A method for operating a coal production analysis system
Through cluster-based feature grouping and coal demand prediction combined with ARIMA and support vector machine models, and genetic algorithm optimization of the output prediction of extreme learning machines, the problem of incomplete consideration of factors in traditional methods is solved, and the accuracy of coal production analysis is improved and the reliability of corporate decision-making.
Patent Information
- Application Number
- CN202411103575.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional coal production analysis methods are difficult to comprehensively and accurately consider various influencing factors, resulting in inaccurate coal demand and output forecast results and are unable to effectively guide coal production.
The cluster-based feature grouping method is used to select relevant features, combine the ARIMA model and the support vector machine model to predict coal demand, and optimize the limit learning machine to predict coal output through genetic algorithm, comprehensively consider various influencing factors, and monitor production data in real time to Formulate work production plans.
It improves the accuracy and accuracy of coal demand and output forecasts, provides a more reliable basis for production decisions, and enhances the competitiveness of coal enterprises.
Smart Images

Figure FHA0000007503440000011 
Figure FHA0000007503440000022 
Figure FHA0000007503440000032
Abstract
Description
Technical Field
[0001] The invention relates to the field of coal production analysis, and in particular to an operation method of a coal production analysis system. Background Art
[0002] As an important energy resource, coal occupies an important position in the national economy. However, coal production is affected by many factors, such as coal prices, local economic development, climate and season, geological conditions, mining technology and capital investment. Accurately predicting coal demand and output is of great significance for rationally arranging coal production, optimizing resource allocation and improving corporate economic benefits.
[0003] Traditional coal production analysis methods usually fail to comprehensively and accurately consider various influencing factors, resulting in inaccurate forecast results and inability to effectively guide coal production. For example, in coal demand forecasting, only some factors may be considered, while the impact of other important factors is ignored; in coal production forecasting, the models and methods used may not be advanced enough to fully explore the potential information in the data.
[0004] In addition, with the intensification of market competition and the improvement of environmental protection requirements, coal enterprises need to grasp market demand more accurately, optimize production processes, improve production efficiency, and reduce costs to enhance their competitiveness. Therefore, the present invention proposes an operation method of a coal production analysis system, which can comprehensively consider various influencing factors by using artificial intelligence technology, accurately predict coal demand and output, and provide a reliable basis for production decisions of coal enterprises. Summary of the invention
[0005] In view of the problem that the above-mentioned traditional coal production analysis method is usually difficult to comprehensively and accurately consider various influencing factors, resulting in inaccurate prediction results and inability to effectively guide coal production, the present invention proposes an operation method of a coal production analysis system, comprising the following steps:
[0006] S1. First, collect the factors that affect coal demand, including coal prices, local economic development, and climate and seasonal factors. Use clustering-based feature grouping to analyze the relationship between different group features and coal demand, and select features more specifically.
[0007] S2, then the coal demand forecast is performed based on the selected features combined with the ARIMA model and the support vector machine model;
[0008] S3. Collect factors affecting coal production again, including geological conditions, mining technology and capital investment factors, and select features using cluster-based feature grouping.
[0009] S4, then use genetic algorithm to optimize extreme learning machine to predict coal production, and change the values of various influencing factors by predicting production, so as to meet coal demand;
[0010] S5. Finally, monitor the production data in real time and formulate corresponding work production plans.
[0011] Preferably, the steps S1 and S3 use the following steps for clustering-based feature grouping:
[0012] S11. First, perform data preprocessing, handle missing values and outliers, ensure data quality, and perform Z-score standardization on the data;
[0013] S12. Use the hierarchical clustering algorithm to group the features and calculate the distance matrix between the features, using the Pearson correlation coefficient distance: in is the average value of feature x and feature y;
[0014] S13, then select hierarchical clustering, gradually merge or split clusters to generate a dendrogram, and select the shortest distance link method: d(A,B)=min{d(a,b):a∈A,b∈B}, where d(A,B) represents the distance between cluster A and cluster B, and d(a,b) represents the distance between a point a in cluster A and a point b in cluster B;
[0015] S14, then determine the number of groupings k of the features by observing the dendrogram, and assign the features to corresponding groups;
[0016] S15. Then analyze each set of features, evaluate their relationship with the target demand, perform linear regression analysis on each set of features, and obtain the regression coefficient, intercept and determination coefficient. The formula of the linear regression model is: y = β0 + β1x1 + β2x2 + ... + β n x n , where x1,x2...x n is the feature, β0 is the intercept, β1,β2,...β n represents the regression coefficient;
[0017] S16. Finally, based on the results of correlation analysis and regression analysis, select the feature group or features that have the strongest relationship with the target demand.
[0018] Preferably, the step S2 combines the ARIMA model and the support vector machine model, and the specific steps are as follows:
[0019] S21. First, use the ARIMA model to model the linear part, use the ADF test to check whether the data is stable, and perform difference processing if necessary;
[0020] S22, then determine the order of the difference through the ADF test, and use the autocorrelation function and partial autocorrelation function graph to help select the values of the autoregressive order and the moving average order;
[0021] S23. Next, fit the ARIMA model based on the selected parameters and obtain the forecast results.
[0022] S24, then calculate the residual where Y t is the actual value, is the predicted value of the ARIMA model, and the residual N t Contains nonlinear parts;
[0023] S25, then use the residual sequence of the ARIMA model as the input of the support vector machine model, reconstruct the residual sequence, use the residual values of the previous periods as input features, and the current residual value as the target variable;
[0024] S26: Further use the training data to fit the support vector machine model to obtain the predicted value of the residual
[0025] S27. Finally, the prediction results of the ARIMA model And the prediction results of the support vector machine model Add together to get the final predicted value
[0026] Preferably, the step S4 uses a genetic algorithm to optimize the extreme learning machine, and applies the idea of evolutionary computing to the parameter optimization of the extreme learning machine to improve the prediction performance of the model. The specific steps are as follows:
[0027] S41, firstly, parameter encoding is performed to encode the key parameters of the extreme learning machine model into chromosomes in the genetic algorithm, and the parameters include the initial range of hidden layer nodes, hidden layer activation functions, hidden layer weights and biases;
[0028] S42, generating an initial population, randomly generating a group of individuals, each individual representing a parameter setting of an extreme learning machine model;
[0029] S43, fitness evaluation, training the extreme learning machine to randomly generate the weight W and bias b of the hidden layer for each individual extreme learning machine model;
[0030] S44. Calculate the hidden layer output matrix. For a given input data X, calculate the output matrix H of the hidden layer: H = g(XW+b), where g is the activation function, W is the weight vector, and b represents the bias; calculate the output layer weight, and use the least squares method to calculate the output layer weight β: β = H g T, where H gis the generalized inverse matrix of H, T is the target output;
[0031] S45, model evaluation, use the trained extreme learning machine model to predict the test data and get the predicted value Evaluate the model performance by calculating the mean square error, and use the inverse of the mean square error as the fitness value: Where MSE is the mean square error, and ε is a small constant to avoid the denominator being zero;
[0032] S46, then perform the operation of the genetic algorithm, use the roulette method to select individuals according to the fitness value, use multi-point crossover to combine the genes of two parent individuals, generate new individuals, and randomly mutate the genes of the new individuals to introduce new features, replace the old individuals with the newly generated individuals, and update the population;
[0033] S47, the last iteration update repeats the fitness evaluation and genetic operation steps until the maximum number of iterations is reached, the fitness reaches a preset threshold, or the fitness change is less than a certain threshold.
[0034] Compared with the prior art, the advantages and positive effects of the present invention are:
[0035] 1. The clustering-based feature grouping method is used to more specifically select features related to coal demand and output, thereby improving the accuracy of the prediction.
[0036] 2. Combining the ARIMA model and the support vector machine model to forecast coal demand can fully consider the linear and nonlinear parts of the data and improve the accuracy of the forecast.
[0037] 3. Genetic algorithm is used to optimize the extreme learning machine for coal production prediction, and the idea of evolutionary computing is applied to parameter optimization to improve the prediction performance of the model. DETAILED DESCRIPTION
[0038] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0039] 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 than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.
[0040] Embodiment, in order to solve the shortcomings of traditional coal production analysis methods, which are usually difficult to comprehensively and accurately consider various influencing factors, resulting in inaccurate prediction results and inability to effectively guide coal production, the present invention first collects factors affecting coal demand, including coal prices, local economic development conditions, climate and seasons, and uses clustering-based feature grouping to analyze the relationship between different groups of features and coal demand. Specifically, firstly, data preprocessing is performed to handle missing values and outliers, ensure data quality, and perform Z-score standardization on the data; hierarchical clustering algorithm is used to group features, calculate the distance matrix between features, and use the Pearson correlation coefficient distance: in is the average value of feature x and feature y; then hierarchical clustering is selected, clusters are merged or split step by step to generate a dendrogram, and the shortest distance link method is selected: d(A,B)=min{d(a,b):a∈A,b∈B}, where d(A,B) represents the distance between cluster A and cluster B, and d(a,b) represents the distance between a point a in cluster A and a point b in cluster B; then the number of groupings k of the features is determined by observing the dendrogram or other methods, and the features are assigned to the corresponding groups; then each group of features is analyzed to evaluate their relationship with the target requirements, and each group of features is subjected to linear regression analysis to obtain the regression coefficient, intercept and determination coefficient. The formula of the linear regression model is: y=β0+β1x1+β2x2+...+β n x n , where x1,x2...x n is the feature, β0 is the intercept, β1,β2,...β n Represents the regression coefficient; finally, based on the results of correlation analysis and regression analysis, select the feature group or feature that has the strongest relationship with the target demand. Use clustering technology to select features and select a more appropriate feature subset to improve the prediction effect.
