CN115081687A - Knowledge-guided multi-source information fusion-based blast furnace gas prediction method - Google Patents
Knowledge-guided multi-source information fusion-based blast furnace gas prediction method Download PDFInfo
- Publication number
- CN115081687A CN115081687A CN202210606561.0A CN202210606561A CN115081687A CN 115081687 A CN115081687 A CN 115081687A CN 202210606561 A CN202210606561 A CN 202210606561A CN 115081687 A CN115081687 A CN 115081687A
- Authority
- CN
- China
- Prior art keywords
- blast furnace
- furnace gas
- data
- model
- generation amount
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000004927 fusion Effects 0.000 title claims abstract description 16
- 230000008569 process Effects 0.000 claims abstract description 21
- 230000002159 abnormal effect Effects 0.000 claims abstract description 16
- 238000010606 normalization Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 10
- 230000007246 mechanism Effects 0.000 claims abstract description 9
- 238000012360 testing method Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 5
- 230000003044 adaptive effect Effects 0.000 claims abstract description 4
- 238000003062 neural network model Methods 0.000 claims abstract description 4
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 239000007789 gas Substances 0.000 claims description 131
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 15
- 239000001301 oxygen Substances 0.000 claims description 15
- 229910052760 oxygen Inorganic materials 0.000 claims description 15
- 239000003034 coal gas Substances 0.000 claims description 13
- 238000002485 combustion reaction Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000007664 blowing Methods 0.000 claims description 7
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 6
- 239000002912 waste gas Substances 0.000 claims description 6
- 229910000831 Steel Inorganic materials 0.000 claims description 5
- 239000010959 steel Substances 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000007405 data analysis Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 3
- 238000011835 investigation Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000010561 standard procedure Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 abstract description 4
- 238000005516 engineering process Methods 0.000 description 5
- 238000007418 data mining Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B5/00—Making pig-iron in the blast furnace
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Chemical & Material Sciences (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Manufacturing & Machinery (AREA)
- Operations Research (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Materials Engineering (AREA)
- Metallurgy (AREA)
- Organic Chemistry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a knowledge-guided multi-source information fusion-based blast furnace gas prediction method, which comprises the following steps of S1: acquiring instantaneous value data of blast furnace gas generation amount and relevant influence factors of the blast furnace gas generation amount, and storing the instantaneous value data as a data file; s2: carrying out operations such as exception removal, normalization preprocessing and the like on the data file data of the S1, and constructing an experimental data set; s3: performing attention mechanism processing and adaptive learning weight operation on the S2 data set input features; s4: inputting the characteristics processed in the step S3 into a neural network model, and respectively training and testing a blast furnace gas generation model and a consumption model; s5: judging whether the prediction result is abnormal, and correcting and optimizing the model; the weight of different information source data is learned in a self-adaptive manner, so that the model is helped to pay attention to key information; improving the model prediction performance by treating different historical moments distinctively; the invention further corrects and optimizes the model by combining with the production flow knowledge, the process background and other guiding rules, and guides the model to actively mine the process information in the data.
Description
Technical Field
The invention belongs to the technical field of blast furnace gas prediction, and particularly relates to a knowledge-guided multi-source information fusion-based blast furnace gas prediction method.
Background
The problems of high energy consumption, low efficiency, heavy pollution and the like exist in the steel production, and the coal gas prediction technology is developed to effectively solve the existing problems. The technology assists workers to reasonably schedule by predicting the gas generation amount and the gas consumption amount, reduces gas diffusion and improves production efficiency. Meanwhile, the gas prediction technology can help steel enterprises to realize the optimal configuration and the effective utilization of energy, and the transition from the straightforward management and the empirical management to the refined and quantitative management is completed.
In the traditional gas prediction method, the input characteristics are single, and the model learning is limited and restricted to a certain extent. How to perform multi-source information fusion and help a model to enhance the prediction capability by utilizing diversified data is an important part of a blast furnace gas prediction technology. Because the data at different historical moments have different contribution degrees to the prediction result, how to make the model learn the weights corresponding to different historical time units is very necessary and is also a key step for realizing low error rate prediction; therefore, the invention provides a knowledge-guided multi-source information fusion-based blast furnace gas prediction method.
