CN118011835A - Blast furnace ironmaking control system based on machine learning - Google Patents
Blast furnace ironmaking control system based on machine learning Download PDFInfo
- Publication number
- CN118011835A CN118011835A CN202410411006.1A CN202410411006A CN118011835A CN 118011835 A CN118011835 A CN 118011835A CN 202410411006 A CN202410411006 A CN 202410411006A CN 118011835 A CN118011835 A CN 118011835A
- Authority
- CN
- China
- Prior art keywords
- data
- blast furnace
- model
- furnace ironmaking
- parameter
- 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.)
- Granted
Links
- 238000010801 machine learning Methods 0.000 title claims abstract description 16
- 239000011159 matrix material Substances 0.000 claims abstract description 16
- 230000035772 mutation Effects 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 10
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 238000000034 method Methods 0.000 claims description 30
- 230000008569 process Effects 0.000 claims description 22
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 12
- 229910052742 iron Inorganic materials 0.000 claims description 8
- 238000005192 partition Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000004519 manufacturing process Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 239000002994 raw material Substances 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000004140 cleaning Methods 0.000 claims description 5
- 230000005856 abnormality Effects 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 238000007621 cluster analysis Methods 0.000 claims description 3
- 238000002485 combustion reaction Methods 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 3
- 239000000446 fuel Substances 0.000 claims description 3
- 239000012535 impurity Substances 0.000 claims description 3
- 230000007774 longterm Effects 0.000 claims description 3
- 239000000463 material Substances 0.000 claims description 3
- 238000011524 similarity measure Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 2
- 238000000638 solvent extraction Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 abstract description 8
- 230000001976 improved effect Effects 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 abstract description 4
- 230000009471 action Effects 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Manufacture Of Iron (AREA)
Abstract
The invention discloses a blast furnace ironmaking control system based on machine learning, which comprises a data acquisition module, a data preprocessing module, a blast furnace ironmaking prediction model building module, a model super-parameter setting module and a blast furnace ironmaking control module. The invention belongs to the technical field of data processing, in particular to a blast furnace ironmaking control system based on machine learning, and the scheme combines clustering and fuzzy theory, is based on clustering analysis and fuzzy division matrix calculation, fully considers the characteristics of time sequence data, and improves the abstract capacity and prediction accuracy of the model on the data characteristics; by dividing elite groups and parallel groups, nonlinear parameters are designed to improve the searching efficiency; based on the position movement and mutation strategies, the population diversity is kept for global searching, local searching can be carried out around the individuals, and better super-parameter combinations can be found; the effect of super-parameter optimization is improved, and the performance of the model is optimized.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a blast furnace ironmaking control system based on machine learning.
Background
The blast furnace ironmaking control system based on machine learning is a technical direction for optimizing an automatic control system in an ironmaking process. Ironmaking is a process of reducing iron ore to iron, and in this complex process, a control system needs to monitor and adjust various parameters and variables to ensure stability, efficiency and product quality of the ironmaking process. However, the general blast furnace ironmaking control system has the problems that the model has poor characterization capability on the data characteristics, and the model cannot understand the internal structure of the data, so that the adaptability of the model is poor and the prediction precision is low; the general blast furnace ironmaking control system has the problems of improper model super-parameter setting and low local optimization and convergence rate in super-parameter searching.
Disclosure of Invention
Aiming at the problems that the model has poor characterization capability on the data characteristics and the model cannot understand the internal structure of the data, so that the adaptability of the model is poor and the prediction precision is low in the conventional blast furnace ironmaking control system, the scheme is combined with the clustering and fuzzy theory to better identify the data clustering characteristics and the fuzzy rule, and based on the clustering analysis and the fuzzy division matrix calculation, more flexible and comprehensive data characteristics are provided for the model, the characteristics of time sequence data are fully considered, the abstract capability and the prediction accuracy of the model on the data characteristics are improved, and positive effects are played in blast furnace ironmaking temperature prediction; aiming at the problems of improper model super-parameter setting and low local optimization and convergence rate of super-parameter searching in a general blast furnace ironmaking control system, the scheme improves the searching efficiency by dividing elite groups and parallel groups and designing nonlinear parameters; based on the position movement and mutation strategies, the population diversity is kept for global searching, local searching can be carried out around the individuals, and better super-parameter combinations can be found; the effect of super-parameter optimization is improved, and the performance of the model is optimized.
