CN114721263B - Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm - Google Patents
Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm Download PDFInfo
- Publication number
- CN114721263B CN114721263B CN202210256441.2A CN202210256441A CN114721263B CN 114721263 B CN114721263 B CN 114721263B CN 202210256441 A CN202210256441 A CN 202210256441A CN 114721263 B CN114721263 B CN 114721263B
- Authority
- CN
- China
- Prior art keywords
- data
- model
- decomposing furnace
- intelligent
- machine learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000005457 optimization Methods 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 42
- 239000004568 cement Substances 0.000 title claims abstract description 29
- 238000010801 machine learning Methods 0.000 title claims abstract description 29
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 26
- 230000006870 function Effects 0.000 claims abstract description 26
- 238000004364 calculation method Methods 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 24
- 235000012054 meals Nutrition 0.000 claims abstract description 22
- 239000002994 raw material Substances 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 10
- 238000003860 storage Methods 0.000 claims abstract description 8
- 230000008859 change Effects 0.000 claims abstract description 5
- 239000002245 particle Substances 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 12
- 239000003245 coal Substances 0.000 claims description 10
- 238000004519 manufacturing process Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 10
- 238000010219 correlation analysis Methods 0.000 claims description 9
- VHUUQVKOLVNVRT-UHFFFAOYSA-N Ammonium hydroxide Chemical compound [NH4+].[OH-] VHUUQVKOLVNVRT-UHFFFAOYSA-N 0.000 claims description 8
- 235000011114 ammonium hydroxide Nutrition 0.000 claims description 8
- 238000002347 injection Methods 0.000 claims description 8
- 239000007924 injection Substances 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000004140 cleaning Methods 0.000 claims description 6
- 238000002485 combustion reaction Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 238000012821 model calculation Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- 238000012847 principal component analysis method Methods 0.000 claims description 4
- 230000003313 weakening effect Effects 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 3
- 239000000243 solution Substances 0.000 claims description 3
- 238000011425 standardization method Methods 0.000 claims description 3
- 239000003344 environmental pollutant Substances 0.000 claims description 2
- 231100000719 pollutant Toxicity 0.000 claims description 2
- 230000007547 defect Effects 0.000 abstract description 3
- 230000001105 regulatory effect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a cement decomposing furnace intelligent regulation and control method based on machine learning and intelligent optimization algorithm, which comprises the following steps: collecting the running condition data of the decomposing furnace from the DCS system in real time and transmitting the running condition data to a special server database for storage; performing a series of operations on the historical operation data, and establishing a training set and a testing set for calculating the decomposing rate of NOx and raw meal of the decomposing furnace; training the model by using a machine learning method, establishing a calculation model of the decomposition rate of raw materials in the decomposing furnace and the NOx emission amount at the outlet of the decomposing furnace, and verifying the accuracy of the model by a test set; constructing the output of a decomposing furnace key variable calculation model established based on a machine learning method as an objective function through an intelligent optimization algorithm, and carrying out parameter optimization to obtain a better reference value of the adjustable operation quantity under the objective working condition; and the model is adaptively updated according to the performance change of the model, so that the accuracy of a calculation result is ensured. The invention can provide guidance for cement plant operators to regulate and control, and make up for the defect of manually regulating operation parameters.
Description
Technical Field
The invention relates to cement production, in particular to a cement decomposing furnace intelligent regulation and control method based on machine learning and intelligent optimization algorithm.
Background
The adjustment of the operating parameters of the decomposing furnace has an important influence on the decomposition of raw materials and the emission of NOx, and is an important object of production regulation. In the cement production process, the system working condition of the decomposing furnace is complex and changeable, on one hand, the decomposing furnace is influenced by the gas-solid flow and the combustion mechanism in the furnace, on the other hand, the decomposing furnace is also greatly related to the working states of the preheater and the rotary kiln, and at present, most cement factories adopt manual operation as a main mode and automatic control as an auxiliary mode for the combustion control of the decomposing furnace, and the decomposing furnace particularly depends on the experience of operators, so that the decomposing furnace has the problems of high energy consumption, high exhaust emission, high temperature fluctuation, poor clinker quality and the like. Therefore, a set of cement decomposing furnace intelligent regulation technology is established, and the cement decomposing furnace intelligent regulation technology has very important practical significance.
Disclosure of Invention
The invention aims to: the invention aims to provide an intelligent regulation and control method for a cement decomposing furnace based on machine learning and intelligent optimization algorithm, so that guidance is provided for regulation and control of operators in cement plants, and the defect of manual regulation of operation parameters is overcome.
