CN116343932A - Method, device and system for predicting micro-reaction activity of catalyst of catalytic cracking device - Google Patents
Method, device and system for predicting micro-reaction activity of catalyst of catalytic cracking device Download PDFInfo
- Publication number
- CN116343932A CN116343932A CN202111565086.9A CN202111565086A CN116343932A CN 116343932 A CN116343932 A CN 116343932A CN 202111565086 A CN202111565086 A CN 202111565086A CN 116343932 A CN116343932 A CN 116343932A
- Authority
- CN
- China
- Prior art keywords
- data
- micro
- catalyst
- catalytic cracking
- abnormal
- 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
- 239000003054 catalyst Substances 0.000 title claims abstract description 177
- 230000000694 effects Effects 0.000 title claims abstract description 115
- 238000004523 catalytic cracking Methods 0.000 title claims abstract description 89
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 89
- 238000000034 method Methods 0.000 title claims abstract description 86
- 230000002159 abnormal effect Effects 0.000 claims abstract description 107
- 238000013528 artificial neural network Methods 0.000 claims abstract description 72
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 51
- 230000000306 recurrent effect Effects 0.000 claims abstract description 28
- 238000007781 pre-processing Methods 0.000 claims abstract description 15
- 210000002569 neuron Anatomy 0.000 claims description 90
- 230000011218 segmentation Effects 0.000 claims description 30
- 238000012216 screening Methods 0.000 claims description 26
- 238000012545 processing Methods 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 20
- 230000003197 catalytic effect Effects 0.000 claims description 19
- 238000012847 principal component analysis method Methods 0.000 claims description 18
- 230000009467 reduction Effects 0.000 claims description 18
- 229910052760 oxygen Inorganic materials 0.000 claims description 14
- 238000012360 testing method Methods 0.000 claims description 12
- 229910052717 sulfur Inorganic materials 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 9
- 210000004205 output neuron Anatomy 0.000 claims description 9
- 230000005856 abnormality Effects 0.000 claims description 8
- 238000005315 distribution function Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 6
- 239000002245 particle Substances 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 5
- 238000011946 reduction process Methods 0.000 claims 1
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 239000000126 substance Substances 0.000 abstract description 2
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 description 26
- 238000011069 regeneration method Methods 0.000 description 23
- 239000002994 raw material Substances 0.000 description 20
- 230000008929 regeneration Effects 0.000 description 20
- 230000008569 process Effects 0.000 description 19
- 229910052757 nitrogen Inorganic materials 0.000 description 13
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 8
- 239000001301 oxygen Substances 0.000 description 8
- 239000002002 slurry Substances 0.000 description 8
- 238000009826 distribution Methods 0.000 description 7
- 238000012549 training Methods 0.000 description 7
- 238000004140 cleaning Methods 0.000 description 6
- 239000004215 Carbon black (E152) Substances 0.000 description 5
- NINIDFKCEFEMDL-UHFFFAOYSA-N Sulfur Chemical compound [S] NINIDFKCEFEMDL-UHFFFAOYSA-N 0.000 description 5
- 238000006477 desulfuration reaction Methods 0.000 description 5
- 230000023556 desulfurization Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 229930195733 hydrocarbon Natural products 0.000 description 5
- 150000002430 hydrocarbons Chemical class 0.000 description 5
- 239000007788 liquid Substances 0.000 description 5
- 239000002184 metal Substances 0.000 description 5
- 229910052751 metal Inorganic materials 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 239000011593 sulfur Substances 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 239000003795 chemical substances by application Substances 0.000 description 3
- 230000009849 deactivation Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 239000002283 diesel fuel Substances 0.000 description 3
- 230000008030 elimination Effects 0.000 description 3
- 238000003379 elimination reaction Methods 0.000 description 3
- 239000003546 flue gas Substances 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000009257 reactivity Effects 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000009533 lab test Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000002547 anomalous effect Effects 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/10—Analysis or design of chemical reactions, syntheses or processes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Chemical & Material Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Crystallography & Structural Chemistry (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Analytical Chemistry (AREA)
- Production Of Liquid Hydrocarbon Mixture For Refining Petroleum (AREA)
Abstract
The invention provides a method, a device and a system for predicting micro-reaction activity of a catalytic cracking device catalyst, which belong to the field of chemical industry and artificial intelligence. The method comprises the following steps: preprocessing variable data to obtain a processed sample set; constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set; optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model; and predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst. Aiming at the problems of low data quantity and low frequency of the related micro-reactivity of the existing catalyst, the abnormal data and the compensation missing data are identified by utilizing a local abnormal factor algorithm based on characteristic attributes, and a catalyst micro-reactivity prediction model is established based on a recurrent RBF neural network, so that the prediction of the micro-reactivity of the catalyst is realized.
Description
Technical Field
The invention belongs to the field of chemical industry and artificial intelligence, and particularly relates to a catalytic cracking device catalyst micro-reaction activity prediction method, a catalytic cracking device catalyst micro-reaction activity prediction device and a catalytic cracking device catalyst micro-reaction activity prediction system.
Background
The catalyst in the catalytic cracking device is the catalyst with the largest usage amount in the petroleum processing process, and the number of the FCC catalysts scrapped in China is huge. The micro-reaction activity of the catalyst is not provided with an on-line detection instrument, so that the micro-reaction activity value cannot be obtained in real time, the load, the raw materials and the production scheme of the catalytic cracking device are frequently adjusted, the detection frequency of the reaction activity of the catalyst is low, the analysis is delayed, the fresh catalyst is replenished and the discharge amount of the spent catalyst is quantitatively controlled by experience in the production process, the dynamic adjustment cannot be realized according to the real-time micro-reaction activity value of the catalyst, and the catalyst unit consumption and the activity of the balancing agent cannot be kept relatively stable. Therefore, research on a real-time prediction method of the micro-reaction activity of the catalyst of the catalytic cracking device is needed to be carried out, and the intelligent level of the catalytic cracking device is improved.
In order to obtain the reaction activity of the catalytic cracking unit catalyst, cui Yufeng and the like analyze the mathematical relationship between the activity of the cracking catalyst and the micro-reaction activity, deduce a dynamic equation of the hydrothermal deactivation of the catalyst, establish a mathematical model of the micro-reaction activity of the industrial device balancing agent, realize the simulation and prediction of the micro-reaction activity of the catalytic cracking unit balancing catalyst, and guide and determine the parameters of the device through the simulation calculation of the micro-reaction activity of the balancing agent. Deng Mingbo and the like establish the mathematical relationship between the catalytic cracking catalyst activity and the micro-inverse activity by using a mathematical derivation method, determine the hydrothermal deactivation kinetic parameters through the hydrothermal deactivation experimental data of the catalyst, and realize the simulation of the equilibrium activity and the prediction of the equilibrium activity of the catalyst. Although the above research methods can all obtain the reactivity of the catalyst, the method relies on mathematical mechanism model operation to obtain the reactivity of the catalyst, and the calculation process is complex and takes a long time.