The technical scheme adopted by the invention is as follows: the invention provides a machine learning-based blast furnace ironmaking control system, which comprises a data acquisition module, a data preprocessing module, a blast furnace ironmaking prediction model building module, a model super-parameter setting module and a blast furnace ironmaking control module, wherein the data acquisition module is used for acquiring data of a blast furnace ironmaking control system;
The data acquisition module acquires blast furnace operation data, raw material data, blast furnace production data, environment data and time sequences;
the data preprocessing module performs data cleaning, data conversion, standardization processing and data set division on the acquired data;
The blast furnace ironmaking prediction model establishment module combines clustering and fuzzy theory, obtains fuzzy time sequence data based on cluster analysis and fuzzy partition matrix calculation, and predicts by using a two-way long-short-term memory network model;
The model super-parameter setting module designs nonlinear parameters by dividing elite groups and parallel groups; performing super-parameter search setting based on the position movement and mutation strategies;
The blast furnace ironmaking control module performs blast furnace ironmaking control through the established blast furnace ironmaking prediction model and the data acquired in real time.
Further, in the data acquisition module, the blast furnace operation data comprise furnace temperature, pressure, gas flow rate and material input speed; the raw material data includes components, impurity content and fuel combustion characteristics of iron ore; the blast furnace production data includes yield, product quality, and exhaust emission data; the environmental data includes climate conditions, temperature and humidity; the temperature in the furnace is used as a data tag.
Further, in the data preprocessing module, the data cleaning is to process the missing value, the abnormal value and the repeated value; the data conversion is to convert the data into a vector form; the normalization processing is to convert the data into a normalized time sequence data set based on a maximum and minimum normalization method; the partitioning of the data set is to partition the data set into a training set and a testing set.
Further, the blast furnace ironmaking prediction model building module specifically comprises the following contents:
initializing, namely taking the time sequence data set as time sequence data input by a model; the initial clustering prototype element set V (0) is used for reflecting control parameter configuration under different operation conditions and is used for model initialization;
The element of the cluster prototype is calculated, ;/>Is the initialized i-th clustering prototype element, which is equivalent to x i; n is the number of elements; the clustering prototype elements are expressed as follows:
;
Wherein x i is the i-th cluster prototype element; t i is the i-th element in the original time-series data; min { T } and max { T } are time series minimum and maximum values, respectively;
updating the clustering prototype element by the following formula:
;
;
;
;
;
Where a ij (t) is the balance factor at time t and a ij (0) =1, 、AndThe i-th clustering prototype elements at the time t+1, the time t and the time t-1 are respectively; And The j-th clustering prototype elements at the time t and the time t-1 are respectively; f (·) is a similarity measure function; Is the Euclidean distance; lambda is the adjustment parameter; μ is the average parameter; i and j are different elements;
Acquiring cluster set, repeatedly updating cluster prototype elements until Where ε is the distance threshold; each element in V now converges to the prototype element of the group that contains it; obtaining a cluster set P, denoted asP 1、p2 and p k are the 1 st, 2 nd and k th cluster centers, respectively; p j is the j-th cluster center;
calculating a fuzzy division matrix of the time sequence to obtain a division matrix U with the size of N rows and k columns, wherein the formula is as follows:
;
Wherein μ ji is a fuzzy division matrix element; x i is the i-th cluster prototype element; x c is the c-th cluster prototype element;
Calculating fuzzy time sequence elements to obtain fuzzy time sequences F= { F 1,F2,…,FN},F1、F2 and F N, wherein the F= { F 1,F2,…,FN},F1、F2 and F N are the 1 st, 2 nd and N th fuzzy time sequence elements respectively; the formula used is as follows:
;
Wherein F i is the i-th fuzzy time series element;
Predicting by using a two-way long-short-term memory network model; the bidirectional long-short-term memory network is a circulating neural network which is specially used for processing long-term dependency in sequence data, and can integrate the information of past and future time steps, so that the network can bidirectionally learn the modes in the sequence; the core of the LSTM architecture is its memory cell, which adds three key gates: an input gate, an output gate, and a forget gate; each gate has a respective function to enhance the learning process; the output of the model element is expressed as follows:
;
;
;
;
;
wherein i t、ft and o t are the outputs of the input gate, the forget gate and the output gate, respectively; Is a sigmoid activation function; a i、Af、Ag and a o are weight matrices of input gates, forget gates, cell state update and output gates, respectively; r i、Rf、Rg and R o are respectively cyclic weight matrices of input gates, forget gates, cell state update and output gates; alpha i、αf、αg and alpha o are the bias of the input gate, the forget gate, the cell state update and the output gate, respectively; tan h (·) is a hyperbolic tangent function; c t-1 and t h-1 are the cell state and the hidden state, respectively, of the previous time step; g t is a candidate value.