The technical scheme is as follows: the invention discloses a cement decomposing furnace intelligent regulation and control method based on machine learning and intelligent optimization algorithm, which comprises the following steps:
(1) Collecting the running condition data of the decomposing furnace from the DCS system in real time and transmitting the running condition data to a special server database for storage;
(2) Performing data cleaning, feature selection, dimension reduction and data standardization on the historical operation data, and establishing a training set and a testing set for calculating the decomposing rate of NOx and raw meal of the decomposing furnace;
(3) Training the model by using a machine learning method, establishing a key variable calculation model for measuring combustion efficiency and pollutants, and verifying the accuracy of the model by a test set;
(4) Constructing the output of a decomposing furnace key variable calculation model established based on a machine learning method as an objective function through an intelligent optimization algorithm, and carrying out parameter optimization to obtain a better reference value of the adjustable operation quantity under the objective working condition;
(5) And the model is adaptively updated according to the performance change of the model, so that the accuracy of a calculation result is ensured.
The historical operation data in the step (2) refer to historical operation original data of a kiln of a cement plant for more than one month, the historical operation original data comprises parameter data adjustable by operators and parameter data not adjustable, the parameter data adjustable comprise raw material blanking amount, coal injection amount, tertiary air amount and ammonia water amount in the cement production process, and the parameter data not adjustable comprise tertiary air temperature, decomposing furnace outlet pressure, kiln tail temperature, kiln tail pressure and coal property.
The step (2) of data cleaning refers to deleting missing values and abnormal value unsteady values generated by faults of sensors or other equipment in historical operation data; the method comprises the following steps:
(2.1) if the missing value in the data only has the missing of the discontinuous time point, linear interpolation filling is carried out by using the recorded values of the time points before and after; if the continuous missing value exists, deleting the measured value in the time period from the system database;
(2.2) judging abnormal values in the data according to the Leider criterion, wherein the abnormal values exist in the data:
wherein the method comprises the steps ofThe sigma is the standard deviation of a certain variable;
(2.3) for unsteady values in the data, because of large fluctuation of a dynamic process, the running parameters cannot truly reflect the running conditions in the decomposing furnace, and a sliding window method is adopted for steady state detection: firstly, selecting a proper window length N, determining the fluctuation condition of running data in the window, and then moving the window backwards to determine the next group of data until all the data are detected; the judgment standard of the stable working condition is as follows:
wherein x is max For a maximum value of an operating parameter within the sliding window, x min Is the minimum value of a certain operation parameter in the sliding window range, x r Is rated under rated working condition, v t Is a stability threshold;
(2.4) carrying out feature selection on factors influencing the NOx emission amount and the raw meal decomposition rate, calculating the degree of interdependence between each parameter and a target value through correlation analysis, eliminating the influence of the parameter with lower correlation, and reducing the calculated amount of the model; selecting characteristic variables with the greatest correlation with NOx emission and raw meal decomposition rate except parameters adjustable by operators such as raw meal blanking amount, coal injection amount, tertiary air amount, ammonia water consumption and the like based on a mutual information correlation analysis method;
(2.5) reducing the dimension of the selected data, and mapping the data points in the original high-dimension space into the low-dimension space by adopting a certain mapping method, so that redundant information and noise in the data are reduced, and the calculated amount of the model is further reduced; performing dimension reduction treatment on the data by using a principal component analysis method, and selecting principal components with contribution degree to the target variable being more than 90%;
(2.6) scaling the data by data normalization to fall within a specified interval, thereby avoiding weakening the effect of smaller numerical parameters in the comprehensive analysis process due to different amounts and orders of magnitude of the data; the data processing is completed by using a min-max standardization method, and the calculation method is as follows:
wherein X is norm Is normalized value, X max 、X min 、X mean Respectively the maximum value, the minimum value and the average value of the data;
(2.7) calculating the hysteresis of each variable, namely how long it takes to influence the change of the controlled object; recording each variable delay time as tau i I=1, 2, 3..n, where N is the number of variables obtained after the processing in the foregoing steps, and by calculating the delay time of each variable, selecting data in each variable delay time period, and completing timing matching of the data; and x (t) is the moment t, and the model object calculates an array formed by the required parameters:
x(t)＝[y(t-1),...,y(t-τ y ),u 1 (t),u 1 (t-1),...,u 1 (t-τ 1 ),...,u N (t),u N (t-1),...,u N (t-τ N )] T
τ in the above formula i The value of (2) is solved by a particle swarm algorithm in which the delay time (τ 1 ,τ 2 ,...,τ p ,τ y ) As a variable needing optimization in an integer space, 0 is equal to or less than tau i ≤τ max ,τ max Maximum delay time for a variable; taking the prediction error as a minimum fitness function, and searching the optimal delay through the minimum fitness function in an optimization process;
(2.8) from the above-mentioned steps (2.1) to (2.7), it is necessary to construct data sets for calculation of NOx emission amount and raw meal decomposition rate, and for the NOx emission amount and raw meal decomposition rate, due to differences in influencing factors, two data sets, respectively { y } 1 (t),x 1 (t) } and { y } 2 (t),x 2 (t)}。
The feature selection in the step (2) is to calculate the degree of interdependence between each parameter and the target value through correlation analysis, eliminate the influence of the parameter with lower correlation, reduce the calculated amount of the model, and select the feature with the largest correlation based on the method of mutual information correlation analysis; the dimension reduction is to map data points in an original high-dimensional space into a low-dimensional space by adopting a certain mapping method, reduce redundant information and noise in the data, further reduce the calculated amount of a model, perform dimension reduction processing on the data by a principal component analysis method, and select principal components with contribution degree of more than 90% to a target variable; the data normalization refers to scaling data to fall into a specific interval, so that the effect of weakening smaller numerical parameters in the comprehensive analysis process due to different data size classes and orders of magnitude is avoided, and the data processing is completed by using a min-max normalization method.
