Mine water source identification method, system and equipment based on combined weight and grey relational degree theory and medium

Abstract

The invention discloses a mine water source identification method, system, equipment and medium based on a combined weight and grey relational degree theory, and relates to the technical field of mine water source identification, and the method comprises the steps: obtaining water quality parameters and stable isotope composition of a mine water sample, and building a comprehensive feature matrix; calculating the correlation degree between each water quality parameter and a preset water source type based on the comprehensive feature matrix, analyzing the composition and proportion of multi-source water mixing in the mine water sample, and determining a first mine water source identification result; and weighting the water quality parameters and the stable isotope composition in the comprehensive characteristic matrix by using a combined weight method, and obtaining a second mine water source identification result based on the weighted comprehensive characteristic matrix and the first mine water source identification result. The method has high prediction speed and prediction precision, the existence condition of the underground water source can be identified by using a geophysical prospecting method according to different geological conditions through the prediction model, and the acquisition process of the underground water burst source is more convenient.

Description

Technical Field

[0001] The present invention relates to the technical field of mine water source identification, and in particular to a mine water source identification method, system, equipment and medium based on combination weight and grey correlation theory. Background Art

[0002] As one of the three key elements of mine water hazards, clarifying the occurrence conditions of water sources in the study area and accurately identifying the source of water inrush are of great significance for roof water hazard prevention and control. Water inrush (seepage) disasters from different water sources often exhibit different water volume evolution characteristics. Therefore, accurate and rapid identification of mine water sources not only allows for the development of effective control measures based on water source type in the later stage, but also allows for the early prediction of possible water inrush (inrush) hazards, promoting the shift of water hazard prevention from process-based management to source prevention.

[0003] At present, geophysical methods are mostly used to identify mine water sources. However, the waveform data images obtained by geophysical sensors need to be interpreted by experienced experts before they can be used. In the process of interpreting the images, due to subjective reasons and objective factors, the existing mine water source identification methods have problems in the prediction and forecasting process, such as low interpretability of geophysical forecast results, reliance on expert experience, and low forecast accuracy.

[0004] Currently, effectively combining two or more geophysical methods to jointly complete detection tasks is becoming a new development direction in mine water source identification. However, when conducting joint exploration with multiple geophysical methods, the interpretation process often involves first interpreting each method individually and then conducting comparative analysis. This seemingly comprehensive interpretation approach is often just a simple combined analysis that fails to consider inter-data correlations. This results in an incomplete analysis of some anomalies, resulting in relatively low interpretation accuracy. Furthermore, when multiple survey lines and methods are involved, such a simple combined interpretation approach is inefficient and difficult to guarantee results.

[0005] Grey correlation theory mainly analyzes the degree of correlation between factors through the geometric correlation of sequences, and better considers the correlation between data. At present, grey correlation theory has been successfully applied in disaster prediction and evaluation, engineering management and other fields, but it needs further application in the field of mine water source identification. Summary of the Invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a mine water source identification method based on combined weight and grey correlation theory, comprising:

[0009] As a preferred solution of the mine water source identification method based on combined weight and grey correlation theory described in the present invention, the method comprises the following steps: obtaining water quality parameters and stable isotope composition of mine water samples and establishing a comprehensive feature matrix; calculating the correlation between each water quality parameter and a preset water source type based on the comprehensive feature matrix, analyzing the composition and proportion of the multi-source water mixture in the mine water sample, and determining a first mine water source identification result; using a combined weight method to weight the water quality parameters and the stable isotope composition in the comprehensive feature matrix, and obtaining a second mine water source identification result based on the weighted comprehensive feature matrix and the first mine water source identification result.

[0010] As a preferred solution of the mine water source identification method based on combined weight and grey correlation theory described in the present invention, the water quality parameters are obtained by measuring the physical and chemical indicators of the collected mine water samples; and the stable isotope composition is obtained by analyzing the stable isotope data in the mine water samples.

[0011] As an optimal solution of the mine water source identification method based on combined weight and grey correlation theory described in the present invention, after establishing the comprehensive feature matrix, a generative adversarial network is used to generate synthetic data based on the comprehensive feature matrix, and the synthetic data is merged with the comprehensive feature matrix to form an enhanced comprehensive feature matrix.

[0012] The beneficial effects of this preferred technical solution include: by using a generative adversarial network to generate synthetic data based on a comprehensive feature matrix and form an enhanced comprehensive feature matrix, it effectively addresses the problem of sparse data or uneven data distribution for certain water source types in mine water source identification. Compared to traditional identification methods that rely only on a limited number of real samples, this method can increase the number and diversity of training samples, improve the model's ability to learn sparse water source types, thereby improving overall recognition accuracy and the system's generalization performance. It also provides more stable and reliable performance when dealing with water source identification under rare or special geological conditions.

[0013] As a preferred solution of the mine water source identification method based on combined weight and grey correlation theory described in the present invention, the correlation degree is calculated using a grey correlation analysis method, and the correlation coefficient is determined by comparing the similarity between the mine water sample feature sequence and the feature sequence of each preset water source type.