[0041] Considering that the next step is to predict the demand, the traditional ARIMA model has high requirements for data and requires the data to be stable, otherwise differential processing is required, which may lead to information loss. In addition, it can only capture linear relationships, and the processing ability for nonlinear relationships is limited. Therefore, the present invention introduces the support vector machine model, which has advantages in processing nonlinear relationships. The two models have complementary advantages, and the combination of the two can achieve good results. The specific operation is as follows: first, use the ARIMA model to model the linear part, use the ADF test to detect whether the data is stable, and perform differential processing if necessary; then determine the order of the difference through the ADF test, and use the autocorrelation function and partial autocorrelation function graph to help select the values of the autoregressive order and the moving average order; secondly, fit the ARIMA model based on the selected parameters, and obtain the prediction results. Then calculate the residual where Y t is the actual value, is the predicted value of the ARIMA model, and the residual N t It includes nonlinear parts. Then, the residual sequence of the ARIMA model is used as the input of the support vector machine model to reconstruct the residual sequence, using the residual values of the previous periods as input features and the current residual value as the target variable. The training data is further used to fit the support vector machine model to obtain the predicted value of the residual. Finally, the prediction results of the ARIMA model are And the prediction results of the support vector machine model Add together to get the final predicted value
[0042] It is further necessary to make a prediction for the production plan. Considering that the extreme learning machine randomly generates the weights and biases of the hidden layer during use, the stability and generalization ability of the model are affected. Therefore, the present invention proposes to use a genetic algorithm to optimize the extreme learning machine. The specific operations are as follows: first, parameter encoding is performed, and the key parameters of the extreme learning machine model are encoded as chromosomes in the genetic algorithm. The parameters include the initial range of hidden layer nodes, hidden layer activation functions, hidden layer weights and biases; then an initial population is generated, and a group of individuals are randomly generated, each individual represents a parameter setting of an extreme learning machine model; then fitness evaluation is performed, and the extreme learning machine is trained to randomly generate the weights W and bias b of the hidden layer for each individual extreme learning machine model; the hidden layer output matrix is calculated, and for a given input data X, the output matrix H of the hidden layer is calculated: H = g(XW+b), where g is the activation function, W is the weight vector, and b represents the bias; the output layer weight is calculated, and the weight β of the output layer is calculated using the least squares method: β = H g T, where H g is the generalized inverse matrix of H, and T is the target output; then the model is evaluated, and the trained extreme learning machine model is used to predict the test data to obtain the predicted value Evaluate the model performance by calculating the mean square error, and use the inverse of the mean square error as the fitness value: Where MSE is the mean square error, and ε is a small constant to avoid the situation where the denominator is zero; next, the genetic algorithm is operated, and the roulette method is used to select individuals according to the fitness value, and the genes of two parent individuals are combined by multi-point crossover to generate new individuals, and the genes of the new individuals are randomly mutated to introduce new features, and the newly generated individuals are replaced by the old individuals to update the population; finally, the iterative update repeats the fitness evaluation, genetic operation and other steps until the maximum number of iterations is reached, the fitness reaches the preset threshold or the fitness change is less than a certain threshold.
[0043] Finally, after completing the production forecast, a series of steps are required to monitor the production data in real time and complete the production plan on time. First, deploy sensors and data acquisition equipment to continuously obtain the key parameters of the production line; transmit these data to the central data processing platform, and use data analysis tools for real-time monitoring and processing; set up an early warning mechanism to promptly detect and notify abnormal situations based on the production plan and set thresholds; use the manufacturing execution system for production management to ensure real-time tracking and dynamic adjustment of the production plan, optimize resource allocation and production scheduling; finally, implement performance evaluation, analyze production data and plan completion, continuously improve the production process, ensure the production plan is completed on time, and achieve efficient production management.
[0044] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.