The parameters for measuring the combustion efficiency and pollution in the step (3) refer to a measured value of the decomposition rate of raw materials in the decomposing furnace and a monitored value of the NOx emission of the outlet of the decomposing furnace.
The step (3) specifically comprises the following steps:
(3.1) establishing a relation model of the NOx emission amount of the decomposing furnace outlet and each parameter and a relation model of the decomposing rate of the raw materials in the decomposing furnace and each parameter on a training set by using a LightGBM machine learning method:
(3.2) optimizing important parameters such as learning_rate, max_depth, num_leave and the like in the LightGBM model by adopting a particle swarm optimization, so that higher model calculation accuracy is realized;
(3.3) in the process of optimizing the parameters of the LightGBM model by using a particle swarm optimization algorithm, optimizing the independent variables by using the learning_rate, the max_depth and the num_leave, and using the root mean square error calculated by the corresponding LightGBM model as an adaptability function, wherein the optimization process meets the following constraint: max_depth and num_leave are integers greater than 0; the calculated global optimal solution is the parameter value of the model, the model is retrained by using the group of parameters, and the calculation models of NOx and raw meal decomposition rate are respectively built;
(3.4) after initializing the population, constructing a model for the learning_rate, max_depth and num_leave corresponding to each particle in the population, and training the model by using a K-fold cross-validation (K-fold Cross Validation) method; selecting global optimal model parameters of the current population by taking the root mean square error of the model as an objective function, and updating the population by taking the current global optimal parameters as a reference; training a model on the basis of the new population, and calculating a model error; judging whether the algorithm converges or not, if not, continuing to update the population, and training a new model; and if the algorithm converges, outputting the model optimal parameters obtained by calculation at the moment.
The step (4) specifically comprises the following steps:
(4.1) the production regulation of the cement decomposing furnace is a multi-objective optimization problem, and the emission of NOx needs to be reduced while ensuring higher raw material decomposition rate;
(4.2) carrying out weighted summation on the raw meal decomposition rate and the NOx emission amount, constructing an objective function of operation parameter optimization, and converting the multi-objective optimization problem into a single-objective optimization problem;
F obj ＝w 1 F NOx -w 2 F η +w 3 ε
wherein w is 1 ,w 2 ,w 3 Weights for each item; epsilon is an overtemperature punishment item of the outlet temperature of the decomposing furnace; f (F) NOx And F is equal to η Calculated from the established LightGBM; the goals and constraints of the optimization are:
minF obj
x i,min ≤x i ≤x i,max (i＝1...m)
and (4.3) using a particle swarm intelligent optimization algorithm, taking a constructed objective function as an fitness function, taking raw material blanking quantity, coal injection quantity, tertiary air quantity and ammonia water quantity as the population of the particle swarm algorithm, and iteratively searching optimal operation parameters as guidance for regulation and control by operators.
The step (5) specifically comprises the following steps: and (3) for the working condition newly added into the database, after data preprocessing, if the model calculation result and the actual value are larger than a certain threshold value, selecting a plurality of groups of data recently, repeating the steps (2) - (4), and reproducing and establishing a new model until the model error meets the condition.
A computer storage medium having stored thereon a computer program which when executed by a processor implements the above-described intelligent regulation method for cement-decomposing furnaces based on machine learning and intelligent optimization algorithms.
The computer equipment comprises a storage, a processor and a computer program which is stored in the storage and can be run on the processor again, wherein the intelligent regulation and control method of the cement decomposing furnace based on the machine learning and intelligent optimization algorithm is realized when the processor executes the computer program.
The beneficial effects are that: compared with the prior art, the invention has the following advantages: the invention can reduce the emission of NOx while maintaining the decomposition rate of raw materials in the cement decomposing furnace, provides guidance for production regulation and control of operators, compensates for the defect that the operators rely on experience to regulate parameters, and assists the efficient low-pollution production in the cement industry.
Drawings
FIG. 1 is a schematic block diagram of a flow of the present invention;
FIG. 2 is a schematic block diagram of data preprocessing of the present invention;
FIG. 3 is a schematic block diagram of the modeling of the present invention;
fig. 4 is a schematic block diagram of intelligent optimization of operating parameters according to the present invention.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings.
