A decision-making method and system for resource utilization of multi-source coal-based solid waste

Abstract

The present invention belongs to the field of resource utilization of coal-based solid waste, and specifically relates to a decision-making method and system for the resource utilization direction of multi-source coal-based solid waste. In order to solve the problem of industrial information asymmetry in the process of resource utilization of coal-based solid waste, the present invention is based on a random forest model of machine learning. By collecting and analyzing the resource and environmental attribute data of multi-source coal-based solid waste and the technical data of solid waste resource utilization, a random forest classification model is established using a machine learning method to achieve the corresponding classification of solid waste resource and environmental attribute data and resource utilization direction technology direction. The decision maker inputs a solid waste resource and environmental attribute data, and the system outputs one or more resource utilization directions. There are many specific resource processing technologies under each direction, which are combined with the requirements of the economic and environmental protection policies of the technology demander and the specific requirements of the solid waste resource processing technology for the physical and chemical properties of the treated solid waste as a decision reference.

Description

Technical Field

[0001] The present invention belongs to the field of resource utilization of coal-based solid waste, and specifically relates to a decision-making method and system for the direction of resource utilization of multi-source coal-based solid waste. Background Art

[0002] Multi-source coal-based solid waste mainly includes solid waste generated during coal mining and processing. The comprehensive utilization rate of coal-based solid waste is only about 60%, and the potential for resource utilization has not been fully tapped. In the process of promoting the resource utilization of coal-based solid waste, there is a problem of industrial information asymmetry. There are many types of technologies for resource treatment and utilization of coal-based solid waste. It has become a difficult problem for technology demanders to choose the utilization direction and detailed technical solutions for a type of coal-based solid waste. This requires a decision-making method and system for the resource utilization direction of multi-source coal-based solid waste, which can provide decision-making support for technology demanders to select resource utilization directions and technologies, promote the resource utilization of coal-based solid waste, and improve the comprehensive utilization rate. Summary of the invention

[0003] The purpose of the present invention is to solve the problem of industrial information asymmetry in the field of multi-source coal-based solid waste resource utilization, and to provide a decision-making method and system for the direction of multi-source coal-based solid waste resource utilization.

[0004] To solve the above problems, the present invention is based on a random forest model of machine learning. By collecting and analyzing the resource and environmental attribute data of multi-source coal-based solid waste and the technical data of solid waste resource utilization, a random forest classification model is established using the method of machine learning to achieve the corresponding classification of solid waste resource and environmental attribute data and resource utilization direction and technical direction. The decision maker inputs a type of solid waste resource and environmental attribute data, and the system outputs one or more resource utilization directions. There are many specific resource processing technologies under each direction. Combined with the economic and environmental protection policies and other requirements of the technology demander and the specific requirements of the solid waste resource processing technology for the physical and chemical properties of the treated solid waste, it serves as a decision-making reference.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A decision-making method for the resource utilization direction of multi-source coal-based solid waste includes the following steps:

[0007] Step 1, obtaining a data set: obtaining resource and environmental attribute information of multi-source coal-based solid waste and technical direction information of solid waste resource utilization, taking the resource and environmental attribute information of multi-source coal-based solid waste as a data set, and dividing it into a training set and a test set;

[0008] Step 2, selecting characteristic attributes from the acquired data set: analyzing the environmental attribute information of solid waste resources, and selecting three types of characteristic attributes for solid waste resource utilization, including material science characteristics, chemical composition and structural characteristics, and mineral phase and composition characteristics;

[0009] Step 3, data preprocessing: if the data under a set of feature attributes have more than 20% missing values, delete this set of data, and fill in the missing values ​​below 20% with the mean; perform consistency check on the data under the same feature attribute, convert text-based feature attributes into numerical values, convert the classification labels of the technical direction of solid waste resource utilization into classification values, and perform unique hot encoding on unrelated feature attributes and classification labels;

[0010] Step 4, build a random forest classifier model: take the data under the characteristic attributes in the training set as input, take the classification label of the technical direction of solid waste resource utilization as output, build a random forest classifier model, standardize the data under the characteristic attributes in the test set with zero mean, import it into the established model, test the generalization of the model, adjust the random forest parameters, and improve the model to realize the classification decision on the direction of solid waste resource utilization;

[0011] Step 5, obtaining one or more classification labels of solid waste resource utilization, each classification label represents a direction of solid waste resource utilization.