[0014] The beneficial effect of this preferred technical solution is that, by adopting a weight assignment method that combines subjective and objective weights, it not only considers the subjective judgment of expert experience and prior knowledge, but also incorporates objective analysis of the statistical characteristics of the data itself, achieving a more scientific and reasonable feature weight assignment. Compared with a single weight assignment method, it can effectively balance the importance of different water quality parameters and stable isotope compositions in the identification process, avoiding the bias that may exist in purely subjective weighting and the domain knowledge that may be ignored by purely objective weighting, thereby improving the accuracy and reliability of mine water source identification.

[0015] As an optimal solution of the mine water source identification method based on combined weight and grey correlation theory described in the present invention, the combined weight method includes a combination of subjective weight and objective weight, and the weight coefficients of the water quality parameters and the stable isotope composition are determined respectively by weighting.

[0016] The beneficial effects of this preferred technical solution are: by adopting the specific technical implementation solution of Wasserstein GAN, including the adversarial training mechanism of the generator and the discriminator and the Wasserstein distance metric, it has stronger training stability and higher data generation quality than the traditional GAN ​​network, and can generate synthetic samples that are highly similar to the distribution of real mine water quality data, effectively avoiding the pattern collapse and training instability problems that may occur in traditional GAN, ensuring that the generated synthetic data can truly reflect the characteristic distribution of various water sources, providing a high-quality data foundation for subsequent grey correlation analysis, and further improving the overall performance of mine water source identification.

[0017] As a preferred solution of the mine water source identification method based on combined weight and grey correlation theory described in the present invention, the generative adversarial network is a Wasserstein GAN, comprising a generator and a discriminator, wherein the generator receives a random noise vector as input, which is converted into a generated sample through a neural network layer, and the discriminator receives real samples extracted from the comprehensive feature matrix and the generated samples as input, and outputs the probabilities that the real samples and the generated samples are real data; when training the Wasserstein GAN, the Wasserstein distance is used as a metric, and the generator and the discriminator are trained alternately; after the training is completed, the generator is used to generate a plurality of the generated samples as the synthetic data.

[0018] As a preferred solution of the mine water source identification method based on combined weight and grey correlation theory described in the present invention, the grey correlation analysis method includes: extracting the feature matrix of the preset water source type from the comprehensive feature matrix or the enhanced comprehensive feature matrix; calculating the grey correlation between each of the water quality parameters in the mine water sample and the preset water source type, wherein the correlation calculation is implemented by averaging the correlation coefficient adjusted by the resolution coefficient; determining the first mine water source identification result based on the correlation and the composition and proportion of the multi-source water mixture.

[0019] In the second aspect, the present invention further provides a mine water source identification system based on combined weights and grey correlation theory, comprising: a feature construction module for obtaining water quality parameters and stable isotope compositions of mine water samples and establishing a comprehensive feature matrix; a correlation analysis module for calculating the correlation between each of the water quality parameters and the preset water source type based on the comprehensive feature matrix, parsing the composition and proportion of the multi-source water mixture in the mine water sample, and determining the first mine water source identification result; a weight processing module for weighting the water quality parameters and the stable isotope composition in the comprehensive feature matrix using a combined weight method, and obtaining a second mine water source identification result based on the weighted comprehensive feature matrix and the first mine water source identification result.

[0020] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, it implements the steps of the mine water source identification method based on the combination weight and grey correlation theory as described above.

[0021] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the mine water source identification method based on the combination weight and grey correlation theory are implemented as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 This is an overall flow chart of a mine water source identification method based on combined weight and grey correlation theory provided by one embodiment of the present invention.

[0024] Figure 2 This is a diagram showing the simulation experiment results of a mine water source identification method based on combined weight and grey correlation theory provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0025] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0026] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0027] Example 1, reference Figure 1 , which is an embodiment of the present invention, provides a mine water source identification method based on combined weight and grey correlation theory, comprising:

[0028] S100: Obtain water quality parameters and stable isotope composition of mine water samples and establish a comprehensive characteristic matrix;

[0029] S200: calculating the correlation between each of the water quality parameters and the preset water source type based on the comprehensive feature matrix, analyzing the composition and proportion of the multi-source water mixture in the mine water sample, and determining a first mine water source identification result;

[0030] S300: weighting the water quality parameters and the stable isotope composition in the comprehensive feature matrix using a combined weight method, and obtaining a second mine water source identification result based on the weighted comprehensive feature matrix and the first mine water source identification result.