Further, the model super-parameter setting module specifically includes the following contents:
initializing, namely establishing a parameter searching space based on the model super parameters, and randomly initializing parameter searching population positions; taking the prediction accuracy of the model established based on the parameter individuals to the test set as an individual fitness value; the search populations are arranged in descending order of fitness value, and the first 50% are used as elite groups Representing elite group individual positions; the rest are parallel groups, use/>Representing parallel sets of individual positions;
nonlinear parameters were designed using the following formula:
;
Wherein a is a nonlinear parameter; t is the number of search iterations; maxT is the maximum number of iterations; Is the average fitness value of the population;
The inertial weight parameters were designed using the following formula:
;
wherein ω is an inertial weight parameter; Is an inertial constant; /(I) Is a random disturbance term;
the position movement strategy is designed with the following formula:
;
;
;
In the method, in the process of the invention, Is the optimal position of the population; /(I)Is the optimal position of the history of the individual; a and C are parameters for introducing randomness and controlling the degree of movement, respectively; p, r and l are all random numbers belonging to 0 to 1, independently of each other; b is a logarithmic parameter; maxω is the maximum inertial weight; /(I)Is the optimal position of the parallel group of the T-th iteration;
a mutation strategy was introduced and was used for the parallel group, the formula used was as follows:
;
;
In the method, in the process of the invention, Is a post-mutation position; sm is a mutation value; f i (T), fmin (T) and fmax (T) are the individual fitness value, the population worst fitness value and the population best fitness value at the T-th iteration, respectively;
Searching and judging, namely presetting a searching threshold value, and establishing a model based on the super parameters represented by the individual positions when the individual fitness value is higher than the fitness threshold value; when the maximum iteration times are reached, the population position search is reinitialized; otherwise, the repartitioning group continues searching.
Further, the blast furnace ironmaking control module is used for setting blast furnace ironmaking prediction model super parameters based on the parameter positions searched by the model super parameter setting module; and (3) collecting blast furnace ironmaking data in real time, and controlling the ironmaking process in real time based on the model predicted temperature in the furnace so as to prevent the iron making abnormality caused by overhigh temperature in the furnace.
By adopting the scheme, the beneficial effects obtained by the invention are as follows:
(1) Aiming at the problems that the representation capability of a model for data features is poor, the model cannot understand the inherent structure of data, so that the adaptability of the model is poor and the prediction accuracy is low in a general blast furnace ironmaking control system, the scheme combines the clustering and fuzzy theory, better identifies the data clustering features and fuzzy rules, provides more flexible and comprehensive data features for the model based on clustering analysis and fuzzy partition matrix calculation, fully considers the characteristics of time series data, improves the abstraction capability and the prediction accuracy of the data features of the model, and plays a positive role in blast furnace ironmaking temperature prediction.
(2) Aiming at the problems of improper model super-parameter setting and low local optimization and convergence rate of super-parameter searching in a general blast furnace ironmaking control system, the scheme improves the searching efficiency by dividing elite groups and parallel groups and designing nonlinear parameters; based on the position movement and mutation strategies, the population diversity is kept for global searching, local searching can be carried out around the individuals, and better super-parameter combinations can be found; the effect of super-parameter optimization is improved, and the performance of the model is optimized.
Drawings
FIG. 1 is a schematic diagram of a machine learning based blast furnace ironmaking control system provided by the invention;
FIG. 2 is a schematic flow diagram of a blast furnace ironmaking prediction model building module;
Fig. 3 is a schematic flow chart of the model hyper-parameter setting module.
The accompanying drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the invention; all other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "top," "bottom," "inner," "outer," and the like indicate orientation or positional relationships based on those shown in the drawings, merely to facilitate describing the present invention and simplify the description, and do not indicate or imply that the system or element being referred to must have a particular orientation, be constructed and operate in a particular orientation, and thus should not be construed as limiting the present invention.