Claims
1. A method for operating a coal production analysis system, characterized in that: The following steps are involved: S1. First, collect the factors that affect coal demand, including coal prices, local economic development, and climate and seasonal factors. Use clustering-based feature grouping to analyze the relationship between different group features and coal demand, and select features more specifically. S2, then the coal demand forecast is performed based on the selected features combined with the ARIMA model and the support vector machine model; S3. Collect factors affecting coal production again, including geological conditions, mining technology and capital investment factors, and select features using cluster-based feature grouping. S4, then use genetic algorithm to optimize extreme learning machine to predict coal production, and change the values of various influencing factors by predicting production, so as to meet coal demand; S5. Finally, monitor the production data in real time and formulate corresponding work production plans; The steps S1 and S3 use the clustering-based feature grouping steps as follows: S11. First, perform data preprocessing, handle missing values and outliers, ensure data quality, and perform Z-score standardization on the data; S12. Use the hierarchical clustering algorithm to group the features and calculate the distance matrix between the features, using the Pearson correlation coefficient distance: in is the average value of feature x and feature y; S13, then select hierarchical clustering, gradually merge or split clusters to generate a dendrogram, and select the shortest distance link method: d(A,B)=min{d(a,b):a∈A,b∈B}, where d(A,B) represents the distance between cluster A and cluster B, and d(a,b) represents the distance between a point a in cluster A and a point b in cluster B; S14, then determine the number of groupings k of the features by observing the dendrogram, and assign the features to corresponding groups; S15. Then analyze each set of features, evaluate their relationship with the target demand, perform linear regression analysis on each set of features, and obtain the regression coefficient, intercept and determination coefficient. The formula of the linear regression model is: y = β0 + β1x1 + β2x2 + ... + β n x n , where x1,x2...x n is the feature, β0 is the intercept, β1,β2,...β n represents the regression coefficient; S16. Finally, based on the results of correlation analysis and regression analysis, select the feature group or features that have the strongest relationship with the target demand; The step S4 uses a genetic algorithm to optimize the extreme learning machine, and applies the idea of evolutionary computing to the parameter optimization of the extreme learning machine to improve the prediction performance of the model. The specific steps are as follows: S41, firstly, parameter encoding is performed to encode the key parameters of the extreme learning machine model into chromosomes in the genetic algorithm, and the parameters include the initial range of hidden layer nodes, hidden layer activation functions, hidden layer weights and biases; S42, generating an initial population, randomly generating a group of individuals, each individual representing a parameter setting of an extreme learning machine model; S43, fitness evaluation, training the extreme learning machine to randomly generate the weight W and bias b of the hidden layer for each individual extreme learning machine model; S44, calculating the hidden layer output matrix, for a given input data X, calculating the output matrix H of the hidden layer: H = g(XW+b), where g is the activation function, W is the weight vector, and b represents the bias; Calculate the weight of the output layer and use the least squares method to calculate the weight β of the output layer: β = H g T, where H g is the generalized inverse matrix of H, T is the target output; S45, model evaluation, use the trained extreme learning machine model to predict the test data and get the predicted value Evaluate the model performance by calculating the mean square error, and use the inverse of the mean square error as the fitness value: Where MSE is the mean square error, and ε is a small constant to avoid the denominator being zero; S46, then perform the operation of the genetic algorithm, use the roulette method to select individuals according to the fitness value, use multi-point crossover to combine the genes of two parent individuals, generate new individuals, and randomly mutate the genes of the new individuals to introduce new features, replace the old individuals with the newly generated individuals, and update the population; S47, the last iteration updates and repeats the fitness evaluation and genetic operation steps until the maximum number of iterations is reached, the fitness reaches a preset threshold, or the fitness change is less than a certain threshold.
2. The method for operating a coal production analysis system according to claim 1, characterized in that: The step S2 combines the ARIMA model and the support vector machine model, and the specific steps are as follows: S21. First, use the ARIMA model to model the linear part, use the ADF test to check whether the data is stable, and perform difference processing if necessary; S22, then determine the order of the difference through the ADF test, and use the autocorrelation function and partial autocorrelation function graph to help select the values of the autoregressive order and the moving average order; S23. Next, fit the ARIMA model based on the selected parameters and obtain the forecast results. S24, then calculate the residual where Y t is the actual value, is the predicted value of the ARIMA model, and the residual N t Contains nonlinear parts; S25, then use the residual sequence of the ARIMA model as the input of the support vector machine model, reconstruct the residual sequence, use the residual values of the previous periods as input features, and the current residual value as the target variable; S26: Further use the training data to fit the support vector machine model to obtain the predicted value of the residual S27. Finally, the prediction results of the ARIMA model And the prediction results of the support vector machine model Add together to get the final predicted value