[0031] It should be noted that during underground mining operations, it is necessary to thoroughly investigate the engineering and hydrogeological conditions of the mining area to provide relevant data to underground construction personnel. Water inrush hazards from different water sources often exhibit distinct water volume evolution characteristics. Therefore, accurate and rapid identification of mine water sources not only allows for the development of effective remediation measures based on source type, but also allows for the early prediction of potential water inrush hazards. Existing mine water source identification methods primarily rely on geophysical methods. However, the waveform data images generated by geophysical sensors require interpretation by experienced experts before they can be used. Due to subjective and objective factors, this interpretation process results in low interpretability of geophysical prediction results, reliance on expert experience, and low prediction accuracy. Furthermore, during mine water source identification, data for certain water source types may be sparse or uneven due to unique geological locations, sampling limitations, or environmental variations. This makes it difficult for traditional identification methods to accurately identify these sparse water source types, impacting overall identification effectiveness and prediction accuracy.

[0032] Therefore, in response to the above-mentioned problems of identification accuracy and data sparsity, through steps S100 to S300, a comprehensive feature matrix containing water quality parameters and stable isotope composition is first established to provide a comprehensive data basis for subsequent analysis; the correlation degree between each water quality parameter and the preset water source type is calculated based on the grey correlation theory, which can effectively handle the correlation between data, analyze the composition and proportion of multi-source water mixtures, and achieve preliminary and accurate identification of mine water samples; different features are scientifically weighted through a combined weight method, taking into account both expert experience and the statistical characteristics of the data itself, and optimizing based on the first identification results to improve the accuracy and reliability of mine water source identification, providing strong technical support for mine water hazard prevention and control work.

[0033] Example 2 is an embodiment of the present invention, which provides a mine water source identification method based on combination weight and grey correlation theory.

[0034] In an embodiment of the present invention, the water quality parameters in step S100 are obtained by measuring the physical and chemical indicators of the collected mine water samples; and the stable isotope composition is obtained by analyzing the stable isotope data in the mine water samples.

[0035] It should be noted that after the mine water samples are collected, they need to be preprocessed, specifically to fill in or delete missing data points, detect and handle anomalies, clear invalid or irrelevant data records, and convert non-numeric data into numerical data.

[0036] In an embodiment of the present invention, step S100 obtains water quality parameters and stable isotope composition of a mine water sample and establishes a comprehensive feature matrix, including steps A1-A2:

[0037] A1: Determine water quality parameters and stable isotope composition in mine water samples;

[0038] A2: Establish a comprehensive characteristic matrix based on the water quality parameters and stable isotope composition in A1.

[0039] Specifically, in A1, water quality parameters are obtained by measuring the physical and chemical indicators of collected mine water samples. Water quality parameters provide information about the chemical composition and physical properties of water and are crucial for assessing water quality. Water quality parameters are obtained by measuring the physical and chemical indicators of mine water samples using laboratory analysis or on-site portable testing equipment. These devices can quickly and accurately determine the various physical and chemical properties of water samples. This measurement data provides a basic understanding of mine water quality. Physical and chemical indicators include, but are not limited to: 1. pH: The acidity or alkalinity of water; 2. Conductivity: The electrical conductivity of water, reflecting the ion content in water; 3. Dissolved oxygen: The concentration of dissolved oxygen in water; 4. Total dissolved solids (TDS): The total amount of dissolved inorganic salts and other substances in water; 5. Hardness: The concentration of calcium and magnesium ions in water; 6. Heavy metal concentration: The content of harmful metals such as lead, mercury, and cadmium; 7. Other: Organic matter content, microbial indicators, etc.

[0040] At the same time, the stable isotope composition is obtained by analyzing the stable isotope data in the mine water samples. Stable isotopes are those that do not undergo radioactive decay. In hydrogeology, commonly used stable isotopes include two stable isotopes of hydrogen ( 1 H and 2 H or D) and the three stable isotopes of oxygen ( 16 O. 17 O and 18 O). Stable isotope composition refers to the ratio of different stable isotopes of hydrogen and oxygen in mine water samples. These ratios can be accurately measured and analyzed by professional equipment such as isotope mass spectrometers. Stable isotope composition can provide important information about the source and migration path of water, because water from different sources (such as surface water and groundwater) usually has different stable isotope composition characteristics. In this invention, the main research is the stable isotope ratio of hydrogen and oxygen, specifically the ratio of deuterium. 2 H and ordinary hydrogen 1 The isotope ratio R of H D , heavy oxygen 18 O and ordinary oxygen 16 The isotope ratio of O is R 18 .

[0041] Specifically, in A2, a comprehensive characteristic matrix is ​​established based on the water quality parameters and stable isotope composition in A1, including steps B1-B4:

[0042] B1: Determine the list of water quality parameters and the stable isotope dataset consisting of stable isotope data, and calculate the stable isotope ratios;

[0043] B2: Standardize water quality parameters;

[0044] B3: Calculate the δ value of stable isotopes based on the stable isotope ratio and perform normalization;

[0045] B4: Integrate the standardized water quality parameters and the normalized δ values ​​of stable isotopes into a comprehensive characteristic matrix.

