A4: Divide the search interval using a partition coefficient to obtain two points ψ1 and ψ2;

[0095] Step A5: Calculate the feasibility corresponding to ψ1 and ψ2 respectively;

[0096] Step A6: Update the search interval;

[0097] Step A7: Calculate the width υ of the search interval;

[0098] Step A8: If the width υ is less than or equal to the precision ε, the search interval is used as the mining rate range; if the width υ is greater than the precision ε, return to step A3.

[0099] In the above step A1, the search interval [m, n] and the precision ε are preset by those skilled in the art according to actual conditions.

[0100] In the above step A2, the search sequence

[0101] In the above step A3, the method for determining the partition coefficient F(r) is: calculating the initial partition coefficient δ′, Subtract the initial partition coefficient from each value in the search sequence to obtain the numerical difference; sort each numerical difference from small to large, obtain the value in the search sequence corresponding to the numerical difference at the front, and use it as the partition coefficient F(r).

[0102] In the above step A4, In the formula, F(r-2) is the value of the first two digits of the search sequence before the partition coefficient F(r), and F(r-1) is the value of the first digit of the search sequence before the partition coefficient F(r).

[0103] In the above step A5, the method for obtaining the feasibility is: the reserve data, distribution data, mode label and the numerical value corresponding to the point are used as analysis data, and the analysis data is input into the trained rate estimation model to predict the corresponding feasibility; the training process of the rate estimation model includes:

[0104] Collect y groups of analysis data in advance, set corresponding feasibility for each of the y groups of analysis data, y is an integer greater than 1, and convert the analysis data and the corresponding feasibility into a corresponding set of feature vectors; the feasibility corresponding to the analysis data is determined by a technician in the historical mineral resource estimation process, collects y groups of analysis data, analyzes each group of analysis data in turn based on actual experience, evaluates the feasibility score of each group of analysis data, and sets the corresponding feasibility for the y groups of analysis data in turn as the corresponding feasibility;

[0105] Each set of feature vectors is used as the input of the rate estimation model. The rate estimation model uses a set of prediction feasibility corresponding to each set of analysis data as output, and uses the actual feasibility corresponding to each set of analysis data as the prediction target. The actual feasibility is the feasibility corresponding to the analysis data set in advance. The training target is to minimize the sum of the prediction errors of all analysis data. The calculation formula of the prediction error is: where η w is the prediction error, w is the group number of the eigenvector corresponding to the analysis data, is the prediction feasibility corresponding to the wth group of analysis data, ρ w is the actual feasibility corresponding to the wth group of analysis data; the rate estimation model is trained until the sum of the prediction errors reaches convergence and the training is stopped.

[0106] The above rate estimation model is specifically a deep neural network model; it includes an input layer, a hidden layer and an output layer; each hidden layer includes multiple neurons, each neuron is connected to the neurons in the next layer, and the connection contains weights, which determine the importance and influence of data transmission in the neural network; an activation function is applied to each neuron between the hidden layer and the output layer, and the activation function introduces nonlinearity, allowing the network to learn more complex patterns and features.

[0107] In the above step A6, the method for updating the search interval is: marking the feasibility corresponding to ψ1 as the first feasibility The feasibility corresponding to ψ2 is marked as the second feasibility like The search interval is updated to Right now like The search interval is updated to Right now

[0108] In the above step A7, the width υ is obtained by subtracting the minimum value from the maximum value of the search interval.

[0109] Data collection module for collecting mineral impact data.

[0110] Mineral impact data include historical market data and environmental data; historical market data include historical supply and demand data and historical price data;

[0111] Historical supply and demand data refers to the production and consumption of mineral resources at historical moments, which helps to understand the supply and demand of mineral resources in the market. When the production is less than the consumption, it means that the market demand for mineral resources is large, so the mining rate needs to be increased, and vice versa. Historical supply and demand data is obtained through mineral supply and demand reports issued by mineral industry research institutions, or mineral resource production and demand reports issued by mineral industry associations. Historical price data refers to the transaction price of mineral resources at historical moments. The higher the transaction price, the greater the economic benefits of mining mineral resources, so the mining rate needs to be increased, and vice versa. Historical price data is obtained through mineral trading platforms or mineral price websites.

