A method for constructing a carbon value-added valuation prediction model of a tailings pond area and application thereof

Abstract

The application of "a tailing pond area carbon value increment valuation method and application thereof", relating to tailing pond carbon sink calculation technology, comprising the following steps: S1. Constructing a carbon value increment valuation prediction model, S11. In the first to be evaluated tailing pond, adding pioneer carbon source in the surface tailings thereof, and randomly sampling the surface tailings of the first to be evaluated tailing pond at different time points; S12. Detecting and characterizing the samples of all sampling points at all sampling time points to obtain the carbon value increment data set of the tailing pond area; S13. Data processing of the carbon value increment data set: obtaining the training set and the test set; S14. Obtaining a plurality of initial models based on machine learning algorithms, setting the initial hyperparameter search range for each initial model; training and testing each initial model using the training set and the test set to obtain the optimal carbon value increment valuation prediction model and determine the key parameters affecting the carbon value increment valuation; the method can select the most suitable carbon value increment valuation prediction model by learning the sampling data of the target tailing pond. When encountering new situations, the model and its parameters can be automatically adjusted to adapt to new data environment and tasks.

Description

Technical Field

[0001] This invention relates to soil remediation technology for tailings pond areas, and more particularly to a method for determining carbon increment in tailings pond areas and its application. Background Technology

[0002] As an important component of mining activities, the carbon sequestration function of tailings ponds still faces many challenges in research. The theoretical foundation for tailings pond carbon sequestration research is weak, lacking systematic research methods and tools. For example, there is a lack of quantitative calculation and prediction methods for changes in carbon sequestration in tailings ponds, leading to inaccurate carbon sequestration assessment results. Furthermore, the research and application of tailings pond carbon sequestration enhancement technologies are relatively lagging.

[0003] Accurate assessment of the carbon sequestration function of tailings ponds is essential for providing a scientific basis for environmental governance and ecological restoration of tailings ponds. Therefore, developing a method to quickly and accurately calculate the carbon increment of tailings pond areas is of great significance. Summary of the Invention

[0004] In view of the shortcomings and needs in the above-mentioned fields, the present invention provides a method for carbon increment determination in tailings pond areas, aiming to improve the ecological environment quality of tailings pond areas and promote soil remediation and vegetation restoration of tailings waste sites.

[0005] This method involves using carbon-based microbial agents as a pioneer carbon source, mixing them into the surface tailings of the tailings pond area, and using various characterization techniques to quantitatively analyze the properties of the surface tailings of the tailings pond area before and after the application of the pioneer carbon source. By integrating the obtained target tailings pond area carbon value-added dataset, the data is input into a preset carbon value-added prediction model to obtain the prediction results of the carbon value-added amount of the tailings pond area.

[0006] Specifically, the technical solution of the present invention is as follows: A first aspect of the present invention provides a method for determining the carbon increment in a tailings dam area, characterized by comprising the following steps: S1. Construct a carbon value-added prediction model. S11. In the first tailings pond to be evaluated, a pioneer carbon source is added to its surface tailings, and multiple random samples are taken from the surface tailings of the first tailings pond to be evaluated at different time points; the different time points refer to 10 to 20 time points between 1 day before the addition of the pioneer carbon source and 5 to 150 days after the addition of the pioneer carbon source. S12. Detect and characterize the samples from all sampling points at all sampling time points to obtain the carbon added value dataset of the tailings pond area; The carbon value-added dataset for the tailings pond area includes the following parameter types: percentage of large aggregates, porosity, cation exchange capacity, pH, available nitrogen content, available phosphorus content, available potassium content, Bacillus abundance, available heavy metal content, heavy metal ion radius, particulate organic carbon content, and mineral-bound organic carbon content. S13. Perform data processing on the carbon added value dataset: The data processing includes data cleaning, Pearson correlation analysis, and data partitioning to obtain the dataset for the machine learning model; Data cleaning refers to removing abnormal data; Pearson correlation analysis is used to exclude collinear data; The data partitioning refers to dividing the dataset of the machine learning model into a training set and a test set; S14. Obtain multiple initial models based on machine learning algorithms, and set the initial hyperparameter search range for each initial model; The training and test sets are used to train and test each initial model; the initial models include: random forest regression model, multilayer perceptron regression model, support vector regression model, and gradient boosting regression model; The training process includes adjusting and optimizing the number of hyperparameter search ranges for each initial model using a gridded search method, and obtaining the optimal hyperparameter combination for each initial model by combining ten-fold cross-validation. The initial models are iterated and screened based on the optimal combination of hyperparameters of each initial model, and the optimal model obtained is the carbon value increment prediction model. S2. Perform feature importance analysis based on the running results of the carbon value increment prediction model to determine the key parameters affecting the carbon value increment.