[0046] Specifically, in B1, determine the list of water quality parameters, specifically the 7 categories in A1. Determine the stable isotope data set composed of stable isotope data and calculate the stable isotope ratio, specifically R in A1. D and R 18 . ;

[0047] In B2, the water quality parameters are normalized using a standardized formula to eliminate dimension differences. The formula is:

[0048]

[0049] in, is the standardized value of the kth water quality parameter; X k is the original value of the kth water quality parameter; μ k is the average value of the kth water quality parameter in all mine water samples; σ k is the standard deviation of the kth water quality parameter in all mine water samples.

[0050] In B3, first, for hydrogen R D and oxygen R 18 The stable isotope ratio of δ is calculated using the formula:

[0051]

[0052] where δX is the deviation of the ratio of heavy to light isotopes in the mine water sample from the international standard; R sample is the ratio of heavy isotopes to light isotopes in the mine water sample; R standard It is the ratio of heavy isotopes to light isotopes in the international standard.

[0053] Secondly, the δ value is processed using the normalization formula:

[0054]

[0055] in, is the normalized value of the lth stable isotope δ value; δX lis the original value of the lth stable isotope δ value; δX min Indicates the minimum value of the stable isotope δ value set, δX max is the maximum value of the δ value of this stable isotope.

[0056] In B4, the standardized water quality parameters and the normalized stable isotope δ values ​​are integrated into a comprehensive feature matrix X, where each row represents a mine water sample and each column represents a feature (water quality parameter or stable isotope δ value), expressed as:

[0057]

[0058] Among them, in this comprehensive feature matrix X, through x ij Represents each element, x ij is the value of the jth feature of the i-th mine water sample, such as x 11 It means i=1, j=1; m is the number of samples; n is the number of features.

[0059] In an optional embodiment, the water quality parameters of the mine water sample obtained in step S100 can also be continuously and automatically monitored by an online monitoring system, including the use of multi-parameter water quality online monitors, flow-through electrochemical sensors and other equipment to achieve real-time dynamic acquisition of water quality parameters.

[0060] In another optional embodiment, the comprehensive feature matrix established in step S100 may also use the independent component analysis (ICA) method to separate independent components in the mixed signal, or use the factor analysis method to identify potential common factors to improve the effectiveness of feature expression.

[0061] In an embodiment of the present invention, after establishing a comprehensive feature matrix, a generative adversarial network is used to generate synthetic data based on the comprehensive feature matrix, and the synthetic data is merged with the comprehensive feature matrix to form an enhanced comprehensive feature matrix.

[0062] In the existing mine water source identification process, the data of certain water source types may be sparse or unbalanced due to the following reasons: 1. Unique mine location: Some mines are located in special geological locations, and certain water source types are relatively rare, resulting in a small number of samples, making it difficult for the model to capture the characteristics of these few samples during training. 2. Sampling limitations: Due to sampling difficulties or insufficient historical data, the number of water samples of certain types is limited. 3. Environmental changes: As the climate or geological environment changes, samples of some water source types may become scarce or change. 4. The number of samples of different water source types may be seriously unbalanced, and samples of the majority category may dominate the model's learning process, causing the model to favor the majority category when identifying.

[0063] To solve the above three problems, the present invention introduces generative adversarial networks (GANs). In this embodiment, Wasserstein GAN (WGAN) is mainly used to generate synthetic data. To address the problem of data sparsity, the present invention generates synthetic samples, especially for samples of sparse categories, which can significantly balance the data distribution and enhance the learning ability of the model. In addition, the generated synthetic data can help the model better understand the characteristics of the data distribution and reduce its sensitivity to data fluctuations, thereby improving the stability and accuracy of the gray relational calculation. Finally, by generating synthetic samples close to the real data distribution, the diversity of the samples is increased, so that the model performs better when processing data in complex environments.

[0064] Generative Adversarial Networks (GANs) are a framework that includes two adversarial models: a generator and a discriminator. The generator is responsible for generating synthetic data that is similar to the distribution of real data, while the discriminator is responsible for distinguishing between the generated data and the real data. Through this adversarial training, the generator continuously improves the quality of the generated data and is eventually able to generate synthetic data that is very similar to the real samples. Wasserstein GAN is used in this embodiment. Wasserstein GAN (WGAN) is a variant of GANs that uses the Wasserstein distance to measure the difference between the distribution of generated data and the distribution of real data. Compared with traditional generative adversarial networks, WGAN has the following advantages: More stable training: 1. Wasserstein distance provides a smoother feedback signal, reducing the training instability problem common in traditional GANs. 2. High-quality generation: WGAN is able to generate higher quality synthetic data, especially when processing sparse or unbalanced data.

[0065] The generation of synthetic data includes steps C1-C6:

[0066] C1: Identify known water source types;

[0067] C2: Use the comprehensive feature matrix X constructed in B4 as the input of the Wasserstein GAN. Build a training set for each water source type so that the generator can learn the characteristic distribution of each water source type.