Aiming at the problem of difficult prediction of the catalyst reaction activity, li Fangcheng and the like, a catalyst prediction method of a denitration device is provided, a multilayer SCR catalyst life prediction model is established based on an SVM algorithm, and the life of the catalyst is predicted through the change of the relative activity of the catalyst. Tang Shijie and the like, and by comparing established curve fitting, grey prediction, BP neural network and grey neural network prediction models, experimental results show that the modeling accuracy of the BP neural network exceeds that of other methods. Li Debo and the like establish a mathematical model for optimizing the SCR catalyst replacement period and strategy of the coal-fired power plant, research the catalyst replacement strategy and evaluate the benefits of the catalyst. Liu Xin provides a method for predicting the activity of a catalyst in an oil refining process based on a BP network equally, and adjusts the dosage of the catalyst through the change of the activity of the catalyst and the change trend of related process parameters, thereby improving the production efficiency and ensuring the realization of stable operation in an optimal state. The research method realizes the prediction of the catalyst reactivity by a data driving method such as a neural network and the like, and can acquire the micro-reactivity of the catalyst in real time based on the data information such as raw materials, products and the like. Compared with the traditional mathematical mechanism model calculation, the data driving calculation method improves the calculation precision and shortens the time. However, the above data driving method has a large calculation error, and there is room for further improvement. In addition, the micro-reactivity of the catalyst has large difference in data acquisition frequency of related variables of a reaction regeneration system, and the data is difficult to clean, so that the establishment of a prediction model is often more difficult.
At present, research on a prediction method of the micro-reactivity of the catalyst of the catalytic cracking device is not mature, the problems of large data acquisition frequency difference, abnormal data and missing data of related variables further increase modeling difficulty, and process variables related to the micro-reactivity of the catalyst of the catalytic cracking device cannot be completely mastered, so that how to better process a data set to obtain accurate related variables and how to improve prediction accuracy are difficulties still needed to be solved at present.
Disclosure of Invention
The embodiment of the invention aims to provide a method, a device and a system for predicting micro-reactivity of a catalytic cracking device catalyst, so as to at least solve the problem of improving prediction precision.
In order to achieve the above object, a first aspect of the present invention provides an intelligent prediction method for micro-reactivity of a catalytic cracking unit catalyst, the method comprising:
preprocessing variable data to obtain a processed sample set;
constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
optimizing the catalytic initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
And predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
Optionally, the preprocessing the variable data to obtain a processed sample set includes:
acquiring all variable data;
determining upper and lower limit interval values of variable data;
segmenting the variable data by adopting a window segmentation method;
identifying an abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor algorithm based on the characteristic attribute, and removing the abnormal data;
compensating missing data in variable data after abnormal data are removed;
normalizing the compensated variable data to obtain normalized data;
performing dimension reduction processing on the normalized data to obtain an input variable data set;
and selecting a preset number of data sets from the input variable data sets as a sample set.
Optionally, the determining the upper and lower limit interval values of each variable data includes:
analyzing the running state of the variable data;
acquiring upper and lower limit interval values of variable data according to the running state;
the operating state includes: low load operating conditions, medium load operating conditions and full load operating conditions.
Optionally, segmenting the variable data by using a window segmentation method includes:
setting the data set of the variable data to be D= { X 1 ,X 2 ,X 3 ,…,X j The j-th vector is X j ={x j1 ,x j2 ,…,x jn N is the total amount of data;
dividing the variable data by using a clustering algorithm, and determining a segmentation threshold value;
determining a window size s according to the segmentation threshold value, the data quantity of the variable data and the running state of the variable data;
the window is slid from the first data of each vector to obtain the data segment S.
Optionally, identifying the abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor algorithm based on the characteristic attribute, and removing the abnormal data comprises the following steps:
the kth distance between the data segment S and the data segment O is calculated, and the formula is as follows:
wherein I is the number of attributes, f (S i ) Is the i-th dimension attribute value of the data segment S, f (O i ) Is the i-th dimensional attribute value of the data segment O, the attribute value comprising: entropy, mean, maximum and peak intervals;
wherein, the Entropy value Entropy is expressed as:
where m= {1,2, …, M }, M is the number of subspaces, p (u) m ) Is the (u) m The probability distribution function of each subspace has the expression:
k(u m ) Is the subspace u where m The number of data contained;
calculating local anomaly factors of the data segment S, wherein the local anomaly factor evaluation function of the data segment S is as follows:
The local reachable densities of the data segments S are:
therein, lrd k (S) is the local reachable density of the data segment S, lrd k (O) is the local reachable density of data segment O, L k (S) is the set of all data segments in the data set D whose distance of the data segment S does not exceed its kth distance, LOF k (S) is a local anomaly factor value for data segment S; sigma reach-distance (S, O) is all neighborhood data L within data segment S k The sum of the reachable distances of (S);
according to local abnormality factor LOF k Value, classifying each data segment;
identifying abnormal data in the data segments of different grades by utilizing a local abnormal factor algorithm;
and screening all the identified abnormal data, and eliminating the abnormal data.
Optionally, compensating missing data in variable data after abnormal data is removed includes:
screening related variables of the missing data;
screening the first E variables in the arrangement of the contribution rate from large to small from the related variables according to a principal component analysis method to serve as input variables of an RBF neural network missing compensation model;
establishing an RBF neural network loss compensation model, wherein an hidden layer structure of the RBF neural network loss compensation model is determined to be E+2 nodes according to input variables and an experimental construction method; the output layer is the compensation output of the missing data;
Performing optimization calculation on the RBF neural network loss compensation model by using a particle swarm algorithm to obtain an optimized RBF neural network loss compensation model;
and carrying out compensation variable soft measurement by adopting the optimized RBF neural network missing compensation model to obtain a compensation value of missing data.
Optionally, the performing the dimension reduction processing on the normalized data to obtain an input variable data set includes:
performing dimension reduction processing on the normalized data by adopting a partial least square method;
and analyzing the variables with the correlation coefficients and the contribution rates larger than the threshold values by using a principal component analysis method to form an input variable data set.
Optionally, constructing an initial prediction model of micro-reactivity of the catalytic cracking catalyst according to the sample set, including:
constructing a catalytic cracking catalyst micro-reaction activity basic model based on a recurrent RBF neural network, wherein the catalytic cracking catalyst micro-reaction activity basic model comprises the following components:
input layer: the layer contains n input variables u i (t),i=1,2,…,n;
Hidden layer: the hidden layer contains J neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is:
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)];
wherein y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
Optionally, the method further comprises:
after a target prediction model of micro-reactivity of the catalytic cracking catalyst is obtained, evaluating the performance of the target prediction model according to root mean square error and precision, wherein the RMSE expression is as follows:
the calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
The second aspect of the invention provides an intelligent prediction device for micro-reaction activity of a catalytic cracking device catalyst, which comprises: a controller for:
preprocessing variable data to obtain a processed sample set;
constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
Optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
and predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
Optionally, the preprocessing the variable data to obtain a processed sample set includes:
acquiring all variable data;
determining upper and lower limit interval values of variable data;
segmenting the variable data by adopting a window segmentation method;
identifying an abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor algorithm based on the characteristic attribute, and removing the abnormal data;
compensating missing data in variable data after abnormal data are removed;
normalizing the compensated variable data to obtain normalized data;
performing dimension reduction processing on the normalized data to obtain an input variable data set;
and selecting a preset number of data sets from the input variable data sets as a sample set.
Optionally, the determining the upper and lower limit interval values of each variable data includes:
analyzing the running state of the variable data;
acquiring upper and lower limit interval values of variable data according to the running state;
The operating state includes: low load operating conditions, medium load operating conditions and full load operating conditions.