[0007] Preferably, the tailings dam area carbon increment determination method is characterized in that the multi-point random sampling in S11 includes sampling at 80 to 150 random sampling points.

[0008] Preferably, the method for determining the carbon increment of the tailings dam area is characterized in that, in step S13, the abnormal data removal includes: By calculating the Z-score for each data point, data with a Z-score greater than 3, and data with a Z-score of 0 at some sampling points due to lack of data acquisition; The data splitting is performed using the train_test_split function built into the scikit-learn package to divide the dataset of the machine learning model, with a training set to test set ratio of 7:3; In S13, the Pearson correlation analysis showed r < 0.6.

[0009] Preferably, the tailings dam area carbon increment determination method is characterized in that the optimization results in S15 include the optimal hyperparameter combination, translatable variance, mean absolute error, mean square error, median absolute error, and coefficient of determination.

[0010] Preferably, the method for determining the carbon increment of the tailings dam area is characterized by setting the following initial hyperparameter search ranges for each preliminary model: Random forest regression model: 'n_estimators': range(1, 1000, 10), 'max_depth': range(1, 1000, 10), 'min_samples_split': range(2, 10, 1), 'min_samples_leaf': range(1, 10, 1), 'max_features': ['auto', 'sqrt', 'log2']; Multilayer perceptron regression model: 'hidden_layer_sizes': [(50,), (100,), (100, 100),(50, 50, 50)], 'activation': ['tanh', 'relu'], 'alpha': range(0.0001, 0.01,0.0001), 'learning_rate': ['constant', 'invscaling', 'adaptive']; Support Vector Regression Model: 'C': range(0.1, 100, 0.1), 'epsilon': range(0.01,0.5, 0.01), 'gamma': ['scale', 'auto']; Gradient boosting regression model: 'n_estimators': range(1, 1000, 10), 'max_depth': range(1, 1000, 10), 'min_samples_split': range(2, 10, 1), 'min_samples_leaf': range(1, 10, 1), 'max_features': ['auto', 'sqrt', 'log2'].

[0011] Preferably, in the tailings pond area carbon value determination method, the different time points mentioned in S11 refer to 1 day before the addition of the pioneer carbon source, and the 5th, 10th, 20th, 30th, 40th, 50th, 60th, 70th, 80th, 90th, 100th, 120th or 150th day thereafter.

[0012] Preferably, in any of the above-mentioned methods for determining carbon value increment in tailings pond areas, the pioneer carbon source is a carbon-based microbial inoculant, prepared by the following steps: (1) Obtaining biochar; (2) Iron modification of biochar: The biochar was immersed in a saturated FeCl3 solution at a mass ratio of 1:15 to 25 for 10 to 24 h; then the biochar was dried at 105 °C for 10 to 14 h, heated to 400 to 600 °C, and kept in a muffle furnace under pure nitrogen for 50 to 70 min; after cooling, the iron-modified biochar was washed several times with deionized water until pH=7±0.3, and then completely dried at 25±2 °C; (3) Iron-modified biochar is placed in a Bacillus suspension to adsorb Bacillus, thereby obtaining the carbon-based microbial agent; Adding a pioneer carbon source to the surface tailings refers to uniformly spraying a carbon-based microbial agent into the surface tailings at a depth of 0-20 cm, with an addition amount of 100-500g of the carbon-based microbial agent per square meter of surface tailings.

[0013] 8. The method for determining carbon value increment in tailings pond area according to claim 7, characterized in that the biochar is made from rice husks, which are dried, ground, and sieved through a 2 mm sieve, mixed with an equal mass of sodium hydroxide, and then placed in a sealed iron container. It was then placed in a tube furnace for pyrolysis, with a heating rate set at 10 °C·min. 1 The pyrolysis temperature was 400℃ for 4 hours; after pyrolysis, the biochar was obtained by cooling to room temperature.