Embodiment one:
referring to fig. 1, the machine learning-based blast furnace ironmaking control system provided by the invention comprises a data acquisition module, a data preprocessing module, a blast furnace ironmaking prediction model building module, a model super-parameter setting module and a blast furnace ironmaking control module;
The data acquisition module acquires blast furnace operation data, raw material data, blast furnace production data, environment data and time sequences; and sending the data to a data preprocessing module;
The data preprocessing module receives the data sent by the data acquisition module; carrying out data cleaning, data conversion, standardization processing and data set division on the collected data; the data is sent to a blast furnace ironmaking prediction model building module;
The blast furnace ironmaking prediction model building module receives the data sent by the data preprocessing module; combining the clustering and fuzzy theory, obtaining fuzzy time sequence data based on cluster analysis and fuzzy partition matrix calculation, and predicting by using a two-way long-short-term memory network model; the data is sent to a model super-parameter setting module;
The model super-parameter setting module is used for receiving the data sent by the blast furnace ironmaking prediction model building module; designing nonlinear parameters by dividing elite groups and parallel groups; performing super-parameter search setting based on the position movement and mutation strategies; and sending the data to a blast furnace ironmaking control module;
the blast furnace ironmaking control module receives the data sent by the model super-parameter setting module; and carrying out blast furnace ironmaking control through the established blast furnace ironmaking prediction model and the data acquired in real time.
Embodiment two:
Referring to fig. 1, the embodiment is based on the above embodiment, and in the data acquisition module, the blast furnace operation data includes the furnace temperature, pressure, gas flow rate, and material input speed; the raw material data includes the composition of iron ore, impurity content and fuel combustion characteristics; blast furnace production data including yield, product quality, and exhaust emission data; environmental data includes climate conditions, temperature and humidity; the temperature in the furnace is used as a data tag.
Embodiment III:
Referring to fig. 1, the embodiment is based on the above embodiment, in the data preprocessing module, the data cleansing is to process the missing values, the outliers and the repeated values; the data conversion is to convert the data into a vector form; the normalization process is to convert the data into a normalized time series data set based on a maximum and minimum normalization method; dividing the data set is to divide the data set into a training set and a testing set.
Embodiment four:
Referring to fig. 1 and 2, the blast furnace ironmaking prediction model building module according to the embodiment specifically includes the following:
initializing, namely taking the time sequence data set as time sequence data input by a model; the initial clustering prototype element set V (0) is used for reflecting control parameter configuration under different operation conditions and is used for model initialization;
The element of the cluster prototype is calculated, ;/>Is the initialized i-th clustering prototype element, which is equivalent to x i; n is the number of elements; the clustering prototype elements are expressed as follows:
;
Wherein x i is the i-th cluster prototype element; t i is the i-th element in the original time-series data; min { T } and max { T } are time series minimum and maximum values, respectively;
updating the clustering prototype element by the following formula:
;
;
;
;
;
Where a ij (t) is the balance factor at time t and a ij (0) =1, 、AndThe i-th clustering prototype elements at the time t+1, the time t and the time t-1 are respectively; And The j-th clustering prototype elements at the time t and the time t-1 are respectively; f (·) is a similarity measure function; Is the Euclidean distance; lambda is the adjustment parameter; μ is the average parameter; i and j are different elements;
Acquiring cluster set, repeatedly updating cluster prototype elements until Where ε is the distance threshold; each element in V now converges to the prototype element of the group that contains it; obtaining a cluster set P, denoted asP 1、p2 and p k are the 1 st, 2 nd and k th cluster centers, respectively; p j is the j-th cluster center;
calculating a fuzzy division matrix of the time sequence to obtain a division matrix U with the size of N rows and k columns, wherein the formula is as follows:
;
Wherein μ ji is a fuzzy division matrix element; x i is the i-th cluster prototype element; x c is the c-th cluster prototype element;
Calculating fuzzy time sequence elements to obtain fuzzy time sequences F= { F 1,F2,…,FN},F1、F2 and F N, wherein the F= { F 1,F2,…,FN},F1、F2 and F N are the 1 st, 2 nd and N th fuzzy time sequence elements respectively; the formula used is as follows:
;
Wherein F i is the i-th fuzzy time series element;
Predicting by using a two-way long-short-term memory network model; the bidirectional long-short-term memory network is a circulating neural network which is specially used for processing long-term dependency in sequence data, and can integrate the information of past and future time steps, so that the network can bidirectionally learn the modes in the sequence; the core of the LSTM architecture is its memory cell, which adds three key gates: an input gate, an output gate, and a forget gate; each gate has a respective function to enhance the learning process; the output of the model element is expressed as follows:
;
;
;
;
;
wherein i t、ft and o t are the outputs of the input gate, the forget gate and the output gate, respectively; Is a sigmoid activation function; a i、Af、Ag and a o are weight matrices of input gates, forget gates, cell state update and output gates, respectively; r i、Rf、Rg and R o are respectively cyclic weight matrices of input gates, forget gates, cell state update and output gates; alpha i、αf、αg and alpha o are the bias of the input gate, the forget gate, the cell state update and the output gate, respectively; tan h (·) is a hyperbolic tangent function; c t-1 and t h-1 are the cell state and the hidden state, respectively, of the previous time step; g t is a candidate value.