As shown in FIG. 1, the intelligent regulation and control method of the cement decomposing furnace based on the machine learning and intelligent optimization algorithm comprises the steps of DCS historical data acquisition, data preprocessing, raw meal decomposition rate and NOx emission prediction model establishment, working condition parameter optimization and the like.
Step 1: collecting the running condition data of the decomposing furnace from the DCS system in real time and transmitting the running condition data to a special server database for storage;
step 2: preprocessing historical operation data, and constructing a data set required by model calculation;
step 3: training a model by using a LightGBM machine learning method, establishing a calculation model of the decomposition rate of raw materials in the decomposing furnace and the NOx emission amount at the outlet of the decomposing furnace, and verifying the accuracy of the model through a test set;
step 4: for real-time working condition data, constructing the output of a decomposing furnace key variable calculation model established based on a machine learning method into an objective function through an intelligent optimization algorithm, carrying out parameter optimization to obtain a better reference value of an adjustable operation quantity under a target working condition, and outputting the better reference value as an adjustment suggestion;
step 5: and for the real-time working condition data, if the error between the predicted value and the actual value calculated according to the established model is larger than a set threshold epsilon, reading the data from the historical database again and repeating the steps, establishing a new model, and updating in a self-adaptive manner to ensure the accuracy of a calculation result.
As shown in fig. 2, the flow of data preprocessing is: data cleaning, feature selection, data dimension reduction, data standardization, time sequence matching and data set construction.
The data cleaning includes missing value processing, outlier processing, and unsteady value processing.
And (3) carrying out linear interpolation filling on the missing values in the data by using the recorded values of the front and rear time points.
Judging the abnormal value existing in the data according to the Laida criterion:
wherein the method comprises the steps ofThe average value of a certain variable in the data, and sigma is the standard deviation of the certain variable. For outliers, linear interpolation padding is performed using recorded values at the previous and subsequent time points.
Because the dynamic process data has large fluctuation, the running parameters can not truly reflect the running conditions in the decomposing furnace, and a sliding window method is adopted for steady state detection: the appropriate window length N is first selected, fluctuations in the running data within the window are determined, and the window is then moved back to determine the next set of data until all data is detected. The judgment standard of the stable working condition is as follows:
wherein x is max For a maximum value of an operating parameter within the sliding window, x min Is the minimum value of a certain operation parameter in the sliding window range, x r Is rated under rated working condition, v t Is the stability threshold.
Based on the mutual information correlation analysis (MI) method, characteristic variables with the greatest correlation with NOx emission and raw meal decomposition rate are selected except parameters adjustable by operators such as raw meal blanking amount, coal injection amount, tertiary air amount, ammonia water consumption and the like.
And (3) performing dimension reduction treatment on the data by using a Principal Component Analysis (PCA) method, and selecting principal components with contribution degree of more than 90% to the target variable.
The data standardization processing is completed by using a min-max standardization method, and the calculation method is as follows:
wherein X is norm Is normalized value, X max 、X min 、X mean The maximum, minimum and average values of the data, respectively.
Calculating the delay time of each variable and recording the delay time of each variable as tau i I=1, 2, 3..n, where N is the one obtained by the preceding stepAnd (3) calculating the delay time of each variable, selecting the data in each variable delay time period, and completing the time sequence matching of the data. And x (t) is the moment t, and the model object calculates an array formed by the required parameters:
x(t)＝[y(t-1),...,y(t-τ y ),u 1 (t),u 1 (t-1),...,u 1 (t-τ 1 ),...,u N (t),u N (t-1),...,u N (t-τ N )] T
as shown in fig. 3, the variable delay time is calculated using a particle swarm algorithm.
Delay time (τ) 1 ,τ 2 ,...,τ p ,τ y ) As a variable needing optimization in an integer space, 0 is equal to or less than tau i ≤τ max ,τ max Is the maximum delay time of the variable. The optimization process searches for the best delay by minimizing the objective function with the prediction error as the objective function.
Firstly, initializing population parameters, and then initializing delay time T= [ tau ] 1 ,τ 2 ,...τ N ,τ y ]. And constructing different data sets for the delay time corresponding to each particle in the population, and training the LightGBM model on the data sets. And selecting global optimal delay time T by taking the root mean square error of the model as an objective function, and updating the population by taking the current global optimal delay time as a reference. And reestablishing the data set on the basis of the new population, training, and calculating the model error. Judging whether the algorithm converges or not, if not, continuing updating the population, establishing a data set and training; if the algorithm converges, outputting the calculated global optimal delay time T opt ＝[τ1,τ2,...τN,τy]opt。
Constructing data sets for calculation of NOx emission amount and raw material decomposition rate, for which two data sets are required to be constructed due to different influencing factors, respectively { y } 1 (t),x 1 (t) } and { y } 2 (t),x 2 (t)}。
For the established data set, the data set is divided into a training set and a test set, the data in the training set is used for model establishment, the data in the test set is used for model verification, and the proportion of the training set and the test set is respectively 70% and 30%.