[0112] Environmental data refers to the natural environment information around the ore body, such as water sources, air quality, soil conditions, vegetation coverage, etc.; it is used to restrict mineral resource mining activities and ensure that no environmental damage is caused during the mining process. If mining activities will bring environmental risks (such as water pollution, air quality deterioration, etc.), it is necessary to adjust the mining method or reduce the mining rate to comply with environmental protection requirements and reduce ecological damage; environmental data is obtained through remote sensing technology and satellite images, geographic information systems and environmental monitoring systems (such as water quality monitoring stations, meteorological stations, etc.).

[0113] The mining optimization module optimizes the mining plan based on mineral impact data and mining cycle.

[0114] Input historical supply and demand data into the supply and demand forecasting model to predict future supply and demand data, which includes the production and consumption of mineral resources during the mining cycle; input historical price data into the price forecasting model to predict future price data, which includes the transaction price of mineral resources during the mining cycle; both the supply and demand forecasting model and the price forecasting model are LSTM models in the recurrent neural network model; the LSTM model is an existing technology, and the specific training process will not be described in detail here; based on future supply and demand data and future price data, the mining plan within the mining cycle is optimized, and the mining plan includes the mining rate and mining method.

[0115] like Figure 2 As shown, the steps to optimize the mining plan include:

[0116] Step B1: construct N mining sets, set different digital labels for different mining sets, and mark them as set labels, and construct a mining matrix based on all set labels;

[0117] Step B2: preset the constraint table capacity and the maximum number of iterations, and determine the evaluation function;

[0118] Step B3: Randomly select a set label from the mining matrix as the current solution, add it to the constraint table, and take the current solution as the optimal solution;

[0119] Step B4: According to the mining matrix, obtain the set labels adjacent to the current solution and use them as adjacent solutions;

[0120] Step B5: Calculate the evaluation value of each adjacent solution;

[0121] Step B6: Determine whether each adjacent solution exists in the constraint table; if so, mark the corresponding adjacent solution as a constraint solution; if not, mark the corresponding adjacent solution as a candidate solution;

[0122] Step B7: Mark the candidate solution with the largest evaluation value as the optimal candidate solution. If the evaluation value of the constraint solution is greater than the evaluation value of the optimal candidate solution, update the optimal candidate solution to the constraint solution. If the evaluation value of the constraint solution is less than or equal to the evaluation value of the optimal candidate solution, delete the constraint solution.

[0123] Step B8: Update the current solution to the optimal candidate solution, and determine whether to update the optimal solution; add all candidate solutions to the constraint table;

[0124] Step B9: Determine whether the number of iterations is equal to the maximum number of iterations; if so, the iteration is terminated and the mining set corresponding to the set label corresponding to the optimal solution is obtained; if not, the number of iterations is increased by one and the process returns to step B4.

[0125] In the above step B1, a mining method is randomly selected from all mining methods, and a value is randomly selected from the mining rate range to construct a mining set. A total of N mining sets are constructed, and the N mining sets are all different.

[0126] In the above step B2, the constraint table capacity and the maximum number of iterations are preset by those skilled in the art according to actual conditions; the expression of the evaluation function is: In the formula, f is the evaluation value, jx is the economic benefit, hy is the environmental impact, are all weight coefficients; the specific values ​​of the weight coefficients can be set according to actual conditions. The weight coefficients reflect the degree of influence of economic benefits and environmental impact on the evaluation value. Technical personnel in this field can preset corresponding weight coefficients according to the actual degree of influence of economic benefits and environmental impact on the evaluation value, so as to accurately evaluate the comprehensive benefits of mining methods and mining rates, that is, to achieve a balance between economic benefits and environmental impacts, so that the mining plan can meet the economic goals while minimizing the negative impact on the environment; the calculation of the evaluation value is a dimensionless calculation.

[0127] The method for obtaining economic benefits and environmental impacts is as follows: according to future supply and demand data and future price data, predicted supply and demand data and predicted price data are obtained; the predicted supply and demand data is one of the future supply and demand data, the predicted price data is one of the future price data, and the predicted supply and demand data and the predicted price data correspond to the same time; the predicted supply and demand data, predicted price data and set labels are used as test data, the test data are input into the trained benefit prediction model, and the corresponding economic benefits are predicted; the environmental data and set labels are used as research data, the research data are input into the trained impact prediction model, and the corresponding environmental impact is predicted; the training process of the benefit prediction model and the impact prediction model are consistent with the training process of the rate estimation model, and both are neural network models.

[0128] In the above step B8, the method for determining whether to update the optimal solution is: if the evaluation value of the optimal candidate solution is greater than the evaluation value of the optimal solution, the optimal solution is updated to the optimal candidate solution; if the evaluation value of the optimal candidate solution is less than or equal to the evaluation value of the optimal solution, the optimal solution is not updated.

[0129] This embodiment obtains mineral characteristic data to preliminarily evaluate the mining method; and based on the mining method, it realizes dynamic and accurate estimation of the mining rate range and mining cycle of mineral resources; at the same time, it integrates multi-source heterogeneous data, including environmental data and historical market data, to realize dynamic optimization of mining methods and mining rates within the mining cycle; it not only overcomes the limitations of traditional static estimation methods and enhances the dynamic management and optimization capabilities of mineral resources, but also can achieve a balance between economic benefits and environmental impacts by optimizing mining plans, thereby improving the economic benefits and environmental protection level of mineral resources, thereby improving the scientificity and sustainability of mineral development.

[0130] Example 2

[0131] The present application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable codes, and when the computer-readable codes are executed by the one or more processors, the mineral resource estimation system based on multi-source heterogeneous big data fusion as described above may be executed.

[0132] The method or system according to the implementation mode of the present application can also be implemented with the help of the architecture of the electronic device shown in the present application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. A storage device in the electronic device, such as a ROM or a hard disk, can store a mineral resource estimation system based on multi-source heterogeneous big data fusion provided in the present application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in the present application is only exemplary. When implementing different devices, one or more components in the electronic device shown in the present application may be omitted according to actual needs.

[0133] Example 3

[0134] One embodiment of the present application discloses a computer-readable storage medium. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are executed by a processor, a mineral resource estimation system based on multi-source heterogeneous big data fusion according to an embodiment of the present application described with reference to the above figures can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0135] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: a mineral resource estimation system based on multi-source heterogeneous big data fusion. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0136] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0137] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A mineral resource estimation system based on multi-source heterogeneous big data fusion, characterized in that: include: Data acquisition module, used to acquire mineral characteristic data; Data analysis module, used to analyze mineral characteristic data and evaluate mining methods; The cycle estimation module estimates the mining cycle based on the mining method; Data collection module, used to collect mineral impact data; The mining optimization module optimizes the mining plan based on mineral impact data and mining cycle.

2. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 1 is characterized in that: The mineral characteristic data include reserve data, grade data and distribution data; The method for evaluating the mining method includes: Different digital labels are set for different mining methods and marked as method labels; the mineral feature data is input into the trained method evaluation model, the corresponding method label is output, and the corresponding mining method is obtained according to the method label; the training process of the method evaluation model includes: The method evaluation model adopts a random forest regression model, collects a group of historical data in advance, and the historical data includes mineral characteristic data and method labels corresponding to the mineral characteristic data; samples b tree subsets from the a group of historical data, each tree subset includes c groups of historical data, and trains each decision tree according to the b tree subsets; wherein a and b are both integers greater than 1, and b<a, a=bc; Initialize model parameters, including the number of decision trees b, the maximum depth of each decision tree d, and the minimum number of samples of internal nodes u; Use a tree subset to recursively construct the first decision tree, and then use the remaining b-1 tree subsets to train b-1 decision trees in turn; the generation process of each decision tree is independent and the same; finally, a random forest containing b decision trees is constructed; Calculate the loss function value on b decision trees; input the mineral characteristic data into each decision tree, record the decision output, which is the method label output by each decision tree and mark it as the prediction label; calculate the mean square error between the average value of all prediction labels and the true value as the loss function value, and the true value is the method label corresponding to the mineral characteristic data input into the decision tree; reduce the loss function value through parameter tuning, and select the model hyperparameter with the smallest loss function value as the model hyperparameter of the method evaluation model.

3. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 2 is characterized in that: The method of recursively constructing a first decision tree using a tree subset comprises: Divide the c group of historical data in a tree subset into a training set α and a validation set β, α p is the pth group of historical data in the training set α, p∈[1,α′], α′<c, α′ is the number of historical data in the training set α; use feature t h Represents a data in the mineral characteristic data, calculates the hth characteristic t h The information gain rate Z(α,t h ); For each feature t h The corresponding information gain rate Z(α,t h ) is calculated and the maximum information gain rate Z(α,t h ) corresponds to the feature t h As an internal node, and according to the maximum information gain rate Z(α,t h ) corresponds to the feature t h Divide the training set α into decision subsets, with the same amount of historical data in each decision subset; For each maximum information gain rate Z(α,t h ) is divided into decision subsets, and the algorithm recursive process is repeated until the number of historical data in the decision subset is less than or equal to the minimum number of samples u of internal nodes, or the number of recursions is greater than or equal to the maximum depth d of each decision tree, and the recursion ends; the trained decision tree model is evaluated using the validation set β, and finally the construction of the first decision tree model is completed.

4. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 3 is characterized in that: Z(α,t h ) is: In the formula, h∈[1,k], k is the number of data in the mineral characteristic data, αg is the training set α according to the characteristic t h The decision subsets divided, g∈[1,k′], k′ is an integer greater than 1, H(αg) is the information entropy of αg, H(α) is the information entropy of the training set α, where the calculation methods of H(α) and H(αg) include: In the formula, αg p Indicates the mode label corresponding to the pth group of historical data in αg.

5. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 4 is characterized in that: The method for estimating the mining cycle includes: According to the reserve data, distribution data and mining method, the mining rate range is estimated; the attenuation factor q is preset; the mining rate change expression is: V(t) = V min +(V max -V min )×e -qt ; In the formula, V(t) is the mining rate corresponding to the mining time t, V min is the minimum mining rate in the mining rate range, V max is the maximum mining rate in the mining rate range, e is a natural constant, and t is the mining time; the expression of reserve data is: In the formula, S is the reserve data, and T is the mining cycle. The numerical solution method is used to solve the value of the mining cycle from the expression of the reserve data.

6. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 5 is characterized in that: The step of estimating the production rate range comprises: Step A1: Set the search interval [m,n] and precision ε; Step A2: Determine the search sequence F(x); Step A3: Determine the partition coefficient F(r) from the search sequence; Step A4: Divide the search interval using a partition coefficient to obtain two points ψ1 and ψ2; Step A5: Calculate the feasibility corresponding to ψ1 and ψ2 respectively; Step A6: Update the search interval; Step A7: Calculate the width υ of the search interval; Step A8: If the width υ is less than or equal to the precision ε, the search interval is used as the mining rate range; if the width υ is greater than the precision ε, return to step A3.

7. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 6 is characterized in that: In step A2, the search sequence In step A3, the method for determining the partition coefficient F(r) is: calculating the initial partition coefficient δ′, Subtract the initial partition coefficient from each value in the search sequence to obtain the value difference; sort each value difference from small to large, obtain the value in the search sequence corresponding to the first value difference, and use it as the partition coefficient F(r); In the step A4, Where, F(r-2) is the value of the first two digits of the search sequence before the partition coefficient F(r), and F(r-1) is the value of the first digit of the search sequence before the partition coefficient F(r); In step A5, the feasibility is obtained by using the reserve data, distribution data, mode labels and numerical values ​​corresponding to the points as analysis data, inputting the analysis data into the trained rate estimation model, and predicting the corresponding feasibility; In step A6, the method for updating the search interval is: marking the feasibility corresponding to ψ1 as the first feasibility The feasibility corresponding to ψ2 is marked as the second feasibility like The search interval is updated to Right now like The search interval is updated to Right now In step A7, the width υ is obtained by subtracting the minimum value from the maximum value of the search interval.

8. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 7 is characterized in that: The training process of the rate estimation model includes: Collect y groups of analysis data in advance, set corresponding feasibility for each of the y groups of analysis data, where y is an integer greater than 1, and convert the analysis data and the corresponding feasibility into a corresponding set of feature vectors; Each group of feature vectors is used as the input of the rate estimation model. The rate estimation model takes a group of prediction feasibility corresponding to each group of analysis data as output, and takes the actual feasibility corresponding to each group of analysis data as the prediction target, where the actual feasibility is the pre-set feasibility corresponding to the analysis data; minimizing the sum of prediction errors of all analysis data is used as the training target; the rate estimation model is trained until the sum of prediction errors converges and the training is stopped; the rate estimation model is a deep neural network model.

9. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 8 is characterized in that: The mineral impact data includes historical market data and environmental data; the historical market data includes historical supply and demand data and historical price data; the historical supply and demand data is the production and consumption of mineral resources at historical moments; the historical price data is the transaction price of mineral resources at historical moments; Input historical supply and demand data into the supply and demand forecasting model to forecast future supply and demand data, including the production and consumption of mineral resources within the mining cycle; input historical price data into the price forecasting model to forecast future price data, including the transaction price of mineral resources within the mining cycle; both the supply and demand forecasting model and the price forecasting model are LSTM models in the recurrent neural network model; optimize the mining plan within the mining cycle based on future supply and demand data and future price data, including mining rate and mining method; The steps of optimizing the mining plan include: Step B1: construct N mining sets, set different digital labels for different mining sets, and mark them as set labels, and construct a mining matrix based on all set labels; Step B2: preset the constraint table capacity and the maximum number of iterations, and determine the evaluation function; Step B3: Randomly select a set label from the mining matrix as the current solution, add it to the constraint table, and take the current solution as the optimal solution; Step B4: According to the mining matrix, obtain the set labels adjacent to the current solution and use them as adjacent solutions; Step B5: Calculate the evaluation value of each adjacent solution; Step B6: Determine whether each adjacent solution exists in the constraint table; if so, mark the corresponding adjacent solution as a constraint solution; if not, mark the corresponding adjacent solution as a candidate solution; Step B7: Mark the candidate solution with the largest evaluation value as the optimal candidate solution. If the evaluation value of the constraint solution is greater than the evaluation value of the optimal candidate solution, update the optimal candidate solution to the constraint solution. If the evaluation value of the constraint solution is less than or equal to the evaluation value of the optimal candidate solution, delete the constraint solution. Step B8: Update the current solution to the optimal candidate solution, and determine whether to update the optimal solution; add all candidate solutions to the constraint table; Step B9: Determine whether the number of iterations is equal to the maximum number of iterations; if so, the iteration is terminated and the mining set corresponding to the set label corresponding to the optimal solution is obtained; if not, the number of iterations is increased by one and the process returns to step B4.

10. The mineral resource estimation system based on multi-source heterogeneous big data fusion according to claim 9 is characterized in that: In the step B1, a mining method is randomly selected from all mining methods, and a value is randomly selected from the mining rate range to construct a mining set, and a total of N mining sets are constructed, and the N mining sets are all different; In step B2, the expression of the evaluation function is: In the formula, f is the evaluation value, jx is the economic benefit, hy is the environmental impact, All are weight coefficients; The method for obtaining economic benefits and environmental impacts is as follows: according to future supply and demand data and future price data, the predicted supply and demand data and the predicted price data are obtained; the predicted supply and demand data is one of the future supply and demand data, the predicted price data is one of the future price data, and the predicted supply and demand data and the predicted price data correspond to the same time; the predicted supply and demand data, the predicted price data and the set labels are used as test data, the test data are input into the trained benefit prediction model, and the corresponding economic benefits are predicted; the environmental data and the set labels are used as research data, the research data are input into the trained impact prediction model, and the corresponding environmental impact is predicted; the training process of the benefit prediction model and the impact prediction model are consistent with the training process of the rate estimation model, and both are neural network models; In step B8, the method for determining whether to update the optimal solution is as follows: if the evaluation value of the optimal candidate solution is greater than the evaluation value of the optimal solution, the optimal solution is updated to the optimal candidate solution; if the evaluation value of the optimal candidate solution is less than or equal to the evaluation value of the optimal solution, the optimal solution is not updated.

Images