Optionally, segmenting the variable data by using a window segmentation method includes:
setting the data set of the variable data to be D= { X 1 ,X 2 ,X 3 ,…,X j The j-th vector is X j ={x j1 ,x j2 ,…,x jn N is the total amount of data;
dividing the variable data by using a clustering algorithm, and determining a segmentation threshold value;
determining a window size s according to the segmentation threshold value, the data quantity of the variable data and the running state of the variable data;
the window is slid from the first data of each vector to obtain the data segment S.
Optionally, identifying the abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor algorithm based on the characteristic attribute, and removing the abnormal data comprises the following steps:
the kth distance between the data segment S and the data segment O is calculated, and the formula is as follows:
wherein I is the number of attributes, f (S i ) Is the i-th dimension attribute value of the data segment S, f (O i ) Is the i-th dimensional attribute value of the data segment O, the attribute value comprising: entropy, mean, maximum and peak intervals;
wherein, the Entropy value Entropy is expressed as:
where m= {1,2, …, M }, M is the number of subspaces, p (u) m ) Is the (u) m The probability distribution function of each subspace has the expression:
k(u m ) Is the subspace u where m The number of data contained;
calculating local anomaly factors of the data segment S, wherein the local anomaly factor evaluation function of the data segment S is as follows:
the local reachable densities of the data segments S are:
therein, lrd k (S) is the local reachable density of the data segment S, lrd k (O) is the local reachable density of data segment O, L k (S) is the set of all data segments in the data set D whose distance of the data segment S does not exceed its kth distance, LOF k (S) is a local anomaly factor value for data segment S; sigma reach-distance (S, O) is all neighborhood data L within data segment S k The sum of the reachable distances of (S);
according to local abnormality factor LOF k Value, classifying each data segment;
identifying abnormal data in the data segments of different grades by utilizing a local abnormal factor algorithm;
and screening all the identified abnormal data, and eliminating the abnormal data.
Optionally, compensating missing data in variable data after abnormal data is removed includes:
screening related variables of the missing data;
screening the first E variables in the arrangement of the contribution rate from large to small from the related variables according to a principal component analysis method to serve as input variables of an RBF neural network missing compensation model;
establishing an RBF neural network loss compensation model, wherein an hidden layer structure of the RBF neural network loss compensation model is determined to be E+2 nodes according to input variables and an experimental construction method; the output layer is the compensation output of the missing data;
Performing optimization calculation on the RBF neural network loss compensation model by using a particle swarm algorithm to obtain an optimized RBF neural network loss compensation model;
and carrying out compensation variable soft measurement by adopting the optimized RBF neural network missing compensation model to obtain a compensation value of missing data.
Optionally, the performing the dimension reduction processing on the normalized data to obtain an input variable data set includes:
performing dimension reduction processing on the normalized data by adopting a partial least square method;
and analyzing the variables with the correlation coefficients and the contribution rates larger than the threshold values by using a principal component analysis method to form an input variable data set.
Optionally, constructing an initial prediction model of micro-reactivity of the catalytic cracking catalyst according to the sample set, including:
constructing a catalytic cracking catalyst micro-reaction activity basic model based on a recurrent RBF neural network, wherein the catalytic cracking catalyst micro-reaction activity basic model comprises the following components:
input layer: the layer contains n input variables u i (t),i=1,2,…,n;
Hidden layer: the hidden layer contains J neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is:
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)];
wherein y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
Optionally, the controller is further configured to:
after a target prediction model of micro-reactivity of the catalytic cracking catalyst is obtained, evaluating the performance of the target prediction model according to root mean square error and precision, wherein the RMSE expression is as follows:
the calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
In a third aspect, the present invention provides an intelligent prediction system for micro-reactivity of a catalytic cracking unit catalyst, the system comprising:
the data processing module is used for preprocessing variable data to obtain a processed sample set;
The initial prediction model construction module is used for constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
the initial prediction model optimization module is used for optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
and the data prediction module is used for predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
In another aspect, the present invention provides a machine-readable storage medium having stored thereon instructions for causing a machine to perform the above-described intelligent prediction method of catalytic cracker catalyst microreactor activity.
Through the technical scheme, the intelligent prediction method for the micro-reaction activity of the catalyst of the catalytic cracking device is provided, and the defects that the micro-reaction activity of the catalyst cannot be predicted in real time, the micro-reaction activity of the catalyst and the related variable data acquisition frequency of a reaction regeneration system are large in difference, and the data are difficult to clean are overcome.
According to the method, the data segments are divided by utilizing a window segmentation method, abnormal data are identified by a local abnormal factor algorithm based on characteristic attributes, the correlation between complete data segments of different dimensional variables and missing data segments is established, missing data are compensated, a catalyst micro-reaction activity prediction model is established based on a recursive RBF neural network, and the prediction of the catalyst micro-reaction activity is realized.
Additional features and advantages of embodiments of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings are included to provide a further understanding of embodiments of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain, without limitation, the embodiments of the invention. In the drawings:
FIG. 1 is a flow chart of an intelligent prediction method for micro-reactivity of a catalytic cracking unit catalyst according to an embodiment of the present invention;
FIG. 2 is a block diagram of an intelligent prediction system for micro-reactivity of a catalytic cracking unit catalyst according to an embodiment of the present invention.
Detailed Description
The following describes specific embodiments of the present invention in detail with reference to the drawings. It should be understood that the detailed description and specific examples, while indicating and illustrating the invention, are not intended to limit the invention.
FIG. 1 is a flow chart of an intelligent prediction method for micro-reactivity of a catalytic cracking unit catalyst according to an embodiment of the present invention. As shown in fig. 1, the method includes:
step one: preprocessing variable data to obtain a processed sample set, wherein the method specifically comprises the following steps:
1) All variable data are acquired. In the application, the variable data are analyzed and obtained through an online instrument or laboratory test to obtain reactor data, regenerator data, desulfurization and denitrification facility data, product distribution data, treatment facility raw material data and the like during the operation of the catalytic cracking device, and specifically: regeneration temperature, regeneration pressure, catalyst inventory, gasoline yield, liquid hydrocarbon yield, raw material sulfur, nitrogen, metal content, fresh feed, reaction pressure, reaction temperature, feed preheating temperature, raw material nitrogen content of diesel yield, regenerator oxygen content, regenerator dense phase inventory, regeneration main air volume, riser slurry feed, riser upper temperature, outlet flue gas temperature, total feed volume, regenerator bottom dense phase temperature, regenerator dilute phase section pressure and the like.
2) Determining upper and lower limit interval values of variable data, and analyzing the running state of the variable data; acquiring upper and lower limit interval values of variable data according to the running state; the operating state includes: low load operating conditions (load less than 80%), medium load operating conditions (load 80% to 95%) and full load operating conditions (load greater than 95%).
3) The variable data is segmented by adopting a window segmentation method, and in the embodiment, the segmentation process comprises the following steps:
Setting the data set of the variable data to be D= { X 1 ,X 2 ,X 3 ,…,X j The j-th vector is X j ={x j1 ,x j2 ,…,x jn N is the total amount of data;
dividing the variable data by using a clustering algorithm, and determining a segmentation threshold value;
determining a window size s according to the segmentation threshold value, the data quantity of the variable data and the running state of the variable data;
the window is slid from the first data of each vector to obtain the data segment S.
