CN116383636A - Coal mill fault early warning method based on PCA and LSTM fusion algorithm - Google Patents
Coal mill fault early warning method based on PCA and LSTM fusion algorithm Download PDFInfo
- Publication number
- CN116383636A CN116383636A CN202310334054.0A CN202310334054A CN116383636A CN 116383636 A CN116383636 A CN 116383636A CN 202310334054 A CN202310334054 A CN 202310334054A CN 116383636 A CN116383636 A CN 116383636A
- Authority
- CN
- China
- Prior art keywords
- coal mill
- early warning
- fault early
- real
- model
- 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
- 239000003245 coal Substances 0.000 title claims abstract description 315
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 43
- 230000004927 fusion Effects 0.000 title claims abstract description 35
- 238000012512 characterization method Methods 0.000 claims abstract description 58
- 238000000513 principal component analysis Methods 0.000 claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 45
- 230000015654 memory Effects 0.000 claims abstract description 27
- 239000011159 matrix material Substances 0.000 claims description 31
- 238000012360 testing method Methods 0.000 claims description 21
- 238000010586 diagram Methods 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 13
- 230000009467 reduction Effects 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 10
- 239000013598 vector Substances 0.000 claims description 8
- 230000007787 long-term memory Effects 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 abstract description 11
- 238000013528 artificial neural network Methods 0.000 description 17
- 230000000875 corresponding effect Effects 0.000 description 11
- 239000003921 oil Substances 0.000 description 11
- 238000010298 pulverizing process Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 6
- 210000004027 cell Anatomy 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 5
- 238000010248 power generation Methods 0.000 description 5
- 230000006403 short-term memory Effects 0.000 description 5
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 239000003638 chemical reducing agent Substances 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000000227 grinding Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 201000007902 Primary cutaneous amyloidosis Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000009529 body temperature measurement Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000011049 filling Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000010687 lubricating oil Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000003801 milling Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 208000014670 posterior cortical atrophy Diseases 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000007789 sealing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Disintegrating Or Milling (AREA)
Abstract
The application discloses a coal mill fault early warning method based on PCA and LSTM fusion algorithm, comprising the following steps: acquiring a plurality of historical operating state parameters of a coal mill in a normal operating condition time period, and carrying out principal component analysis on the parameters to obtain a coal mill state characterization parameter data set; constructing a coal mill fault early warning model based on a long-short-period memory network, and performing iterative training on the coal mill fault early warning model by utilizing a coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model; acquiring a plurality of real-time running state parameters of the coal mill, and carrying out principal component analysis on the plurality of real-time running state parameters to obtain real-time state characterization parameters of the coal mill; and inputting the real-time state characterization parameters of the coal mill into a fully trained coal mill fault early warning model to obtain a fault early warning result of the coal mill. According to the invention, the relevant parameters of the faults of the coal mill are accurately and effectively predicted based on the relevant variables, so that the monitoring and fault early warning of the real-time running state of the coal mill are realized.
Description
Technical Field
The invention relates to the technical field of thermal power generation system fault early warning, in particular to a coal mill fault early warning method, device, electronic equipment and computer readable storage medium based on PCA and LSTM fusion algorithm.
Background
The power generation amount of thermal power generation accounts for about 70% of the total power generation amount of China. The coal mill is used as core equipment of a coal-fired power plant pulverizing system, and the operation health state of the coal mill is closely related to the safe and stable operation of the power plant. Raw coal conveyed from a coal yard is subjected to procedures such as screening and coal washing, but various impurities and foreign matters are unavoidably contained, a coal feeder and a coal mill are in a high-load running state for a long time, a coal pulverizing system is in a bad running environment, the working process is complex, various faults of the coal pulverizing system can be caused, the power generation output of a coal-fired power plant can be directly influenced by the faults of the coal pulverizing system, and unnecessary economic loss is caused. Therefore, the monitoring and fault early warning of the running state of the coal mill are of great significance in guaranteeing the safe running of the coal-fired power plant.
In the prior art, most power plants mainly realize part state monitoring by installing sensors at key parts of a coal pulverizing system, for example, parameters of important measuring points such as primary air quantity at an inlet, current of a coal mill, outlet temperature and the like are monitored, on-site operators manually judge the running state of the coal mill according to the numerical value of each sensor displayed by a monitoring system, and the future running trend of the coal mill is analyzed based on experience. Because the coal mill has a plurality of measuring points, the number of the measuring points displayed on the interface of the monitoring system is tens, and operators can hardly grasp the overall operation state and the future change trend of the coal mill in real time through the state early warning of each measuring point.
Therefore, a coal mill fault early warning method based on a PCA and LSTM fusion algorithm is required to be provided, and the problems of high labor cost and low judgment accuracy in the prior art are solved because the coal mill fault early warning is realized by manually judging the running state by field personnel according to the monitoring data of the coal pulverizing system and analyzing the future running trend of the coal mill based on experience.
Disclosure of Invention
In view of the foregoing, it is necessary to provide a method, a device, an electronic device and a computer readable storage medium for early warning of coal mill faults based on a fusion algorithm of PCA and LSTM, so as to solve the technical problems of large workload, high labor cost and low judgment accuracy in the prior art caused by manually judging the running state and the future running trend of the coal mill according to the monitoring data of a pulverizing system by field personnel.
In order to solve the problems, the invention provides a coal mill fault early warning method based on a PCA and LSTM fusion algorithm, which comprises the following steps:
acquiring a plurality of historical operation state parameters of a coal mill in a normal operation working condition time period, and performing principal component analysis on the historical operation state parameters to obtain a coal mill state characterization parameter data set;
Constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the coal mill fault early warning model by utilizing the coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model;
acquiring a plurality of real-time running state parameters of the coal mill, and performing principal component analysis on the real-time running state parameters to obtain real-time state characterization parameters of the coal mill;
and inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early-warning model to obtain a fault early-warning result of the coal mill.
Further, performing principal component analysis on the plurality of historical operating state parameters to obtain a coal mill state characterization parameter data set, including:
obtaining a high-dimensional feature matrix of the coal mill according to the historical operation state parameters;
performing decentration on the high-dimensional feature matrix to obtain a standardized high-dimensional feature matrix of the coal mill;
solving a covariance matrix of the standardized high-dimensionality feature matrix;
calculating eigenvalues of the covariance matrix and eigenvectors corresponding to the eigenvalues;
and sequencing the characteristic values, and performing dimension reduction processing on the real-time running state parameters according to the characteristic vectors corresponding to the sequenced characteristic values to obtain a coal mill state characterization parameter data set.
Further, a coal mill fault early warning model based on a long-term and short-term memory network is constructed, and the method comprises the following steps:
stacking and combining one input layer, two LSTM layers, two Dropout layers and one full-connection layer which are sequentially connected to obtain the coal mill fault early warning model;
the input layer is used for inputting multidimensional input data of a preset time step;
the LSTM layer is used for predicting and obtaining multidimensional output data of a preset output step length according to the multidimensional input data of the preset time step length;
the Dropout layer is used for reducing the overfitting probability;
the full connection layer is used for extracting the associated characteristics of the output data of the Dropout layer through nonlinear change and outputting the output data with preset mapping dimension.
Further, the iteration training is performed on the coal mill fault early warning model by using the coal mill state characterization parameter data set to obtain a coal mill fault early warning model with complete training, and the method comprises the following steps:
dividing the coal mill state characterization parameter data set into a training set and a testing set;
training the coal mill fault early warning model by using the training set, and calculating a loss value and a prediction error of the trained model;
Judging whether the model meets a preset accuracy standard according to the loss value and the prediction error of the trained model;
when the model does not meet the preset accuracy standard, adjusting the super parameters of the model, and continuing training the model until the model meets the preset accuracy standard, so as to obtain a fully trained coal mill fault early warning model;
and determining an early warning threshold value of the fault early warning model of the coal mill with complete training according to the test set.
Further, determining an early warning threshold of the fully trained coal mill fault early warning model according to the test set includes:
inputting the test set into the fully trained coal mill fault early warning model to obtain a residual sequence;
and analyzing the residual sequence by using a mean value-standard deviation control diagram method to obtain upper and lower residual limit values, and setting the upper and lower residual limit values as a coal mill fault early warning threshold value.
Further, inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early warning model to obtain a fault early warning result of the coal mill, including:
inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early warning model to obtain a real-time residual value;
Judging whether the real-time residual value exceeds the fault early warning threshold value of the coal mill;
and when the real-time residual value exceeds the fault early warning threshold value of the coal mill, determining that the coal mill will fail.
Further, before the principal component analysis is performed on the plurality of real-time operation state parameters, the method further includes: preprocessing the real-time running state parameters;
the pretreatment comprises the following steps: and carrying out outlier rejection, missing value completion and normalization processing on the running state parameters.
The invention also provides a coal mill fault early warning device based on the PCA and LSTM fusion algorithm, which comprises:
the data set establishing module is used for acquiring a plurality of historical operation state parameters of the coal mill in a normal operation working condition time period, and carrying out principal component analysis on the historical operation state parameters to obtain a coal mill state characterization parameter data set;
the model training module is used for constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the coal mill fault early warning model by utilizing the coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model;
the real-time data acquisition module is used for acquiring a plurality of real-time running state parameters of the coal mill and carrying out principal component analysis on the real-time running state parameters to obtain real-time state characterization parameters of the coal mill;
And the early warning module is used for inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early warning model to obtain a fault early warning result of the coal mill.
The invention also provides electronic equipment, which comprises a processor and a memory, wherein the memory is stored with a computer program, and when the computer program is executed by the processor, the method for early warning the fault of the coal mill based on the fusion algorithm of PCA and LSTM is realized.
The invention also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for early warning the fault of the coal mill based on the fusion algorithm of PCA and LSTM is realized.
Compared with the prior art, the invention has the beneficial effects that: firstly, performing principal component analysis on historical operating state data of a coal mill under normal operating conditions to obtain a coal mill state characterization parameter data set; secondly, constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the fault early warning model by utilizing a coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model; thirdly, acquiring real-time running state parameters of the coal mill and carrying out principal component analysis on the real-time running parameters to obtain real-time state characterization parameters of the coal mill; and finally, inputting the real-time state characterization parameters into a fully trained coal mill fault early-warning model to obtain a fault early-warning result of the coal mill. According to the method, the main component analysis is adopted to effectively reduce the dimensions of a plurality of variables influencing the safe operation of the coal mill, and the original high-dimensional related variables are converted into low-dimensional uncorrelated variables, so that the time and space complexity of a subsequent deep learning algorithm are reduced; by constructing a fault early warning model based on a long-short-term memory neural network, deep feature extraction is carried out on the relation before and after the time of operation data, and relevant parameters of the faults of the coal mill are accurately and effectively predicted based on related variables, so that the monitoring and fault early warning of the real-time operation state of the coal mill are realized.
Drawings
FIG. 1 is a schematic flow chart of an embodiment of a coal mill fault early warning method based on a PCA and LSTM fusion algorithm provided by the invention;
FIG. 2 is a schematic diagram of one embodiment of a single timing cell of an LSTM network;
FIG. 3 is a schematic diagram of one embodiment of a LSTM network multi-step prediction stacking architecture;
FIG. 4 is a schematic diagram of an embodiment of a data stream transmission structure of a coal mill fault early warning model according to the present invention;
FIG. 5 is a graph of temperature prediction for one embodiment of a coal pulverizer outlet under normal operating conditions provided by the present invention;
FIG. 6 is a regression graph of a loss function for one embodiment of a coal pulverizer fault warning model provided by the present invention;
FIG. 7 is a schematic flow chart diagram of an embodiment of training and real-time monitoring of a coal mill fault early warning model provided by the invention;
FIG. 8 is a schematic diagram showing a prediction contrast of one embodiment of a different machine learning model according to the present invention;
FIG. 9 is a schematic diagram illustrating a part of an embodiment of a residual error warning at a fault time according to the present invention;
FIG. 10 is a schematic structural diagram of an embodiment of a coal pulverizer fault early warning device based on a PCA and LSTM fusion algorithm provided by the invention;
fig. 11 is a schematic structural diagram of an embodiment of an electronic device according to the present invention.
Detailed Description
Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form a part hereof, and together with the description serve to explain the principles of the invention, and are not intended to limit the scope of the invention.
Prior to the description of the embodiments, related terms of the present application will be explained first.
PCA: principal Component Analysis, principal component analysis, is the most widely used data dimension reduction algorithm. The main idea of PCA is to map n-dimensional features onto k-dimensions, which are completely new orthogonal features, also called principal components, and are k-dimensional features reconstructed on the basis of the original n-dimensional features. PCA works by sequentially finding a set of mutually orthogonal axes from the original space, the selection of which is closely related to the data itself.
LSTM: the Long Short-Term Memory network is a recurrent neural network which is further developed on the basis of a recurrent neural network and is widely applied to the prediction of time sequence data.
The embodiment of the invention provides a coal mill fault early warning method based on a PCA and LSTM fusion algorithm, as shown in fig. 1, fig. 1 is a flow diagram of the coal mill fault early warning method based on the PCA and LSTM fusion algorithm, comprising the following steps:
Step S101: acquiring a plurality of historical operation state parameters of a coal mill in a normal operation working condition time period, and performing principal component analysis on the historical operation state parameters to obtain a coal mill state characterization parameter data set;
step S102: constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the coal mill fault early warning model by utilizing the coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model;
step S103: acquiring a plurality of real-time running state parameters of the coal mill, and performing principal component analysis on the real-time running state parameters to obtain real-time state characterization parameters of the coal mill;
step S104: and inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early-warning model to obtain a fault early-warning result of the coal mill.
According to the coal mill fault early warning method based on the PCA and LSTM fusion algorithm, firstly, main component analysis is carried out on historical operation state data of a coal mill under normal operation conditions to obtain a coal mill state characterization parameter data set; secondly, constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the fault early warning model by utilizing a coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model; thirdly, acquiring real-time running state parameters of the coal mill and carrying out principal component analysis on the real-time running parameters to obtain real-time state characterization parameters of the coal mill; and finally, inputting the real-time state characterization parameters into a fully trained coal mill fault early-warning model to obtain a fault early-warning result of the coal mill. The method of the embodiment adopts principal component analysis to effectively reduce the dimensions of a plurality of variables affecting the safety, converts the original high-dimensional related variables into low-dimensional uncorrelated variables, and reduces the time and space complexity of the subsequent deep learning algorithm; by constructing a fault early warning model based on a long-short-term memory neural network, deep feature extraction is carried out on the relation before and after the time of operation data, and relevant parameters of the faults of the coal mill are accurately and effectively predicted based on related variables, so that the monitoring and fault early warning of the real-time operation state of the coal mill are realized.
In order to comprehensively analyze common faults of the coal mill, firstly, operation records and historical data of the coal mill of the coal-fired power plant in the past three years are analyzed and summarized, and working experience of on-site operators and overhaulers is combined to conclude that the common faults of the MPS170 HP-II coal mill are coal blockage and less coal/coal breakage.
The reasons for the coal blocking fault of the coal mill are that the coal feeding amount is overlarge, the moisture of raw coal is higher, the coal quality is not matched with the output, and the like, when the fault occurs, the outlet temperature is usually reduced, the primary air pressure difference is increased, the outlet pressure is reduced, when the blockage is lighter, the current of the coal mill is increased, and when the blockage is serious, the current is reduced; the reasons for the coal-less/coal-breaking faults of the coal mill may be blockage of a coal dropping pipe, blockage of an outlet of the coal feeder, idling of a belt of the coal feeder and the like, and the faults usually cause the influences of the rise of outlet temperature, the reduction of primary air pressure difference, the reduction of current and the like.
According to the analysis, faults such as coal blockage, coal shortage/coal breakage and the like can influence the outlet temperature, the primary air pressure difference, the primary air quantity and the coal mill current, and the method of the embodiment can establish a coal mill fault early warning mechanism based on predicted values of the coal mill outlet temperature, the primary air pressure difference and the primary air quantity at future moments due to the fact that the current value fluctuation is large.
As a specific example, in step S101, based on the mechanism analysis and expert experience, the historical operating state parameters of the normal operation period of the coal mill include 22 parameters related to the failure of the coal mill, as shown in table 1:
table 1 coal mill fault related parameters
| Sequence number | Variable name/unit | Sequence number | Variable name/ |
| 1 | Coal quantity feedback signal/t/h of coal feeder | 12 | Oil temperature/DEGC of oil tank of hydraulic station of coal mill |
| 2 | Current feedback signal/A of coal feeder | 13 | Position feedback/%of inlet air quantity regulating door |
| 3 | Differential pressure between sealing wind and primary wind/kPa | 14 | Inlet damper position feedback/% |
| 4 | Air-powder mixture pressure at outlet/kPa | 15 | Inlet primary air pressure/ |
| 5 | Inlet primary air flow/T/H | 16 | Motor bearing temperature/°c |
| 6 | Inlet primary air temperature/°c | 17 | Motor coil temperature/°c |
| 7 | Primary air inlet/outlet pressure difference/kPa | 18 | Speed reducer input shaft bearing temperature/°c |
| 8 | Coal mill current/A | 19 | Speed reducer gear box oil sump oil temperature/°c |
| 9 | Outlet temperature/° |
20 | Oil temperature/°c of oil groove of thrust bearing of |
| 10 | Milling oil pressure/MPa | 21 | Oil pressure/MPa before oil distributor |
| 11 | Grinding roller reaction force oil pressure/MPa | 22 | Raw coal bin level |
As can be seen from table 1, the number of parameters related to the failure of the coal mill is numerous, which inevitably increases the time and space complexity of the subsequent deep learning algorithm, so that it is necessary to recombine the indexes with certain correlation into a new set of mutually independent comprehensive indexes to replace the original indexes, and the information of the original variables is reflected as much as possible by fewer comprehensive variables.
As a preferred embodiment, in step S101, the principal component analysis is performed on the plurality of historical operating state parameters to obtain a coal mill state characterization parameter data set, including:
obtaining a high-dimensional feature matrix of the coal mill according to the historical operation state parameters;
performing decentration on the high-dimensional feature matrix to obtain a standardized high-dimensional feature matrix of the coal mill;
solving a covariance matrix of the standardized high-dimensionality feature matrix;
calculating eigenvalues of the covariance matrix and eigenvectors corresponding to the eigenvalues;
and sequencing the characteristic values, and performing dimension reduction processing on the real-time running state parameters according to the characteristic vectors corresponding to the sequenced characteristic values to obtain a coal mill state characterization parameter data set.
As a specific example, a specific procedure for performing Principal Component Analysis (PCA) of the operation state parameters is described below with a formula.
The method reduces an n-dimensional feature matrix into a k-dimensional feature matrix, and comprises the following steps:
the first step: decentralizing (removing the mean value) and normalizing all the characteristics; firstly, obtaining the average value of elements in an n-dimensional feature matrix
Subtracting the corresponding mean value from each element to obtain a standardized high-dimensional feature matrix;
And a second step of: solving a covariance matrix C; assuming the data dimension is n, the covariance represents the degree of correlation of the two variables, x, y being positively correlated if x increases and negatively correlated if y increases. After the eigenvector matrix is obtained by the covariance matrix, the direction with the maximum variance is determined, and the direction is the direction of the main component, so that the conversion of the coordinate system is completed.
Covariance matrix:
wherein covariance:
and a third step of: the eigenvalue λ of the covariance matrix C and the corresponding eigenvector u (one eigenvector for each eigenvalue): the formula is: cu=λu, for N eigenvalues λ, each λ i Corresponding feature vector u i The feature value lambda is sorted from large to small to select the top k of the largest.
Fourth step: projecting the original features onto the selected feature vectors to obtain k-dimensional new features; and selecting the first k maximum eigenvalues and corresponding eigenvectors to perform data projection, thereby realizing the reduction of the n-dimensional eigenvalue matrix into a k-dimensional eigenvalue matrix.
For each sample X i The original characteristics are that
The new feature after projection is
The calculation formula is as follows:
the dimension reduction process is further described below with reference to examples, wherein normal operation data of the coal mill is selected and input into the PCA model, and low-dimension feature vectors are obtained to represent dimension data.
As a specific embodiment, the high-dimensional data { x (1), x (2), x (3), x (4), x (5) & x (22) & gt of the coal mill is first subjected to decentralization, so that the origin of coordinates is in the data center, covariance is calculated according to row vectors by using a numpy library cov function in python, the eigenvalue of the covariance matrix C and the corresponding eigenvector matrix R are obtained, the PCA model outputs eigenvalues from large to small and related eigenvector eigenvectors, and finally 14-dimensional high-dimensional data is simplified into 6-dimensional data { x } ～ (1),x ～ (2),x ～ (3),x ～ (4),x ～ (5),x ～ (6) From the variance contribution ratio, the feature that the cumulative contribution ratio of the principal component has exceeded 95% can be seen, as shown in table 2. That is, more than 95% of the features of the previous 14-dimensional data can be demonstrated with the current 6-dimensional data.
TABLE 2PCA dimension reduction eigenvalues and variance contribution rates
The original data vector after the dimension reduction has the following characteristics:
(1) The information amount is measured only by variance and is not affected by factors other than the data set.
(2) The main components are orthogonal, so that the factors of interaction among the original data components of the coal mill can be eliminated.
(3) The feature value decomposition method is fast in dimension reduction, simple and efficient in processing of the original data, and convenient for subsequent LSTM model processing.
As a specific embodiment, in step S102, a coal mill fault early warning model is constructed based on a long and short term memory network. The long-term memory network and the short-term memory network are the memory networks which are good at handling long-time dependence in deep learning, the core of the long-term memory network is a special structure, as shown in fig. 2, fig. 2 shows the structure of an LSTM network, and the calculation formula of each LSTM unit is as follows:
f t ＝σ(W f ×[h t-1 ,x t ]+b f ) (1)
i t ＝σ(W i ×[h t-1 ,x t ]+b i ) (2)
o t ＝σ(W o ×[h t-1 ,x t ]+b o ) (5)
h t ＝o t ×tanh(C t ) (6)
Wherein, the forgetting threshold is represented by i t Representing the input threshold value of the input signal,
representing cell information, C at the last time t Indicating the state of the cell (where the cycle takes place), o t Represents the output threshold, h t Represents the output of the current cell, h t-1 Representing the output of the last time cell.
As shown in fig. 3, fig. 3 is a diagram of an LSTM multi-step predictive stacking architecture, which expresses the specific structure of data expansion along time steps. The current moment is marked as the t moment, and the specific output of the LSTM neural network predicted value depends on the historical values of all relevant characteristic indexes in the previous period, namely the values of the characteristic principal components after the dimension reduction in the last moment.
Assuming that the outlet temperature of the coal mill is selected as a prediction index of the LSTM neural network, the input of the LSTM multi-step prediction model is denoted as X= { X 1 ,X 2 ,X 3 ,X 4 ,...,X t For designing LSTM neural network input as multi-step long structure, setting time step as S and output step as L, then for input
Wherein F is i Representing principal component values after dimension reduction of 6 PCAs at moment i, the model outputs T= { T t+1 ,T t+2 ,...,T t+L And the predicted value of the outlet temperature of the coal mill from 1 to L after the t moment is shown.
The number of layers and the number of neurons of the neural network can influence the model accuracy, and in general, the number of layers and the number of neurons can be increased to improve the model fitting effect, but the fitting phenomenon can occur.
In order to achieve the best effect of the model, as a preferred embodiment, the construction of the coal mill fault early warning model based on the long-term and short-term memory network comprises the following steps:
stacking and combining one input layer, two LSTM layers, two Dropout layers and one full-connection layer which are sequentially connected to obtain the coal mill fault early warning model;
the input layer is used for inputting multidimensional input data of a preset step time step;
the LSTM layer is used for predicting and obtaining multidimensional output data of a preset output step length according to the multidimensional input data of the preset time step length;
the Dropout layer is used for reducing the overfitting probability;
the full connection layer is used for extracting the associated characteristics of the output data of the Dropout layer through nonlinear change and outputting the output data with preset mapping dimension.
As a specific embodiment, as shown in fig. 4, fig. 4 is a schematic diagram of a data flow transmission structure of the fault early warning model of the coal mill. To construct an input data format conforming to the LSTM model, data needs to be converted into 3 dimensions when flowing, where the dimensions of the input layer data are (None, 24, 6), none is the size of the batch-size, 24 is the number of input steps, and 6 is the feature dimension. The number of LSTM Layer neurons is set to 5, and 2 Dropout layers are used for reducing the overfitting probability, and after the LSTM layers are arranged, the Dropout rate is set to 0.2. And the Dense layer is used as an output layer to extract the correlation characteristics of the previous layer of data through nonlinear change, and finally, the dimension result is mapped to an output space to output (None, 1) dimension data.
As a preferred embodiment, the iterative training is performed on the coal mill fault early warning model by using the coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model, which includes:
dividing the coal mill state characterization parameter data set into a training set and a testing set;
training the coal mill fault early warning model by using the training set, and calculating a loss value and a prediction error of the trained model;
judging whether the model meets a preset accuracy standard according to the loss value and the prediction error of the trained model;
when the model does not meet the preset accuracy standard, adjusting the super parameters of the model, and continuing training the model until the model meets the preset accuracy standard, so as to obtain a fully trained coal mill fault early warning model;
and determining an early warning threshold value of the fault early warning model of the coal mill with complete training according to the test set.
As a preferred embodiment, determining the early warning threshold of the fully trained coal mill fault early warning model according to the test set includes:
inputting the test set into the fully trained coal mill fault early warning model to obtain a residual sequence;
And analyzing the residual sequence by using a mean value-standard deviation control diagram method to obtain upper and lower residual limit values, and setting the upper and lower residual limit values as a coal mill fault early warning threshold value.
As a specific example, when model training is performed, the batch processing size is set to 128, the iteration number epoch is set to 100, the learning rate is continuously attenuated, the initial value thereof is set to 0.01,
and (3) performing model error verification by using the test set data, and evaluating the training model based on the root mean square error and the decision coefficient. Training and optimizing by using an Adam optimization algorithm until the model loss is minimum, and finally obtaining and storing a coal mill fault early warning model with the best prediction effect by continuously adjusting and optimizing the super parameters of the LSTM neural network.
And (3) calculating a residual sequence between an output prediction result and an actual value after the test set sequence is input into the coal mill fault early warning model, analyzing the obtained residual sequence by using a mean value-standard deviation control diagram method to obtain residual upper and lower limit values of corresponding parameters, and setting the residual upper and lower limit values as early warning thresholds of the coal mill. The residual threshold algorithm formulas are shown in formulas (7) and (8), wherein mu is a sequence mean value, and sigma is a sequence standard deviation.
Early warning threshold:
UCL＝μ+3·σ (7)
residual threshold:
LCL＝μ-3·σ (8)
the following describes a specific training procedure by means of a specific practical example:
the historical data of 18006 coal mills of the unit of 0 th day of 16 th year of 2021 to 2 nd day of 4 th year of 2021 are selected for training and testing, wherein the data of 67% of the training set before selection is 12045, and the data of the test set after selection is 5932. After 100 iterations, the model is basically and accurately fitted, the predicted result and the residual result of the historical data are shown in fig. 5, and the predicted data are basically matched with the historical actual measurement data. The regression curves of the test set and the training set loss function are shown in fig. 6, and it can be seen that the fitting effect of the model is better along with the increase of training, the loss values of the training set and the test set are gradually reduced, and the loss of the test set is not obviously reduced after the 50 th training. The final PCA-LSTM temperature prediction model was found to have a coefficient of 0.982 and a root mean square error of 0.320.
In order to realize the early warning of the future running state of the coal mill, besides selecting the outlet temperature of the coal mill as the prediction index of the coal mill fault early warning model, a coal mill fault early warning mechanism can be established based on the obtained prediction values of the outlet temperature of the coal mill, the primary air pressure difference and the primary air volume at the future moment.
Because the working environment of the pulverizing system is bad, the sensor additionally arranged on the equipment has the possibility of sudden failure, and the network fluctuation of the data transmission in the factory can also cause the loss or abnormality of the data. In order to improve the validity of the data, as a preferred embodiment, before performing principal component analysis on a plurality of the real-time operation state parameters, the method further includes: preprocessing the real-time running state parameters;
the pretreatment comprises the following steps: and carrying out outlier rejection, missing value completion and normalization processing on the running state parameters.
As a specific example, an outlier is primarily referred to as a significant abnormal rise or fall beyond the limit or value of the measured value. Because the data set has the temperature, flow and other slow-change signals and current and other instantaneous signals, the abnormal value judging method corresponding to the different types of measuring points is set. The method adopted by the missing value complementation is mean value filling, and the dimension of the processed data is 14-dimensional data.
As a specific example, for the coal mill outlet temperature values, since each coal mill has three outlet temperature measurement points, the following is specified at the time of actual measurement: if the limit value of the outlet temperatures of the three coal mills is less than 1 ℃, taking the average value of the three outlet temperatures as the outlet temperature of the coal mill; if the difference is greater than 1 ℃, the intermediate value of the three temperature values is selected as the outlet temperature of the coal mill.
As a preferred embodiment, inputting the real-time state characterization parameter of the coal mill into the well-trained coal mill fault early-warning model to obtain the fault early-warning result of the coal mill, including:
inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early warning model to obtain a real-time residual value;
judging whether the real-time residual value exceeds the fault early warning threshold value of the coal mill;
and when the real-time residual value exceeds the fault early warning threshold value of the coal mill, determining that the coal mill will fail.
As a specific embodiment, the method for setting the threshold value of the mean value-standard deviation control chart is used for setting the threshold value of actual operation parameters of coal mills with different loads, establishing the upper limit and the lower limit of the threshold value under the normal working condition of the coal mill, adjusting the fault early-warning model of the coal mill under different loads, carrying out residual analysis on the temperature predicted value and the actual value of the fault early-warning model of the coal mill, and setting the early-warning threshold value.
As a specific embodiment, the real-time data is input into the established coal mill fault early warning model after the PCA is used for dimension reduction, the predicted value and the real-time value residual error are judged, if the predicted value and the real-time value residual error exceed the threshold value, the fault is diagnosed, and otherwise, the equipment is in a normal state.
As shown in fig. 7, fig. 7 shows an overall flowchart of the method of the present embodiment from the model building and training process to fault pre-warning of real-time data using a fully trained coal mill fault pre-warning model (PCA-LSTM neural network model).
It can be understood that besides the early warning of the coal breakage fault of the coal with less coal, the method can also early warn the faults of coal blockage of the coal mill, abrasion of the grinding roller of the coal mill, abnormal lubricating oil pressure and the like. The fault judgment is mainly realized according to the predicted value and the actual value residual error of fault-associated measuring points such as an outlet temperature value, a primary air quantity value, a primary air pressure difference value, a grinding oil pressure value and the like.
In order to demonstrate the effect of the invention, the fault early warning model of the invention is compared with three common neural networks.
As a specific embodiment, the coal mill fault early warning model (expressed by PCA-LSTM), LSTM, RNN and traditional BP neural network are respectively used for carrying out modeling experiments on the same data set, the four neural networks all use a double-layer hidden layer structure, grid super parameters are basically controlled to be consistent, the final prediction result is shown in figure 8, the evaluation indexes of different models are shown in table 3, and R is shown in the table 3 2 The influence caused by different dimensions is avoided, and the accuracy is higher when the model is closer to 1.
As apparent from FIG. 8 and Table 3, the PCA-LSTM neural network has better prediction accuracy than other deep learning models, and R thereof 2 Reaching 0.982, rmse is only 0.320, bp neural network is less effective, presumably it is not possible to perform better analysis of time series data.
TABLE 3 comparison of evaluation indexes of different deep learning models
| Model name | R 2 | RMSE |
| BP neural network | 0.127 | 3.149 |
| RNN neural network | 0.850 | 1.416 |
| LSTM multi-step predictive model | 0.907 | 1.287 |
| PCA-LSTM prediction model | 0.982 | 0.320 |
To further illustrate the practicability of the fault early warning of the present invention, the fault early warning effect of the present invention is shown by a specific embodiment
As a specific embodiment, early warning experiments are carried out by combining the data of the related faults of the coal loss of the medium-speed coal mill of a certain 350MW thermal power generating unit. The inspection of the power plant related heat engine maintenance report can be as follows: 2021, 4, 19, 00:03: the 00 mill outlet temperature rises rapidly causing tripping due to the coal mill belt slipping causing less coal. The data division training set of the data of the year 2021, the month 4, the month 17, the year 2021, the month 4, the day 19 and the test set are input into the application to establish a PCA-LSTM multi-step prediction model, and after the application of the PCA-LSTM model and the threshold method are fused, a diagnosis result shows that the algorithm can send out an early warning signal at the 8384 th sample point of the test set, namely the year 2021, the month 4, the day 19, the 00:01:10, and the SIS system can send out an early warning signal at the 5358 th sample point, namely the SIS system 110s can identify a fault signal in advance. As shown in fig. 9, fig. 9 shows a partial diagram of the residual warning at the time of failure.
The embodiment also provides a coal mill fault early warning device based on the PCA and LSTM fusion algorithm, as shown in fig. 10, the coal mill fault early warning device 1000 based on the PCA and LSTM fusion algorithm includes:
the data set establishing module 1001 is configured to obtain a plurality of historical operating state parameters of a normal operating condition period of the coal mill, and perform principal component analysis on the plurality of historical operating state parameters to obtain a coal mill state characterization parameter data set;
the model training module 1002 is configured to construct a coal mill fault early warning model based on a long-short-term memory network, and perform iterative training on the coal mill fault early warning model by using the coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model;
the real-time data acquisition module 1003 is configured to acquire a plurality of real-time operation state parameters of the coal mill, and perform principal component analysis on the plurality of real-time operation state parameters to obtain real-time state characterization parameters of the coal mill;
and the early warning module 1004 is configured to input the real-time state characterization parameter of the coal mill into the fully trained coal mill fault early warning model to obtain a fault early warning result of the coal mill.
As shown in fig. 1100, the present invention further provides an electronic device 1100, which may be a mobile terminal, a desktop computer, a notebook computer, a palm computer, a server, and other computing devices. The electronic device includes a processor 1101, a memory 1102, and a display 1103.
The processor 1101 may be, in some embodiments, a central processing unit (Central Processing Unit, CPU), microprocessor or other data processing chip for executing program code or processing data stored in the memory 1102, such as a coal pulverizer fault warning method program based on a PCA and LSTM fusion algorithm.
The display 1103 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, or the like in some embodiments. The display 1103 is for displaying information on a computer device and for displaying a visual user interface. The components 1101-1103 of the computer devices communicate with each other through a system bus.
The embodiment also provides a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when the computer program is executed by a processor, the method for early warning of the fault of the coal mill based on the PCA and LSTM fusion algorithm is realized.
According to the computer readable storage medium and the computing device provided by the embodiments of the present invention, the specific description of the method for early warning the failure of the coal mill based on the fusion algorithm of PCA and LSTM can be referred to, and the method has similar advantages as the method for early warning the failure of the coal mill based on the fusion algorithm of PCA and LSTM, and is not repeated herein.
The invention discloses a coal mill fault early warning method, a device, electronic equipment and a computer readable storage medium based on a PCA and LSTM fusion algorithm, which are characterized in that firstly, main component analysis is carried out on historical operation state data of a coal mill under normal operation conditions to obtain a coal mill state characterization parameter data set; secondly, constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the fault early warning model by utilizing a coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model; thirdly, acquiring real-time running state parameters of the coal mill and carrying out principal component analysis on the real-time running parameters to obtain real-time state characterization parameters of the coal mill; and finally, inputting the real-time state characterization parameters into a fully trained coal mill fault early-warning model to obtain a fault early-warning result of the coal mill.
According to the method, the main component analysis is adopted to effectively reduce the dimensions of a plurality of variables affecting the safety, and the original high-dimensional related variables are converted into low-dimensional uncorrelated variables, so that the time and space complexity of a subsequent deep learning algorithm are reduced; by constructing a fault early warning model based on a long-short-term memory neural network, deep feature extraction is carried out on the relation before and after the time of operation data, and relevant parameters of the faults of the coal mill are accurately and effectively predicted based on related variables, so that the monitoring and fault early warning of the real-time operation state of the coal mill are realized.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention.
Claims (10)
1. A coal mill fault early warning method based on a PCA and LSTM fusion algorithm is characterized by comprising the following steps of;
acquiring a plurality of historical operation state parameters of a coal mill in a normal operation working condition time period, and performing principal component analysis on the historical operation state parameters to obtain a coal mill state characterization parameter data set;
constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the coal mill fault early warning model by utilizing the coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model;
acquiring a plurality of real-time running state parameters of the coal mill, and performing principal component analysis on the real-time running state parameters to obtain real-time state characterization parameters of the coal mill;
and inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early-warning model to obtain a fault early-warning result of the coal mill.
2. The coal mill fault early warning method based on the fusion algorithm of PCA and LSTM according to claim 1, wherein the main component analysis is performed on the plurality of historical operating state parameters to obtain a coal mill state characterization parameter data set, comprising:
obtaining a high-dimensional feature matrix of the coal mill according to the historical operation state parameters;
performing decentration on the high-dimensional feature matrix to obtain a standardized high-dimensional feature matrix of the coal mill;
solving a covariance matrix of the standardized high-dimensionality feature matrix;
calculating eigenvalues of the covariance matrix and eigenvectors corresponding to the eigenvalues;
and sequencing the characteristic values, and performing dimension reduction processing on the real-time running state parameters according to the characteristic vectors corresponding to the sequenced characteristic values to obtain a coal mill state characterization parameter data set.
3. The coal mill fault early warning method based on the PCA and LSTM fusion algorithm according to claim 1, wherein the construction of the coal mill fault early warning model based on the long-term memory network comprises the following steps:
stacking and combining one input layer, two LSTM layers, two Dropout layers and one full-connection layer which are sequentially connected to obtain the coal mill fault early warning model;
The input layer is used for inputting multidimensional input data of a preset time step;
the LSTM layer is used for predicting and obtaining multidimensional output data of a preset output step length according to the multidimensional input data of the preset time step length;
the Dropout layer is used for reducing the overfitting probability;
the full connection layer is used for extracting the associated characteristics of the output data of the Dropout layer through nonlinear change and outputting the output data with preset mapping dimension.
4. The coal mill fault early warning method based on the fusion algorithm of PCA and LSTM according to claim 1, wherein the iterative training of the coal mill fault early warning model by using the coal mill state characterization parameter data set is performed to obtain a fully trained coal mill fault early warning model, comprising:
dividing the coal mill state characterization parameter data set into a training set and a testing set;
training the coal mill fault early warning model by using the training set, and calculating a loss value and a prediction error of the trained model;
judging whether the model meets a preset accuracy standard according to the loss value and the prediction error of the trained model;
when the model does not meet the preset accuracy standard, adjusting the super parameters of the model, and continuing training the model until the model meets the preset accuracy standard, so as to obtain a fully trained coal mill fault early warning model;
And determining an early warning threshold value of the fault early warning model of the coal mill with complete training according to the test set.
5. The coal mill fault pre-warning method based on a PCA and LSTM fusion algorithm of claim 4, wherein determining pre-warning thresholds of the trained complete coal mill fault pre-warning model from the test set comprises:
inputting the test set into the fully trained coal mill fault early warning model to obtain a residual sequence;
and analyzing the residual sequence by using a mean value-standard deviation control diagram method to obtain upper and lower residual limit values, and setting the upper and lower residual limit values as a coal mill fault early warning threshold value.
6. The method for early warning of a coal mill fault based on a fusion algorithm of PCA and LSTM according to claim 5, wherein the step of inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early warning model to obtain the fault early warning result of the coal mill comprises the following steps:
inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early warning model to obtain a real-time residual value;
judging whether the real-time residual value exceeds the fault early warning threshold value of the coal mill;
And when the real-time residual value exceeds the fault early warning threshold value of the coal mill, determining that the coal mill will fail.
7. The coal mill fault pre-warning method based on the fusion algorithm of PCA and LSTM according to claim 1, wherein before the main component analysis is performed on the plurality of real-time operation state parameters, the method further comprises: preprocessing the real-time running state parameters;
the pretreatment comprises the following steps: and carrying out outlier rejection, missing value completion and normalization processing on the running state parameters.
8. The utility model provides a coal pulverizer trouble early warning device based on PCA and LSTM fusion algorithm which characterized in that includes:
the data set establishing module is used for acquiring a plurality of historical operation state parameters of the coal mill in a normal operation working condition time period, and carrying out principal component analysis on the historical operation state parameters to obtain a coal mill state characterization parameter data set;
the model training module is used for constructing a coal mill fault early warning model based on a long-short-term memory network, and performing iterative training on the coal mill fault early warning model by utilizing the coal mill state characterization parameter data set to obtain a fully trained coal mill fault early warning model;
The real-time data acquisition module is used for acquiring a plurality of real-time running state parameters of the coal mill and carrying out principal component analysis on the real-time running state parameters to obtain real-time state characterization parameters of the coal mill;
and the early warning module is used for inputting the real-time state characterization parameters of the coal mill into the fully trained coal mill fault early warning model to obtain a fault early warning result of the coal mill.
9. An electronic device, comprising a processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements a coal mill fault pre-warning method based on a PCA and LSTM fusion algorithm as claimed in any one of claims 1-7.
10. A computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when the computer program is executed by a processor, the method for early warning of coal mill faults based on a fusion algorithm of PCA and LSTM is realized according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310334054.0A CN116383636A (en) | 2023-03-30 | 2023-03-30 | Coal mill fault early warning method based on PCA and LSTM fusion algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310334054.0A CN116383636A (en) | 2023-03-30 | 2023-03-30 | Coal mill fault early warning method based on PCA and LSTM fusion algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116383636A true CN116383636A (en) | 2023-07-04 |
Family
ID=86972614
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310334054.0A Pending CN116383636A (en) | 2023-03-30 | 2023-03-30 | Coal mill fault early warning method based on PCA and LSTM fusion algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116383636A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117172232A (en) * | 2023-11-02 | 2023-12-05 | 深圳市迪博企业风险管理技术有限公司 | Audit report generation method, audit report generation device, audit report generation equipment and audit report storage medium |
| CN117312965A (en) * | 2023-11-30 | 2023-12-29 | 国网辽宁省电力有限公司 | Unknown fault self-learning method for GIS high-voltage isolating switch |
| CN117404678A (en) * | 2023-10-23 | 2024-01-16 | 国能长源荆门发电有限公司 | Online coal blending system for relieving slag formation of high-alkali coal burned by boiler |
| CN117491793A (en) * | 2023-12-29 | 2024-02-02 | 国网浙江省电力有限公司电力科学研究院 | Hydrogen electric coupling system comprehensive performance test method, device and medium |
| CN117992859A (en) * | 2024-04-03 | 2024-05-07 | 华侨大学 | Early-stage fault early-warning and identifying method and device for electromechanical equipment provided with SCADA system |
-
2023
- 2023-03-30 CN CN202310334054.0A patent/CN116383636A/en active Pending
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117404678A (en) * | 2023-10-23 | 2024-01-16 | 国能长源荆门发电有限公司 | Online coal blending system for relieving slag formation of high-alkali coal burned by boiler |
| CN117404678B (en) * | 2023-10-23 | 2024-05-31 | 国能长源荆门发电有限公司 | Online coal blending system for relieving slag formation of high-alkali coal burned by boiler |
| CN117172232A (en) * | 2023-11-02 | 2023-12-05 | 深圳市迪博企业风险管理技术有限公司 | Audit report generation method, audit report generation device, audit report generation equipment and audit report storage medium |
| CN117172232B (en) * | 2023-11-02 | 2024-01-26 | 深圳市迪博企业风险管理技术有限公司 | Audit report generation method, audit report generation device, audit report generation equipment and audit report storage medium |
| CN117312965A (en) * | 2023-11-30 | 2023-12-29 | 国网辽宁省电力有限公司 | Unknown fault self-learning method for GIS high-voltage isolating switch |
| CN117491793A (en) * | 2023-12-29 | 2024-02-02 | 国网浙江省电力有限公司电力科学研究院 | Hydrogen electric coupling system comprehensive performance test method, device and medium |
| CN117491793B (en) * | 2023-12-29 | 2024-05-10 | 国网浙江省电力有限公司电力科学研究院 | Hydrogen electric coupling system comprehensive performance test method, device and medium |
| CN117992859A (en) * | 2024-04-03 | 2024-05-07 | 华侨大学 | Early-stage fault early-warning and identifying method and device for electromechanical equipment provided with SCADA system |
| CN117992859B (en) * | 2024-04-03 | 2024-06-07 | 华侨大学 | Early-stage fault early-warning and identifying method and device for electromechanical equipment provided with SCADA system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116383636A (en) | Coal mill fault early warning method based on PCA and LSTM fusion algorithm | |
| CN110441065B (en) | Gas turbine on-line detection method and device based on LSTM | |
| CN110320892B (en) | Sewage treatment equipment fault diagnosis system and method based on L asso regression | |
| CN108320043B (en) | Power distribution network equipment state diagnosis and prediction method based on electric power big data | |
| CN110738274A (en) | nuclear power device fault diagnosis method based on data driving | |
| CN110763929A (en) | Intelligent monitoring and early warning system and method for convertor station equipment | |
| CN111537219B (en) | Fan gearbox performance detection and health assessment method based on temperature parameters | |
| KR101178235B1 (en) | Prediction and fault detection method and system for performance monitoring of plant instruments using principal component analysis, response surface method, Fuzzy Support Vector Regression and Generalized Likelihood Ratio Test | |
| CN110083860B (en) | Industrial fault diagnosis method based on relevant variable selection | |
| CN110490496B (en) | Method for screening sensitive variables influencing product quality in complex industrial process based on stepwise reduction | |
| CN108921230A (en) | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network | |
| CN113036913B (en) | Method and device for monitoring state of comprehensive energy equipment | |
| CN111797533B (en) | Nuclear power device operation parameter abnormity detection method and system | |
| CN108764305A (en) | A kind of improved colony intelligence machine learning fault diagnosis system | |
| CN111340110A (en) | Fault early warning method based on industrial process running state trend analysis | |
| CN111273125A (en) | RST-CNN-based power cable channel fault diagnosis method | |
| CN115730191A (en) | Attention mechanism-based coal mill fault early warning method | |
| CN114519923A (en) | Intelligent diagnosis and early warning method and system for power plant | |
| CN112036087A (en) | Multi-strategy fused nuclear power key equipment fault diagnosis method and system | |
| CN114997309A (en) | Water feed pump fault early warning method and device | |
| CN201035376Y (en) | Failure diagnosis device under small sample conditional in the process of manufacturing production | |
| CN105741184B (en) | Transformer state evaluation method and device | |
| CN116028849B (en) | Emulsion pump fault diagnosis method based on depth self-coding network | |
| CN112036496A (en) | Nuclear power device fault diagnosis method and system | |
| Raptodimos et al. | An artificial neural network approach for predicting the performance of ship machinery equipment |
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 |