By executing the operation, aiming at the problems that the representation capability of the model for the data characteristics is poor, the model cannot understand the internal structure of the data, so that the adaptability of the model is poor and the prediction accuracy is low, the scheme combines the clustering and fuzzy theory, better identifies the data clustering characteristics and fuzzy rules, provides more flexible and comprehensive data characteristics for the model based on the clustering analysis and fuzzy partition matrix calculation, fully considers the characteristics of time series data, improves the abstract capability and the prediction accuracy of the data characteristics of the model, and plays a positive role in the blast furnace ironmaking temperature prediction.
Fifth embodiment:
referring to fig. 1 and 3, the blast furnace ironmaking control module according to the above embodiment specifically includes the following:
initializing, namely establishing a parameter searching space based on the model super parameters, and randomly initializing parameter searching population positions; taking the prediction accuracy of the model established based on the parameter individuals to the test set as an individual fitness value; the search populations are arranged in descending order of fitness value, and the first 50% are used as elite groups Representing elite group individual positions; the rest are parallel groups, use/>Representing parallel sets of individual positions;
nonlinear parameters were designed using the following formula:
;
Wherein a is a nonlinear parameter; t is the number of search iterations; maxT is the maximum number of iterations; Is the average fitness value of the population;
The inertial weight parameters were designed using the following formula:
;
wherein ω is an inertial weight parameter; Is an inertial constant; /(I) Is a random disturbance term;
the position movement strategy is designed with the following formula:
;
;
;
In the method, in the process of the invention, Is the optimal position of the population; /(I)Is the optimal position of the history of the individual; a and C are parameters for introducing randomness and controlling the degree of movement, respectively; p, r and l are all random numbers belonging to 0 to 1, independently of each other; b is a logarithmic parameter; maxω is the maximum inertial weight; /(I)Is the optimal position of the parallel group of the T-th iteration;
a mutation strategy was introduced and was used for the parallel group, the formula used was as follows:
;
;
In the method, in the process of the invention, Is a post-mutation position; sm is a mutation value; f i (T), fmin (T) and fmax (T) are the individual fitness value, the population worst fitness value and the population best fitness value at the T-th iteration, respectively;
Searching and judging, namely presetting a searching threshold value, and establishing a model based on the super parameters represented by the individual positions when the individual fitness value is higher than the fitness threshold value; when the maximum iteration times are reached, the population position search is reinitialized; otherwise, the repartitioning group continues searching.
By executing the operations, the scheme improves the searching efficiency by dividing elite groups and parallel groups and designing nonlinear parameters aiming at the problems of improper model super-parameter setting and low local optimization and convergence speed of super-parameter searching in a general blast furnace ironmaking control system; based on the position movement and mutation strategies, the population diversity is kept for global searching, local searching can be carried out around the individuals, and better super-parameter combinations can be found; the effect of super-parameter optimization is improved, and the performance of the model is optimized.
Example six:
referring to fig. 1, in this embodiment, based on the foregoing embodiment, a blast furnace ironmaking control module sets blast furnace ironmaking prediction model super parameters based on parameter positions searched by a model super parameter setting module; and (3) collecting blast furnace ironmaking data in real time, and controlling the ironmaking process in real time based on the model predicted temperature in the furnace so as to prevent the iron making abnormality caused by overhigh temperature in the furnace.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made hereto without departing from the spirit and principles of the present invention.
The invention and its embodiments have been described above with no limitation, and the actual construction is not limited to the embodiments of the invention as shown in the drawings. In summary, if one of ordinary skill in the art is informed by this disclosure, a structural manner and an embodiment similar to the technical solution should not be creatively devised without departing from the gist of the present invention.