[0068] C3: Build WGAN model;

[0069] C4: train the WGAN model;

[0070] C5: After training is completed, the generator is used to generate a large number of samples as synthetic data;

[0071] C6: Merge the generated large number of samples with the original comprehensive feature matrix X to form an enhanced comprehensive feature matrix X'.

[0072] It should be further understood that the aforementioned water source types such as "surface water, shallow groundwater" are intended to illustrate the "preset water source types" applicable to the present invention, rather than to constitute a limitation thereto. The core advantage of the present invention lies in the versatility of its methodology, which can adapt to water source identification tasks under different geological conditions. In specific applications, those skilled in the art can flexibly determine the "preset water source types" to be identified and carry out targeted divisions based on the actual hydrogeological conditions of the area where the mine is located, historical exploration data or expert experience. For example, in a typical coal mining area in North China, the preset water source types can be specifically divided into Quaternary loose layer pore water, Shanxi Group sandstone fissure water, and Taiyuan Group and Ordovician limestone karst water. The technical solution proposed in the present invention can adaptively perform feature learning and correlation analysis on these specifically divided water source types, thereby achieving accurate identification.

[0073] It should be noted that in C2, the data for each water source type here refers to the water quality parameters and stable isotope data corresponding to each water source type.

[0074] In C3, Generator: The generator model receives a random noise vector z as input and transforms it into a generated sample X_gen through a series of neural network layers, which can also be called a synthetic sample X_gen.

[0075] Discriminator: The discriminator receives the real samples extracted from the comprehensive feature matrix X and the generated samples X_gen as input, and outputs the probability that they are real data.

[0076] The loss function L of the generator G for:

[0077]

[0078] Where z is a random noise vector, which is the input of the generator. This noise usually comes from a simple distribution (such as normal distribution or uniform distribution) and is used as the initial input of the generator. z (z) is the probability distribution of the random noise vector z. G(z) is the output of the generator, i.e., the generated sample X_gen. The generator G converts the random noise vector z into a synthetic sample with a distribution similar to the real data. D(·) is the output of the discriminator, representing the probability that the input data is a real sample. In a WGAN, the output of the discriminator is not a probability, but a real number that measures the similarity between the input sample and the real data. is the expected value operator, which means that the noise vector z is taken from its distribution p zThe expectation of all possible values ​​of (z).

[0079] The loss function L of the discriminator D for:

[0080]

[0081] Among them, x is a real sample, which comes from the comprehensive feature matrix X and is used to train the discriminator so that it can distinguish between real data and generated data; p data is the distribution of the real sample x, which describes the probability distribution of the real sample in the feature space; p gen is the distribution of generated samples X_gen, which is the distribution learned by the generator G model; λ is the weight of the gradient penalty, which controls the influence of the gradient penalty term (Gradient Penalty). The gradient penalty is used to constrain the gradient of the discriminator, ensuring that it has smooth behavior in the transition region between real samples and generated samples, thereby improving the stability of training. GP is the gradient penalty term, which is used to constrain the gradient of the discriminator to prevent the discriminator from over-learning. Its specific form is:

[0082]

[0083] in, The sample obtained by interpolation between the real sample x and the generated sample X_gen; is the gradient of the discriminator with respect to the input.

[0084] Preferably, the generator's loss function L G The main driving force is that the generator generates samples that are as "real" as possible, and the loss function L of the discriminator is D The discrimination ability is improved by distinguishing between real samples and generated samples, and the stability of training is improved by the gradient penalty term.

[0085] Specifically, in C4, the generator and discriminator are trained alternately, so that the generator can generate synthetic data with a distribution similar to that of real samples. The Wasserstein distance is used as a metric to ensure the stability of the training process and the quality of the generated samples.

[0086] Specifically, in C5-C6, after training is complete, the generator generates a large number of generated samples X_gen (especially for water source types with a small number of samples) as synthetic data. This large number of generated samples X_gen is merged with the original comprehensive feature matrix X to form an enhanced comprehensive feature matrix X'. It should be noted that the ratio of synthetic data to real data should be appropriate to avoid overfitting of the model due to excessive synthetic data.

[0087] In an optional embodiment, the generative adversarial network can also adopt a conditional generative adversarial network (cGAN), which introduces the water source type label as conditional information, so that the generator can generate more targeted synthetic samples for specific water source types.

[0088] In another optional embodiment, synthetic data generation can also use a variational autoencoder (VAE) to generate new samples with similar statistical properties by learning the potential representation distribution of the data. This is suitable for situations where the number of samples is extremely small.

[0089] In the embodiment of the present invention, the calculation of the correlation degree in step S200 adopts the grey correlation analysis method, and the correlation coefficient is determined by comparing the similarity between the characteristic sequence of the mine water sample and the characteristic sequence of each preset water source type, including steps D1-D4:

[0090] D1: Extract the feature matrix of known water source types from the enhanced comprehensive feature matrix X';

[0091] D2: Calculate the grey correlation between each water quality parameter in the mine water sample and the known water source type;

[0092] D3: Analyze the composition and proportion of multi-source water mixtures in mine water samples based on the comprehensive characteristic matrix;

[0093] D4: Determine the first mine water source identification result based on the correlation and the composition and proportion of the multi-source water mixture.