FIG. 4 is a method of building a model of NOx emissions and raw meal degradation rate predictions using the LightGBM machine learning method.
By utilizing a LightGBM machine learning method, a relation model of the NOx emission amount of the decomposing furnace outlet and each parameter and a relation model of the decomposing rate of raw materials in the decomposing furnace and each parameter are established on a training set:
in order to achieve higher accuracy of the model, the parameters of the LightGBM need to be adjusted. And optimizing important parameters such as learning_rate, max_depth, num_leave and the like in the LightGBM model by adopting a particle swarm algorithm, so that higher model calculation accuracy is realized.
In the process of optimizing the parameters of the LightGBM model by using a particle swarm optimization algorithm, the independent variables are optimized by using the learning_rate, the max_depth and the num_leave, and the root mean square error calculated by the corresponding LightGBM model is used as an fitness function, so that the following constraint is satisfied in the optimization process: max_depth and num_leave are integers greater than 0. The calculated global optimal solution is the parameter value of the model, the model is retrained by using the group of parameters, and the calculation models of NOx and raw meal decomposition rate are respectively built.
After initializing the population, a model is built for each particle in the population corresponding to learning_rate, max_depth, num_leave, and the model is trained using a K-fold cross-validation (K-fold Cross Validation) method. And selecting global optimal model parameters of the current population by taking the root mean square error of the model as an objective function, and updating the population by taking the current global optimal parameters as a reference. Training a model on the basis of the new population, and calculating model errors. Judging whether the algorithm converges or not, if not, continuing to update the population, and training a new model; and if the algorithm converges, outputting the model optimal parameters obtained by calculation at the moment.
The production regulation of cement decomposing furnaces is a multi-objective optimization problem, and it is required to reduce the emission of NOx while ensuring a higher raw material decomposition rate.
And carrying out weighted summation on the raw material decomposition rate and the NOx emission amount, constructing an objective function of the operation parameter optimization, and converting the multi-objective optimization problem into a single-objective optimization problem.
F obj ＝w 1 F NOx -w 2 F η +w 3 ε
Wherein: w (w) 1 ,w 2 ,w 3 Weights for each item; epsilon is an overtemperature punishment item of the outlet temperature of the decomposing furnace; f (F) NOx And F is equal to η Calculated from the established LightGBM. The goals and constraints of the optimization are:
minF obj
x i,min ≤x i ≤x i,max (i＝1...m)
the particle swarm intelligent optimization algorithm is used, a constructed objective function is used as an fitness function, the raw material blanking amount, the coal injection amount, the tertiary air amount and the ammonia water amount are used as the population of the particle swarm algorithm, and the optimal operation parameters are searched for iteratively and used as guidance for regulation and control of operators.
Claims (8)
1. The intelligent regulation and control method for the cement decomposing furnace based on the machine learning and intelligent optimization algorithm is characterized by comprising the following steps of:
(1) Collecting the running condition data of the decomposing furnace from the DCS system in real time and transmitting the running condition data to a special server database for storage;
(2) Performing data cleaning, feature selection, dimension reduction and data standardization on the historical operation data, and establishing a training set and a testing set for calculating the decomposing rate of NOx and raw meal of the decomposing furnace;
(3) Training the model by using a machine learning method, establishing a key variable calculation model for measuring combustion efficiency and pollutants, and verifying the accuracy of the model by a test set;
(3.1) establishing a relation model of the NOx emission amount of the decomposing furnace outlet and each parameter and a relation model of the decomposing rate of the raw materials in the decomposing furnace and each parameter on a training set by using a LightGBM machine learning method:
(3.2) optimizing important parameters in the LightGBM model by adopting a particle swarm optimization, so that higher model calculation accuracy is realized, wherein the important parameters comprise learning_rate, max_depth and num_leave;
(3.3) in the process of optimizing the parameters of the LightGBM model by using a particle swarm optimization algorithm, optimizing the independent variables by using the learning_rate, the max_depth and the num_leave, and using the root mean square error calculated by the corresponding LightGBM model as an adaptability function, wherein the optimization process meets the following constraint: max_depth and num_leave are integers greater than 0; the calculated global optimal solution is the parameter value of the model, the model is retrained by using the group of parameters, and the calculation models of NOx and raw meal decomposition rate are respectively built;