[0014] Another aspect of the present invention provides an application of the tailings dam area carbon increment determination method according to any one of claims 1 to 8, characterized in that, (1) For new tailings ponds to be evaluated, add a pioneer carbon source to the surface tailings of the pond area; (2) At any time point T after the addition of the pioneer carbon source, 5 to 10 sampling points were randomly selected from the surface soil of the tailings pond area for sampling. (3) Obtain the carbon value-added prediction model and key parameters established based on the first tailings dam to be evaluated in the carbon value-added determination method of any one of claims 1 to 7; (4) The samples are tested and characterized to obtain the values ​​of the key parameters, which are used as input variables for the carbon value-added prediction model; (5) Input the input variables into the carbon value-prediction model to obtain the predicted carbon value of the tailings dam to be evaluated at any time point Tn during the 150-day period after time point T.

[0015] Preferably, the application is characterized in that, at the time point Tn where the predicted carbon increment is obtained, random sampling and detection characterization are performed on the new tailings dam area to be evaluated to obtain the measured carbon increment, and the predicted carbon increment is compared with the measured value to verify the accuracy of the carbon increment prediction model. When the explanatory variance is ≤0.6, the mean absolute error is ≥3, the mean square error is ≥3, the median absolute error is ≥3, and the coefficient of determination is ≤0.6, the prediction accuracy is deemed unsatisfactory and the model parameters need to be updated. The method for updating the model parameters is as follows: Based on the new tailings dam to be evaluated, proceed to steps S11-S14 of any of the methods in claims 1-8 to obtain a carbon value-added prediction model with updated parameters; then proceed to step S2 to obtain updated key parameters. The new tailings pond to be evaluated refers to other tailings ponds that are different from the first tailings pond to be evaluated, or to the original target tailings ponds whose prediction accuracy of the original carbon value increment prediction model has decreased due to environmental changes, and whose model parameters need to be updated.

[0016] Technical advantages of the present invention: This invention provides a method for constructing an optimal carbon increment prediction model for a target tailings dam and then tracking and determining the carbon sequestration capacity of the tailings dam based on this model. This method can select the most suitable carbon increment prediction model by learning from sampled data of the target tailings dam. It can automatically adjust the model and its parameters to adapt to new data environments and tasks when encountering new situations; whereas traditional methods require manual intervention to adjust parameters. The carbon increment prediction method of this invention can identify new patterns that have never been encountered before, while traditional calculation methods can usually only handle cases that are explicitly programmed to be processed.

[0017] The carbon increment determination method for tailings ponds provided by this invention provides a scientific basis for the ecological restoration and carbon sink function enhancement and research of tailings pond areas. The carbon increment determination method of this invention provides carbon-based microbial agents, which have a significant heavy metal passivation effect and can significantly improve the soil structure and environmental quality of tailings pond areas.

[0018] The method of this invention is simple to operate, has a wide range of applications, and produces accurate results. It is of great value for the research on environmental management and sustainable development of lead-zinc tailings pond areas. Attached Figure Description

[0019] Figure 1 The preparation process of carbon-based microbial inoculants; Figure 2 The process of constructing a carbon added value prediction model. Detailed Implementation

[0020] The technical solution of the present invention will be described in an exemplary, clear, and complete manner below with reference to some embodiments thereof. However, these exemplary embodiments should not be regarded as limiting the scope of protection of the present invention.

[0021] Example 1. Preparation and application of carbon-based microbial inoculants (1) Obtaining biochar: Rice husks from farmland around the tailings pond area are collected, dried, ground and passed through a 2 mm sieve. An equal mass of sodium hydroxide is added and mixed evenly before being placed into a sealed iron crucible. Then, pyrolysis is carried out in a tube furnace at a heating rate of 10℃ / min and a pyrolysis temperature of 400℃ for 4 hours. After pyrolysis, the biochar is obtained by cooling to room temperature.

[0022] (2) Iron modification of biochar: The biochar was immersed in a saturated FeCl3 solution at a mass ratio of 1:20 for 24 hours; the treated biochar was dried at 105℃ for 12 hours, heated to 500℃ and kept in a muffle furnace under pure nitrogen for 1 hour; after cooling, the iron-modified biochar was thoroughly washed and completely dried at 25±2℃.