4) And identifying the abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor (LOF) algorithm based on the characteristic attribute, and removing the abnormal data. In this embodiment, the process of eliminating abnormal data includes the steps of:
the kth distance between the data segment S and the data segment O is calculated, and the formula is as follows:
wherein I is the number of attributes, f (S i ) Is thatThe i-th dimension attribute value of the data segment S, f (O i ) Is the i-th dimensional attribute value of the data segment O, the attribute value comprising: entropy, mean, maximum and peak intervals; since the entropy value can more accurately reflect the distribution state of the data, the entropy value is used as one of key indexes of the data attribute.
Wherein, the Entropy value Entropy is expressed as:
where m= {1,2, …, M }, M is the number of subspaces, p (u) m ) Is the (u) m The probability distribution function of each subspace has the expression:
k(u m ) Is the subspace u where m The number of data contained;
calculating local anomaly factors of the data segment S, wherein the local anomaly factor evaluation function of the data segment S is as follows:
the local reachable densities of the data segments S are:
therein, lrd k (S) is the local reachable density of the data segment S, lrd k (O) is the local reachable density of data segment O, L k (S) is the set of all data segments in the data set D whose distance of the data segment S does not exceed its kth distance, LOF k (S) is a local anomaly factor value for data segment S; sigma reach-distance (S, O) is all neighborhood data L within data segment S k The sum of the reachable distances of (S); local anomaly factor LOF of comparison data k Value, when LOF k The larger the data, the greater the likelihood of anomalies in the data.
According to local abnormality factor LOF k Value, classifying each data segment;
identifying abnormal data in the data segments of different grades by utilizing a local abnormal factor algorithm;
and screening all the identified abnormal data, and eliminating the abnormal data.
5) And compensating missing data in variable data after abnormal data are removed. In this embodiment, compensating for missing data includes the steps of:
screening relevant variables of the missing data, and in the embodiment, screening relevant variables of the missing data by analyzing information of the missing data;
Screening the first E variables in the arrangement from large to small contribution rate from the related variables according to a principal component analysis method to serve as input variables of a Radial Basis Function (RBF) neural network missing compensation model;
establishing an RBF neural network loss compensation model, wherein an hidden layer structure of the RBF neural network loss compensation model is determined to be E+2 nodes according to input variables and an experimental construction method; the output layer is the compensation output of the missing data;
performing optimization calculation on the RBF neural network loss compensation model by using a particle swarm algorithm to obtain an optimized RBF neural network loss compensation model;
and carrying out compensation variable soft measurement by adopting the optimized RBF neural network missing compensation model to obtain a compensation value of missing data. The compensation process can solve the problem of data missing in the data segment caused by less acquisition frequency and data rejection.
6) And carrying out normalization processing on the compensated variable data to obtain normalized data. In the embodiment, the dimension reduction processing is performed on the normalized data by adopting a Partial Least Squares (PLS) method, so that the influence of dimension and magnitude difference in the model training process is eliminated; and analyzing the variables with the correlation coefficients and the contribution rates larger than the threshold values by using a principal component analysis method to form an input variable data set.
7) And selecting a preset number of data sets from the input variable data sets as a sample set. The problem that the difference of the threshold range between the data of different interval sections in the variable data set is large is solved through the processing, and meanwhile, the problem that data is missing or the difference of the data dimension between different variables is large in the data acquisition process is solved.
Step two: constructing an initial prediction model of micro-reactivity of the catalytic cracking catalyst according to the sample set, wherein the initial prediction model comprises the following steps:
constructing a catalytic cracking catalyst micro-reaction activity basic model based on a recurrent RBF neural network, wherein the catalytic cracking catalyst micro-reaction activity basic model comprises the following components:
input layer: the layer contains n input variables u i (t),i=1,2,…,n;
Hidden layer: the hidden layer contains J neurons, and the output expression of each hidden layer neuron is:
wherein the method comprises the steps of,c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is:
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)];
wherein y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
Step three: optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
step four: and predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
After a target prediction model of micro-reactivity of the catalytic cracking catalyst is obtained, evaluating the performance of the target prediction model according to root mean square error and precision, wherein the RMSE expression is as follows:
the calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
Example 1
(1) Acquiring data and sorting and cleaning
In this embodiment, reactor data, regenerator data, desulfurization and denitrification facility data, product distribution data, treatment facility raw material data, and the like during the operation of the catalytic cracker are obtained by on-line instrument or laboratory analysis, wherein the parameters include the following 50 items, each item is 500, and the method mainly comprises: catalyst activity value, regeneration temperature, regeneration pressure, catalyst inventory, gasoline yield, liquid hydrocarbon yield, raw material sulfur, nitrogen, metal content, fresh feed amount, reaction pressure, reaction temperature, feed preheating temperature, raw material nitrogen content of diesel oil yield, regenerator oxygen content, regenerator dense phase inventory, regeneration main air quantity, riser slurry feed amount, riser upper temperature, outlet flue gas temperature, total feed amount, regenerator bottom dense phase temperature, regenerator dilute phase section pressure and the like.
And then analyzing based on the data statistics result in the data acquisition time period, wherein in the embodiment, the variable data acquisition time period is in a full-load state, and the upper and lower limit interval values of each variable data are confirmed.
Segmenting the variable data by adopting a window segmentation method, and setting the data set of the catalytic cracking anti-regeneration system variable as D= { X 1 ,X 2 ,X 3 ,…,X 50 The j-th vector is X j ={x j1 ,x j2 ,…,x j500 The window size is set to S, and the window slides from the first data of each vector to obtain a data segment S.
And identifying the abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor (LOF) algorithm based on the characteristic attribute, and removing the abnormal data.
And aiming at missing data in a data segment caused by less data acquisition frequency and data elimination, adopting an RBF neural network to compensate the continuous missing data. Firstly, screening relevant variables of the missing data as input, screening the first 7 variables with larger contribution rate according to a principal component analysis method, and outputting the compensation output of the missing data by an output layer. The data set is finally 480 pieces obtained through the data cleaning method.
The data is then subjected to a normalization process,
and eliminating the influence of the dimension and the order of magnitude difference on the model training process aiming at the threshold range among different interval data in the abnormal data set. The normalized data is subjected to dimension reduction processing by adopting a Partial Least Squares (PLS), and the acquired data is analyzed by utilizing a principal component analysis method, so that 7 input variables which are based on the correlation coefficient and the contribution rate of the variables are obtained: regenerator oxygen content, riser slurry feed, riser upper temperature, raw material nitrogen content, total feed, catalyst inventory, and main air volume for regeneration. The output variable to be collected is the micro-reaction activity value of the catalyst. The 480 groups of data are divided into two parts: of these, 250 groups served as training samples and 230 groups served as test samples.
(2) Constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set
A recurrent RBF neural network is utilized to design a catalytic cracking catalyst micro-reaction activity prediction model,
input layer: the layer contains 7 input variables u i (t),i=1,2,…,7;
Hidden layer: the hidden layer contains 4 neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)]
y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
(3) And optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model. And evaluating the performance of the target prediction model according to the root mean square error and the accuracy, wherein the RMSE expression is as follows:
The calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
In this example, the root mean square error and the accuracy of the prediction model are evaluated as shown in table 1, and it can be seen from table 1 that the error and the accuracy of the catalyst micro-activity are within a reasonable range.
Table 1 prediction results for different algorithms
Example two
(1) Acquiring data and sorting and cleaning
In this embodiment, reactor data, regenerator data, desulfurization and denitrification facility data, product distribution data, treatment facility raw material data, and the like during the operation of the catalytic cracker are obtained by on-line instrument or laboratory analysis, wherein the following 40 parameters are included, each of which is 400, and mainly include: catalyst activity value, regeneration temperature, regeneration pressure, catalyst inventory, gasoline yield, liquid hydrocarbon yield, raw material sulfur, nitrogen, metal content, fresh feed amount, reaction pressure, reaction temperature, feed preheating temperature, raw material nitrogen content for diesel oil yield, regenerator oxygen content, regenerator dense phase inventory, regeneration main air volume, riser slurry feed amount, riser upper temperature and the like.
And then analyzing based on the data statistics result in the data acquisition time period, wherein in the embodiment, the variable data acquisition time period is in a full-load state, and the upper and lower limit interval values of each variable data are confirmed.
Segmenting the variable data by adopting a window segmentation method, and setting the data set of the catalytic cracking anti-regeneration system variable as D= { X 1 ,X 2 ,X 3 ,…,X 40 The j-th vector is X j ={x j1 ,x j2 ,…,x j400 The window size is set to S, and the window slides from the first data of each vector to obtain a data segment S.
And identifying the abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor (LOF) algorithm based on the characteristic attribute, and removing the abnormal data.
And aiming at missing data in a data segment caused by less data acquisition frequency and data elimination, adopting an RBF neural network to compensate the continuous missing data. Firstly, screening relevant variables of the missing data as input, screening the first 7 variables with larger contribution rate according to a principal component analysis method, and outputting the compensation output of the missing data by an output layer. The data set is finally obtained to be 380 by the data cleaning method.
The data is then subjected to a normalization process,
and eliminating the influence of the dimension and the order of magnitude difference on the model training process aiming at the threshold range among different interval data in the abnormal data set. The normalized data is subjected to dimension reduction processing by adopting a Partial Least Squares (PLS), and the acquired data is analyzed by utilizing a principal component analysis method, so that 6 input variables which are based on the correlation coefficient and the contribution rate of the variables are obtained: regenerator oxygen content, riser slurry feed, riser upper temperature, raw material nitrogen content, total feed, catalyst inventory, and main air volume for regeneration. The output variable to be collected is the micro-reaction activity value of the catalyst. The 380 selected data are divided into two parts: of these, 200 groups served as training samples and 180 groups served as test samples.
(2) Constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set
A recurrent RBF neural network is utilized to design a catalytic cracking catalyst micro-reaction activity prediction model,
input layer: the layer contains 6 input variables u i (t),i=1,2,…,6;
Hidden layer: the hidden layer contains 3 neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)]
y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
(3) And optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model. And evaluating the performance of the target prediction model according to the root mean square error and the accuracy, wherein the RMSE expression is as follows:
The calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
In this example, the root mean square error and the accuracy of the prediction model are evaluated as shown in table 2, and it can be seen from table 2 that the error and the accuracy of the catalyst micro-activity are within a reasonable range.
Table 2 prediction results for different algorithms
Example III
(1) Acquiring data and sorting and cleaning
In this embodiment, reactor data, regenerator data, desulfurization and denitrification facility data, product distribution data, treatment facility raw material data, and the like during the operation of the catalytic cracker are obtained by on-line instrument or laboratory analysis, wherein the parameters include the following 30 items, each item is 300, and mainly include: catalyst activity value, regeneration temperature, regeneration pressure, catalyst inventory, gasoline yield, liquid hydrocarbon yield, raw material sulfur, nitrogen, metal content, fresh feed amount, reaction pressure, reaction temperature, feed preheating temperature, raw material nitrogen content for diesel oil yield, regenerator oxygen content, regenerator dense phase inventory, regeneration main air volume, riser slurry feed amount, riser upper temperature and the like.
And then analyzing based on the data statistics result in the data acquisition time period, wherein in the embodiment, the variable data acquisition time period is in a full-load state, and the upper and lower limit interval values of each variable data are confirmed.
Segmenting the variable data by adopting a window segmentation method, and setting the data set of the catalytic cracking anti-regeneration system variable as D= { X 1 ,X 2 ,X 3 ,…,X 30 The j-th vector is X j ={x j1 ,x j2 ,…,x j300 The window size is set to S, and the window slides from the first data of each vector to obtain a data segment S.
And identifying the abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor (LOF) algorithm based on the characteristic attribute, and removing the abnormal data.
And aiming at missing data in a data segment caused by less data acquisition frequency and data elimination, adopting an RBF neural network to compensate the continuous missing data. Firstly, screening related variables of the missing data as input, screening the first 5 variables with larger contribution rate according to a principal component analysis method, and outputting the compensation output of the missing data by an output layer. The data set obtained by the data cleaning method is 290 bars.
The data is then subjected to a normalization process,
and eliminating the influence of the dimension and the order of magnitude difference on the model training process aiming at the threshold range among different interval data in the abnormal data set. The normalized data is subjected to dimension reduction processing by adopting a Partial Least Squares (PLS), and the acquired data is analyzed by utilizing a principal component analysis method, so that 6 input variables which are based on the correlation coefficient and the contribution rate of the variables are obtained: regenerator oxygen content, riser slurry feed, riser upper temperature, raw material nitrogen content, total feed, catalyst inventory, and main air volume for regeneration. The output variable to be collected is the micro-reaction activity value of the catalyst. The 290 groups of data are divided into two parts: of these, 150 groups served as training samples and 140 groups served as test samples.
(2) Constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set
A recurrent RBF neural network is utilized to design a catalytic cracking catalyst micro-reaction activity prediction model,
input layer: the layer contains 5 input variables u i (t),i=1,2,…,5;
Hidden layer: the hidden layer contains 3 neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)]
y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
(3) And optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model. And evaluating the performance of the target prediction model according to the root mean square error and the accuracy, wherein the RMSE expression is as follows:
The calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
In this example, the root mean square error and the accuracy of the prediction model are evaluated as shown in table 3, and it can be seen from table 3 that the error and the accuracy of the catalyst micro-activity are within a reasonable range.
TABLE 3 prediction results for different algorithms
The second aspect of the invention provides an intelligent prediction device for micro-reaction activity of a catalytic cracking device catalyst, which comprises: a controller for:
preprocessing variable data to obtain a processed sample set;
constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
and predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
In one embodiment, the preprocessing the variable data to obtain a processed sample set includes:
acquiring all variable data; in the application, the variable data are analyzed and obtained through an online instrument or laboratory test to obtain reactor data, regenerator data, desulfurization and denitrification facility data, product distribution data, treatment facility raw material data and the like during the operation of the catalytic cracking device, and specifically: regeneration temperature, regeneration pressure, catalyst inventory, gasoline yield, liquid hydrocarbon yield, raw material sulfur, nitrogen, metal content, fresh feed, reaction pressure, reaction temperature, feed preheating temperature, raw material nitrogen content of diesel yield, regenerator oxygen content, regenerator dense phase inventory, regeneration main air volume, riser slurry feed, riser upper temperature, outlet flue gas temperature, total feed volume, regenerator bottom dense phase temperature, regenerator dilute phase section pressure and the like.
Determining upper and lower limit interval values of variable data, and analyzing the running state of the variable data; acquiring upper and lower limit interval values of variable data according to the running state; the operating state includes: low load operating conditions (load less than 80%), medium load operating conditions (load 80% to 95%) and full load operating conditions (load greater than 95%).
Segmenting the variable data by adopting a window segmentation method;
identifying an abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor algorithm based on the characteristic attribute, and removing the abnormal data;
compensating missing data in variable data after abnormal data are removed;
normalizing the compensated variable data to obtain normalized data;
performing dimension reduction processing on the normalized data to obtain an input variable data set;
and selecting a preset number of data sets from the input variable data sets as a sample set. The problem that the difference of the threshold range between the data of different interval sections in the variable data set is large is solved through the processing, and meanwhile, the problem that data is missing or the difference of the data dimension between different variables is large in the data acquisition process is solved.
In one embodiment, the determining the upper and lower limit interval values of each variable data includes:
analyzing the running state of the variable data;
acquiring upper and lower limit interval values of variable data according to the running state;
the operating state includes: low load operating conditions, medium load operating conditions and full load operating conditions.
In one embodiment, segmenting the variable data using window segmentation includes:
setting the data set of the variable data to be D= { X 1 ,X 2 ,X 3 ,…,X j The j-th vector is X j ={x j1 ,x j2 ,…,x jn N is the total amount of data;
dividing the variable data by using a clustering algorithm, and determining a segmentation threshold value;
determining a window size s according to the segmentation threshold value, the data quantity of the variable data and the running state of the variable data;
the window is slid from the first data of each vector to obtain the data segment S.
In one embodiment, identifying and culling anomalous data segments using a local anomaly factor (LOF) algorithm based on a characteristic attribute includes:
the kth distance between the data segment S and the data segment O is calculated, and the formula is as follows:
wherein I is the number of attributes, f (S i ) Is the i-th dimension attribute value of the data segment S, f (O i ) Is a data segmentAn i-th dimensional attribute value of O, the attribute value comprising: entropy, mean, maximum and peak intervals; since the entropy value can more accurately reflect the distribution state of the data, the entropy value is used as one of key indexes of the data attribute.
Wherein, the Entropy value Entropy is expressed as:
where m= {1,2, …, M }, M is the number of subspaces, p (u) m ) Is the (u) m The probability distribution function of each subspace has the expression:
k(u m ) Is the subspace u where m The number of data contained;
calculating local anomaly factors of the data segment S, wherein the local anomaly factor evaluation function of the data segment S is as follows:
the local reachable densities of the data segments S are:
therein, lrd k (S) is the local reachable density of the data segment S, lrd k (O) is the local reachable density of data segment O, L k (S) is the set of all data segments in the data set D whose distance of the data segment S does not exceed its kth distance, LOF k (S) is a local anomaly factor value for data segment S; sigma reach-distance (S, O) is all neighborhood data L within data segment S k The sum of the reachable distances of (S); local anomaly factor LOF of comparison data k Value, when LOF k The larger the data is, the more alien the data occursThe greater the likelihood of being constant.
According to local abnormality factor LOF k Value, classifying each data segment;
identifying abnormal data in the data segments of different grades by utilizing a local abnormal factor algorithm;
and screening all the identified abnormal data, and eliminating the abnormal data.
In one embodiment, compensating missing data in variable data after eliminating abnormal data includes:
Screening relevant variables of the missing data, and in the embodiment, screening relevant variables of the missing data by analyzing information of the missing data;
screening the first E variables in the arrangement of the contribution rate from large to small from the related variables according to a principal component analysis method to serve as input variables of an RBF neural network missing compensation model;
establishing an RBF neural network loss compensation model, wherein an hidden layer structure of the RBF neural network loss compensation model is determined to be E+2 nodes according to input variables and an experimental construction method; the output layer is the compensation output of the missing data;
performing optimization calculation on the RBF neural network loss compensation model by using a particle swarm algorithm to obtain an optimized RBF neural network loss compensation model;
and carrying out compensation variable soft measurement by adopting the optimized RBF neural network missing compensation model to obtain a compensation value of missing data. The compensation process can solve the problem of data missing in the data segment caused by less acquisition frequency and data rejection.
In one embodiment, the performing the dimension reduction processing on the normalized data to obtain an input variable data set includes:
performing dimension reduction processing on the normalized data by adopting a partial least square method (PLS);
And analyzing the variables with the correlation coefficients and the contribution rates larger than the threshold values by using a principal component analysis method to form an input variable data set.
In one embodiment, constructing an initial prediction model of catalytic cracking catalyst microreaction activity from the sample set comprises:
constructing a catalytic cracking catalyst micro-reaction activity basic model based on a recurrent RBF neural network, wherein the catalytic cracking catalyst micro-reaction activity basic model comprises the following components:
input layer: the layer contains n input variables u i (t),i=1,2,…,n;
Hidden layer: the hidden layer contains J neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is:
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)];
wherein y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
Optionally, the controller is further configured to:
after a target prediction model of micro-reactivity of the catalytic cracking catalyst is obtained, evaluating the performance of the target prediction model according to root mean square error and precision, wherein the RMSE expression is as follows:
the calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
FIG. 2 is a block diagram of an intelligent prediction system for micro-reactivity of a catalytic cracking unit catalyst according to an embodiment of the present invention. As shown in fig. 2, the system includes:
the data processing module is used for preprocessing variable data to obtain a processed sample set;
the initial prediction model construction module is used for constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
the initial prediction model optimization module is used for optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
and the data prediction module is used for predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
In another aspect, the present invention provides a machine-readable storage medium having stored thereon instructions for causing a machine to perform the above-described intelligent prediction method of catalytic cracker catalyst microreactor activity.
Those skilled in the art will appreciate that all or part of the steps in a method for implementing the above embodiments may be implemented by a program stored in a storage medium, where the program includes several instructions for causing a single-chip microcomputer, chip or processor (processor) to perform all or part of the steps in a method according to the embodiments of the invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The alternative embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the embodiments of the present invention are not limited to the specific details of the above embodiments, and various simple modifications may be made to the technical solutions of the embodiments of the present invention within the scope of the technical concept of the embodiments of the present invention, and all the simple modifications belong to the protection scope of the embodiments of the present invention. In addition, the specific features described in the above embodiments may be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the various possible combinations of embodiments of the invention are not described in detail.
In addition, any combination of the various embodiments of the present invention may be made, so long as it does not deviate from the idea of the embodiments of the present invention, and it should also be regarded as what is disclosed in the embodiments of the present invention.
Claims (20)
1. An intelligent prediction method for micro-reactivity of a catalytic cracking unit catalyst is characterized by comprising the following steps:
preprocessing variable data to obtain a processed sample set;
constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
and predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
2. The intelligent prediction method for micro-reactivity of a catalytic cracker catalyst according to claim 1, wherein the preprocessing of variable data to obtain a processed sample set comprises:
acquiring all variable data;
determining upper and lower limit interval values of variable data;
segmenting the variable data by adopting a window segmentation method;
identifying an abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor algorithm based on the characteristic attribute, and removing the abnormal data;
Compensating missing data in variable data after abnormal data are removed;
normalizing the compensated variable data to obtain normalized data;
performing dimension reduction processing on the normalized data to obtain an input variable data set;
and selecting a preset number of data sets from the input variable data sets as a sample set.
3. The intelligent prediction method for micro-reactivity of catalyst of catalytic cracker according to claim 2, wherein the determining the upper and lower limit interval values of each variable data comprises:
analyzing the running state of the variable data;
acquiring upper and lower limit interval values of variable data according to the running state;
the operating state includes: low load operating conditions, medium load operating conditions and full load operating conditions.
4. The intelligent prediction method for micro-reactivity of catalyst of catalytic cracker according to claim 2, wherein the segmentation of the variable data by window segmentation method comprises:
setting the data set of the variable data to be D= { X 1 ,X 2 ,X 3 ,…,X j The j-th vector is X j ={x j1 ,x j2 ,…,x jn N is the total amount of data;
dividing the variable data by using a clustering algorithm, and determining a segmentation threshold value;
Determining a window size s according to the segmentation threshold value, the data quantity of the variable data and the running state of the variable data;
the window is slid from the first data of each vector to obtain the data segment S.
5. The intelligent prediction method for micro-reactivity of a catalytic cracker catalyst according to claim 2, wherein the identifying of the abnormal data segment and the abnormal data in the abnormal data segment by using a local abnormality factor algorithm based on characteristic attributes, and the rejecting of the abnormal data comprises:
the kth distance between the data segment S and the data segment O is calculated, and the formula is as follows:
wherein I is the number of attributes, f (S i ) Is the i-th dimension attribute value of the data segment S, f (O i ) Is the i-th dimensional attribute value of the data segment O, the attribute value comprising: entropy, mean, maximum and peak intervals;
wherein, the Entropy value Entropy is expressed as:
where m= {1,2, …, M }, M is the number of subspaces, p (u) m ) Is the (u) m The probability distribution function of each subspace has the expression:
k(u m ) Is the subspace u where m The number of data contained;
calculating local anomaly factors of the data segment S, wherein the local anomaly factor evaluation function of the data segment S is as follows:
the local reachable densities of the data segments S are:
therein, lrd k (S) is the local reachable density of the data segment S, lrd k (O) is the local reachable density of data segment O, L k (S) is the set of all data segments in the data set D whose distance of the data segment S does not exceed its kth distance, LOF k (S) is a local anomaly factor value for data segment S; sigma reach-distance (S, O) is all neighborhood data L in data segment S k The sum of the reachable distances of (S);
according to local abnormality factor LOF k Value, classifying each data segment;
identifying abnormal data in the data segments of different grades by utilizing a local abnormal factor algorithm;
and screening all the identified abnormal data, and eliminating the abnormal data.
6. The intelligent prediction method for micro-reactivity of a catalytic cracker catalyst according to claim 2, wherein compensating missing data in variable data after eliminating abnormal data comprises:
screening related variables of the missing data;
screening the first E variables in the arrangement of the contribution rate from large to small from the related variables according to a principal component analysis method to serve as input variables of an RBF neural network missing compensation model;
establishing an RBF neural network loss compensation model, wherein an hidden layer structure of the RBF neural network loss compensation model is determined to be E+2 nodes according to input variables and an experimental construction method; the output layer is the compensation output of the missing data;
Performing optimization calculation on the RBF neural network loss compensation model by using a particle swarm algorithm to obtain an optimized RBF neural network loss compensation model;
and carrying out compensation variable soft measurement by adopting the optimized RBF neural network missing compensation model to obtain a compensation value of missing data.
7. The intelligent prediction method for micro-reactivity of a catalytic cracker catalyst according to claim 2, wherein the performing a dimension reduction process on the normalized data to obtain an input variable data set comprises:
performing dimension reduction processing on the normalized data by adopting a partial least square method;
and analyzing the variables with the correlation coefficients and the contribution rates larger than the threshold values by using a principal component analysis method to form an input variable data set.
8. The intelligent prediction method for micro-reactivity of a catalytic cracking unit catalyst according to claim 2, wherein constructing an initial prediction model for micro-reactivity of a catalytic cracking catalyst according to the sample set comprises:
constructing a catalytic cracking catalyst micro-reaction activity basic model based on a recurrent RBF neural network, wherein the catalytic cracking catalyst micro-reaction activity basic model comprises the following components:
input layer: the layer contains n input variables u i (t),i=1,2,…,n;
Hidden layer: the hidden layer contains J neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is:
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)];
wherein y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
9. The intelligent prediction method for micro-reactivity of a catalytic cracker catalyst according to claim 1, wherein the method further comprises:
after a target prediction model of micro-reactivity of the catalytic cracking catalyst is obtained, evaluating the performance of the target prediction model according to root mean square error and precision, wherein the RMSE expression is as follows:
the calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
10. An intelligent prediction device for micro-reactivity of a catalyst of a catalytic cracking device is characterized by comprising: a controller for:
preprocessing variable data to obtain a processed sample set;
constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
and predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
11. The intelligent prediction apparatus for micro-reactivity of a catalytic cracker catalyst according to claim 10, wherein the preprocessing of the variable data to obtain a processed sample set comprises:
acquiring all variable data;
determining upper and lower limit interval values of variable data;
segmenting the variable data by adopting a window segmentation method;
identifying an abnormal data segment and abnormal data in the abnormal data segment by adopting a local abnormal factor algorithm based on the characteristic attribute, and removing the abnormal data;
Compensating missing data in variable data after abnormal data are removed;
normalizing the compensated variable data to obtain normalized data;
performing dimension reduction processing on the normalized data to obtain an input variable data set;
and selecting a preset number of data sets from the input variable data sets as a sample set.
12. The intelligent prediction apparatus for micro-reactivity of catalyst in catalytic cracker according to claim 11, wherein the determining the upper and lower limit interval values of each variable data comprises:
analyzing the running state of the variable data;
acquiring upper and lower limit interval values of variable data according to the running state;
the operating state includes: low load operating conditions, medium load operating conditions and full load operating conditions.
13. The intelligent prediction apparatus for micro-reactivity of catalyst in catalytic cracker according to claim 11, wherein the segmentation of the variable data by window segmentation method comprises:
setting the data set of the variable data to be D= { X 1 ,X 2 ,X 3 ,…,X j The j-th vector is X j ={x j1 ,x j2 ,…,x jn N is the total amount of data;
dividing the variable data by using a clustering algorithm, and determining a segmentation threshold value;
Determining a window size s according to the segmentation threshold value, the data quantity of the variable data and the running state of the variable data;
the window is slid from the first data of each vector to obtain the data segment S.
14. The intelligent prediction apparatus for micro-reactivity of a catalytic cracker catalyst according to claim 11, wherein the method for identifying the abnormal data segment and the abnormal data in the abnormal data segment by using a local abnormality factor algorithm based on the characteristic attribute, and rejecting the abnormal data comprises:
the kth distance between the data segment S and the data segment O is calculated, and the formula is as follows:
wherein I is the number of attributes, f (S i ) Is the i-th dimension attribute value of the data segment S, f (O i ) Is the i-th dimensional attribute value of the data segment O, the attribute value comprising: entropy, mean, maximum and peak intervals;
wherein, the Entropy value Entropy is expressed as:
where m= {1,2, …, M }, M is the number of subspaces, p (u) m ) Is the (u) m The probability distribution function of each subspace has the expression:
k(u m ) Is the subspace u where m The number of data contained;
calculating local anomaly factors of the data segment S, wherein the local anomaly factor evaluation function of the data segment S is as follows:
the local reachable densities of the data segments S are:
therein, lrd k (S) is the local reachable density of the data segment S, lrd k (O) is the local reachable density of data segment O, L k (S) is the set of all data segments in the data set D whose distance of the data segment S does not exceed its kth distance, LOF k (S) is a local anomaly factor value for data segment S; sigma reach-distance (S, O) is all neighborhood data L in data segment S k The sum of the reachable distances of (S);
according to local abnormality factor LOF k Value, classifying each data segment;
identifying abnormal data in the data segments of different grades by utilizing a local abnormal factor algorithm;
and screening all the identified abnormal data, and eliminating the abnormal data.
15. The intelligent prediction apparatus for micro-reactivity of a catalytic cracker catalyst according to claim 11, wherein compensating for missing data in variable data after excluding abnormal data comprises:
screening related variables of the missing data;
screening the first E variables in the arrangement of the contribution rate from large to small from the related variables according to a principal component analysis method to serve as input variables of an RBF neural network missing compensation model;
establishing an RBF neural network loss compensation model, wherein an hidden layer structure of the RBF neural network loss compensation model is determined to be E+2 nodes according to input variables and an experimental construction method; the output layer is the compensation output of the missing data;
Performing optimization calculation on the RBF neural network loss compensation model by using a particle swarm algorithm to obtain an optimized RBF neural network loss compensation model;
and carrying out compensation variable soft measurement by adopting the optimized RBF neural network missing compensation model to obtain a compensation value of missing data.
16. The intelligent prediction apparatus for micro-reactivity of a catalytic cracker catalyst according to claim 11, wherein the dimension reduction processing of the normalized data to obtain an input variable data set comprises:
performing dimension reduction processing on the normalized data by adopting a partial least square method;
and analyzing the variables with the correlation coefficients and the contribution rates larger than the threshold values by using a principal component analysis method to form an input variable data set.
17. The intelligent prediction apparatus for catalytic cracking unit catalyst microreaction activity according to claim 11, wherein constructing an initial prediction model for catalytic cracking catalyst microreaction activity according to the sample set comprises:
constructing a catalytic cracking catalyst micro-reaction activity basic model based on a recurrent RBF neural network, wherein the catalytic cracking catalyst micro-reaction activity basic model comprises the following components:
input layer: the layer contains n input variables u i (t),i=1,2,…,n;
Hidden layer: the hidden layer contains J neurons, and the output expression of each hidden layer neuron is:
wherein c j (t) is the center vector of the jth neuron, σ j (t) is the width of the jth neuron, the input vector h of the jth hidden layer neuron j (t) is:
h j (t)=[u 1 (t),u 2 (t),u 3 (t),u 4 (t),v j (t)×y(t-1)];
wherein y (t-1) is the output of the recurrent RBF neural network at time t-1, v j (t) outputting feedback connection weights of the neurons and the jth hidden layer neurons at the moment t;
output layer:
wherein w (t) is the connection weight vector of the hidden layer and the output layer, θ (t) is the output vector of the hidden layer neuron, θ j (t) is the output of the jth hidden layer neuron, w j And (t) is the connection weight of the j-th hidden layer neuron and the output neuron, and y (t) is the output of the recurrent RBF neural network at the moment t.
18. The intelligent prediction apparatus for micro-reactivity of catalyst for catalytic cracking unit according to claim 10, wherein the controller is further configured to:
after a target prediction model of micro-reactivity of the catalytic cracking catalyst is obtained, evaluating the performance of the target prediction model according to root mean square error and precision, wherein the RMSE expression is as follows:
the calculation formula of the prediction precision is as follows:
where z=1, 2, …, Z is the number of test samples, y d (t) is the desired output of the catalyst microreactor activity and y (t) is the actual output of the catalyst microreactor activity.
19. An intelligent prediction system for micro-reactivity of a catalytic cracking unit catalyst, which is characterized by comprising:
the data processing module is used for preprocessing variable data to obtain a processed sample set;
the initial prediction model construction module is used for constructing an initial prediction model of micro-reaction activity of the catalytic cracking catalyst according to the sample set;
the initial prediction model optimization module is used for optimizing the initial prediction model by adopting a rapid descent algorithm to obtain a catalytic cracking catalyst micro-reaction activity target prediction model;
and the data prediction module is used for predicting the preprocessed data to be detected by using the target prediction model to obtain a predicted value of the micro-reaction activity of the catalyst.
20. A machine-readable storage medium having instructions stored thereon for causing a machine to perform the intelligent prediction method of micro-reactivity of a catalytic cracking unit catalyst according to any of claims 1-9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111565086.9A CN116343932A (en) | 2021-12-20 | 2021-12-20 | Method, device and system for predicting micro-reaction activity of catalyst of catalytic cracking device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111565086.9A CN116343932A (en) | 2021-12-20 | 2021-12-20 | Method, device and system for predicting micro-reaction activity of catalyst of catalytic cracking device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116343932A true CN116343932A (en) | 2023-06-27 |
Family
ID=86874702
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111565086.9A Pending CN116343932A (en) | 2021-12-20 | 2021-12-20 | Method, device and system for predicting micro-reaction activity of catalyst of catalytic cracking device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116343932A (en) |
-
2021
- 2021-12-20 CN CN202111565086.9A patent/CN116343932A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109814513B (en) | Catalytic cracking unit optimization method based on data model | |
Ge et al. | Quality prediction for polypropylene production process based on CLGPR model | |
CN109002686B (en) | Multi-grade chemical process soft measurement modeling method capable of automatically generating samples | |
CN109396576B (en) | Deep learning-based electric spark machining stability and energy consumption state optimization decision system and decision method | |
CN109670625B (en) | NOx emission concentration prediction method based on unscented Kalman filtering least square support vector machine | |
CN102693451A (en) | Method for predicting ammonia process flue gas desulphurization efficiency based on multiple parameters | |
CN111237988B (en) | Control method and system for subway vehicle-mounted air conditioning unit | |
CN113724800A (en) | Catalytic cracking unit simulation prediction method based on molecular level mechanism model and big data technology | |
CN112489733A (en) | Octane number loss prediction method based on particle swarm algorithm and neural network | |
CN112364560B (en) | Intelligent prediction method for working hours of mine rock drilling equipment | |
CN108510180B (en) | Method for calculating performance interval of production equipment | |
CN112446597A (en) | Storage box quality evaluation method, system, storage medium, computer equipment and application | |
CN113440990A (en) | EMD-LSTM based outlet SO2Concentration prediction method | |
CN114358116A (en) | Oil-immersed transformer fault diagnosis method and system and readable storage medium | |
CN112465239A (en) | Desulfurization system operation optimization method based on improved PSO-FCM algorithm | |
CN112990546A (en) | Chemical plant power transformer fault prediction method based on particle swarm and neural network | |
CN116599396A (en) | Intelligent control method and system for generator set | |
CN114764682B (en) | Rice safety risk assessment method based on multi-machine learning algorithm fusion | |
CN117150677A (en) | Method and device for determining type selection parameters of electric dust collector | |
CN115169453A (en) | Hot continuous rolling width prediction method based on density clustering and depth residual error network | |
CN111199304B (en) | Multi-target combustion optimization method based on data-driven fusion strategy | |
CN107193992A (en) | A kind of 220kV main transformer condition evaluation prediction methods based on decision Tree algorithms | |
CN100370453C (en) | Rare-earth cascade extraction separation component content soft measuring method | |
CN106647285A (en) | Catalyst activity detecting method based on soft measurement technology | |
CN1327376C (en) | Soft measuring meter moduling method based on supporting vector machine |
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 |