(3.4) after initializing the population, constructing a model for learning_rate, max_depth and num_leave corresponding to each particle in the population, and training the model by using a K-fold cross-validation K-fold Cross Validation method; selecting global optimal model parameters of the current population by taking the root mean square error of the model as an objective function, and updating the population by taking the current global optimal parameters as a reference; training a model on the basis of the new population, and calculating a model error; judging whether the algorithm converges or not, if not, continuing to update the population, and training a new model; if the algorithm converges, outputting the model optimal parameters obtained by calculation at the moment;
(4) Constructing the output of a decomposing furnace key variable calculation model established based on a machine learning method as an objective function through an intelligent optimization algorithm, and carrying out parameter optimization to obtain a better reference value of the adjustable operation quantity under the objective working condition;
(4.1) the production regulation of the cement decomposing furnace is a multi-objective optimization problem, and the emission of NOx needs to be reduced while ensuring higher raw material decomposition rate;
(4.2) carrying out weighted summation on the raw meal decomposition rate and the NOx emission amount, constructing an objective function of operation parameter optimization, and converting the multi-objective optimization problem into a single-objective optimization problem;
F obj ＝w 1 F NOx -w 2 F η +w 3 ε
wherein w is 1 ,w 2 ,w 3 Weights for each item; epsilon is an overtemperature punishment item of the outlet temperature of the decomposing furnace; f (F) NOx And F is equal to η Calculated from the established LightGBM; the goals and constraints of the optimization are:
minF obj
x i,min ≤x i ≤x i,max (i＝1…m)
(4.3) using a particle swarm intelligent optimization algorithm, taking a constructed objective function as an fitness function, taking raw material blanking quantity, coal injection quantity, tertiary air quantity and ammonia water consumption as the population of the particle swarm algorithm, and iteratively searching optimal operation parameters as guidance for regulation and control by operators;
(5) And the model is adaptively updated according to the performance change of the model, so that the accuracy of a calculation result is ensured.
2. The intelligent regulation and control method for the cement decomposing furnace based on the machine learning and intelligent optimization algorithm according to claim 1, wherein the historical operation data in the step (2) refer to historical operation raw data of a cement plant kiln for more than one month, the historical operation raw data comprise operator adjustable parameter data and non-adjustable parameter data, the adjustable parameter data comprise raw material blanking amount, coal injection amount, tertiary air amount and ammonia water amount in the cement production process, and the non-adjustable parameter data comprise tertiary air temperature, decomposing furnace outlet pressure, kiln tail temperature, kiln tail pressure and coal type property.
3. The intelligent regulation and control method for the cement decomposing furnace based on the machine learning and intelligent optimization algorithm according to claim 1, wherein the data cleaning in the step (2) means to delete missing values and abnormal value unsteady values generated by sensor or other equipment faults in historical operation data; the method comprises the following steps:
(2.1) if the missing value in the data only has the missing of the discontinuous time point, linear interpolation filling is carried out by using the recorded values of the time points before and after; if the continuous missing value exists, deleting the measured value in the time period from the system database;
(2.2) judging abnormal values in the data according to the Leider criterion, wherein the abnormal values exist in the data:
wherein the method comprises the steps ofThe sigma is the standard deviation of a certain variable;
(2.3) for unsteady values in the data, because of large fluctuation of a dynamic process, the running parameters cannot truly reflect the running conditions in the decomposing furnace, and a sliding window method is adopted for steady state detection: firstly, selecting a proper window length N, determining the fluctuation condition of running data in the window, and then moving the window backwards to determine the next group of data until all the data are detected; the judgment standard of the stable working condition is as follows:
wherein x is max For a maximum value of an operating parameter within the sliding window, x min Is the minimum value of a certain operation parameter in the sliding window range, x r Is rated under rated working condition, v t Is a stability threshold;
(2.4) carrying out feature selection on factors influencing the NOx emission amount and the raw meal decomposition rate, calculating the degree of interdependence between each parameter and a target value through correlation analysis, eliminating the influence of the parameter with lower correlation, and reducing the calculated amount of the model; based on a mutual information correlation analysis method, selecting characteristic variables which are the greatest in correlation with NOx emission and raw meal decomposition rate except parameters adjustable by operators, wherein the parameters adjustable by the operators comprise raw meal blanking amount, coal injection amount, tertiary air amount and ammonia water consumption;
(2.5) reducing the dimension of the selected data, and mapping the data points in the original high-dimension space into the low-dimension space by adopting a certain mapping method, so that redundant information and noise in the data are reduced, and the calculated amount of the model is further reduced; performing dimension reduction treatment on the data by using a principal component analysis method, and selecting principal components with contribution degree to the target variable being more than 90%;