[0023] (3) Prepare a bacterial suspension by preparing the Bacillus agent, and then add biochar directly to the bacterial suspension (the mass ratio of bacterial suspension to iron-modified biochar is 5:1), mix well so that the Bacillus agent is loaded onto the iron-modified biochar by adsorption, and obtain the carbon-based microbial agent. (4) Apply a pioneer carbon source to the tailings pond: break up the large and agglomerated tailings, remove the garbage remaining on the surface of the pond area, and use a deep turner to add the carbon-based microbial agent to the 0-20 cm surface tailings at a dosage of 100-500g per square meter.

[0024] The Bacillus agent required in this invention is a conventional commercially available agent; the Bacillus agent used in this embodiment was purchased from the Shanghai Microbiology Preservation Center and contains Bacillus subtilis and Bacillus spp. (the effective live bacteria ratio of the two bacteria in the above bacterial suspension is 50:1), with catalog numbers ATCC6633 and SHBCC D17966AS 1.232, respectively.

[0025] Example 2. Constructing a carbon value-added prediction model Step 1. Sample the tailings pond area at different time points (1 day before the application of the pioneer carbon source, and 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 120 or 150 days after the addition of the pioneer carbon source); Step 2 involves obtaining detection data of tailings samples at all time points and sampling points through a series of detection and characterization techniques. This data includes aggregate composition, porosity, cation exchange capacity, pH, available nitrogen content, available phosphorus content, available potassium content, microbial abundance index, microbial diversity index, available heavy metal content, soil particulate organic carbon content, and mineral-bound organic carbon content, in order to construct the carbon value-added dataset for the tailings pond area. The characterization techniques include: Aggregate composition: The mass proportions of soil aggregates at various levels were determined using dry and wet sieving methods. Untreated tailings samples were peeled off along their natural surface into blocks with a diameter of 10 mm; after air drying, they were sieved through sieves of different aperture sizes.

[0026] Based on the percentage of each grade of material obtained from dry sieving, the air-dried samples obtained from dry sieving are mixed in equal proportions, fully moistened, and then sieved. Aggregates are separated by the action of water flow, and after rinsing, drying, and weighing, the mass ratio of each grade of soil aggregates is determined.

[0027] Porosity was determined using X-ray computed tomography. Basic tailings sample images were acquired using X-ray computed tomography (CT) scans; then, image reconstruction was performed using phase-sensitive X-ray image processing and tomographic reconstruction software, and each tailings sample was analyzed as a 32-bit TIF grayscale image; the porosity was quantitatively analyzed using ImageJ software.

[0028] Cation exchange capacity was determined using the hexaamminecobalt trichloride spectrophotometric method. Take a tailings sample that has been treated with a 2 mm dry sieve, mix it with a solution of hexaamminecobalt trichloride, centrifuge after equilibration, and measure the absorbance of the supernatant.

[0029] pH was determined using the deionized water leaching method. Take an appropriate amount of tailings sample, add deionized water at a solid-liquid ratio of 1:2.5, shake to mix, let stand, and use a pH meter to measure the pH value of the supernatant to reflect the acid-base status of the tailings.

[0030] Available nitrogen, phosphorus, and potassium were determined using standard methods: The available nitrogen in the soil was determined by alkaline hydrolysis-diffusion method, the available phosphorus in the soil was determined by hydrochloric acid-hydrofluoric acid digestion-molybdenum antimony colorimetric method, and the available potassium was determined by ammonium acetate extraction-flame photometry method to assess the nutrient element content.

[0031] Microbial richness index and microbial diversity index: Sequence reads obtained using 16S rRNA gene sequencing technology were used to estimate the relative abundance of Bacillus spp. in the entire microbial community.

[0032] The available content of heavy metals was determined using the diethylenetriaminepentaacetic acid extraction method. The available heavy metals in the tailings were extracted by leaching with a diethylenetriaminepentaacetic acid-calcium chloride-triethanolamine buffer solution, and their contents were determined by inductively coupled plasma atomic emission spectrometry.

[0033] Soil particulate organic carbon and mineral-bound organic carbon were determined using an elemental analyzer: The tailings samples were extracted and analyzed using the sodium hexametaphosphate extraction method. They were sieved into different particle sizes using a 53μm sieve, and the elemental content of each part was determined by an elemental analyzer.

[0034] The initial data at all sampling points at all time points are obtained to form a carbon value-added dataset. A sample of the data in the dataset is shown in Table 1 below.

[0035] Table 1. Centralized Data Composition of Carbon Value Added (Example)

[0036] Step 3: Process the carbon added value dataset. Data processing includes data cleaning, Pearson analysis, and partitioning.

[0037] The data cleaning process includes outlier removal, which removes outliers (values ​​with a Z score greater than 3 are considered outliers by calculating the Z score of each data point) and values ​​that are 0 at a certain sampling point because no data was obtained. Collinearity was excluded by Pearson correlation analysis (r<0.6); The partitioning process includes using the built-in train_test_split function in the scikit-learn package to divide the original dataset at each time point into a training set and a test set; The random seed is fixed at 42 to ensure repeatability, and the training set:test set ratio is 7:3.

[0038] The statistical values ​​of the training or test set data are shown in Table 2 below, which represents the input feature dataset.

[0039] Table 2. Description of the input feature dataset for the carbon value increment prediction model.

[0040] Step 4. Use initial models with different machine learning algorithms, including random forest regression model, multilayer perceptron regression model, support vector regression model and gradient boosting regression model; The initial algorithm hyperparameter search range is set for each initial model, as shown in Table 2.

[0041] Table 3 Initial Algorithm Hyperparameter Search Range

[0042] Then, the initial models are trained and tested using the training and test sets obtained in step 3 to determine the performance of each initial model.

[0043] This paper employs a random forest regression model algorithm, utilizing ensemble learning to construct multiple decision trees as weak learners. During training, a subset of samples and features are randomly selected for training to ensure model diversity and generalization ability. Each decision tree performs regression prediction on the data, and the predictions from all decision trees are arithmetically averaged to obtain the final output of the random forest regression model, thus enabling prediction of the test data. The algorithm employing a multilayer perceptron regression model involves constructing a neural network comprising an input layer, a hidden layer, and an output layer. Activation functions are used to nonlinearly transform the data between layers, and backpropagation is employed to optimize the network weights, enabling the model to fit complex nonlinear relationships. During training, the predictive ability of the multilayer perceptron regression model is improved by minimizing the difference between predicted and actual values.

[0044] The algorithm employs a support vector regression model, seeking an optimal hyperplane that minimizes the sum of distances from all sample points to that hyperplane. Simultaneously, slack variables and penalty coefficients are introduced to allow for a certain degree of data deviation. During the optimization process, regression prediction of the tested data is achieved by minimizing the total loss function and maximizing the interval width, thereby improving the predictive ability of the support vector regression model.

[0045] This algorithm employs a gradient boosting regression model, iteratively constructing multiple weak learners (in this example, decision trees), with each learner attempting to correct the errors of the previous learner. During training, the model's predictive performance is gradually improved by optimizing the gradient direction of the loss function. Finally, the predictions from all weak learners are summed to obtain the final output of the gradient boosting regression model, thus enabling the prediction of the test data.

[0046] The initial hyperparameters of each initial model were adjusted and optimized using a gridded search method, and the optimal hyperparameter combination for each initial model was obtained by combining it with ten-fold cross-validation (as shown in Table 4). The prediction accuracy of each initial learning model under the optimal hyperparameter combination is evaluated using five indicators: interpretable variance, mean absolute error, mean squared error, median absolute error, and coefficient of determination, as described in Table 3.

[0047] Table 4. Optimization results and carbon increment prediction accuracy of each initial model

[0048] As shown in Table 4, the gradient boosting regression model outperforms the other models in all five evaluation indicators. Therefore, the gradient boosting regression model is adopted as the carbon breeding value prediction model.

[0049] The initialization process for the gradient boosting regression model is as follows:

[0050] Where L(y) i , γ ) is the loss function, y i It is the actual value. γ It is a model prediction.

[0051] The iterative process for determining the number of trees is as follows: (1). Calculate the residuals:

[0052] in F m-1 It is the predicted value of the model after the (m-1)th iteration.

[0053] (2). Using residual r min Using this as the target variable, a new regression tree is trained to obtain the leaf node regions of the m-th tree. R jm ,in j=1, 2,…, J J is the number of leaf nodes: (3). For each leaf node j Find the optimal value that minimizes the loss function. :

[0054] (4). Update the model:

[0055] in It is an indicator function, when x is in the leaf node region. R jm The value is 1 if the time is within the range, and 0 otherwise.

[0056] Step 6. Perform feature importance analysis on the input feature dataset described in Table 3 by running the above model. The analysis results show that the key parameters affecting the accuracy of carbon increment prediction in tailings pond areas are: cation exchange capacity (cmol / kg), pH, available nitrogen (mg / kg), available phosphorus (mg / kg), Bacillus spp. (%), heavy metal electronegativity, and heavy metal ion radius (pm). Their influence weights on the model prediction results are listed in Table 5. Table 5. Characteristic Importance Analysis of Carbon Value Added Prediction.

[0057] In subsequent references, it is only necessary to randomly sample and detect key parameters (data from a time section) at a single point in time as input variables for the model to obtain the predicted carbon increment of the tailings dam.

[0058] In different situations, such as when facing different tailings ponds, or when the environment of the tailings pond used to build the model changes significantly, and the prediction accuracy of the original model decreases, under the framework of the method established in this invention, staff can provide an updated carbon value-added dataset based on the new situation to perform the above steps 2 to 5 and obtain an updated prediction model.

[0059] Example 3. Application of carbon value-added determination in the same tailings pond area In Example 2, after constructing a carbon breeding value prediction model and obtaining key parameters, the dynamics of carbon sink levels before and after adding a pioneer carbon source at a certain stage were monitored.

[0060] S1. Sampling and testing: On the same day (time point T) after adding the pioneer carbon source, multiple random sampling points (5-8 sampling points) are carried out in the tailings pond area; according to the key parameters affecting the carbon increment of the target tailings pond area determined in step 6 of Example 2, the samples are tested and characterized to obtain the input variables of the carbon increment prediction model; The input variables at time point T are shown in Table 6. Table 6. Description of the input feature dataset for carbon value increment prediction (input variables)

[0061] S2. Input the data (input variables) listed in Table 6 into the carbon value-prediction model-gradient boosting regression model constructed in Example 2, and run the calculation to obtain the carbon value-added data of the tailings pond area at sampling time point T, as shown in Table 6. Then, predict the carbon value-added for 10 to 150 days after adding the pioneer carbon source according to the time series. The prediction results for some time points are shown in Tables 7 to 9.

[0062] Table 6 Actual values ​​of carbon sequestration data for day T at time point 6

[0063] Table 7. Predicted carbon increment (carbon sink data) 10 days after time point T based on key parameters.

[0064] Table 8. Carbon increment (carbon sink data) forecast data 90 days after time point T

[0065] Table 9. Carbon increment (carbon sink data) forecast data 150 days after time point T

[0066] Subsequently, random sampling was performed at the aforementioned prediction time points to obtain measured values, which were then compared and analyzed with the predicted values ​​to test the accuracy of the model's predictions.

[0067] Table 10 Comparison and Analysis of Simulation Results and Measured Values

[0068] Table 10 shows that the carbon increment prediction method provided by this invention is accurate and reliable with small error and a determination coefficient of over 0.9, which can accurately predict the carbon increment of tailings ponds.

[0069] The carbon increment determination method provided by this invention can significantly reduce the workload of traditional monitoring methods. When there are major changes in the tailings dam environment, only an updated tailings dam carbon increment dataset is needed to obtain a model and key parameter combination that is more suitable for the new situation and can be used for prediction under the new situation.

[0070] Example 4. Effects of different external carbon sources on carbon increment in tailings ponds. Comparison of the carbon-based microbial inoculant of the present invention with existing conventional soil amendment microbial inoculants.

[0071] Microbial agents for conventional soil amendment: biological organic fertilizer purchased from Hebei Xiangniu Fertilizer Co., Ltd.

[0072] S1. Conventional microbial agents are used as the primary carbon source; S2. Sampling and testing: Two samplings were conducted on the tailings pond area before and after the application of the pioneer carbon source, and the samples from each sampling point were quantitatively analyzed. S3. Process the quantitative analysis data.