[0012] Furthermore, the consistency check in step 3 includes checking the value range of the data, correcting the data that exceeds the range of the order of magnitude, and maintaining the data unit format consistent.

[0013] Furthermore, the specific steps of establishing the random forest classifier model in step 4 are:

[0014] (1) Set the range of random forest parameters, the number of trees is between 100 and 200, the number of tree features is the square root of all feature attributes of the data set, and the depth of the tree is between 10 and 100;

[0015] (2) The data in the training set are normalized with zero mean, and the grid search method is used to divide the training set data into ten parts through the ten-fold cross-validation method. Nine of them are used as training data and one is used as validation data in turn. The mean of the classification accuracy of the ten results is used as the standard for judging the parameter tree model. The parameter model with the highest accuracy is selected as the optimal hyperparameter ratio. The training set after zero mean normalization is then imported into the random forest model with the optimal hyperparameter for training to obtain the random forest classifier model.

[0016] A decision-making system for the resource utilization direction of multi-source coal-based solid waste, including a data acquisition module, a data storage management module, a model building module, and a decision classification module;

[0017] The data acquisition module collects resource and environmental attribute information data of multi-source coal-based solid waste, resource utilization direction data of multi-source coal-based solid waste, and multiple resource utilization technical solutions under each utilization direction, and transmits the collected data to the data storage management module;

[0018] The data storage management module stores the data from the data acquisition module and performs addition, deletion, modification and query of the data at any time;

[0019] The classification model construction module preprocesses the resource and environmental attribute information data of multi-source coal-based solid waste and the resource utilization direction data of multi-source coal-based solid waste in the data storage management module. The resource and environmental attribute information data of multi-source coal-based solid waste is used as a data set, which is divided into a training set and a test set. The data under the characteristic attributes in the training set is used as input, and the classification label of the technical direction of solid waste resource utilization is used as output. A random forest classifier model is established, and the data under the characteristic attributes in the test set is subjected to zero-mean normalization processing and imported into the established model. The generalization of the model is tested, the random forest parameters are adjusted, and the model is improved to realize the classification decision on the direction of solid waste resource utilization;

[0020] The classification model building module transmits the established random forest classifier to the decision classification module, inputs the resource and environmental attributes of coal-based solid waste into the decision classification module, outputs the label classification of the resource utilization direction of coal-based solid waste, and multiple resource utilization technical solutions under each utilization direction.

[0021] Further, the resource and environmental attribute information data of the multi-source coal-based solid waste and the resource utilization direction data of the multi-source coal-based solid waste in the data storage management module are pre-processed, specifically:

[0022] If the data under a set of characteristic attributes have more than 20% missing values, this set of data will be deleted, and the missing values ​​below 20% will be filled with the mean; the data under the same characteristic attribute will be checked for consistency, the text-type characteristic attributes will be converted into numerical types, the classification labels of the technical direction of solid waste resource utilization will be converted into classification values, and unrelated characteristic attributes and classification labels will be uniquely encoded.

[0023] Furthermore, the consistency check includes checking the value range of the data, correcting the data that exceeds the range of the order of magnitude, and maintaining the data unit format consistent.

[0024] Furthermore, the specific steps of establishing the random forest classifier model are:

[0025] (1) Set the range of random forest parameters, the number of trees is between 100 and 200, the number of tree features is the square root of all feature attributes of the data set, and the depth of the tree is 10 to 100;

[0026] (2) The data in the training set are normalized with zero mean, and the grid search method is used to divide the training set data into ten parts through the ten-fold cross-validation method. Nine of them are used as training data and one is used as validation data in turn. The mean of the classification accuracy of the ten results is used as the standard for judging the parameter tree model. The parameter model with the highest accuracy is selected as the optimal hyperparameter ratio. The training set after zero mean normalization is then imported into the random forest model with the optimal hyperparameter for training to obtain the random forest classifier model.

[0027] Compared with the prior art, the present invention has the following advantages:

[0028] The present invention can intelligently realize the classification of the resource utilization directions of multi-source coal-based solid waste, provide decision makers with a clear direction for handling coal-based solid waste, and combine with the coal-based solid waste treatment technical solutions under related directions to quickly make auxiliary decision-making opinions on the resource utilization of coal-based solid waste, play a guiding role in the preliminary investigation stage of the resource utilization of coal-based solid waste, greatly shorten the time for formulating plans, clarify the direction and path of solid waste utilization, and promote the resource utilization of coal-based solid waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flow chart of a decision-making method for resource utilization of multi-source coal-based solid wastes according to the present invention;

[0030] Figure 2 It is a schematic diagram of the decision-making system for the resource utilization direction of multi-source coal-based solid waste of the present invention. DETAILED DESCRIPTION

[0031] Example 1

[0032] like Figure 1 As shown, a decision-making method for the resource utilization direction of multi-source coal-based solid waste includes the following steps:

[0033] Step 1, obtain the data set: obtain the resource and environmental attribute information of multi-source coal-based solid waste and the technical direction information of solid waste resource utilization through online data query and project practice test, and use the resource and environmental attribute information of multi-source coal-based solid waste as the data set, which is divided into 80% training set and 20% test;

[0034] Step 2, select characteristic attributes from the acquired data set: analyze the environmental attribute information of solid waste resources, and select three types of characteristic attributes for solid waste resource utilization, including material science characteristics, chemical composition and structural characteristics, and mineral phase and composition characteristics. Material science characteristics include particle size, particle density, and particle morphology. Chemical composition characteristics include carbon content, water content, loss on ignition, calorific value, ash content, total sulfur content, and oxide content data such as silicon dioxide, ferric oxide, aluminum oxide, sulfur oxide, calcium oxide, magnesium oxide, and potassium oxide, as well as heavy metal content data such as nickel, arsenic, chromium, copper, lead, mercury, zinc, manganese, and cadmium. Mineral phases include crystallinity percentage and crystal phase data. Technical information on the direction of coal-based solid waste resource utilization includes the direction label of the technology, which includes seven labels: fuel power generation, engineering building materials, soil reclamation, mine filling product purification, product auxiliary materials, and ecological governance.

[0035] Step 3, data preprocessing: If the data under a set of feature attributes have more than 20% missing values, delete this set of data, and use the mean to fill in the missing values ​​below 20%; check the consistency of the data under the same feature attribute, including checking the value range of the data, correcting the data that exceeds the range of the sum of the magnitudes, and keeping the data unit format consistent. Convert text-based feature attributes to numerical types, convert the classification labels of the technical direction of solid waste resource utilization to classification values, and perform unique hot encoding on unrelated feature attributes and classification labels;

[0036] Step 4, build a random forest classifier model: take the data under the characteristic attributes in the training set as input, take the classification label of the solid waste resource utilization technology direction as output, build a random forest classifier model, standardize the data under the characteristic attributes in the test set with zero mean, import it into the established model, test the generalization of the model, adjust the random forest parameters, and improve the model to realize the classification decision of the solid waste resource utilization direction. The evaluation indicators of the test set are Hamming loss function, 0-1 error rate, coverage, sorting loss and average precision;

[0037] The specific steps to build a random forest classifier model are:

[0038] (1) Set the range of random forest parameters, the number of trees is between 100 and 200, the number of tree features is the square root of all feature attributes of the data set, and the depth of the tree is between 10 and 100;

[0039] (2) The data in the training set are normalized with zero mean, and the grid search method is used to divide the training set data into ten parts through the ten-fold cross-validation method. Nine of them are used as training data and one is used as validation data in turn. The mean of the classification accuracy of the ten results is used as the standard for judging the hyperparameter model. The model with the highest accuracy is selected as the optimal hyperparameter ratio. The training set after zero mean normalization is then imported into the random forest model with the optimal hyperparameter for training to obtain the random forest classifier model.

[0040] Step 5, obtaining one or more classification labels of solid waste resource utilization, each classification label represents a direction of solid waste resource utilization.

[0041] like Figure 2 As shown, a decision-making system for resource utilization of multi-source coal-based solid waste includes a data acquisition module, a data storage management module, a model building module, and a decision classification module;

[0042] The data acquisition module collects resource and environmental attribute information data of multi-source coal-based solid waste, resource utilization direction data of multi-source coal-based solid waste, and multiple resource utilization technical solutions under each utilization direction, and transmits the collected data to the data storage management module;

[0043] The data storage management module stores the data from the data acquisition module and performs addition, deletion, modification and query of the data at any time;

[0044] The classification model building module pre-processes the resource and environmental attribute information data of the multi-source coal-based solid waste and the resource utilization direction data of the multi-source coal-based solid waste in the data storage management module:

[0045] If the data under a set of characteristic attributes have more than 20% missing values, this set of data will be deleted, and the missing values ​​below 20% will be filled with the mean; the data under the same characteristic attribute will be checked for consistency, the text-type characteristic attributes will be converted into numerical types, the classification labels of the technical direction of solid waste resource utilization will be converted into classification values, and unrelated characteristic attributes and classification labels will be uniquely encoded.

[0046] The resource and environmental attribute information data of multi-source coal-based solid waste is used as a data set, which is divided into 80% training set and 20% test set. The data under the characteristic attributes in the training set is used as input, and the classification label of the technical direction of solid waste resource utilization is used as output. A random forest classifier model is established, and the data under the characteristic attributes in the test set is processed by zero mean standardization and imported into the established model to test the generalization of the model, adjust the random forest parameters, and improve the model to achieve classification decisions on the direction of solid waste resource utilization;

[0047] The specific steps to build a random forest classifier model are:

[0048] (1) Set the range of random forest parameters, the number of trees is between 100 and 200, the number of tree features is the square root of all feature attributes of the data set, and the depth of the tree is 10 to 100;

[0049] (2) The data in the training set are normalized with zero mean, and the grid search method is used to divide the training set data into ten parts through the ten-fold cross-validation method. Nine of them are used as training data and one is used as validation data in turn. The mean of the classification accuracy of the ten results is used as the standard for judging the parameter tree model. The parameter model with the highest accuracy is selected as the optimal hyperparameter ratio. The training set after zero mean normalization is then imported into the random forest model with the optimal hyperparameter for training to obtain the random forest classifier model.

[0050] The classification model construction module transmits the established random forest classifier to the decision classification module, inputs the resource and environmental attributes of coal-based solid waste into the decision classification module, outputs the label classification of the resource utilization direction of coal-based solid waste, and multiple resource utilization technical solutions under each utilization direction.

[0051] During the construction of this model, we analyzed and compared the construction schemes based on various machine learning classification algorithms, and compared the advantages and disadvantages of the model from the classification model evaluation indicators such as accuracy, precision, recall rate and F1-score:

[0052]

[0053] In the comparison between the model constructed based on random forest and the models constructed by other typical classification algorithms, the model constructed based on random forest is superior to other models in classification accuracy, which proves that the model has good classification effect and high classification accuracy. In terms of various indicators such as precision, recall rate, F1-Score, it is also better than other typical models, which fully proves the superiority of the classification model for resource utilization direction of coal-based solid waste constructed based on random forest.

Claims

1. A decision-making method for the resource utilization direction of multi-source coal-based solid waste, characterized in that: The following steps are involved: Step 1, obtaining a data set: obtaining resource and environmental attribute information of multi-source coal-based solid waste and technical direction information of solid waste resource utilization, taking the resource and environmental attribute information of multi-source coal-based solid waste as a data set, and dividing it into a training set and a test set; Step 2, selecting characteristic attributes from the acquired data set: analyzing the environmental attribute information of solid waste resources, and selecting three types of characteristic attributes for solid waste resource utilization, including material science characteristics, chemical composition and structural characteristics, and mineral phase and composition characteristics; Step 3, data preprocessing: if the data under a set of feature attributes have more than 20% missing values, delete this set of data, and fill in the missing values ​​below 20% with the mean; perform consistency check on the data under the same feature attribute, convert text-based feature attributes into numerical types, convert the classification labels of the technical direction of solid waste resource utilization into classification values, and perform unique hot encoding on unrelated feature attributes and classification labels; Step 4, build a random forest classifier model: take the data under the characteristic attributes in the training set as input, take the classification label of the technical direction of solid waste resource utilization as output, build a random forest classifier model, standardize the data under the characteristic attributes in the test set with zero mean, import it into the established model, test the generalization of the model, adjust the random forest parameters, and improve the model to realize the classification decision on the direction of solid waste resource utilization; Step 5, obtaining one or more classification labels of solid waste resource utilization, each classification label representing a direction of solid waste resource utilization; The specific steps of establishing the random forest classifier model in step 4 are: (1) Set the range of random forest parameters. The number of trees is between 100 and 200. The number of tree features is the square root of all feature attributes of the data set. The depth of the tree is between 10 and 100. (2) The data in the training set are normalized with zero mean, and the grid search method is used to divide the training set data into ten parts through the ten-fold cross-validation method. Nine of them are used as training data and one is used as validation data in turn. The mean of the classification accuracy of the ten results is used as the standard for judging the parameter tree model. The parameter model with the highest accuracy is selected as the optimal hyperparameter ratio. The training set after zero mean normalization is then imported into the random forest model with the optimal hyperparameter for training to obtain the random forest classifier model.

2. A decision-making method for resource utilization direction of multi-source coal-based solid waste according to claim 1, characterized in that: The consistency check in step 3 includes checking the value range of the data, correcting the data that exceeds the range of the order of magnitude, and maintaining the data unit format consistent.

3. A decision-making system for the resource utilization direction of multi-source coal-based solid waste, characterized in that: It includes data acquisition module, data storage management module, model building module and decision classification module; The data acquisition module collects resource and environmental attribute information data of multi-source coal-based solid waste, resource utilization direction data of multi-source coal-based solid waste, and multiple resource utilization technical solutions under each utilization direction, and transmits the collected data to the data storage management module; The data storage management module stores the data from the data acquisition module and performs addition, deletion, modification and query of the data at any time; The model building module preprocesses the resource and environmental attribute information data of multi-source coal-based solid waste and the resource utilization direction data of multi-source coal-based solid waste in the data storage management module. The resource and environmental attribute information data of multi-source coal-based solid waste is used as a data set, which is divided into a training set and a test set. The data under the characteristic attributes in the training set is used as input, and the classification label of the technical direction of solid waste resource utilization is used as output. A random forest classifier model is established, and the data under the characteristic attributes in the test set is subjected to zero-mean normalization processing and imported into the established model. The generalization of the model is tested, the random forest parameters are adjusted, and the model is improved to realize the classification decision on the direction of solid waste resource utilization; The model building module transmits the established random forest classifier to the decision classification module, inputs the resource and environmental attributes of the coal-based solid waste into the decision classification module, and outputs the resource utilization direction label classification of the coal-based solid waste, as well as multiple resource utilization technical solutions under each utilization direction; The specific steps of establishing the random forest classifier model are: (1) Set the range of random forest parameters, the number of trees is between 100 and 200, the number of tree features is the square root of all feature attributes of the data set, and the depth of the tree is 10 to 100; (2) The data in the training set are normalized with zero mean, and the grid search method is used to divide the training set data into ten parts through the ten-fold cross-validation method. Nine of them are used as training data and one is used as validation data in turn. The mean of the classification accuracy of the ten results is used as the standard for judging the parameter tree model. The parameter model with the highest accuracy is selected as the optimal hyperparameter ratio. The training set after zero mean normalization is then imported into the random forest model with the optimal hyperparameter for training to obtain the random forest classifier model.

4. A decision-making system for resource utilization direction of multi-source coal-based solid waste according to claim 3, characterized in that: The preprocessing of the resource and environmental attribute information data of the multi-source coal-based solid waste and the resource utilization direction data of the multi-source coal-based solid waste in the data storage management module is specifically as follows: If the data under a set of characteristic attributes have more than 20% missing values, this set of data will be deleted, and the missing values ​​below 20% will be filled with the mean; the data under the same characteristic attribute will be checked for consistency, the text-based characteristic attributes will be converted into numerical types, the classification labels of the technical direction of solid waste resource utilization will be converted into classification values, and unrelated characteristic attributes and classification labels will be uniquely encoded.

5. A decision-making system for resource utilization direction of multi-source coal-based solid waste according to claim 4, characterized in that: The consistency check includes checking the value range of the data, correcting the data that exceeds the range of the order of magnitude, and maintaining the data unit format consistent.

Images