Disclosure of Invention
The invention aims to provide a knowledge-guided multisource information fusion-based blast furnace gas prediction method, so as to solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme: a blast furnace gas prediction method based on knowledge-guided multi-source information fusion is disclosed, and S1: acquiring instantaneous value data of blast furnace gas generation amount (consumption) and relevant influence factors of the blast furnace gas generation amount (consumption) from a real-time database, and storing the instantaneous value data as a data file;
s2: preprocessing operations such as exception removal and normalization are carried out on the data file data stored in the S1, and an experimental data set is constructed;
s3: performing attention mechanism processing and adaptive learning weight operation on input features of the S2 data set, wherein data of different input factors and different historical moments can obtain differentiated weights;
s4: inputting the characteristics processed in the step S3 into a neural network model, and respectively training and testing a blast furnace gas generation model and a blast furnace gas consumption model;
s5: and (4) judging whether the prediction result is abnormal or not by combining expert knowledge and professional background, and correcting and optimizing the model.
Preferably, the process of step S1 includes:
s11: the database collects various blast furnace gas related data in the steel system in real time, the collection frequency is one second, in order to meet the actual requirement, the step adopts down sampling, namely the average value of the original instantaneous value data in one minute is taken as the instantaneous value data of the minute, and meanwhile, the method also reduces the noise of the original data;
s12: in order to avoid the influence of irrelevant factors on the blast furnace gas generation amount (consumption), reduce the characteristic dimension and enhance the interpretability of the model, the Pearson correlation coefficient is respectively calculated for the influencing factors of the blast furnace gas generation amount (consumption) in the step, and the most key influencing factors are screened out as input characteristics; the specific calculation formula of the Pearson correlation coefficient is as follows:
the experimental data set is subjected to correlation analysis, the key influence factors of the blast furnace gas generation amount are the air supply amount and the air supply oxygen content, and the key influence factors of the blast furnace gas consumption amount are the waste gas temperature and the waste gas oxygen content.
Preferably, the process of step S2 includes:
s21: abnormal value detection: the total X of the experimental data of the blast furnace gas generation amount (consumption amount) is subjected to normal distribution, and the step adopts a Lauda criterion method to detect abnormal values; assuming that mu and sigma respectively represent the mathematical expectation and standard deviation of the experimental data set, the probability that the experimental data values are more or less than (mu +3 sigma) is very small since the experimental data values are more or less likely to occur in the range of (mu-3 sigma, mu +3 sigma), so this step detects the experimental data values more or less than (mu +3 sigma) as abnormal values, rejects them, and fills up the abnormal data using the data mean or linear interpolation;
s22: data normalization: the data normalization problem is an important problem in feature vector expression in data mining, when different features are listed together, small data on absolute numerical values are 'eaten' by large data due to the expression mode of the features, and what we need to do at this time is to normalize the feature data to ensure that each feature is treated equally by a classifier. The data is normalized by standard methods, i.e.
Wherein x is max Is the maximum value, x, in the feature dataset min Is the minimum value in the feature data set;
s23: in the step, an experimental data set is divided into a training set, a verification set and a test set according to the proportion of 70%, 15% and 15% respectively; after the influence factors are screened according to S12, each sample feature in the step comprises three information factor historical quantities of t time units before the current moment; for blast furnace gas generation amount prediction, (sample characteristics, sample labels) are (historical data of blast furnace gas generation amount, air supply volume and air supply oxygen content in t time units, and blast furnace gas generation amount after fifteen minutes); for blast furnace gas consumption prediction, (sample characteristics, sample label) is (blast furnace gas consumption per t time units, history data of exhaust gas temperature and exhaust gas oxygen content, and blast furnace gas consumption after fifteen minutes).
Preferably, the process of step S3 includes:
s31: taking the prediction of the blast furnace gas generation amount as an example, let V be (V is) the historical blast furnace gas generation amount in t time units 1 ，v 2 ，v 3 ，...v t ) T Using the formula:
H＝σ(WV)；
V′＝H⊙V；
calculating to obtain a blast furnace gas historical occurrence quantity characteristic V' based on an attention mechanism, wherein the dimension of a matrix W is txt, sigma represents a Sigmoid activation function, and a indicates a product of elements; other influencing factors in the blast furnace gas generation amount prediction and related input data of the blast furnace gas consumption amount are processed in the same way;
s32: taking the blast furnace gas generation amount prediction as an example, assuming that the historical generation amount characteristic of the blast furnace gas t time unit after the step S31 is V ', the blowing air amount characteristic is S ', and the blowing oxygen content characteristic is Y ', the integrated characteristic after splicing is as follows:
C＝||βV′，γS′，ηY′||；
the model can better balance and fuse multi-source information by adaptively learning beta, gamma and eta parameters, and the larger the parameter value is, the larger the contribution of the information to model prediction is; the same is true for blast furnace gas consumption prediction.