[0073] Monitoring data show that the carbon-based microbial agent provided by this invention has a significantly different effect on tailings ponds compared with existing conventional soil amendment microbial agents, mainly in terms of the passivation effect on heavy metals, as detailed in Tables 11 and 12 below.

[0074] Table 11 Carbon increase in tailings pond area before and after adding the carbon-based microbial agent of the present invention.

[0075] Table 12 Bio-organic Fertilizer

[0076] Compared with existing conventional soil amendment microbial agents, the carbon-based microbial agent provided by this invention has a more significant effect on heavy metal passivation.

[0077] In summary, the above-described technical solutions of the present invention are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. All equivalent structural transformations made using the contents of the present invention's specification and drawings under the technical concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A method for determining the carbon increment value of a tailings dam area, characterized in that, Includes the following steps: S1. Construct a carbon value-added prediction model. S11. In the first tailings pond to be evaluated, a pioneer carbon source is added to its surface tailings, and multiple random samples are taken from the surface tailings of the first tailings pond to be evaluated at different time points; the different time points refer to 10 to 20 time points between 1 day before the addition of the pioneer carbon source and 5 to 150 days after the addition of the pioneer carbon source. S12. Detect and characterize the samples from all sampling points at all sampling time points to obtain the carbon added value dataset of the tailings pond area; The carbon value-added dataset for the tailings pond area includes the following parameter types: percentage of large aggregates, porosity, cation exchange capacity, pH, available nitrogen content, available phosphorus content, available potassium content, Bacillus abundance, available heavy metal content, heavy metal ion radius, particulate organic carbon content, and mineral-bound organic carbon content. S13. Perform data processing on the carbon added value dataset: The data processing includes data cleaning, Pearson correlation analysis, and data partitioning to obtain the dataset for the machine learning model; Data cleaning refers to removing abnormal data; Pearson correlation analysis is used to exclude collinear data; The data partitioning refers to dividing the dataset of the machine learning model into a training set and a test set; S14. Obtain multiple initial models based on machine learning algorithms, and set the initial hyperparameter search range for each initial model; The training and test sets are used to train and test each initial model; The initial models include: random forest regression model, multilayer perceptron regression model, support vector regression model, and gradient boosting regression model; The training process includes adjusting and optimizing the number of hyperparameter search ranges for each initial model using a gridded search method, and obtaining the optimal hyperparameter combination for each initial model by combining ten-fold cross-validation. The initial models are iterated and screened based on the optimal combination of hyperparameters of each initial model, and the optimal model obtained is the carbon value increment prediction model. S2. Perform feature importance analysis based on the running results of the carbon value increment prediction model to determine the key parameters affecting the carbon value increment.

2. The method for determining the carbon increment of a tailings dam area according to claim 1, characterized in that, The multi-point random sampling described in S11 includes sampling from 80 to 150 random sampling points.

3. The method for determining the carbon increment of a tailings dam area according to claim 1, characterized in that, The abnormal data to be removed as described in S13 includes: By calculating the Z-score for each data point, data with a Z-score greater than 3, and data with a Z-score of 0 at some sampling points due to lack of data acquisition; The data splitting is performed using the train_test_split function built into the scikit-learn package to divide the dataset of the machine learning model, with a training set to test set ratio of 7:3; In S13, the Pearson correlation analysis showed r < 0.

6.

4. The method for determining the carbon increment of a tailings dam area according to claim 1, characterized in that, The optimization results described in S15 include the optimal combination of hyperparameters, the explanatory variance, the mean absolute error, the mean squared error, the median absolute error, and the coefficient of determination.

5. The method for determining the carbon increment of a tailings dam area according to claim 1, characterized in that, The initial hyperparameter search ranges for each initial model are set as follows: Random forest regression model: 'n_estimators': range(1, 1000, 10), 'max_depth': range(1, 1000, 10), 'min_samples_split': range(2, 10, 1), 'min_samples_leaf': range(1, 10, 1), 'max_features': ['auto', 'sqrt', 'log2']; Multilayer perceptron regression model: 'hidden_layer_sizes': [(50,), (100,), (100, 100), (50,50, 50)], 'activation': ['tanh', 'relu'], 'alpha': range(0.0001, 0.01,0.0001), 'learning_rate': ['constant', 'invscaling', 'adaptive']; Support Vector Regression Model: 'C': range(0.1, 100, 0.1), 'epsilon': range(0.01, 0.5,0.01), 'gamma': ['scale', 'auto']; Gradient boosting regression model: 'n_estimators': range(1, 1000, 10), 'max_depth': range(1, 1000, 10), 'min_samples_split': range(2, 10, 1), 'min_samples_leaf': range(1, 10, 1), 'max_features': ['auto', 'sqrt', 'log2'].

6. The tailings dam area carbon value determination method according to claim 1, wherein the different time points mentioned in S11 refer to 1 day before the addition of the pioneer carbon source, and the 5th, 10th, 20th, 30th, 40th, 50th, 60th, 70th, 80th, 90th, 100th, 120th or 150th day thereafter.

7. The method for determining the carbon increment of a tailings dam area according to any one of claims 1 to 6, characterized in that, The pioneer carbon source is a carbon-based microbial inoculant, which is prepared by the following steps: (1) Obtaining biochar; (2) Iron modification of biochar: The biochar was immersed in a saturated FeCl3 solution at a mass ratio of 1:15 to 25 for 10 to 24 hours; then the biochar was dried at 105 °C for 10 to 14 hours, heated to 400 to 600 °C, and kept in a muffle furnace under pure nitrogen for 50 to 70 minutes; after cooling, the iron-modified biochar was washed several times with deionized water until pH=7±0.3, and then completely dried at 25±2 °C; (3) Iron-modified biochar is placed in a Bacillus suspension to adsorb Bacillus, thereby obtaining the carbon-based microbial agent; Adding a pioneer carbon source to the surface tailings refers to uniformly spraying a carbon-based microbial agent into the surface tailings at a depth of 0-20 cm, with an addition amount of 100-500g of the carbon-based microbial agent per square meter of surface tailings.

8. The method for determining the carbon increment of a tailings dam area according to claim 7, characterized in that, The biochar is made from rice husks, which are dried, ground, and sieved through a 2 mm sieve. An equal mass of sodium hydroxide is added, the mixture is thoroughly mixed, and then it is placed in a sealed iron container. It was then placed in a tube furnace for pyrolysis, with a heating rate set at 10 °C·min. 1 The pyrolysis temperature was 400℃ for 4 hours; after pyrolysis, the biochar was obtained by cooling to room temperature.

9. The application of the tailings dam area carbon increment determination method according to any one of claims 1 to 8, characterized in that, (1) For new tailings ponds to be evaluated, add a pioneer carbon source to the surface tailings of the pond area; (2) At any time point T after the addition of the pioneer carbon source, 5 to 10 sampling points were randomly selected from the surface soil of the tailings pond area for sampling. (3) Obtain the carbon value-added prediction model and key parameters established based on the first tailings dam to be evaluated in the carbon value-added determination method of any one of claims 1 to 7; (4) The samples are tested and characterized to obtain the values ​​of the key parameters, which are used as input variables for the carbon value-added prediction model; (5) Input the input variables into the carbon value-prediction model to obtain the predicted carbon value of the tailings dam to be evaluated at any time point Tn during the 150-day period after time point T.

10. The application according to claim 9, characterized in that, At the time point Tn when the predicted carbon increment is obtained, random sampling and characterization are carried out on the new tailings dam area to be evaluated to obtain the measured carbon increment. The predicted carbon increment is compared with the measured value to verify the accuracy of the carbon increment prediction model. When the explanatory variance is ≤0.6, the mean absolute error is ≥3, the mean square error is ≥3, the median absolute error is ≥3, and the coefficient of determination is ≤0.6, the prediction accuracy is deemed unsatisfactory and the model parameters need to be updated. The method for updating the model parameters is as follows: Based on the new tailings dam to be evaluated, proceed to steps S11-S14 of any of the methods in claims 1-8 to obtain a carbon value-added prediction model with updated parameters; then proceed to step S2 to obtain updated key parameters. The new tailings pond to be evaluated refers to other tailings ponds that are different from the first tailings pond to be evaluated, or to the original target tailings ponds whose prediction accuracy of the original carbon value increment prediction model has decreased due to environmental changes, and whose model parameters need to be updated.

Images