[0094] First, in D1, extract the feature matrix of the known water source type. From the enhanced comprehensive feature matrix X' generated by C6, extract the corresponding feature matrix X according to the known water source type determined in C1 i Each feature matrix X i The dataset contains raw data related to the water source type (i.e., water quality parameters and stable isotope data) and synthetic data enhanced by WGAN. These feature matrices will be used in subsequent grey relational analysis to determine the correlation between the mine water samples and each water source type.

[0095] Then, for each water source type feature matrix X extracted i Standardization is performed to ensure that the dimensions of each characteristic parameter are consistent for the subsequent grey relational degree calculation. Similarly, the characteristic matrix X_sample of the mine water sample is also standardized.

[0096] In D2, for each mine water sample, the grey correlation formula is used to calculate its correlation with each known water source type.

[0097] For example, the grey relational degree can be calculated using the following formula:

[0098]

[0099] Among them, γ(x0(k),x i (k)) is the mine water sample sequence x0 and the i-th preset water source type sequence x i The correlation coefficient on the kth feature; ρ is the resolution coefficient, usually set to 0.5, which is used to adjust the resolution of the correlation coefficient; min i min k |x0(k)-x i (k)| is the minimum difference between the two levels; max i max k |x0(k)-x i (k)| is the maximum difference between the two levels.

[0100] The correlation degree is obtained by averaging the correlation coefficients of each feature:

[0101]

[0102] Where n is the total number of features; γ(x0,x i ) is the overall correlation between the mine water sample and the i-th water source type.

[0103] It should be noted that the grey correlation analysis method mainly analyzes the degree of correlation between factors through the geometric correlation of the sequence, which better considers the correlation between data. Compared with the traditional correlation analysis, it has the advantages of low sample size requirements and no need for typical distribution assumptions. It is particularly suitable for complex situations of small and medium-sized samples and multiple factors in mine water source identification.

[0104] In an optional embodiment, the correlation degree may be calculated in step S200 by using an improved grey correlation analysis method, by introducing a weight factor to differentiate the importance of different features, or by using a dynamic resolution coefficient to adaptively adjust according to the data distribution characteristics, so as to further improve the accuracy of the correlation degree calculation.

[0105] In another optional embodiment, a linear mixing model can also be used to analyze the composition and proportion of the multi-source water mixture in step S200. By establishing a group of end-member mixing equations, the mixing proportion of each water source is solved using the least squares method or the non-negative least squares method.

[0106] In an embodiment of the present invention, the combined weight method in step S300 includes a combination of subjective weights and objective weights, and weight coefficients are determined for the water quality parameters and the stable isotope composition respectively through a weighted manner.

[0107] Exemplarily, the implementation of the combined weight method may include steps E1-E3:

[0108] E1: Determine subjective weights using the analytic hierarchy process;

[0109] E2: Calculate objective weights using the entropy weight method;

[0110] E3: The subjective weight and objective weight are combined by weighted average to obtain the final weight coefficient.

[0111] Specifically, in E2, the formula for calculating the objective weight using the entropy weight method is:

[0112]

[0113] Where w j is the objective weight of the jth feature; E j is the information entropy of the jth feature; m is the total number of features.

[0114] Information entropy E j The calculation formula is:

[0115]

[0116] Where n is the total number of samples; p ij is the relative frequency of the i-th sample on the j-th feature, and the calculation formula is:

[0117]

[0118] Where x ij is the value of the i-th sample on the j-th feature after standardization.

[0119] In E3, the formula for calculating the combination weight is:

[0120]

[0121] Where, is the final weight of the j-th feature; is the subjective weight; is the objective weight; α is the combination coefficient, which is usually 0.3-0.7.

[0122] It should be noted that the combined weighting method, by employing a weighting method that combines subjective and objective weights, not only considers the subjective judgment of expert experience and prior knowledge, but also incorporates objective analysis of the statistical characteristics of the data itself, achieving a more scientific and reasonable feature weighting. Compared with a single weighting method, this scheme can effectively balance the importance of different water quality parameters and stable isotope compositions in the identification process, avoiding the potential bias of purely subjective weighting and the domain knowledge that may be overlooked by purely objective weighting, significantly improving the accuracy and reliability of mine water source identification.

[0123] In an optional embodiment, the subjective weights determined in step S300 may also be determined using an expert scoring method, by inviting multiple hydrogeological experts to score the importance of each feature, and calculating the average value as the subjective weight to reduce the influence of individual expert subjective bias.

[0124] In another optional embodiment, the weight combination in step S300 can also adopt the product normalization method, by calculating the geometric mean of the subjective weight and the objective weight, or using an optimization algorithm to dynamically determine the optimal combination coefficient, and adjust the weight combination strategy according to the actual recognition effect feedback.

[0125] It should be noted that, through the comprehensive technical solution described above, this invention constructs a mine water source identification system that integrates data enhancement, correlation analysis, and weight optimization. Compared to traditional geophysical exploration methods, this solution addresses the issues of strong reliance on expert experience and low forecast accuracy. Through WGAN data enhancement technology, it effectively addresses the data sparsity issue for certain water source types. By combining grey correlation theory with a combined weighting method, it achieves accurate identification of multi-source water mixing and scientifically assigns feature weights, providing strong technical support for mine water hazard prevention and control.

[0126] Example 3 is an embodiment of the present invention, which provides a mine water source identification system based on combination weight and grey correlation theory.

[0127] Feature construction module, used to obtain water quality parameters and stable isotope composition of mine water samples and establish a comprehensive feature matrix;

[0128] The correlation analysis module is used to calculate the correlation between each water quality parameter and the preset water source type based on the comprehensive feature matrix, analyze the composition and proportion of the multi-source water mixture in the mine water sample, and determine the first mine water source identification result;

[0129] The weight processing module is used to use the combined weight method to weight the water quality parameters and stable isotope composition in the comprehensive feature matrix, and obtain the second mine water source identification result based on the weighted comprehensive feature matrix and the first mine water source identification result.

[0130] Example 4, reference Figure 2 , which is an embodiment of the present invention, provides a mine water source identification method based on combined weight and grey correlation theory. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0131] like Figure 2As shown, in this embodiment, 12 different locations of a mine are selected as test points (numbered 1-12), and corresponding mine water samples are obtained for water source identification verification. The experimental setting includes a baseline sample group and a test sample group, wherein the baseline sample group includes water area samples and water-free area samples. The water area samples are standard water samples collected from geological areas with confirmed water content, including known water source types such as groundwater and surface water. The water-free area samples are comparative samples collected from areas with confirmed water-free or interfering water sources. The test sample group includes test 1 samples and test 2 samples. Test 1 samples are unknown samples collected from actual water-containing areas and are used to verify the method's ability to identify real water samples. Test 2 samples are unknown samples collected from actual water-free areas and are used to verify the method's anti-interference ability and discrimination accuracy.

[0132] All collected samples were processed according to the method in Example 2. First, water quality parameters were obtained, and physical and chemical indicators such as pH value, conductivity, dissolved oxygen, total dissolved solids, hardness, and heavy metal concentration of each sample were measured. At the same time, stable isotope composition was obtained, and δ2H and δ 18 O stable isotope ratio. Data standardization was then performed, using the standardization formula to normalize the water quality parameters and the stable isotope δ values. Finally, a comprehensive feature matrix was established, integrating the standardized water quality parameters and stable isotope δ values ​​into a 12×n-dimensional comprehensive feature matrix.

[0133] Based on the established comprehensive feature matrix, the feature matrix was extracted from the benchmark samples of the water area and the waterless area as the reference sequence. For each test point sample, the grey correlation degree between it and the benchmark samples of the water area and the waterless area was calculated. The grey correlation degree calculation formula was used, in which the resolution coefficient was set to 0.5. The calculated grey correlation degree was then converted into the feature similarity distance as Figure 2 The "distance Ω·m" value on the Y axis. The preliminary water source type determination of each test point was determined through correlation analysis, completing the determination of the first mine water source identification result.

[0134] Based on the first identification results, a combined weighting method was used for further optimization. Subjective weights were determined using the analytic hierarchy process (AHP), objective weights were calculated using the entropy weighting method, and combined weights were obtained through weighted averaging. Weight coefficients were assigned to the water quality parameters and stable isotope composition in the comprehensive feature matrix. Based on the weighted comprehensive feature matrix and the first identification results, the second mine water source identification result was obtained, achieving the final and accurate identification of the mine water source.

[0135] from Figure 2From the experimental results, it can be observed that the characteristic distance curve of test 1 (actual water) shows a high degree of similarity with the baseline of the water area at multiple points such as No. 2, 4, 8, and 12, and the characteristic distance value is small, indicating that the gray correlation degree is high and the recognition is accurate. The characteristic distance curve of test 2 (actual no water) maintains consistency with the baseline of the water-free area at most points, and shows obvious differentiation at key nodes such as No. 6 and 10, verifying the anti-interference ability of the method. It is particularly noteworthy that at positions such as No. 6, the characteristic distance of the test sample is between the baseline with water and the baseline without water, indicating that there may be a multi-source water mixture phenomenon at this location, which meets the technical feature of "analyzing the composition and proportion of multi-source water mixture in mine water samples" in the claims.

[0136] Comparison with actual geological exploration results confirmed that the first sample correctly identified 10 of the 12 test points as water-bearing areas, with an accuracy rate of 83.3%. The second sample correctly identified 11 of the 12 test points as water-free areas, with an accuracy rate of 91.7%. The overall identification accuracy reached 87.5%, significantly higher than the 70-75% accuracy of traditional geophysical methods. All four curves showed good continuity and regularity at the 12 test points, demonstrating the excellent stability and reproducibility of the proposed method.

[0137] Experimental results fully demonstrate the technical advantages of the proposed method. WGAN data augmentation technology effectively addresses the issue of insufficient samples for certain water source types. The application of a combined weighting approach makes the distribution of feature weights more scientific and reasonable, avoiding the limitations of single-weight methods. Grey correlation theory exhibits excellent adaptability to complex situations involving small samples and multiple factors, enhancing the method's anti-interference capabilities. More importantly, the proposed method not only qualitatively identifies water source types but also quantitatively analyzes the composition ratios of mixed water sources, providing important technical support for mine water hazard prevention and control.

[0138] It should be noted that this embodiment verifies the effectiveness and reliability of the mine water source identification method based on combined weights and grey correlation theory in practical applications. Compared with traditional geophysical prospecting methods, the present invention solves the problems of strong dependence on expert experience and low prediction accuracy. Through quantitative data analysis and scientific weight distribution, it realizes the accurate identification of mine water sources, and has good engineering application prospects and promotion value.

[0139] Example 5 is the fifth embodiment of the present invention. It is different from the previous four embodiments in that if the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0140] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0141] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0142] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0143] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A mine water source identification method based on combined weight and grey correlation theory is characterized by: include: Obtain water quality parameters and stable isotope composition of mine water samples and establish a comprehensive characteristic matrix; Calculating the correlation between each of the water quality parameters and the preset water source type based on the comprehensive feature matrix, analyzing the composition and proportion of the multi-source water mixture in the mine water sample, and determining a first mine water source identification result; The water quality parameters and the stable isotope composition in the comprehensive feature matrix are weighted using a combined weight method, and a second mine water source identification result is obtained based on the weighted comprehensive feature matrix and the first mine water source identification result.

2. The mine water source identification method based on combined weight and grey correlation theory according to claim 1, characterized in that: The water quality parameters are obtained by measuring the physical and chemical indicators of the collected mine water samples; the stable isotope composition is obtained by analyzing the stable isotope data in the mine water samples.

3. The mine water source identification method based on combined weight and grey correlation theory according to claim 2, characterized in that: After establishing the comprehensive feature matrix, a generative adversarial network is used to generate synthetic data based on the comprehensive feature matrix, and the synthetic data is merged with the comprehensive feature matrix to form an enhanced comprehensive feature matrix.

4. The mine water source identification method based on combined weight and grey relational theory according to claim 3, characterized in that: The correlation degree is calculated using a grey correlation analysis method, and the correlation coefficient is determined by comparing the similarity between the mine water sample feature sequence and the feature sequence of each preset water source type.

5. The mine water source identification method based on combined weight and grey relational theory according to claim 4, characterized in that: The combined weight method includes a combination of subjective weights and objective weights, and weight coefficients are determined for the water quality parameters and the stable isotope composition respectively in a weighted manner.

6. The mine water source identification method based on combined weight and grey correlation theory according to claim 5, characterized in that: The generative adversarial network is a Wasserstein GAN, which includes a generator and a discriminator, wherein the generator receives a random noise vector as input and converts it into a generated sample through a neural network layer; the discriminator receives a real sample extracted from the comprehensive feature matrix and the generated sample as input, and outputs the probability that the real sample and the generated sample are real data; When training the Wasserstein GAN, the generator and the discriminator are trained alternately using the Wasserstein distance as a metric; After the training is completed, the generator is used to generate a plurality of generated samples as the synthetic data.

7. The mine water source identification method based on combined weight and grey relational theory according to claim 6, characterized in that: The grey correlation analysis method includes: extracting the feature matrix of the preset water source type from the comprehensive feature matrix or the enhanced comprehensive feature matrix; calculating the grey correlation between each of the water quality parameters in the mine water sample and the preset water source type, wherein the correlation calculation is achieved by averaging the correlation coefficient adjusted by the resolution coefficient; and determining the first mine water source identification result based on the correlation and the composition and proportion of the multi-source water mixture.

8. A mine water source identification system based on combined weight and grey correlation theory, applying the mine water source identification method based on combined weight and grey correlation theory as claimed in any one of claims 1 to 7, characterized in that: include: Feature construction module, used to obtain water quality parameters and stable isotope composition of mine water samples and establish a comprehensive feature matrix; a correlation analysis module, configured to calculate the correlation between each of the water quality parameters and a preset water source type based on the comprehensive feature matrix, analyze the composition and proportion of the multi-source water mixture in the mine water sample, and determine a first mine water source identification result; A weight processing module is used to use a combined weight method to weight the water quality parameters and the stable isotope composition in the comprehensive feature matrix, and obtain a second mine water source identification result based on the weighted comprehensive feature matrix and the first mine water source identification result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the mine water source identification method based on combined weight and grey correlation theory according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the mine water source identification method based on combined weight and grey relational theory according to any one of claims 1 to 7 are implemented.

Images