(2.6) scaling the data by data normalization to fall within a specified interval, thereby avoiding weakening the effect of smaller numerical parameters in the comprehensive analysis process due to different amounts and orders of magnitude of the data; the data processing is completed by using a min-max standardization method, and the calculation method is as follows:
wherein X is norm Is normalized value, X max 、X min 、X mean Respectively the maximum value, the minimum value and the average value of the data;
(2.7) calculating the hysteresis of each variable, namely how long it takes to influence the change of the controlled object; recording each variable delay time as tau i I=1, 2,3 … N, wherein N is the number of variables obtained after the processing of the foregoing steps, and the data in each variable delay period is selected by calculating the delay time of each variable, so as to complete the time sequence matching of the data; and x (t) is the moment t, and the model object calculates an array formed by the required parameters:
x(t)＝[y(t-1),…,y(t-τ y ),u 1 (t),u 1 (t
-1),…,u 1 (t-τ 1 ),…,u N (t),u N (t-1),…,u N (t-τ N )] T
τ in the above formula i The value of (2) is solved by a particle swarm algorithm in which the delay time (τ 1 ,τ 2 ,…,τ p ,τ y ) As a variable needing optimization in an integer space, 0 is equal to or less than tau i ≤τ max ,τ max Maximum delay time for a variable; taking the prediction error as a minimum fitness function, and searching the optimal delay through the minimum fitness function in an optimization process;
(2.8) from the above-mentioned steps (2.1) to (2.7), it is necessary to construct data sets for calculation of NOx emission amount and raw meal decomposition rate, and for the NOx emission amount and raw meal decomposition rate, due to differences in influencing factors, two data sets, respectively { y } 1 (t),x 1 (t) } and { y } 2 (t),x 2 (t)}。
4. The intelligent regulation and control method for the cement decomposing furnace based on the machine learning and intelligent optimization algorithm according to claim 1, wherein the feature selection in the step (2) means that the degree of interdependence between each parameter and a target value is calculated through correlation analysis, the influence of the parameter with lower correlation is eliminated, the calculated amount of a model is reduced, and the feature with the largest correlation is selected based on the method of mutual information correlation analysis; the dimension reduction is to map data points in an original high-dimensional space into a low-dimensional space by adopting a certain mapping method, reduce redundant information and noise in the data, further reduce the calculated amount of a model, perform dimension reduction processing on the data by a principal component analysis method, and select principal components with contribution degree of more than 90% to a target variable; the data normalization refers to scaling data to fall into a specific interval, so that the effect of weakening smaller numerical parameters in the comprehensive analysis process due to different data size classes and orders of magnitude is avoided, and the data processing is completed by using a min-max normalization method.
5. The intelligent regulation and control method for the cement decomposing furnace based on the machine learning and intelligent optimization algorithm according to claim 1, wherein the parameters for measuring the combustion efficiency and pollution in the step (3) refer to a measured value of the decomposition rate of raw materials in the decomposing furnace and a monitored value of the NOx emission amount at the outlet of the decomposing furnace.
6. The intelligent regulation and control method for the cement-decomposing furnace based on the machine learning and intelligent optimization algorithm according to claim 1, wherein the step (5) is specifically: and (3) for the working condition newly added into the database, after data preprocessing, if the model calculation result and the actual value are larger than a certain threshold value, selecting a plurality of groups of data recently, repeating the steps (2) - (4), and reproducing and establishing a new model until the model error meets the condition.
7. A computer storage medium having stored thereon a computer program which, when executed by a processor, implements the intelligent regulation method for cement-decomposing furnaces based on machine learning and intelligent optimization algorithm as claimed in any one of claims 1 to 6.
8. A computer device comprising a memory, a processor and a computer program stored on the memory and running on the processor again, characterized in that the processor implements the intelligent regulation method of a cement-decomposing furnace based on machine learning and intelligent optimization algorithm according to any one of claims 1-6 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210256441.2A CN114721263B (en) | 2022-03-16 | 2022-03-16 | Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210256441.2A CN114721263B (en) | 2022-03-16 | 2022-03-16 | Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114721263A CN114721263A (en) | 2022-07-08 |
| CN114721263B true CN114721263B (en) | 2024-01-23 |
Family
ID=82238050
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210256441.2A Active CN114721263B (en) | 2022-03-16 | 2022-03-16 | Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114721263B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116224795B (en) * | 2023-03-06 | 2023-11-17 | 北京全应科技有限公司 | Thermoelectric production equipment control method based on machine learning model |