Claims (7)
1. Blast furnace ironmaking control system based on machine learning, its characterized in that: the system comprises a data acquisition module, a data preprocessing module, a blast furnace ironmaking prediction model building module, a model super-parameter setting module and a blast furnace ironmaking control module;
The data acquisition module acquires blast furnace operation data, raw material data, blast furnace production data, environment data and time sequences;
the data preprocessing module performs data cleaning, data conversion, standardization processing and data set division on the acquired data;
The blast furnace ironmaking prediction model establishment module combines clustering and fuzzy theory, obtains fuzzy time sequence data based on cluster analysis and fuzzy partition matrix calculation, and predicts by using a two-way long-short-term memory network model;
The model super-parameter setting module designs nonlinear parameters by dividing elite groups and parallel groups; performing super-parameter search setting based on the position movement and mutation strategies;
The blast furnace ironmaking control module performs blast furnace ironmaking control through the established blast furnace ironmaking prediction model and the data acquired in real time;
Nonlinear parameters are designed in the model super-parameter setting module, and the following formula is used:
;
Wherein a is a nonlinear parameter; t is the number of search iterations; maxT is the maximum number of iterations; Is the average fitness value of the population;
The inertial weight parameters are designed in the model super-parameter setting module, and the formula is as follows:
;
wherein ω is an inertial weight parameter; Is an inertial constant; /(I) Is a random perturbation term.
2. The machine learning based blast furnace ironmaking control system of claim 1, wherein: the blast furnace ironmaking prediction model building module specifically comprises the following contents:
initializing, namely taking the time sequence data set as time sequence data input by a model; the initial clustering prototype element set V (0) is used for reflecting control parameter configuration under different operation conditions and is used for model initialization;
The element of the cluster prototype is calculated, ;/>Is the initialized i-th clustering prototype element, which is equivalent to x i; n is the number of elements; the clustering prototype elements are expressed as follows:
;
Wherein x i is the i-th cluster prototype element; t i is the i-th element in the original time-series data; min { T } and max { T } are time series minimum and maximum values, respectively;
updating the clustering prototype element by the following formula:
;
;
;
;
;
Where a ij (t) is the balance factor at time t and a ij (0) =1, 、/>And/>The i-th clustering prototype elements at the time t+1, the time t and the time t-1 are respectively; /(I)And/>The j-th clustering prototype elements at the time t and the time t-1 are respectively; f (·) is a similarity measure function; /(I)Is the Euclidean distance; lambda is the adjustment parameter; μ is the average parameter; i and j are different elements;
Acquiring cluster set, repeatedly updating cluster prototype elements until Where ε is the distance threshold; each element in V now converges to the prototype element of the group that contains it; obtaining a cluster set P, denoted asP 1、p2 and p k are the 1 st, 2 nd and k th cluster centers, respectively; p j is the j-th cluster center;
calculating a fuzzy division matrix of the time sequence to obtain a division matrix U with the size of N rows and k columns, wherein the formula is as follows:
;
Wherein μ ji is a fuzzy division matrix element; x i is the i-th cluster prototype element; x c is the c-th cluster prototype element;
Calculating fuzzy time sequence elements to obtain fuzzy time sequences F= { F 1,F2,…,FN},F1、F2 and F N, wherein the F= { F 1,F2,…,FN},F1、F2 and F N are the 1 st, 2 nd and N th fuzzy time sequence elements respectively; the formula used is as follows:
;
where F i is the i-th ambiguous time series element.
3. The machine learning based blast furnace ironmaking control system of claim 1, wherein: the blast furnace ironmaking prediction model building module further comprises:
Predicting by using a two-way long-short-term memory network model; the bidirectional long-short-term memory network is a circulating neural network which is specially used for processing long-term dependency in sequence data, and can integrate the information of past and future time steps, so that the network can bidirectionally learn the modes in the sequence; the core of the LSTM architecture is its memory cell, which adds three key gates: an input gate, an output gate, and a forget gate; each gate has a respective function to enhance the learning process; the output of the model element is expressed as follows:
;
;
;
;
;
wherein i t、ft and o t are the outputs of the input gate, the forget gate and the output gate, respectively; Is a sigmoid activation function; a i、Af、Ag and a o are weight matrices of input gates, forget gates, cell state update and output gates, respectively; r i、Rf、Rg and R o are respectively cyclic weight matrices of input gates, forget gates, cell state update and output gates; alpha i、αf、αg and alpha o are the bias of the input gate, the forget gate, the cell state update and the output gate, respectively; tan h (·) is a hyperbolic tangent function; c t-1 and t h-1 are the cell state and the hidden state, respectively, of the previous time step; g t is a candidate value.
4. The machine learning based blast furnace ironmaking control system of claim 1, wherein: the model super-parameter setting module specifically comprises the following contents:
initializing, namely establishing a parameter searching space based on the model super parameters, and randomly initializing parameter searching population positions; taking the prediction accuracy of the model established based on the parameter individuals to the test set as an individual fitness value; the search populations are arranged in descending order of fitness value, and the first 50% are used as elite groups Representing elite group individual positions; the rest are parallel groups, use/>Representing parallel sets of individual positions;
Designing nonlinear parameters;
designing inertia weight parameters;
the position movement strategy is designed with the following formula:
;
;
;
In the method, in the process of the invention, Is the optimal position of the population; /(I)Is the optimal position of the history of the individual; a and C are parameters for introducing randomness and controlling the degree of movement, respectively; p, r and l are all random numbers belonging to 0 to 1, independently of each other; b is a logarithmic parameter; maxω is the maximum inertial weight; /(I)Is the optimal position of the parallel group of the T-th iteration;
a mutation strategy was introduced and was used for the parallel group, the formula used was as follows:
;
;
In the method, in the process of the invention, Is a post-mutation position; sm is a mutation value; f i (T), fmin (T) and fmax (T) are the individual fitness value, the population worst fitness value and the population best fitness value at the T-th iteration, respectively;
Searching and judging, namely presetting a searching threshold value, and establishing a model based on the super parameters represented by the individual positions when the individual fitness value is higher than the fitness threshold value; when the maximum iteration times are reached, the population position search is reinitialized; otherwise, the repartitioning group continues searching.
5. The machine learning based blast furnace ironmaking control system of claim 1, wherein: in the data acquisition module, the blast furnace operation data comprise furnace temperature, pressure, gas flow rate and material input speed; the raw material data includes components, impurity content and fuel combustion characteristics of iron ore; the blast furnace production data includes yield, product quality, and exhaust emission data; the environmental data includes climate conditions, temperature and humidity; the temperature in the furnace is used as a data tag.
6. The machine learning based blast furnace ironmaking control system of claim 1, wherein: the blast furnace ironmaking control module is used for setting the super parameters of the blast furnace ironmaking prediction model based on the parameter positions searched by the model super parameter setting module; and (3) collecting blast furnace ironmaking data in real time, and controlling the ironmaking process in real time based on the model predicted temperature in the furnace so as to prevent the iron making abnormality caused by overhigh temperature in the furnace.
7. The machine learning based blast furnace ironmaking control system of claim 1, wherein: in the data preprocessing module, the data cleaning is to process the missing value, the abnormal value and the repeated value; the data conversion is to convert the data into a vector form; the normalization processing is to convert the data into a normalized time sequence data set based on a maximum and minimum normalization method; the partitioning of the data set is to partition the data set into a training set and a testing set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410411006.1A CN118011835B (en) | 2024-04-08 | Blast furnace ironmaking control system based on machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410411006.1A CN118011835B (en) | 2024-04-08 | Blast furnace ironmaking control system based on machine learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN118011835A true CN118011835A (en) | 2024-05-10 |
CN118011835B CN118011835B (en) | 2024-06-25 |
Family
ID=
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104899463A (en) * | 2015-06-18 | 2015-09-09 | 中南大学 | Blast furnace molten iron silicon content four-classification trend prediction model establishing method and application |
CN105257425A (en) * | 2014-06-11 | 2016-01-20 | 凯文·李·弗里斯特 | Quintuple-effect generation multi-cycle hybrid renewable energy system with integrated energy provisioning, storage facilities and amalgamated control system |
CN105868867A (en) * | 2016-04-25 | 2016-08-17 | 常州英集动力科技有限公司 | Method and system for optimized operation of heating boiler cluster |
CN108764517A (en) * | 2018-04-08 | 2018-11-06 | 中南大学 | A kind of blast furnace molten iron silicon content trend method, equipment and storage medium |
CN108898214A (en) * | 2018-06-29 | 2018-11-27 | 山东师范大学 | A kind of online sequence data prediction technique and device |
CN115758193A (en) * | 2022-11-07 | 2023-03-07 | 国网河北省电力有限公司电力科学研究院 | Distributed energy storage aggregation control method and device |
CN115936061A (en) * | 2022-11-04 | 2023-04-07 | 南京工业职业技术大学 | Thermal power plant flue gas oxygen content soft measurement method and system based on data driving |
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105257425A (en) * | 2014-06-11 | 2016-01-20 | 凯文·李·弗里斯特 | Quintuple-effect generation multi-cycle hybrid renewable energy system with integrated energy provisioning, storage facilities and amalgamated control system |
CN104899463A (en) * | 2015-06-18 | 2015-09-09 | 中南大学 | Blast furnace molten iron silicon content four-classification trend prediction model establishing method and application |
CN105868867A (en) * | 2016-04-25 | 2016-08-17 | 常州英集动力科技有限公司 | Method and system for optimized operation of heating boiler cluster |
CN108764517A (en) * | 2018-04-08 | 2018-11-06 | 中南大学 | A kind of blast furnace molten iron silicon content trend method, equipment and storage medium |
CN108898214A (en) * | 2018-06-29 | 2018-11-27 | 山东师范大学 | A kind of online sequence data prediction technique and device |
CN115936061A (en) * | 2022-11-04 | 2023-04-07 | 南京工业职业技术大学 | Thermal power plant flue gas oxygen content soft measurement method and system based on data driving |
CN115758193A (en) * | 2022-11-07 | 2023-03-07 | 国网河北省电力有限公司电力科学研究院 | Distributed energy storage aggregation control method and device |
Non-Patent Citations (6)
Title |
---|
MINCHAO XIE,等: "Analysis of Industrial Metabolic Flux in Blast Furnace Ironmaking System", IOP CONF. SERIES: EARTH AND ENVIRONMENTAL SCIENCE, vol. 359, 31 December 2019 (2019-12-31), pages 1 - 5 * |
WANGYAN LI ,等: "A data-based soft-sensor approach to estimating raceway depth in ironmaking blast furnaces", POWDER TECHNOLOGY, vol. 390, 31 December 2021 (2021-12-31), pages 529 * |
刘紫微,杨晓忠: "—种基于CS-FCM算法的模糊时间序列预测模型", 应用数学学报, vol. 45, no. 3, 31 May 2022 (2022-05-31), pages 322 - 338 * |
张乾君,等: "基于时间序列聚类的多雷达数据融合", 电讯技术, vol. 59, no. 2, 28 February 2019 (2019-02-28), pages 145 - 150 * |
杜学强: "基于混杂Petri网的钢铁制造流程能量流建模仿真研究", 中国博士学位论文数据库, 31 May 2023 (2023-05-31), pages 1 - 117 * |
马晓峰,钱俊磊: "特殊条件下高炉防灌渣自动控制系统", 冶金自动化, no. 2, 31 December 2010 (2010-12-31), pages 62 - 64 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bayati et al. | MLPSO: a filter multi-label feature selection based on particle swarm optimization | |
Wang et al. | A cluster-based competitive particle swarm optimizer with a sparse truncation operator for multi-objective optimization | |
Zeng et al. | Accurately clustering single-cell RNA-seq data by capturing structural relations between cells through graph convolutional network | |
Bonzo et al. | Clustering panel data via perturbed adaptive simulated annealing and genetic algorithms | |
CN111931418B (en) | Thermal efficiency prediction method of horseshoe flame glass kiln based on evolutionary neural network | |
Guo et al. | Towards accurate and compact architectures via neural architecture transformer | |
CN110851662B (en) | Heterogeneous information network link prediction method based on meta-path | |
CN108734223A (en) | The social networks friend recommendation method divided based on community | |
CN113094822A (en) | Method and system for predicting residual life of mechanical equipment | |
Bian et al. | Subarchitecture ensemble pruning in neural architecture search | |
CN115496263A (en) | Short-term photovoltaic power generation prediction method based on space-time genetic attention network | |
CN117034767A (en) | Ceramic roller kiln temperature prediction method based on KPCA-GWO-GRU | |
CN116629428A (en) | Building energy consumption prediction method based on feature selection and SSA-BiLSTM | |
Bai et al. | Comparative analysis of thermal preference prediction performance in different conditions using ensemble learning models based on ASHRAE Comfort Database II | |
Dai et al. | An opposition-based differential evolution clustering algorithm for emotional preference and migratory behavior optimization | |
Jin et al. | Soft sensor modeling for small data scenarios based on data enhancement and selective ensemble | |
CN118011835B (en) | Blast furnace ironmaking control system based on machine learning | |
CN114067395A (en) | Late stage fusion multi-view clustering method and system based on local maximum alignment | |
CN117454765A (en) | Copper smelting furnace spray gun service life prediction method based on IPSO-BP neural network | |
Ariafar et al. | Faster & More Reliable Tuning of Neural Networks: Bayesian Optimization with Importance Sampling. | |
CN118011835A (en) | Blast furnace ironmaking control system based on machine learning | |
CN100380395C (en) | Supervised classification process of artificial immunity in remote sensing images | |
CN113570161B (en) | Method for constructing stirred tank reactant concentration prediction model based on width transfer learning | |
Ding et al. | Simplified space based neural architecture search | |
Lu et al. | Apso-based optimization algorithm of lstm neural network 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 |