Preferably, the process of step S4 includes:
s41: the neural network is constructed by two fully-connected layers, and the activation functions are both ReLU; the integration characteristic C corresponding to the blast furnace gas generation amount (consumption amount) obtained in step S32 is fed into the neural network, and the blast furnace gas generation amount (consumption amount) after fifteen minutes is predicted, that is:
y i ＝Model(C i )，
wherein, Model is the network Model of the invention, C i As an integrated feature of the ith sample, y i The predicted value of the blast furnace gas for the ith sample.
S42: the neural network adopts an MSE loss function, an Adam optimizer is used for reversely propagating an optimization model, and the Adam optimizer can adjust different learning rates for different parameters, so that the model is rapidly converged; the MSE loss function equation is as follows:
wherein n is the number of training set samples, y' i The blast furnace gas signature value for the ith sample.
Preferably, the process of S5 includes:
s51: according to the technical background, the blast furnace can generate coal gas, and the attached hot blast stove can consume part of the coal gas; the blast furnace consists of three hot blast furnaces, and the coal gas consumed by all the hot blast furnaces of one blast furnace is called the coal gas consumption of the blast furnace;
s52: the gas consumption of the hot blast stove depends on the state of the hot blast stove, the hot blast stove mainly has two states of combustion and air supply, when the hot blast stove is in the combustion state, the blast furnace gas can be consumed, and when the hot blast stove is in the air supply state, the gas consumption is 0;
s53: according to actual investigation and data analysis, the change of the exhaust gas temperature can accurately indicate the state conversion time of the hot blast stove, namely:
combustion wind transfer state: the time when the temperature data reaches the maximum value and begins to fall;
air supply to combustion state: the temperature data passes through a smoother starting rise time after falling from the highest point to the lowest point.
Compared with the prior art, the invention has the beneficial effects that: the invention provides a knowledge-guided multi-source information fusion-based blast furnace gas prediction method, which combines an attention mechanism, self-adaptive learning and deep learning, utilizes the advantages of multi-source information data to realize full-automatic high-accuracy real-time blast furnace gas prediction, and the model input is multi-information, and helps the model to pay attention to key information by self-adaptively learning the weights of different information source data; the model is based on an attention mechanism, and the prediction performance of the model is improved by treating different historical moments distinctively. Meanwhile, the invention further corrects and optimizes the model by combining with the production flow knowledge, the process background and other guiding rules, and guides the model to actively mine the process information in the data.
Drawings
FIG. 1 is a schematic diagram of the present invention;
fig. 2 is a graph showing the consumption amount of hot blast furnace gas and the exhaust gas temperature of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1 to 2, the present invention provides a technical solution: a blast furnace gas prediction method based on knowledge-guided multi-source information fusion is disclosed, and S1: acquiring instantaneous value data of blast furnace gas generation amount (consumption) and relevant influence factors of the blast furnace gas generation amount (consumption) from a real-time database, and storing the instantaneous value data as a data file;
in this embodiment, preferably, the process of step S1 includes:
s11: the database collects various blast furnace gas related data in the steel system in real time, the collection frequency is one second, in order to meet the actual requirement, the step adopts down sampling, namely the average value of the original instantaneous value data in one minute is taken as the instantaneous value data of the minute, and meanwhile, the method also reduces the noise of the original data;
s12: in order to avoid the influence of irrelevant factors on the blast furnace gas generation amount (consumption), reduce the characteristic dimension and enhance the interpretability of the model, the Pearson correlation coefficient is respectively calculated for the influencing factors of the blast furnace gas generation amount (consumption) in the step, and the most key influencing factors are screened out as input characteristics; the specific calculation formula of the Pearson correlation coefficient is as follows:
the experimental data set is subjected to correlation analysis, the key influence factors of the blast furnace gas generation amount are the air supply amount and the air supply oxygen content, and the key influence factors of the blast furnace gas consumption amount are the waste gas temperature and the waste gas oxygen content.
S2: preprocessing operations such as exception removal and normalization are carried out on the data file data stored in the S1, and an experimental data set is constructed;
in this embodiment, preferably, the process of step S2 includes:
s21: abnormal value detection: the total X of the experimental data of the blast furnace gas generation amount (consumption amount) is subjected to normal distribution, and the step adopts a Lauda criterion method to detect abnormal values; assuming that mu and sigma respectively represent the mathematical expectation and standard deviation of the experimental data set, the probability that the experimental data values are more or less than (mu +3 sigma) is very small since the experimental data values are more or less likely to occur in the range of (mu-3 sigma, mu +3 sigma), so this step detects the experimental data values more or less than (mu +3 sigma) as abnormal values, rejects them, and fills up the abnormal data using the data mean or linear interpolation;
s22: data normalization: the data normalization problem is an important problem in feature vector expression in data mining, when different features are listed together, small data on absolute numerical values are eaten by big data due to the expression mode of the features, and what we need to do at this time is to perform normalization processing on the feature data so as to ensure that each feature is treated equally by a classifier. The data is normalized by standard methods, i.e.
Wherein x is max Is the maximum value, x, in the feature dataset min Is the minimum value in the feature data set;
s23: in the step, an experimental data set is divided into a training set, a verification set and a test set according to the proportion of 70%, 15% and 15% respectively; after screening according to the influence factors of S12, each sample feature in the step contains three information factor historical quantities of t time units before the current time; for blast furnace gas generation amount prediction, (sample characteristics, sample labels) are (historical data of blast furnace gas generation amount, air supply volume and air supply oxygen content in t time units, and blast furnace gas generation amount after fifteen minutes); for blast furnace gas consumption prediction, (sample characteristics, sample label) is (blast furnace gas consumption per t time units, history data of exhaust gas temperature and exhaust gas oxygen content, and blast furnace gas consumption after fifteen minutes).
S3: performing attention mechanism processing and adaptive learning weight operation on input features of the S2 data set, wherein data of different input factors and different historical moments can obtain differentiated weights;
in this embodiment, preferably, the process of step S3 includes:
s31: taking the prediction of the blast furnace gas generation amount as an example, suppose t timeThe unit of the historical generation amount of the blast furnace gas is V ═ V (V) 1 ，v 2 ，v 3 ，...v t ) T Using the formula:
H＝σ(WV)；
V′＝H⊙V；
calculating to obtain a blast furnace gas historical occurrence quantity characteristic V' based on an attention mechanism, wherein the dimension of a matrix W is txt, sigma represents a Sigmoid activation function, and a indicates a product of elements; other influencing factors in the blast furnace gas generation amount prediction and related input data of the blast furnace gas consumption amount are processed in the same way;
s32: taking the blast furnace gas generation amount prediction as an example, assuming that the historical generation amount characteristic of the blast furnace gas t time unit after the step S41 is V ', the blowing air quantity characteristic is S ', and the blowing oxygen content characteristic is Y ', the integrated characteristic after splicing is as follows:
C＝||βV′，γS′，ηY′||；
the model can better balance and fuse multi-source information by adaptively learning beta, gamma and eta parameters, and the larger the parameter value is, the larger the contribution of the information to model prediction is; the same is true for blast furnace gas consumption prediction.
S4: inputting the characteristics processed in the step S3 into a neural network model, and respectively training and testing a blast furnace gas generation model and a blast furnace gas consumption model;
in this embodiment, preferably, the process of step S4 includes:
s41: the neural network is constructed by two full connection layers, and the activation functions are both ReLU; the integrated feature C corresponding to the blast furnace gas generation amount (consumption amount) obtained in step S32 is fed into the neural network, and the blast furnace gas generation amount (consumption amount) after fifteen minutes is predicted, that is, the integrated feature C is:
y i ＝Model(C i )，
wherein, Model is the network Model of the invention, C i As an integrated feature of the ith sample, y i The blast furnace gas prediction value for the ith sample.
S42: the neural network adopts an MSE loss function, an Adam optimizer is used for reversely propagating an optimization model, and the Adam optimizer can adjust different learning rates for different parameters, so that the model is rapidly converged; the MSE loss function is formulated as follows:
wherein n is the number of training set samples, y' i The blast furnace gas signature value for the ith sample.
S5: judging whether the prediction result is abnormal or not, and correcting and optimizing the model;
in this embodiment, preferably, the process of S5 includes:
s51: according to the technical background, the blast furnace can generate coal gas, and the attached hot blast stove can consume part of the coal gas; the blast furnace consists of three hot blast furnaces, and the coal gas consumed by all the hot blast furnaces of one blast furnace is called the coal gas consumption of the blast furnace;
s52: the gas consumption of the hot blast stove depends on the state of the hot blast stove, the hot blast stove mainly has two states of combustion and air supply, when the hot blast stove is in the combustion state, the blast furnace gas can be consumed, and when the hot blast stove is in the air supply state, the gas consumption is 0;
s53: according to actual investigation and data analysis, the change of the exhaust gas temperature can accurately indicate the state conversion time of the hot blast stove, namely:
combustion transfer wind state: the moment when the temperature data reaches the maximum value and begins to fall;
air supply to combustion state: the temperature data passes through a smoother starting rise time after falling from the highest point to the lowest point.
The method is inspired by the process knowledge, the exhaust gas temperature after fifteen minutes is predicted by using a multilayer perceptron model, then the state of the hot blast stove after fifteen minutes is judged according to the predicted value of the exhaust gas temperature, and the state of the hot blast stove is used for correcting and optimizing the blast furnace gas consumption prediction model. Specifically, if the hot-blast stove is determined to be in the blowing state fifteen minutes later, the consumption amount of the blast furnace gas is 0, whereas if the hot-blast stove is determined to be in the combustion state fifteen minutes later, the consumption amount thereof is a predicted blast furnace gas amount. This process knowledge is of great significance to the performance prediction of the present invention.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (6)
1. A multi-source information fusion blast furnace gas prediction method based on knowledge guidance is characterized in that:
s1: acquiring instantaneous value data of blast furnace gas generation amount (consumption) and relevant influence factors of the blast furnace gas generation amount (consumption) from a real-time database, and storing the instantaneous value data as a data file;
s2: carrying out exception removal and normalization preprocessing operation on the data file data stored in the S1, and constructing an experimental data set;
s3: performing attention mechanism processing and adaptive learning weight operation on the S2 data set input features;
s4: inputting the characteristics processed in the step S3 into a neural network model, and respectively training and testing a blast furnace gas generation model and a blast furnace gas consumption model;
s5: and (4) judging whether the prediction result is abnormal or not by combining expert knowledge and professional background, and correcting and optimizing the model.
2. The blast furnace gas prediction method based on knowledge-guided multi-source information fusion according to claim 1, characterized in that: the process of step S1 includes:
s11: the database collects the relevant data of various blast furnace gas in the steel system in real time, the collection frequency is one second, in order to meet the actual demand, the step adopts down sampling, namely the average value of the original instantaneous value data in one minute is taken as the instantaneous value data of the minute;
s12: in order to avoid the influence of irrelevant factors on the blast furnace gas generation amount (consumption), reduce the characteristic dimension and enhance the interpretability of the model, the Pearson correlation coefficient is respectively calculated for the influencing factors of the blast furnace gas generation amount (consumption) in the step, and the most key influencing factors are screened out as input characteristics; the specific calculation formula of the Pearson correlation coefficient is as follows:
the experimental data set is subjected to correlation analysis, the key influence factors of the blast furnace gas generation amount are the air supply amount and the air supply oxygen content, and the key influence factors of the blast furnace gas consumption amount are the waste gas temperature and the waste gas oxygen content.
3. The blast furnace gas prediction method based on knowledge-guided multi-source information fusion according to claim 1, characterized in that: the process of step S2 includes:
s21: abnormal value detection: the total X of the experimental data of the blast furnace gas generation amount (consumption amount) is subjected to normal distribution, and the step adopts a Lauda criterion method to detect abnormal values; assuming that mu and sigma respectively represent the mathematical expectation and standard deviation of the experimental data set, the probability that the experimental data values are more or less than (mu +3 sigma) is very small since the experimental data values are more or less likely to occur in the range of (mu-3 sigma, mu +3 sigma), so this step detects the experimental data values more or less than (mu +3 sigma) as abnormal values, rejects them, and fills up the abnormal data using the data mean or linear interpolation;
s22: data normalization: normalizing the feature data to ensure that each feature is treated equally by the classifier; the data is normalized by standard methods, i.e.
Wherein x is max Is the maximum value, x, in the feature dataset min As a characteristic numberA minimum value in the dataset;
s23: in the step, an experimental data set is divided into a training set, a verification set and a test set according to the proportion of 70%, 15% and 15% respectively; after screening according to the influence factors of S12, each sample feature in the step contains three information factor historical quantities of t time units before the current time; for blast furnace gas generation amount prediction, (sample characteristics, sample labels) are (historical data of blast furnace gas generation amount, air supply volume and air supply oxygen content in t time units, and blast furnace gas generation amount after fifteen minutes); for blast furnace gas consumption prediction, (sample characteristics, sample label) is (blast furnace gas consumption per t time units, history data of exhaust gas temperature and exhaust gas oxygen content, and blast furnace gas consumption after fifteen minutes).
4. The blast furnace gas prediction method based on knowledge-guided multi-source information fusion according to claim 1, characterized in that: the process of step S3 includes:
s31: taking the prediction of the blast furnace gas generation amount as an example, the blast furnace gas generation amount in the past is assumed to be V ═ V (V is the amount of the blast furnace gas generated in the past per t time unit) 1 ，v 2 ，v 3 ，...v t ) T Using the formula:
H＝σ(WV)；
V′＝H⊙V；
calculating to obtain a blast furnace gas history occurrence quantity characteristic V' based on an attention mechanism, wherein the dimension of a matrix W is txt, sigma represents a Sigmoid activation function, and the corresponding product indicates an element product; other influencing factors in the blast furnace gas generation amount prediction and related input data of the blast furnace gas consumption amount are processed in the same way;
s32: taking the blast furnace gas generation amount prediction as an example, assuming that the historical generation amount characteristic of the blast furnace gas t time unit after the step S31 is V ', the blowing air amount characteristic is S ', and the blowing oxygen content characteristic is Y ', the integrated characteristic after splicing is as follows:
C＝||βV′，γS′，ηY′||；
the model can better balance and fuse multi-source information by adaptively learning beta, gamma and eta parameters, and the larger the parameter value is, the larger the contribution of the information to model prediction is; the same is true for blast furnace gas consumption prediction.
5. The blast furnace gas prediction method based on knowledge-guided multi-source information fusion according to claim 1, characterized in that: the process of step S4 includes:
s41: the neural network is constructed by two full connection layers, and the activation functions are both ReLU; the integration characteristic C corresponding to the blast furnace gas generation amount (consumption amount) obtained in step S32 is fed into the neural network, and the blast furnace gas generation amount (consumption amount) after fifteen minutes is predicted, that is:
y i ＝Model(C i )，
wherein, Model is the network Model of the invention, C i As an integrated feature of the ith sample, y i The predicted value of the blast furnace gas for the ith sample.
S42: the neural network adopts an MSE loss function, an Adam optimizer is used for reversely propagating an optimization model, and the Adam optimizer can adjust different learning rates for different parameters, so that the model is rapidly converged; the MSE loss function equation is as follows:
wherein n is the number of training set samples, y' i The blast furnace gas signature value for the ith sample.
6. The blast furnace gas prediction method based on knowledge-guided multi-source information fusion according to claim 1, characterized in that: the process of S5 includes:
s51: according to the technical background, the blast furnace can generate coal gas, and the attached hot blast stove can consume part of the coal gas; the blast furnace consists of three hot blast furnaces, and the coal gas consumed by all the hot blast furnaces of one blast furnace is called the coal gas consumption of the blast furnace;
s52: the gas consumption of the hot blast stove depends on the state of the hot blast stove, the hot blast stove mainly has two states of combustion and air supply, when the hot blast stove is in the combustion state, the blast furnace gas can be consumed, and when the hot blast stove is in the air supply state, the gas consumption is 0;
s53: according to actual investigation and data analysis, the change of the exhaust gas temperature can accurately predict the state conversion time of the hot blast stove, namely:
combustion transfer wind state: the moment when the temperature data reaches the maximum value and begins to fall;
air supply to combustion state: the temperature data passes through a smoother starting rise time after falling from the highest point to the lowest point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210606561.0A CN115081687A (en) | 2022-05-31 | 2022-05-31 | Knowledge-guided multi-source information fusion-based blast furnace gas prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210606561.0A CN115081687A (en) | 2022-05-31 | 2022-05-31 | Knowledge-guided multi-source information fusion-based blast furnace gas prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115081687A true CN115081687A (en) | 2022-09-20 |
Family
ID=83248540
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210606561.0A Pending CN115081687A (en) | 2022-05-31 | 2022-05-31 | Knowledge-guided multi-source information fusion-based blast furnace gas prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115081687A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115657448A (en) * | 2022-12-09 | 2023-01-31 | 电信科学技术第五研究所有限公司 | Frequency offset analysis method for clock discipline |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113781110A (en) * | 2021-09-07 | 2021-12-10 | 中国船舶重工集团公司第七0九研究所 | User behavior prediction method and system based on multi-factor weighted BI-LSTM learning |
-
2022
- 2022-05-31 CN CN202210606561.0A patent/CN115081687A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113781110A (en) * | 2021-09-07 | 2021-12-10 | 中国船舶重工集团公司第七0九研究所 | User behavior prediction method and system based on multi-factor weighted BI-LSTM learning |
Non-Patent Citations (1)
| Title |
|---|
| 李志刚等: "CNN-GRU模型预测高炉煤气产生量", 《机械设计与制造》, no. 4, 8 April 2022 (2022-04-08), pages 69 - 72 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115657448A (en) * | 2022-12-09 | 2023-01-31 | 电信科学技术第五研究所有限公司 | Frequency offset analysis method for clock discipline |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111353656B (en) | Steel enterprise oxygen load prediction method based on production plan | |
| Zhang et al. | Wind speed forecasting using a two-stage forecasting system with an error correcting and nonlinear ensemble strategy | |
| CN109146121A (en) | The power predicating method stopped in the case of limited production based on PSO-BP model | |
| CN109685240A (en) | A kind of prediction of 95598 traffic work orders and unusual fluctuation method for early warning based on LSTM deep learning | |
| CN108009667A (en) | A kind of energy demand total amount and structure prediction system | |
| CN110598929B (en) | Wind power nonparametric probability interval ultrashort term prediction method | |
| CN107918368B (en) | The dynamic prediction method and equipment of iron and steel enterprise's coal gas yield and consumption | |
| CN114757104B (en) | Method for constructing hydraulic real-time regulation and control model of series gate group water transfer project | |
| Meng et al. | Forecasting energy consumption based on SVR and Markov model: A case study of China | |
| CN112016616A (en) | High-frequency financial time sequence multi-class prediction method | |
| CN110837939A (en) | Power grid multi-target project screening method and system | |
| CN114678080A (en) | Converter end point phosphorus content prediction model, construction method and phosphorus content prediction method | |
| CN115081687A (en) | Knowledge-guided multi-source information fusion-based blast furnace gas prediction method | |
| CN111680452A (en) | Power grid engineering accurate investment decision simulation method based on full-factor data mining | |
| CN115310685A (en) | Hydroelectric generating set full life cycle cost calculation method based on parameter correction | |
| CN116644562B (en) | New energy power station operation and maintenance cost evaluation system | |
| CN109829115B (en) | Search engine keyword optimization method | |
| Zhu et al. | Application research of the xgboost-svm combination model in quantitative investment strategy | |
| Zhao et al. | Forecasting short-term oil price with a generalised pattern matching model based on empirical genetic algorithm | |
| CN114004414B (en) | Method for predicting energy load of enterprise coupling Copula and LSTM | |
| Zhang | Research of boosting algorithm machine learning in logistics enterprise financial risk prediction | |
| CN113723660A (en) | Specific behavior type prediction method and system based on DNN-LSTM fusion model | |
| Jin et al. | A causal model-based scheduling approach for coke oven gas system in steel industry | |
| Liu et al. | Electricity Consumption Forecasting in Jiangxi Province Based on GRA-PSO-BP | |
| Zhao et al. | Multi-Working Conditions Converter Gas Holders Capacities Prediction Based on Improved Multi-LSTM Transfer Ensemble |
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 |