| CN116629316B (en) * | 2023-07-26 | 2024-03-08 | 无锡雪浪数制科技有限公司 | Object production model training method, device, electronic equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104536396A (en) * | 2014-12-08 | 2015-04-22 | 沈阳工业大学 | Soft measurement modeling method used in cement raw material decomposing process in decomposing furnace |
| CN104765350A (en) * | 2015-04-03 | 2015-07-08 | 燕山大学 | Cement decomposing furnace control method and system based on combined model predicting control technology |
| CN105574264A (en) * | 2015-12-16 | 2016-05-11 | 长春工业大学 | SVR soft measuring method for kiln tail decomposition rate of cement decomposing furnace |
| CN111833970A (en) * | 2020-06-18 | 2020-10-27 | 湖北博华自动化系统工程有限公司 | Construction method and application of cement clinker quality characterization parameter prediction model |
| CN112085196A (en) * | 2020-09-10 | 2020-12-15 | 南京工业大学 | Ammonia injection adjusting method for SCR denitration system based on SCA algorithm optimization BP neural network |
| CN113268871A (en) * | 2021-05-21 | 2021-08-17 | 燕山大学 | Cement chimney NOX prediction method based on multivariable time sequence depth network model |
| CN113589693A (en) * | 2021-07-22 | 2021-11-02 | 燕山大学 | Cement industry decomposing furnace temperature model prediction control method based on neighborhood optimization |
-
2022
- 2022-03-16 CN CN202210256441.2A patent/CN114721263B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104536396A (en) * | 2014-12-08 | 2015-04-22 | 沈阳工业大学 | Soft measurement modeling method used in cement raw material decomposing process in decomposing furnace |
| CN104765350A (en) * | 2015-04-03 | 2015-07-08 | 燕山大学 | Cement decomposing furnace control method and system based on combined model predicting control technology |
| CN105574264A (en) * | 2015-12-16 | 2016-05-11 | 长春工业大学 | SVR soft measuring method for kiln tail decomposition rate of cement decomposing furnace |
| CN111833970A (en) * | 2020-06-18 | 2020-10-27 | 湖北博华自动化系统工程有限公司 | Construction method and application of cement clinker quality characterization parameter prediction model |
| CN112085196A (en) * | 2020-09-10 | 2020-12-15 | 南京工业大学 | Ammonia injection adjusting method for SCR denitration system based on SCA algorithm optimization BP neural network |
| CN113268871A (en) * | 2021-05-21 | 2021-08-17 | 燕山大学 | Cement chimney NOX prediction method based on multivariable time sequence depth network model |
| CN113589693A (en) * | 2021-07-22 | 2021-11-02 | 燕山大学 | Cement industry decomposing furnace temperature model prediction control method based on neighborhood optimization |
Non-Patent Citations (2)
| Title |
|---|
| Prediction of f-CaO content in cement clinker: A novel prediction method based on LightGBM and Bayesian optimization;Xiaochen Hao等;《ES》;1-11 * |
| 基于动态数据维数约简和半监督学习的水泥分解炉生料分解率预报建模;王沼钧;《中国优秀硕士学位论文全文数据库 工程科技I辑》(第3期);正文第21-24、28-41、53、61-63页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114721263A (en) | 2022-07-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114721263B (en) | Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm | |
| CN109190848B (en) | SCR system NO based on time delay predictionxEmission concentration prediction method | |
| CN113433911B (en) | Accurate control system and method for ammonia spraying of denitration device based on accurate concentration prediction | |
| CN113627071B (en) | Coal-fired boiler NO based on whale algorithm optimization long-short-term memory network X Soft measurement method | |
| CN110598929B (en) | Wind power nonparametric probability interval ultrashort term prediction method | |
| CN111829003B (en) | Power plant combustion control system and control method | |
| CN104765350A (en) | Cement decomposing furnace control method and system based on combined model predicting control technology | |
| CN111476422A (en) | L ightGBM building cold load prediction method based on machine learning framework | |
| CN112066355B (en) | Self-adaptive adjusting method of waste heat boiler valve based on data driving | |
| CN112749840B (en) | Method for acquiring energy efficiency characteristic index reference value of thermal power generating unit | |
| CN115016276B (en) | Intelligent water content adjustment and environment parameter Internet of things big data system | |
| CN113469449B (en) | Optimization control method and system for desulfurization system | |
| CN115077640A (en) | Real-time prediction method for main steam flow in urban solid waste incineration process | |
| CN117289652A (en) | Numerical control machine tool spindle thermal error modeling method based on multi-universe optimization | |
| CN109101683B (en) | Model updating method for pyrolysis kettle of coal quality-based utilization and clean pretreatment system | |
| CN110684547A (en) | Optimized control method for biomass pyrolysis carbonization kiln | |
| CN111336828A (en) | Heating furnace temperature controller based on FCM fuzzy time sequence | |
| CN117270387A (en) | SCR denitration system low ammonia escape control method and system based on deep learning | |
| CN113836819B (en) | Bed temperature prediction method based on time sequence attention | |
| CN113488111B (en) | Ammonia injection amount optimization modeling method for SCR denitration system | |
| CN111222708B (en) | Power plant combustion furnace temperature prediction method based on transfer learning dynamic modeling | |
| CN103558762B (en) | The implementation method of the immune genetic PID controller based on graphical configuration technology | |
| CN117574290B (en) | Thermodynamic data anomaly detection and repair method based on mechanism and data collaborative driving | |
| CN105068423B (en) | Method for realizing intelligent parameter identification of steam turbine and speed regulating system thereof by one key | |
| CN114021819B (en) | Thermal power plant nitrogen oxide content prediction method based on fractional gray delay model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |