CN114021449B - Prediction method for coal mill safety evaluation - Google Patents
Prediction method for coal mill safety evaluation Download PDFInfo
- Publication number
- CN114021449B CN114021449B CN202111279815.4A CN202111279815A CN114021449B CN 114021449 B CN114021449 B CN 114021449B CN 202111279815 A CN202111279815 A CN 202111279815A CN 114021449 B CN114021449 B CN 114021449B
- Authority
- CN
- China
- Prior art keywords
- coal mill
- coal
- variable
- principal component
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000003245 coal Substances 0.000 title claims abstract description 170
- 238000011156 evaluation Methods 0.000 title claims abstract description 31
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 50
- 238000000513 principal component analysis Methods 0.000 claims abstract description 27
- 230000015654 memory Effects 0.000 claims abstract description 13
- 239000013598 vector Substances 0.000 claims description 43
- 239000011159 matrix material Substances 0.000 claims description 37
- 238000004140 cleaning Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 7
- 230000007787 long-term memory Effects 0.000 claims description 5
- 230000006403 short-term memory Effects 0.000 claims description 5
- 238000010219 correlation analysis Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000012549 training Methods 0.000 abstract description 6
- 230000002035 prolonged effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 23
- 210000004027 cell Anatomy 0.000 description 14
- 230000004913 activation Effects 0.000 description 5
- 239000002817 coal dust Substances 0.000 description 4
- 239000000843 powder Substances 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Medical Informatics (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Development Economics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Entrepreneurship & Innovation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a prediction method for coal mill safety evaluation, which comprises the steps of extracting key factor variables and corresponding expert scores from a historical database to form an original data set; simplifying the data clustering of the original data set; reducing the dimension of key factor variables under each cluster based on a principal component analysis strategy to obtain principal component factors under each cluster; combining principal component analysis and a long-period memory neural network, taking principal component factors of a training sample as input variables of the long-period memory neural network, taking corresponding coal mill operation safety scores as output variables, and establishing a prediction model for coal mill equipment safety assessment; and evaluating the coal mill operation safety of the coal mill operation data based on the prediction model. The real-time online safety evaluation can be carried out on the running state of the coal mill, the running state of the coal mill is perceived in advance, the intervention work is implemented, the safety is improved, and the service life of the coal mill is prolonged.
Description
Technical Field
The invention relates to the technical field of optimization of system operation evaluation algorithms based on machine learning, in particular to a prediction method for coal mill safety evaluation.
Background
In the prior art, coal mill equipment operation safety evaluation research mainly analyzes the coal mill equipment operation safety from the angle of occurrence of coal mill equipment faults. For example, in fault diagnosis based on a quantitative model, firstly, the fault type of a coal mill under study is determined, secondly, a corresponding fault expression is established according to a certain fault type, and finally, whether the fault occurs or not is judged according to the fault expression, wherein the accurate establishment of the fault expression is a key for judging the running performance of the coal mill; in fault diagnosis based on a signal model, signals generated in the operation of the coal mill are mainly identified according to a sensing and measuring tool, wherein the sensing and measuring tool is a key for judging the operation performance of the coal mill, and in addition, a large number of sensors are installed and used and maintained later, so that excessive cost is generated, and the construction of an economic power plant is not facilitated; in fault diagnosis based on a historical data model, according to historical operation data of coal mill equipment, identifying fault parameters of the coal mill based on a certain intelligent algorithm and deeply mining the fault model of the coal mill equipment.
The research is mainly conducted aiming at a certain type of faults of the coal mill, the nonlinear strong coupling characteristic is arranged among the operation parameters of the coal mill equipment, the change of one parameter often causes the change of other parameters, the occurrence of various faults is caused, and the operation safety of the coal mill equipment cannot be evaluated in an omnibearing manner.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a prediction method for coal mill safety evaluation, which establishes a prediction model for coal mill equipment operation safety evaluation based on Principal Component Analysis (PCA) and long-short-term memory neural network (LSTM) to realize coal mill operation safety evaluation prediction.
The technical scheme adopted by the invention is as follows:
A prediction method for coal mill safety evaluation comprises the following steps:
s1, extracting key factor variables and corresponding expert scores thereof which influence the safe operation of coal mill equipment from a historical operation database to form an original data set;
s2, removing abnormal points in the original data set based on a data cleaning strategy, and clustering by adopting a fuzzy C-means clustering algorithm;
s3, adopting a principal component analysis strategy to reduce the dimension of key factor variables under each cluster, acquiring effective information in the key factor variables under each cluster, and constructing a principal component variable sample set;
The principal component analysis strategy includes:
The sample sequence vector x n of the sample set is centered:
Wherein x n is the nth group of sample sequence vectors in a certain type of clustered sample set, the number of the n sample set sequence vectors, m represents the number of factors influencing the operation safety of coal mill equipment, the subscript j is the jth factor, and x nj、xnj' is data before and after centering respectively;
Constructing a sample centralized dataset X by X nj', and calculating a covariance matrix XX T of the X;
performing eigenvalue decomposition on the covariance matrix XX T, and solving eigenvalues and eigenvectors of the covariance matrix;
selecting the largest k eigenvalues, normalizing the corresponding k eigenvectors, and forming an eigenvector matrix W as a row vector;
Converting the sample set into a new space constructed by k eigenvectors: f k＝WTxn;
Outputting a principal component variable sample set D' = (F 1,F2,...,Fk), wherein the dimension of the sample set is reduced from m to k;
S4, establishing a prediction model by combining principal component analysis and a long-term and short-term memory neural network: the main component variable is used as an input variable of the long-short-period memory neural network, the historical expert score corresponding to the main component variable is used as an output variable of the long-short-period memory neural network, and a coal mill operation safety evaluation prediction model is constructed:
and S5, evaluating the coal mill operation safety of the coal mill operation data based on the coal mill operation safety evaluation prediction model.
The further technical scheme is as follows:
before the key factor variables are subjected to dimensionality reduction based on a principal component analysis strategy, carrying out correlation analysis on the key factor variables, and judging the degree of correlation among the key factor variables based on a Pearson correlation coefficient analysis method, wherein the expression is as follows:
Wherein ρ XY is a pearson correlation coefficient, X, Y represents a sequence vector 1 and a sequence vector 2, respectively; σ X、σY represents the standard deviation of the sequence vector 1 and the sequence vector 2 respectively; cov (X, Y) represents the covariance between sequence vector 1 and sequence vector 2.
The data cleaning strategy adopts the following formula:
rij＝|Vij-Vavj|/Vavj·100％＞20％
wherein V ij represents data corresponding to the j variable of the i sample in the original data set; v avj represents the average value corresponding to the jth variable in the original dataset; r ij represents the percentage of the average value of the jth variable that the residual error between the jth variable and the average value of the jth variable of the ith sample occupies in the original dataset.
Clustering by adopting a fuzzy C-means clustering algorithm, which specifically comprises the following steps:
calculating a clustering center vector:
Wherein v i represents a cluster center vector, i is an ith cluster, i is more than or equal to 1 and less than or equal to c, and c represents a classification number; l represents the iteration number, n represents the total number of sequence vectors of the data set, c is more than or equal to 2 and less than or equal to n, j represents the j-th sequence vector, u j represents the j-th data sample, a ij represents the membership degree of the j-th group of data samples to the i-th cluster, a ij forms a membership degree matrix A (L) of the samples to each cluster, and the upper mark L represents the L-th iteration;
Iterative update a (L+1):
Comparing A (L) with A (L+1) in matrix norm form, stopping iteration if A (L+1)-AL is less than or equal to delta, otherwise taking L=L+1, and continuing to execute iteration; wherein δ represents the allowable error between the norm of the L-th iteration membership matrix and the norm of the L+1-th iteration membership matrix.
The beneficial effects of the invention are as follows:
The method provided by the invention realizes the assessment and prediction of the operation safety of the coal mill, can perform real-time online safety assessment on the operation state of the coal mill, senses the operation state of the coal mill in advance and performs intervention work, and allows the fault state of the coal mill to be adjusted to the safe operation state within a limited time, so that the intervention capability of operators in a coal-fired power plant is improved, and the service life of the coal mill is prolonged. The prediction model for coal mill safety evaluation has the advantages of wide application range, short training time, high prediction precision and high reliability.
The method provided by the invention combines Principal Component Analysis (PCA) and long-short-term memory neural network (LSTM) to establish a prediction model, wherein the prediction model takes principal component variables as input and history expert scores corresponding to the variables as output, so that the overall running condition of the coal mill can be comprehensively estimated, and the problem that nonlinear strong coupling exists between running parameters of the coal mill equipment is effectively solved.
The method extracts the effective information based on principal component analysis, shortens the variable dimension of the matrix of the original data on the basis of retaining the effective information of the original data, reduces the vector of the input end of the neural network, and improves the training speed of the long-term and short-term memory neural network.
According to the invention, the relation between key factors influencing the operation safety of the coal mill and the operation state of the coal mill is obtained through nonlinear fitting, and the operation safety evaluation of the coal mill equipment is obtained in a mathematical modeling mode, so that the method is more scientific and strict, the error evaluation caused by subjective factors of evaluation experts can be avoided, and further the safety problem and the cost loss caused by the error evaluation of the experts are avoided.
Drawings
FIG. 1 is a flow chart of a predictive method for coal pulverizer safety assessment of the present invention.
FIG. 2 shows the results of the key factor variables of the present invention after the data cleansing strategy.
FIG. 3 is a graph showing the clustering results of instantaneous coal feeding amount of the coal feeder, primary air quantity of the inlet of the coal mill, pressure difference between the upper side and the lower side of the grinding bowl and the frequency of the dynamic separator of the coal mill, which are obtained by the clustering treatment of the method.
FIG. 4 is a clustering result of instantaneous coal feeding amount of the coal feeder, opening of a cold and hot primary air baffle of the coal mill and primary air temperature of an inlet and an outlet of the coal mill, which are obtained by clustering treatment of the method.
FIG. 5 is a clustering result of instantaneous coal feeding amount of a coal feeder and inlet and outlet pressure of the coal mill, which are obtained by clustering treatment according to the method of the invention.
FIG. 6 shows the contribution rate and the cumulative contribution rate of each component obtained by the principal component analysis of the method of the present invention.
FIG. 7 is a diagram of a long and short term memory neural network architecture on which the method of the present invention is based.
FIG. 8 is a flow chart of a coal pulverizer operation safety assessment prediction model established by using the PCA-LSTM neural network of the invention.
Detailed Description
The following describes specific embodiments of the present invention with reference to the drawings.
The application relates to a prediction method for coal mill safety evaluation, which can refer to a flow chart of fig. 1 and comprises the following steps:
s1, extracting key factor variables and corresponding expert scores thereof which influence the safe operation of coal mill equipment from a historical operation database to form an original data set;
Specifically, key factors influencing the safe operation of the coal mill equipment are identified by combining the working principle of the coal mill and the fault case of the coal mill, corresponding variables and corresponding expert scores are extracted from a historical operation database, the sampling period is once per minute, and the operation load of the boiler is as short as 350MW-400MW, so that an original data set (original matrix) influencing the operation safety of the coal mill equipment is formed.
S11, analyzing the working principle of the coal mill and identifying key factors affecting the operation safety of the coal mill by using the fault case of the coal mill. Coal mill vibration faults and coal mill fire faults are two common types of faults of coal mill equipment. If the reasons such as loosening and falling of the basic components in the coal mill are not considered, the phenomenon that the coal mill is full of coal or the coal quantity is too small in the coal mill equipment is a direct reason for causing the vibration of the coal mill; when the fire phenomenon occurs in the coal mill system, the outlet temperature of the coal mill often shows a jump type rising phenomenon, the inlet and outlet pressure difference of the coal mill rises rapidly, in addition, the unbalanced input of the cold air quantity and the hot air quantity of the coal mill leads to the excessive high cold air quantity in the coal mill, the coal dust can not be dried to the qualified temperature, the coal dust with larger humidity can not be normally conveyed, the coal dust stays in the coal mill, when the hot air quantity in the coal mill is excessive, the coal dust spontaneously ignites, and when serious, the coal mill explosion is directly caused.
S12, primarily identifying key factor variables influencing the safe operation of the coal mill, wherein the key factor variables mainly comprise: the method comprises the following steps of coal inlet quantity G of a coal mill, primary air quantity M, primary air temperature of an inlet and an outlet of the coal mill (T in,Tout), opening degree of a cold and hot primary air baffle plate of the coal mill (O Pcool,OPhot), primary pressure difference of the inlet and the outlet of the coal mill (P in,Pout), pressure difference of a grinding bowl of the coal mill (Pm) and frequency of a dynamic separator of the coal mill (sepf). Further, a corresponding coal mill facility historical score (sc) is obtained by historical expert evaluation. The acquisition interval is 1min, the acquisition time is one day, and the original matrix of the historical operation data of the coal mill equipment is obtained, and is shown in table 1.
Table 1 partial historical operating data and historical expert scoring for coal mill equipment
S2, removing abnormal points in the original data set based on a data cleaning strategy, and clustering by adopting a fuzzy C-means clustering algorithm (FCM), wherein the method specifically comprises the following steps:
S21, eliminating abnormal points in data in an original matrix based on a data cleaning strategy, wherein the principle of judging that each variable data is an abnormal point is as follows: the residual error between each variable data and the corresponding variable data mean exceeds 20% of the mean, and the variable data can be identified as an abnormal point, so that the data cleaning work is completed. The data cleaning strategy principle is shown as follows:
rij＝|Vij-Vavj|/Vavj·100％＞20％
In the above formula, V ij represents data corresponding to the j variable of the i sample in the original data set; v avj represents the average value corresponding to the jth variable in the original dataset; r ij represents the percentage of the average value of the jth variable that the residual error between the jth variable and the average value of the jth variable of the ith sample occupies in the original dataset.
As shown in fig. 2, the result of each key factor variable after being processed by the data cleaning strategy is shown.
S22, a fuzzy C-means clustering algorithm, which is based on iterative self-organizing data analysis of an objective function, comprises the following steps:
Assuming that n sequence vectors are given, U= { U 1,u2,u3,...ui…,un }, wherein U i＝(xi1,xi2,xi3...,xim)∈Rm, i are the ith clusters, 1.ltoreq.i.ltoreq.c, c is a classification number, 2.ltoreq.c.ltoreq.n; r represents the real number domain;
Initializing, wherein A (0)∈Mc is that A (0) is an initialization membership matrix, and M c is that classifying space for dividing u into c classes;
Number of iterations l=0, 1,2,..when:
In the above formula, v i represents a cluster center vector; n is the total number of sequence vectors representing the data set, j represents the jth sequence vector, u j represents the jth data sample, a ij represents the membership degree of the jth group of data samples to the ith cluster, a ij forms a membership degree matrix A (L) of the samples to each cluster, and the superscript L represents the L-th iteration;
A (L) is iteratively updated to a (L+1) using the following formula:
Comparing A L with A (L+1) in matrix norm form, stopping iteration if A (L+1)-AL is less than or equal to delta, otherwise taking L=L+1, and continuing to execute iteration; wherein delta is the allowable error between the norm of the L-th iterative membership matrix and the norm of the L+1th iterative membership matrix;
Finally, based on the principle of maximum membership, blurring is clarified, and the category of each data in the original matrix is determined.
The Cluster1, cluster2 shown in fig. 1. Different clustering results in the first to fourth clustering cases are shown in fig. 3 to 5. FIG. 3 shows the results of clustering the instantaneous coal feed amount of the coal feeder, the primary air quantity of the inlet of the coal mill, the pressure difference between the upper side and the lower side of the grinding bowl and the frequency of the dynamic separator of the coal mill; FIG. 4 shows the clustering result of instantaneous coal feeding amount of the coal feeder, the opening of a cold and hot primary air baffle of the coal mill and the primary air temperature of an inlet and an outlet of the coal mill; FIG. 5 shows the clustering result of instantaneous coal feeding amount of the coal feeder and inlet and outlet pressure of the coal mill.
And the abnormal points of the original data are removed based on a data cleaning strategy, normal, stable and effective data points are reserved, the data after the data are cleaned are subjected to clustering analysis based on the module C mean value clustering, the complexity of the data is simplified, the subsequent principal component analysis and the construction of a coal mill equipment operation safety evaluation prediction model are facilitated, and the model prediction precision is improved.
S3, adopting a principal component analysis strategy to reduce the dimension of key factor variables under each cluster, acquiring effective information in the key factor variables under each cluster, and constructing a principal component variable sample set, wherein the method specifically comprises the following steps of:
S31, not generally, carrying out correlation analysis on key factor variables before carrying out dimension reduction compression on the key factor variables based on a principal component analysis strategy, and judging the degree of correlation among the key factor variables based on a Pearson correlation coefficient analysis method, wherein the expression is as follows:
in the above formula, ρ XY is a pearson correlation coefficient, and X, Y represents a sequence vector 1 and a sequence vector 2, respectively; σ X、σY represents the standard deviation of the sequence vector 1 and the sequence vector 2 respectively; cov (X, Y) represents the covariance between sequence vector 1 and sequence vector 2.
The above-mentioned key factors influencing the operation safety of the coal mill equipment are numerous, and the mutual coupling degree between the variables is high, so that the difficulty of establishing the evaluation and prediction model of the operation safety of the coal mill equipment is increased, and therefore, the degree of correlation between the influencing factors is firstly judged based on the pearson correlation coefficient analysis method so as to verify the accuracy of the key factor variables preliminarily selected in the step S1. The value range of the pearson correlation coefficient is [ -1,1], the correlation coefficient is 0 to indicate that the correlation between two variables is irrelevant, and the correlation coefficient is close to 1 or-1 to indicate that the correlation between the two variables is stronger. The correlation analysis shows the correlation discrimination results of the factor variables as shown in table 2 below.
TABLE 2 results of correlation determination between key factor variables
As shown in the table 2, the correlation coefficient between the upper and lower pressure differences of the grinding bowl of the coal mill and the instantaneous coal feeding amount of the coal feeder is 0.63 at most, and the correlation coefficient between the upper and lower pressure differences of the grinding bowl and the temperature of the outlet air powder is-0.58; the relative coefficient of the upper and lower pressure difference of the grinding bowl and the opening of the cold air primary air regulating baffle is-0.45; the correlation coefficient between the instantaneous coal feeding amount of the coal feeder and the temperature of the air powder at the outlet of the coal mill reaches-0.85; the correlation coefficient between the primary air quantity of the inlet of the coal mill and the temperature of the air powder of the outlet of the coal mill is-0.14; if the instantaneous coal feeding amount of the coal feeder is too high, if the primary air quantity is not adjusted timely or the opening degree of the folding door is too small, the coal blocking of the coal mill is caused, at the moment, the pressure difference between the upper part and the lower part of the grinding bowl is increased, and the outlet temperature and the primary air quantity of the inlet of the coal mill are reduced. The frequency correlation coefficient of the inlet air temperature of the coal mill and the dynamic separator of the coal mill is 0.67; the frequency of the dynamic separator is an important guarantee for the coal mill to produce the coal powder with qualified fineness, when the dynamic separator is in failure, the pressure difference between the inlet and the outlet of the coal mill is reduced, and the outlet temperature of the coal mill is correspondingly changed. The following is indicated: there is a strong correlation between many influencing factors.
S32, principal component analysis (PCA (Principal Component Analysis)) is to convert high-dimensional related variables into low-dimensional new comprehensive variables which are mutually uncorrelated and can represent effective information of the original data. Based on a principal component analysis strategy, data compression is carried out on a plurality of variables influencing the operation safety of the coal mill, principal component factors (effective information) influencing the safe operation of the coal mill equipment of a plurality of key factors are extracted, the compressed variables are principal component variables, the dimension influencing the operation safety variable of the coal mill is shortened on the basis of keeping the effective information of original data, the number of parameters of an input end of a neural network is reduced, and the speed of training a prediction model of the neural network is improved. The method specifically comprises the following steps:
from the analysis, the key factor variables affecting the operation safety of the coal mill equipment comprise 10 main component analysis specific steps, and a defined sample set D is shown in the following formula:
Wherein n represents the number of sample sequence vectors; m represents the number of key influencing factors, in this embodiment, m is 10; wherein x n1 is the coal inlet G, x n2 of the coal mill and the primary air flows M, x n3 and x n4 are the inlet and outlet primary air temperatures of the coal mill respectively (T in,Tout)、xn5 and x n6 are the cold and hot primary air baffle openings of the coal mill (OP cool,OPhot)、xn7 and x n8 are the inlet and outlet primary pressure differences of the coal mill (P in,Pout)、xn9 and x n10 are the upper and lower pressure differences of a grinding bowl of the coal mill (P m) and the frequency of a dynamic separator of the coal mill (sepf).
The sample sequence vector x n of sample set D is de-centered:
Wherein, subscript j is the j factor, and x nj、xni' is the data before and after centralization respectively;
Constructing a sample de-centering dataset X from X ni', and calculating a covariance matrix XX T of X;
performing eigenvalue decomposition on the covariance matrix XX T, and solving eigenvalues and eigenvectors of the covariance matrix;
Sorting the eigenvalues from large to small, selecting the largest k eigenvectors, and respectively taking the corresponding k eigenvectors as row vectors to form an eigenvector matrix W;
converting the sample set D into a new space constructed by k eigenvectors: f k＝WTxn
Outputting a principal component variable sample set D' = (F 1,F2,...,Fk); where k represents the principal component variable that affects the safety factor of operation of the coal mill equipment, i.e. the sample set dimension is reduced from m to k.
Specifically, a matrix laboratory (MATLAB) simulation platform is used for carrying out principal component analysis on historical operation data of coal mill equipment, and when the variance contribution rate of principal components extracted by PCA reaches more than 95%, the principal components can be considered to replace original data information. The contribution rate and the accumulated contribution rate of each component are shown in fig. 6, and the contribution rate and the accumulated contribution rate of each principal component under the first cluster, the second cluster, the third cluster and the fourth cluster are shown in (a), (b), (c) and (d) in the figure respectively.
S4, establishing a prediction model by combining principal component analysis and a long-term and short-term memory neural network: taking the main component variable as an input variable of the long-short-period memory neural network, and taking a historical expert score corresponding to the main component variable as an output variable of the long-short-period memory neural network to construct a coal mill operation safety evaluation prediction model;
Specifically, referring to fig. 7, a long-short-term memory (LSTM) neural network is formed by connecting a plurality of neural network cells to form a network structure, where R t-1 is the last cell state, h t-1 is the last cell layer output, x t is the current cell input, the LSTM neural network cell structure includes 3 sigmod activation functions and two tanh activation functions, and in fig. 7, sigma is respectively set at the forgetting gate, the update gate and the output gate of the neural network from left to right; the main function of the sigma 1 function is to filter useless information in the cell information of the upper LSTM neural network and retain valid information; the main function of the sigma 2 function is to determine the part of information which needs to be updated in the upper LSTM neural network cell information; the main function of the sigma 3 function is to acquire initial output information; the tanh function is arranged at an update gate and an output gate of the LSTM neural network, and the tanh1 function is used for generating new information; tanh2 updates the neural network cell state mainly at the next moment. The forget gate, update gate and output gate cooperate to control and protect cell status. The LSTM neural network cell update mechanism is as follows:
the forgetting gate is controlled based on a sigmoid function, and has the main functions of filtering useless information in the cell information of the upper LSTM neural network layer and retaining effective information, wherein the operation formula of the forgetting gate is shown as follows:
ft＝σ1(Wf·[Lt-1,xt]+bf)
in the above formula, f t represents information generated by the forgetting gate, σ1 represents an activation function of the forgetting gate layer, W f represents a weight matrix of forgetting gate connection of a previous layer neural network value to a next layer neural network, L t-1 represents a previous layer neural network output, x t represents a current layer neural network input, and b f represents a weight threshold of the forgetting gate connection.
The main function of the update gate is to update information and generate new information based on sigma 2 and tanh1 functions, wherein sigma 2 functions are used for determining the part of information to be updated in the upper LSTM neural network cell information, tanh1 functions are used for generating the new information, and the update gate is calculated as shown in the following formula:
it＝σ2(Wi·[Lt-1,xt]+bi)
b2t＝tanh1(Wc·[Lt-1,xt]+bc)
Rt＝Rt-1·ft+it·b2t
in the above formula, σ2 represents an activation function of an update gate layer, W i represents a weight matrix of update gate confirmation update information, L t-1 represents a previous layer neural network output, x t represents a current layer neural network input, b i represents a weight threshold of update gate confirmation update information, W c represents a weight matrix of update gate connection, b i represents a weight threshold of update gate update information, i t represents information confirming update, b 2t represents new information generated, and R t represents new neuron state.
The output gate is composed of sigma 3 and tanh2 functions, the sigma 3 functions are used for determining initial output information, the tanh2 functions are used for scaling the output information to a (-1, 1) interval and multiplying the output information with the initial output pair by pair, so that the output of the model is obtained, that is, updating of the LSTM neural network cell state is completed, and the output gate is calculated as shown in the following formula:
c3t＝σ3(Wo·[Lt-1,xt]+bo)
Lt＝c3t·tanh(Rt)
In the above formula, σ3 represents an activation function for determining the neuron state output information, W o represents a weight matrix for confirming the output information, L t-1 represents the output of the previous layer of the neural network, x t represents the input of the current layer of the neural network, b o represents a weight threshold for confirming the output information, c 3t represents the neuron state output information, and L t represents the output information of the neural network of the previous layer.
The LSTM cells shown in fig. 1 represent the neural network of each layer, and the evaluator represents the coal mill operation safety evaluation prediction model.
And establishing a coal mill operation safety evaluation prediction model based on the PCA-LSTM neural network, taking a main component variable as an input variable of the long-short-term memory neural network, and taking a historical expert score corresponding to the main component variable as an output variable of the long-short-term memory neural network to establish the coal mill operation safety evaluation prediction model.
The LSTM neural network is provided with the forgetting gate and the updating gate, so that the information is selectively reserved and updated, the learning capacity of the network is enhanced, and the problems of gradient disappearance or gradient explosion and the like are avoided.
S5, establishing a coal mill operation safety evaluation prediction model based on the PCA-LSTM neural network, and evaluating the coal mill operation safety of the coal mill operation data. The flow of the online evaluation is shown in fig. 8, wherein a safety prediction evaluator represents a coal mill operation safety evaluation prediction model.
The prediction value of the operation safety of the coal mill equipment based on PCA-LSTM is compared with the true value to be corresponding, and the data prediction value and the true value under each cluster can be obtained to be relatively close.
On the basis of keeping effective information of original data, the application shortens the dimension of the original matrix and reduces the vector of the input end of the neural network by utilizing principal component analysis, thereby improving the training speed of the neural network.
And (3) in order to further verify the effectiveness of the coal mill safety prediction model established by the PCA-LSTM neural network, establishing the coal mill safety prediction model based on the LSTM neural network, and comparing the calculation time and the relative error of the two prediction models. The results are shown in tables 3 and 4.
Table 3 LSTM and PCA-LSTM predictive model training time costs
TABLE 4 PCA-GRNN and GRNN predictive model relative error
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (3)
1. The prediction method for evaluating the safety of the coal mill is characterized by comprising the following steps of:
s1, extracting key factor variables and corresponding expert scores thereof which influence the safe operation of coal mill equipment from a historical operation database to form an original data set;
The key factor variables affecting the safe operation of the coal mill are identified for the first time, and the key factor variables comprise: the method comprises the following steps of coal inlet amount, primary air quantity, primary air temperature at an inlet and an outlet of a coal mill, opening of a cold primary air baffle of the coal mill, primary pressure difference at an inlet and an outlet of the coal mill, pressure difference between an upper part and a lower part of a grinding bowl of the coal mill and frequency of a dynamic separator of the coal mill; the historical scores of the corresponding coal mill equipment are evaluated and obtained by historical specialists; data acquisition is carried out according to the set acquisition interval and acquisition time length, and a historical operation data original matrix of coal mill equipment is obtained;
s2, removing abnormal points in the original data set based on a data cleaning strategy, and clustering by adopting a fuzzy C-means clustering algorithm;
The data cleaning strategy adopts the following formula:
rij＝|Vij-Vavj|/Vavj·100％＞20％
Wherein V ij represents data corresponding to the j variable of the i sample in the original data set; v avj represents the average value corresponding to the jth variable in the original dataset; r ij represents the percentage of the average value of the jth variable that the residual error between the jth variable and the average value of the jth variable of the ith sample occupies in the original data set;
s3, adopting a principal component analysis strategy to reduce the dimension of key factor variables under each cluster, acquiring effective information in the key factor variables under each cluster, and constructing a principal component variable sample set;
The principal component analysis strategy includes:
The sample sequence vector x n of the sample set is centered:
Wherein x n is the nth group of sample sequence vectors in a certain type of clustered sample set, the number of the n sample set sequence vectors, m represents the number of factors influencing the operation safety of coal mill equipment, the subscript j is the jth factor, and x nj、xnj' is data before and after centering respectively;
Constructing a sample centralized dataset X by X nj', and calculating a covariance matrix XX T of the X;
performing eigenvalue decomposition on the covariance matrix XX T, and solving eigenvalues and eigenvectors of the covariance matrix;
selecting the largest k eigenvalues, normalizing the corresponding k eigenvectors, and forming an eigenvector matrix W as a row vector;
Converting the sample set into a new space constructed by k eigenvectors: f k＝WTxn;
Outputting a principal component variable sample set D' = (F 1,F2,...,Fk), wherein the dimension of the sample set is reduced from m to k;
S4, establishing a prediction model by combining principal component analysis and a long-term and short-term memory neural network: taking the main component variable as an input variable of the long-short-period memory neural network, and taking a historical expert score corresponding to the main component variable as an output variable of the long-short-period memory neural network to construct a coal mill operation safety evaluation prediction model;
and S5, evaluating the coal mill operation safety of the coal mill operation data based on the coal mill operation safety evaluation prediction model.
2. The prediction method for coal mill safety assessment according to claim 1, wherein before the key factor variables are subjected to dimensionality reduction based on a principal component analysis strategy, the key factor variables are subjected to correlation analysis, and the degree of correlation among the key factor variables is judged based on a pearson correlation coefficient analysis method, wherein the expression is as follows:
Wherein ρ XY is a pearson correlation coefficient, X, Y represents a sequence vector 1 and a sequence vector 2, respectively; σ X、σY represents the standard deviation of the sequence vector 1 and the sequence vector 2 respectively; cov (X, Y) represents the covariance between sequence vector 1 and sequence vector 2.
3. The prediction method for coal mill safety assessment according to claim 1, wherein clustering is performed by using a fuzzy C-means clustering algorithm, comprising:
calculating a clustering center vector:
Wherein v i represents a cluster center vector, i is an ith cluster, i is more than or equal to 1 and less than or equal to c, and c represents a classification number; l represents the iteration number, n represents the total number of sequence vectors of the data set, 2.ltoreq.c.ltoreq.n, j represents the jth sequence vector, u j represents the jth data sample, a ij represents the membership degree of the jth group of data samples to the ith cluster, a ij constitutes a membership degree matrix A (L) of the samples to each cluster, and the superscript L represents the L-th iteration;
A (L) is iteratively updated to a (L+1) using the following formula:
Comparing A (L) with A (L+1) in matrix norm form, stopping iteration if A (L+1)-AL is less than or equal to delta, otherwise taking L=L+1, and continuing to execute iteration; wherein δ represents the allowable error between the norm of the L-th iteration membership matrix and the norm of the l+1-th iteration membership matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111279815.4A CN114021449B (en) | 2021-10-29 | 2021-10-29 | Prediction method for coal mill safety evaluation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111279815.4A CN114021449B (en) | 2021-10-29 | 2021-10-29 | Prediction method for coal mill safety evaluation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114021449A CN114021449A (en) | 2022-02-08 |
| CN114021449B true CN114021449B (en) | 2024-05-24 |
Family
ID=80059350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111279815.4A Active CN114021449B (en) | 2021-10-29 | 2021-10-29 | Prediction method for coal mill safety evaluation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114021449B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105528650A (en) * | 2015-12-02 | 2016-04-27 | 江苏省电力公司信息通信分公司 | Machine room temperature and humidity prediction method based on principle component analysis and BP neural network |
| CN109670629A (en) * | 2018-11-16 | 2019-04-23 | 浙江蓝卓工业互联网信息技术有限公司 | Coal-burning boiler thermal efficiency forecast method based on shot and long term Memory Neural Networks |
| CN111308991A (en) * | 2020-03-19 | 2020-06-19 | 中国大唐集团科学技术研究院有限公司华东电力试验研究院 | Coal mill operation fault identification method and application |
| CN111582571A (en) * | 2020-04-30 | 2020-08-25 | 中国电力科学研究院有限公司 | Power grid operation situation sensing method and system with model driving and data driving integrated |
| CN111695601A (en) * | 2020-05-15 | 2020-09-22 | 特变电工西安电气科技有限公司 | Photovoltaic power prediction method, device, equipment and readable storage medium |
| CN111950191A (en) * | 2020-07-07 | 2020-11-17 | 湖南大学 | Rotary kiln sintering temperature prediction method based on hybrid deep neural network |
| CN111966569A (en) * | 2019-05-20 | 2020-11-20 | 中国电信股份有限公司 | Hard disk health degree evaluation method and device and computer readable storage medium |
| CN113408699A (en) * | 2021-06-16 | 2021-09-17 | 中国地质科学院 | Lithology identification method and system based on improved radial basis function neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11933159B2 (en) * | 2020-03-26 | 2024-03-19 | Aspentech Corporation | System and methods for developing and deploying oil well models to predict wax/hydrate buildups for oil well optimization |
-
2021
- 2021-10-29 CN CN202111279815.4A patent/CN114021449B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105528650A (en) * | 2015-12-02 | 2016-04-27 | 江苏省电力公司信息通信分公司 | Machine room temperature and humidity prediction method based on principle component analysis and BP neural network |
| CN109670629A (en) * | 2018-11-16 | 2019-04-23 | 浙江蓝卓工业互联网信息技术有限公司 | Coal-burning boiler thermal efficiency forecast method based on shot and long term Memory Neural Networks |
| CN111966569A (en) * | 2019-05-20 | 2020-11-20 | 中国电信股份有限公司 | Hard disk health degree evaluation method and device and computer readable storage medium |
| CN111308991A (en) * | 2020-03-19 | 2020-06-19 | 中国大唐集团科学技术研究院有限公司华东电力试验研究院 | Coal mill operation fault identification method and application |
| CN111582571A (en) * | 2020-04-30 | 2020-08-25 | 中国电力科学研究院有限公司 | Power grid operation situation sensing method and system with model driving and data driving integrated |
| CN111695601A (en) * | 2020-05-15 | 2020-09-22 | 特变电工西安电气科技有限公司 | Photovoltaic power prediction method, device, equipment and readable storage medium |
| CN111950191A (en) * | 2020-07-07 | 2020-11-17 | 湖南大学 | Rotary kiln sintering temperature prediction method based on hybrid deep neural network |
| CN113408699A (en) * | 2021-06-16 | 2021-09-17 | 中国地质科学院 | Lithology identification method and system based on improved radial basis function neural network |
Non-Patent Citations (1)
| Title |
|---|
| 基于聚类LSTM深度学习模型的主动配电网电能质量预测;翁国庆;龚阳光;舒俊鹏;黄飞腾;;高技术通讯;20200715(第07期);第687-697页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114021449A (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Data-driven mode identification and unsupervised fault detection for nonlinear multimode processes | |
| CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
| CN109670629B (en) | Coal-fired boiler thermal efficiency prediction method based on long-short term memory neural network | |
| CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
| CN109740859A (en) | Transformer condition evaluation and system based on Principal Component Analysis and support vector machines | |
| CN110571792A (en) | Analysis and evaluation method and system for operation state of power grid regulation and control system | |
| CN110413601B (en) | Generator data screening method based on Gauss naive Bayes and prediction error method | |
| Tran et al. | Algorithm for estimating online bearing fault upon the ability to extract meaningful information from big data of intelligent structures | |
| CN113588308A (en) | Coal mill detonation fault early warning and diagnosis method based on LSTM and Stacking fusion algorithm | |
| CN110175682A (en) | A kind of optimization core pivot element analysis fault monitoring method based on Chaos-Particle Swarm Optimization | |
| CN110427019B (en) | Industrial process fault classification method and control device based on multivariate discriminant analysis | |
| Zouari et al. | Fault detection system for centrifugal pumps using neural networks and neuro-fuzzy techniques | |
| Jian et al. | Application of Model‐Based Deep Learning Algorithm in Fault Diagnosis of Coal Mills | |
| CN113050602B (en) | Industrial process fault classification method based on robust semi-supervised discriminant analysis | |
| CN111242204A (en) | Operation and maintenance management and control platform fault feature extraction method | |
| CN112734131B (en) | Fan blade icing state prediction method based on deep learning algorithm | |
| CN114021449B (en) | Prediction method for coal mill safety evaluation | |
| CN113156908B (en) | Multi-working-condition industrial process monitoring method and system with mechanism and data combined fusion | |
| CN115526433A (en) | Power plant reheat flue gas baffle operation prediction method based on integrated hybrid model | |
| CN112949524B (en) | Engine fault detection method based on empirical mode decomposition and multi-core learning | |
| Lange et al. | Predicting particle fineness in a cement mill | |
| Lu et al. | Visual monitoring of industrial operation states based on kernel fisher vector and self-organizing map networks | |
| Mendonça et al. | Fault detection and isolation of industrial processes using optimized fuzzy models | |
| Shang et al. | Sparse slow feature analysis for enhanced control monitoring and fault isolation | |
| Tang et al. | The improvement of remaining useful life prediction for aero-engines by classification and deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |