CN115879369A - Coal mill fault early warning method based on optimized LightGBM algorithm - Google Patents
Coal mill fault early warning method based on optimized LightGBM algorithm Download PDFInfo
- Publication number
- CN115879369A CN115879369A CN202211403260.4A CN202211403260A CN115879369A CN 115879369 A CN115879369 A CN 115879369A CN 202211403260 A CN202211403260 A CN 202211403260A CN 115879369 A CN115879369 A CN 115879369A
- Authority
- CN
- China
- Prior art keywords
- early warning
- parameter
- lightgbm
- optimized
- 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.)
- Pending
Links
- 239000003245 coal Substances 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 32
- 230000009467 reduction Effects 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 239000002245 particle Substances 0.000 claims description 12
- 238000005304 joining Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000001133 acceleration Effects 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 5
- 238000004804 winding Methods 0.000 claims description 4
- YZCKVEUIGOORGS-IGMARMGPSA-N Protium Chemical compound [1H] YZCKVEUIGOORGS-IGMARMGPSA-N 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 229910052731 fluorine Inorganic materials 0.000 claims description 2
- 125000001153 fluoro group Chemical group F* 0.000 claims description 2
- 238000012545 processing Methods 0.000 abstract description 7
- 238000005457 optimization Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 7
- 238000004590 computer program Methods 0.000 description 6
- 238000003860 storage Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 230000007774 longterm Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 2
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000010298 pulverizing process Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000010977 unit operation Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a coal mill fault early warning method based on an optimized LightGBM algorithm, which comprises the steps of collecting historical operation data of parameters of a coal mill from an SIS system; carrying out data preprocessing on the original data, and carrying out data dimension reduction through a Pearson coefficient method; constructing an optimized LightGBM parameter prediction model in a normal state of the coal mill; calculating a predicted value of the early warning parameter and a residual average value thereof as a reference value of the early warning threshold, and setting an early warning threshold range according to various boundary conditions on the basis of the reference value; and (4) periodically acquiring actual operation data and inputting the actual operation data into the trained optimized LightGBM parameter prediction model to obtain an early warning parameter residual error average value in an actual operation state, judging whether the actual operation data exceeds the early warning threshold range, and sending an early warning signal once the actual operation data exceeds the early warning threshold range. The method has the characteristics of high training speed, low memory occupation and high accuracy; according to the method and the device, the main hyper-parameters of the LightGBM model are automatically optimized by adopting the SSA-PSO optimization algorithm, the trouble of manual parameter adjustment is avoided, and the efficiency of the model in processing a large amount of data is improved.
Description
Technical Field
The application relates to the technical field of coal-fired units, in particular to a coal mill fault early warning method based on optimized LightGBM algorithm.
Background
The coal mill is used as a large auxiliary equipment in a coal-fired unit, is a key component of a unit pulverizing system, and the operation condition of the coal mill can influence the operation efficiency of the whole generator set, so that research on the fault of the coal mill is necessary.
Most of the traditional coal-fired unit equipment fault early warning is that single or multiple parameter early warning values are set for early warning according to mechanism modeling or practical experience. The traditional fault early warning method has the problems of false alarm and missed alarm in the ordinary early warning process, and the originally set early warning threshold value is not applicable any more after the self mechanism of the equipment is changed along with the increase of the service time of the equipment, so that the early warning threshold value needs to be recalculated and set.
On the other hand, most domestic coal-fired units accumulate massive historical operating data since the start of production, but most of the operating data are still not effectively utilized. If the running states of various devices can be analyzed and predicted by methods such as data driving and the like, the efficiency of fault early warning of the conventional coal-fired unit can be greatly improved. Therefore, the fault early warning research of the coal-fired unit equipment based on intelligent algorithms such as machine learning and the like is needed.
Disclosure of Invention
Based on the above technical problem, a coal mill fault early warning method based on optimized LightGBM algorithm is provided, and the problems that the early warning threshold value set artificially in the prior art cannot meet the long-term operation requirement and the historical operation data is not effectively utilized are solved.
In order to achieve the above object, the present application provides the following technical solutions:
a coal mill fault early warning method based on an optimized LightGBM algorithm comprises the following steps:
s1, collecting original historical operation data of operation parameters of a coal mill in a normal operation state from an SIS (Small information System);
s2, preprocessing the original historical operation data, and performing data dimension reduction on the preprocessed historical operation data through a Pearson coefficient;
s3, selecting data from historical operating data obtained after data dimensionality reduction according to a preset number, inputting the selected data serving as a training set into a LightGBM model, and training to obtain a LightGBM parameter prediction model in a normal operating state of the coal mill; optimizing the LightGBM parameter prediction model by utilizing an SSA-PSO algorithm to obtain an optimized LightGBM parameter prediction model;
s4, calculating by using the optimized LightGBM parameter prediction model to obtain an early warning parameter prediction value, and calculating to obtain a residual average value of the early warning parameter prediction value; taking the residual average value of the early warning parameter predicted value as an early warning threshold reference value, and setting an early warning threshold range according to the early warning threshold reference value and various boundary conditions;
step S5, regularly collecting actual operation data of the coal mill, inputting the actual operation data into the optimized LightGBM parameter prediction model after data dimension reduction, calculating to obtain an early warning parameter residual error average value in an actual operation state, and judging whether the early warning parameter residual error average value exceeds the early warning threshold range; and outputting an early warning prompt when the average value of the early warning parameter residuals exceeds the early warning threshold range.
Optionally, the operating parameters include hot blast door opening, cold blast door opening, primary wind pressure, inlet temperature, inlet flow, gearbox input bearing temperature, front axle temperature, motor winding temperature, rear axle bearing temperature, coal pulverizer current, coal feeder current, and coal feeder coal quantity.
Optionally, the preprocessing the raw historical operating data includes:
rejecting nonunion in the original historical operating data;
eliminating unsteady and noise values in the original historical operating data;
normalizing the original historical operating data by the following specific formula:
wherein the content of the first and second substances,
vector A representing the x-th group of samples in the operating parameter i A value of (d); />
Represents->
Normalizing the result;
represents vector A i The mean value of (a); max (A) i ) Represents vector A i The maximum value of (a); min (A) i ) Represents vector A i Of the measured value (c).
Optionally, rejecting nonnumbers in the original historical operating data by using an isnan function in MATLAB software; and eliminating unsteady and noise values in the original historical operating data by using a 3 sigma criterion.
Optionally, the performing, by using a pearson coefficient method, data dimensionality reduction on the preprocessed historical operating data includes:
according to the historical operation data obtained after preprocessing, calculating the Pearson coefficient of the early warning parameter and other parameters, wherein the Pearson coefficient has the following calculation formula:
wherein x is i (i =1,2,3,.. Times, n) is the value of some other parameter x in the ith sample, y i For the value of the early warning parameter y in the ith sample,
is the average of some other parameter x over n samples, <' >>
The average value of the early warning parameter y in n samples is shown, and r is a correlation coefficient of some other parameter x and the early warning parameter y;
and selecting historical operating data of a plurality of groups of other parameters with the highest correlation with the early warning parameters according to the preset group number as the input of the step S3.
Optionally, the step S3 specifically includes:
step S301, selecting 300-600 samples as a training set from historical operating data obtained after data dimension reduction, inputting the training set into a LightGBM model, and training to obtain a LightGBM parameter prediction model under a normal operating state of the coal mill;
step S302, initializing the LightGBM parameter prediction model and initializing each hyper-parameter; the hyper-parameters comprise learning rate, tree depth, leaf node number and minimum data number on the leaf;
step S303, setting parameters of an SSA-PSO algorithm; the parameters of the SSA-PSO algorithm comprise a population size N, a dimension d and a maximum iteration number iter max Acceleration factor c 1 And acceleration factor c 2 ；
Step S304, using the hyper-parameters to be optimized as initial particles, setting the maximum position and the minimum position of the particles, initializing the initial position and the speed of each particle, and constructing a current search area;
step S305, calculating the fitness of each particle, and setting X b The individual with the best position in the population is taken as a scout; let X w The individual with the worst position in the population is taken as an adder;
step S306, updating the individuals of the reconnaissance and the participants;
step S307, calculating the optimal individual position;
step S308, judging whether the iteration termination condition of the current search area is met, if so, executing step S309, and if not, returning to step S305;
step S309, judging whether an iteration termination condition of the complete algorithm is met, if so, executing step S310, and if not, returning to step S304;
and S310, outputting a super-parameter optimal combination, and using the super-parameter optimal combination as a parameter of the LightGBM parameter prediction model to obtain an optimized LightGBM parameter prediction model.
Further optionally, in step S306, the scout searches the current search area, and the position update formula is:
wherein the content of the first and second substances,
representing the position of the ith individual scout in the d dimension at the t iteration; α is a random number between 0 and 1; iter max Is the maximum iteration number;
the participator explores according to the guidance of the scout, and the position updating formula is as follows:
v i(d+1) ＝rr·ω·v id +c 1 r 1 (p id -x id )+c 2 r 2 (p gd -x id )
x i(d+1) ＝x id +vi (d+1)
wherein, the first and the second end of the pipe are connected with each other,
the influence factor is used for guiding the individual joining person to update the speed of the individual joining person; omega is a weight coefficient used for adjusting the searching capability of the algorithm; v. of id The velocity of an individual i of an enrollee in d dimension; r is a radical of hydrogen 1 And r 2 Is a random number between 0 and 1; p is a radical of id And p gd Respectively representing an individual optimal solution and a global optimal solution of the individual i of the subscriber on the d dimension; x is the number of id The position of an individual i of the joiner in the d dimension; v. of i(d+1) Updating the speed of the individual i of the subscriber after the updating formula of the corresponding position of the individual i in the dimension d + 1; x is the number of i(d+1) For the enrollee individual i a position x in d +1 dimension according to the previous dimension id And velocity v i(d+1) The updated position.
Further optionally, the iteration termination condition of the current search area is that the maximum iteration number is reached; the iteration termination condition of the complete algorithm is that the maximum iteration times are reached or the optimal solution is not changed;
the maximum number of iterations is 50.
Optionally, the obtaining of the predicted value of the early warning parameter by using the optimized LightGBM parameter prediction model through calculation, and obtaining the average value of the residuals of the predicted value of the early warning parameter through calculation specifically include:
inputting the training set into the optimized LightGBM parameter prediction model, and calculating to obtain an early warning parameter prediction value;
and calculating the residual error between the predicted value and the actual value of the early warning parameter, and calculating the average value of the residual errors of the predicted value of the early warning parameter by a sliding window method.
Further optionally, the periodically acquiring actual operation data of the coal mill, performing data dimension reduction, and inputting the actual operation data into the optimized LightGBM parameter prediction model, and calculating to obtain an average value of residual errors of the early warning parameters in an actual operation state specifically includes:
after actual operation data of the coal mill are collected, selecting multiple groups of actual operation data of other parameters with the highest correlation with the early warning parameters to form an early warning data set, inputting the early warning data set into the optimized LightGBM parameter prediction model to obtain actual early warning parameter predicted values in an actual operation state, and calculating by using a sliding window method to obtain early warning parameter residual error average values in the actual operation state.
The application has at least the following beneficial effects:
1. the embodiment of the application provides a coal mill fault early warning method based on an optimized LightGBM algorithm, the method comprises the steps of firstly preprocessing original data and reducing dimensions of the data, screening a training set and inputting the data into a LightGBM parameter prediction model based on SSAPSO optimization to obtain an early warning parameter residual error, obtaining an early warning threshold reference value after obtaining a residual error average value through calculation by a sliding window method, obtaining a dynamic threshold range by combining various boundary conditions of the coal mill under the current unit working condition, predicting the early warning parameter of the coal mill in the actual operation process and calculating the residual error average value, immediately sending early warning when the residual error average value exceeds the early warning threshold range, and realizing early fault early warning of the coal mill; the method applies the LightGBM (Light Gradient Boosting Machine) algorithm to fault early warning, and has the characteristics of high training speed, low memory occupation and high accuracy; compared with the traditional fault early warning method, the LightGBM algorithm also supports parallel learning, and can ensure the fault prediction precision while processing large-scale data; according to the method and the device, the main hyper-parameters of the LightGBM model are automatically optimized by adopting an SSA-PSO optimization algorithm, the trouble of artificial parameter adjustment is avoided, and the efficiency of the model in processing a large amount of data is improved.
2. Compared with the traditional fault early warning technology, the method and the device have the advantages that historical operation data are utilized, the training set of the parameter prediction model can be updated periodically, the dynamic early warning threshold range is set in cooperation with the current equipment operation condition, and long-term early warning is achieved.
Drawings
Fig. 1 is a schematic flow chart of a coal mill fault early warning method based on optimized LightGBM algorithm according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a residual average value of a predicted value of an early warning parameter of a coal mill under a current working condition according to an embodiment of the application;
FIG. 3 is a schematic diagram of a coal pulverizer fault warning result according to an embodiment of the present application;
fig. 4 is a schematic flow chart of a coal mill fault early warning method based on an optimized LightGBM algorithm according to an embodiment of the present application;
fig. 5 is an internal structural diagram of a computer device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
In one embodiment, as shown in fig. 1, there is provided a coal mill fault early warning method based on optimized LightGBM algorithm, including the following steps:
step S1, collecting original historical operation data of operation parameters of a related coal mill in a normal operation state from an SIS system.
The coal mill is an HP843/Dyn medium-speed coal mill from a unit No. 1 of a certain coal-fired power plant, original historical operating data are from an SIS system of the power plant, and the sampling time interval is 2min. The coal mill operation parameters related in the embodiment of the application comprise the opening degree of a hot blast door, the opening degree of a cold blast door, primary wind pressure, inlet temperature, inlet flow, the input bearing temperature of a gear box, the temperature of a front shaft, the temperature of a motor winding, the temperature of a rear shaft bearing, coal mill current, coal feeder current and coal feeder coal quantity, and the total number of 12 related operation parameters of the medium-speed coal mill; 10000 samples were collected in total.
According to the past fault history of the plant and the fault mechanism analysis of the coal mill, the temperature of a rear shaft bearing of the coal mill is selected as an early warning parameter, and the rest parameters are selected as an input part of a training set according to a subsequent data dimensionality reduction result. It should be noted that, the selection of the operation parameters and the warning parameters is specific to the embodiment, and in other embodiments, the selection may be automatically adjusted according to actual situations.
And S2, preprocessing the original historical operation data, and performing data dimension reduction on the preprocessed historical operation data through a Pearson coefficient method. Data collected from the SIS system are subjected to data preprocessing and Pearson coefficient method dimension reduction processing, and the processed data are used for training a parameter prediction model.
Wherein, the preprocessing of the original historical operating data comprises:
(1) Rejecting nonnumbers (NAN values) in the original historical operating data; specifically, an isnan function in MATLAB software can be used for removing NAN values in original historical operating data;
(2) Eliminating unsteady values and noise values in original historical operating data; specifically, the 3 σ criterion can be utilized to eliminate unsteady and noise values in the original historical operating data;
(3) The original historical operating data is normalized, and the specific formula is as follows:
wherein, the first and the second end of the pipe are connected with each other,
represents the x-th group of samples in the running parameter vector A i A value of (d) above; />
Represents->
Normalizing the result;
represents vector A i The mean value of (a); max (A) i ) Represents vector A i The maximum value of (a); min (A) i ) Represents vector A i Is measured.
In addition, the data dimensionality reduction of the historical operating data obtained after preprocessing by the Pearson coefficient method comprises the following steps:
(1) According to the historical operation data obtained after preprocessing, calculating a Pearson coefficient of the early warning parameter and other parameters, wherein the Pearson coefficient calculation formula is as follows:
wherein x is i (i =1,2,3.., n) is the value of some other parameter x in the ith sample, y i For the value of the early warning parameter y in the ith sample,
is the average of some other parameter x over n samples>
The average value of the early warning parameter y in n samples is shown, and r is a correlation coefficient of some other parameter x and the early warning parameter y;
(2) And selecting historical operating data of a plurality of groups of other parameters with the highest correlation with the early warning parameters according to the preset group number as the input of the step S3.
In other words, for the processed data, several groups of other parameters with the highest correlation among the other parameters and the early warning parameters are screened as input items of a subsequent parameter prediction model training set by calculating pearson coefficients of the other parameters and the early warning parameters, and the selection of the early warning parameters and the input parameters is determined according to models of coal mills of different units, early warning requirements and the like. The number of groups to be selected, which is preset, is also set according to the actual situation.
For example, according to the Pearson correlation coefficient from high to low, the operational parameters of the top five correlation coefficients are selected as the input of the training set, which are the motor winding temperature, the coal mill inlet temperature, the coal mill front bearing temperature, the coal mill inlet flow rate and the coal feeder coal amount. It should be noted that the above-mentioned parameters are the screening result of the present embodiment, and can be adjusted by themselves when the present application is used in other embodiments.
S3, selecting data from historical operating data obtained after data dimensionality reduction according to a preset number, inputting the selected data serving as a training set into a LightGBM model, and training to obtain a LightGBM parameter prediction model in a normal operating state of the coal mill; and optimizing the LightGBM parameter prediction model by utilizing an SSA-PSO algorithm to obtain an optimized LightGBM parameter prediction model.
Namely, selecting partial dimensionality-reduced data as training set input, constructing a LightGBM parameter prediction model under the normal state of the coal mill, and constructing an optimized LightGBM parameter prediction model. The specific steps of step S3 are as follows:
step S301, selecting 300-600 samples as a training set from historical operating data obtained after data dimension reduction, inputting the training set into a LightGBM model, and training to obtain a LightGBM parameter prediction model under a normal operating state of the coal mill;
for example, the first 400 samples in the processed data in step S2 may be selected as a training set to train the LightGBM parameter prediction model;
step S302, initializing a LightGBM parameter prediction model and initializing each hyper-parameter;
the hyper-parameters comprise a learning rate (learning _ rate), a tree depth (max _ depth), a leaf node number (num _ samples), and a minimum data number on a leaf (min _ child _ samples);
step S303, setting parameters of the SSA-PSO algorithm, wherein the parameters of the SSA-PSO algorithm comprisePopulation size N, dimension d, maximum iteration number iter max Acceleration factor c 1 And acceleration factor c 2 Etc.;
step S304, using the hyper-parameters to be optimized as initial particles, setting the maximum position and the minimum position of the particles, initializing the initial position and the speed of each particle, and constructing a current search area;
step S305, calculating the fitness of each particle, and setting X b The individual with the best position in the population is taken as a scout; let X w The individuals with the worst positions in the population are taken as the participants;
step S306, updating the individuals of the reconnaissance and the joining person;
specifically, the scout searches in a large range in the current search area, and the position updating formula is as follows:
wherein, the first and the second end of the pipe are connected with each other,
representing the position of the ith individual scout in the d dimension at the t iteration; α is a random number between 0 and 1; iter max Is the maximum iteration number;
the participator explores according to the guidance of the scout, and the position updating formula is as follows:
v i(d+1) ＝rr·ω·v id +c 1 r 1 (p id -x id )+c 2 r 2 (p gd -x id )
x i(d+1) ＝x id +v i(d+1)
wherein the content of the first and second substances,
the influence factor is used for guiding the individual joining person to update the speed of the individual joining person; omega is a weight coefficient used for adjusting the searching capability of the algorithm; v. of id The velocity of the individual i of the enrollee in the d dimension; r is a radical of hydrogen 1 And r 2 All are random numbers between 0 and 1; p is a radical of id And p gd Respectively representing an individual optimal solution and a global optimal solution of the individual i of the subscriber in the d dimension; x is a radical of a fluorine atom id The positions of the individual i of the participants on the d dimension; v. of i(d+1) Updating the speed of the individual i of the subscriber after the updating formula of the corresponding position of the individual i in the dimension d + 1; x is the number of i(d+1) For the enrollee individual i a position x in d +1 dimension according to the previous dimension id And velocity v i(d+1) An updated position;
step S307, calculating the optimal individual position;
step S308, judging whether the iteration termination condition of the current search area is met, if so, executing step S309, and if not, returning to step S305;
wherein, the iteration termination condition of the current search area is that the maximum iteration times are reached;
step S309, judging whether an iteration termination condition of the complete algorithm is met, if so, executing step S310, and if not, returning to step S304;
wherein, the iteration termination condition of the complete algorithm is that the maximum iteration times is reached or the optimal solution is not changed any more;
and S310, outputting the super-parameter optimal combination, and using the super-parameter optimal combination as a parameter of the LightGBM parameter prediction model to obtain the optimized LightGBM parameter prediction model.
Further, the maximum number of iterations may be set to 50. In this embodiment, after the SSA-PSO algorithm is iterated for 50 times, the obtained hyper-parameters are: learning rate learning _ rate =0.2, tree depth max _ depth =10, number of leaf nodes num _ leaves =1390, minimum data on leaf min _ child _ samples =55.
S4, calculating by using an optimized LightGBM parameter prediction model to obtain an early warning parameter prediction value, and calculating to obtain a residual average value of the early warning parameter prediction value; and taking the residual average value of the early warning parameter predicted value as an early warning threshold reference value, and setting an early warning threshold range according to the early warning threshold reference value and various boundary conditions.
The method for calculating and obtaining the early warning parameter predicted value by using the optimized LightGBM parameter prediction model and calculating and obtaining the residual average value of the early warning parameter predicted value specifically comprises the following steps:
inputting the training set into a trained optimized LightGBM parameter prediction model, and calculating to obtain an early warning parameter prediction value;
and calculating residual errors between the predicted values and the actual values of the early warning parameters, and calculating to obtain a residual error average value of the predicted values of the early warning parameters by a sliding window method.
For example, inputting the training set into the optimized LightGBM parameter prediction model to obtain corresponding rear bearing temperature prediction data, performing residual analysis on the temperature prediction data, namely calculating a difference value between a temperature prediction value and a temperature actual value to obtain residual points, selecting M (M = 20) continuous residual points as the length of a sliding window, sliding the window forwards by 1 residual point each time, and calculating an average value T of all residual points in the window. Fig. 2 is a schematic diagram of a residual average value of a predicted value of an early warning parameter of a coal mill under a current working condition in the embodiment of the application.
It can be easily seen from fig. 2 that the average residual error values are uniformly distributed between (-0.4,0.4), and the absolute value of the residual error value of the bearing temperature can be set to be 0.4 as the early warning threshold range, that is, the range of the dotted line in fig. 2, according to the current unit operation condition and various boundary conditions of the coal pulverizer. It should be noted that the early warning threshold range in this embodiment is set as the early warning threshold setting under the current working condition, and because the computer has limited data processing capability, the setting effect of displaying different early warning thresholds under all working conditions cannot be achieved, and the setting of the early warning threshold range can be determined according to the situation when the method is specifically used.
Step S5, regularly collecting actual operation data of the coal mill, inputting the actual operation data into an optimized LightGBM parameter prediction model after data dimension reduction, calculating to obtain an early warning parameter residual error average value in an actual operation state, and judging whether the early warning parameter residual error average value exceeds an early warning threshold range; and when the residual error average value of the early warning parameter is judged to exceed the early warning threshold range, outputting an early warning prompt.
The method comprises the following steps of periodically acquiring actual operation data of the coal mill, inputting the actual operation data into an optimized LightGBM parameter prediction model after data dimension reduction, and calculating to obtain an early warning parameter residual error average value under an actual operation state, wherein the actual operation data are specifically as follows:
after actual operation data of the coal mill are collected, selecting multiple groups of actual operation data of other parameters with highest relevance with the early warning parameters to form an early warning data set, inputting the early warning data set into an optimized LightGBM parameter prediction model to obtain actual early warning parameter predicted values in an actual operation state, and calculating by using a sliding window method to obtain early warning parameter residual error average values in the actual operation state.
In the step, for actual operation data of the on-site coal mill, selecting the characteristic data which is selected in the step S2 and has high correlation with the rear bearing temperature (early warning parameter) to form an early warning data set, inputting the data set into an optimized LightGBM parameter prediction model to obtain a predicted value of the rear bearing temperature of the coal mill to be early warned currently, calculating a residual average value T in a sliding window, judging whether the residual average value T exceeds an early warning threshold range, and immediately sending an early warning signal once the residual average value T exceeds the early warning threshold range, so that early warning of the fan fault is realized.
Fig. 3 shows the coal mill fault early warning result in this embodiment, the sliding window method selects the sliding window length as M =20, sets the window sliding step length as 1, sets the residual threshold value as |0.4|, and determines that an alarm is effective once when the average residual value of the early warning parameter exceeds the range of (-0.4,0.4), so as to effectively avoid the influence of abnormal points. It can be seen from fig. 3 that the threshold of 0.4 was breached at sample 430 and an alarm was raised.
Another schematic flow diagram of the present application is shown in fig. 4.
Further, in this embodiment, the optimized LightGBM parameter prediction model is compared with the LightGBM parameter prediction model that is not optimized. In order to reasonably evaluate the performance of the model, the application utilizes the average absolute error MAE, the root mean square error RMSE and the average absolute percentage error MAPE as evaluation indexes, and the mathematical expression of the evaluation indexes is as follows:
wherein n is the number of samples, y i Is the actual temperature of the rear bearing, y' i For the rear bearing predicted temperature, table 1 compares the accuracy of the modified LightGBM bearing temperature prediction model with the accuracy of the unmodified LightGBM bearing temperature prediction model.
TABLE 1 comparison of model prediction accuracies
| Evaluation index | LightGBM | SSAPSO optimized LightGBM |
| MAE | 0.254 | 0.206 |
| RMSE | 0.379 | 0.286 |
| MAPE | 1.325 | 0.896 |
The embodiment of the application provides a coal mill fault early warning method based on an optimized LightGBM algorithm, the method comprises the steps of firstly preprocessing original data and reducing dimensions of the data, screening a training set to be input into a LightGBM parameter prediction model based on SSAPSO optimization to obtain an early warning parameter residual error, obtaining an early warning threshold reference value after obtaining a residual error average value through calculation by a sliding window method, obtaining a dynamic threshold range by combining various boundary conditions of the coal mill under the current unit working condition, predicting the early warning parameter of the coal mill in the actual operation process and calculating the residual error average value, immediately sending early warning when the residual error average value exceeds the early warning threshold range, and achieving early fault early warning of the coal mill.
Compared with the traditional fault early warning technology, the method and the device pay attention to the utilization of historical operation data, the training set of the parameter prediction model can be updated regularly, the dynamic early warning threshold range is set in cooperation with the current equipment operation condition, and long-term early warning is achieved.
The method and the device apply a LightGBM (Light Gradient Boosting Machine) algorithm to fault early warning, and the algorithm has the characteristics of high training speed, low memory occupation and high accuracy. Compared with the traditional fault early warning method, the LightGBM algorithm also supports parallel learning, and can ensure the fault prediction precision while processing large-scale data. According to the method and the device, the main hyper-parameters of the LightGBM model are automatically optimized by adopting an SSAPSO optimization algorithm, the trouble of manual parameter adjustment is avoided, and the efficiency of the model in processing a large amount of data is improved.
It should be understood that although the steps in the flowcharts of fig. 1 and 4 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Moreover, at least some of the steps in fig. 1 and 4 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed in turn or alternately with other steps or at least some of the other steps.
In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 5. The computer device comprises a processor, a memory, a communication interface, a display screen and an input device which are connected through a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program to run on the non-volatile storage medium. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement the coal mill fault early warning method based on the optimized LightGBM algorithm. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 5 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, which includes a memory and a processor, wherein the memory stores a computer program, and all or part of the procedures in the method of the above embodiment are involved.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, relating to all or part of the flow in the method of the above embodiment.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), for example.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, and these are all within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A coal mill fault early warning method based on an optimized LightGBM algorithm is characterized by comprising the following steps:
s1, collecting original historical operation data of operation parameters of a coal mill in a normal operation state from an SIS system;
s2, preprocessing the original historical operation data, and performing data dimension reduction on the preprocessed historical operation data through a Pearson coefficient;
s3, selecting data from historical operating data obtained after data dimensionality reduction according to a preset number, inputting the selected data serving as a training set into a LightGBM model, and training to obtain a LightGBM parameter prediction model in a normal operating state of the coal mill; optimizing the LightGBM parameter prediction model by utilizing an SSA-PSO algorithm to obtain an optimized LightGBM parameter prediction model;
s4, calculating by using the optimized LightGBM parameter prediction model to obtain an early warning parameter prediction value, and calculating to obtain a residual average value of the early warning parameter prediction value; taking the residual average value of the early warning parameter predicted value as an early warning threshold reference value, and setting an early warning threshold range according to the early warning threshold reference value and various boundary conditions;
s5, regularly collecting actual operation data of the coal mill, inputting the actual operation data into the optimized LightGBM parameter prediction model after data dimension reduction, calculating to obtain an early warning parameter residual error average value in an actual operation state, and judging whether the early warning parameter residual error average value exceeds the early warning threshold range; and outputting an early warning prompt when the average value of the early warning parameter residuals exceeds the early warning threshold range.
2. The coal pulverizer fault warning method based on the optimized LightGBM algorithm of claim 1, wherein the operational parameters comprise hot blast door opening, cold blast door opening, primary wind pressure, inlet temperature, inlet flow, gearbox input bearing temperature, front axle temperature, motor winding temperature, rear axle bearing temperature, coal pulverizer current, coal feeder current, and coal feeder coal volume.
3. The optimized LightGBM algorithm-based coal mill fault early warning method as claimed in claim 1, wherein the preprocessing the raw historical operational data comprises:
rejecting nonnumbers in the original historical operating data;
eliminating unsteady and noise values in the original historical operating data;
normalizing the original historical operating data by the following specific formula:
wherein, the first and the second end of the pipe are connected with each other,
vector A representing the x-th group of samples in the operating parameter i A value of (d) above; />
Represents->
Normalizing the result; />
Represents vector A i The mean value of (a); max (A) i ) Represents vector A i The maximum value of (a); min (A) i ) Represents vector A i Is measured.
4. The coal mill fault early warning method based on the optimized LightGBM algorithm as claimed in claim 3, wherein the non-numbers in the original historical operating data are removed by using an isnan function in MATLAB software; and eliminating unsteady and noise values in the original historical operating data by using a 3 sigma criterion.
5. The coal mill fault early warning method based on the optimized LightGBM algorithm, according to claim 1, wherein the data dimensionality reduction of the preprocessed historical operation data by the Pearson coefficient method comprises:
according to the historical operation data obtained after preprocessing, calculating the Pearson coefficient of the early warning parameter and other parameters, wherein the Pearson coefficient has the following calculation formula:
wherein x is i (i =1,2,3.., n) is the value of some other parameter x in the ith sample, y i For the value of the early warning parameter y in the ith sample,
is the average of some other parameter x over n samples, <' >>
The average value of the early warning parameter y in n samples is shown, and r is a correlation coefficient of some other parameter x and the early warning parameter y;
and selecting historical operating data of a plurality of groups of other parameters with the highest correlation with the early warning parameters according to the preset group number as the input of the step S3.
6. The coal mill fault early warning method based on the optimized LightGBM algorithm as claimed in claim 1, wherein the step S3 specifically comprises:
step S301, selecting 300-600 samples as a training set from historical operating data obtained after data dimension reduction, inputting the training set into a LightGBM model, and training to obtain a LightGBM parameter prediction model under a normal operating state of the coal mill;
step S302, initializing the LightGBM parameter prediction model and initializing each super parameter; the hyper-parameters comprise learning rate, tree depth, leaf node number and minimum data number on the leaf;
step S303, setting parameters of an SSA-PSO algorithm; the parameters of the SSA-PSO algorithm comprisePopulation size N, dimension d, maximum iteration number iter max Acceleration factor c 1 And acceleration factor c 2 ；
Step S304, using the hyper-parameters to be optimized as initial particles, setting the maximum position and the minimum position of the particles, initializing the initial position and the speed of each particle, and constructing a current search area;
step S305, calculating the fitness of each particle, and setting X b The individuals with the best positions in the population are taken as scouts; let X w The individual with the worst position in the population is taken as an enrollee;
step S306, updating the individuals of the reconnaissance and the joining person;
step S307, calculating the optimal individual position;
step S308, judging whether the iteration termination condition of the current search area is met, if so, executing step S309, and if not, returning to step S305;
step S309, judging whether an iteration termination condition of the complete algorithm is met, if so, executing step S310, and if not, returning to step S304;
and S310, outputting a super-parameter optimal combination, and using the super-parameter optimal combination as a parameter of the LightGBM parameter prediction model to obtain an optimized LightGBM parameter prediction model.
7. The coal mill fault early warning method based on optimized LightGBM algorithm as claimed in claim 6, wherein in step S306, the scout searches the current search area with the location update formula:
wherein the content of the first and second substances,
indicates the number of i scouts in the t iterationThe position of the body in the d dimension; α is a random number between 0 and 1; iter (R) max Is the maximum iteration number;
the participator explores according to the guidance of the scout, and the position updating formula is as follows:
v i(d+1) ＝rr·ω·v id +c 1 r 1 (p id -x id )+c 2 r 2 (p gd -x id )
x i(d+1) ＝x id +v i(d+1)
wherein the content of the first and second substances,
the influence factor is used for guiding the individual joining person to update the speed of the individual joining person; omega is a weight coefficient used for adjusting the searching capability of the algorithm; v. of id The velocity of an individual i of an enrollee in d dimension; r is a radical of hydrogen 1 And r 2 Is a random number between 0 and 1; p is a radical of id And p gd Respectively representing an individual optimal solution and a global optimal solution of the individual i of the subscriber in the d dimension; x is the number of id The positions of the individual i of the participants on the d dimension; v. of i(d+1) Updating the speed of the individual i of the subscriber after the updating formula of the corresponding position of the individual i in the dimension d + 1; x is a radical of a fluorine atom i(d+1) For the enrollee individual i a position x in d +1 dimension according to the previous dimension id And velocity v i(d+1) The updated position.
8. The coal mill fault early warning method based on the optimized LightGBM algorithm as claimed in claim 6, wherein the iteration termination condition of the current search area is that the maximum number of iterations is reached; the iteration termination condition of the complete algorithm is that the maximum iteration times are reached or the optimal solution is not changed any more;
the maximum number of iterations is 50.
9. The coal mill fault early warning method based on the optimized LightGBM algorithm as claimed in claim 1, wherein said calculating to obtain the predicted value of the early warning parameter by using the optimized LightGBM parameter prediction model and calculating to obtain the average value of the residuals of the predicted value of the early warning parameter specifically comprises:
inputting the training set into the optimized LightGBM parameter prediction model, and calculating to obtain an early warning parameter prediction value;
and calculating the residual error between the predicted value and the actual value of the early warning parameter, and calculating the average value of the residual errors of the predicted value of the early warning parameter by a sliding window method.
10. The coal mill fault early warning method based on the optimized LightGBM algorithm as claimed in claim 5, wherein the actual operation data of the coal mill is periodically collected, the actual operation data is input into the optimized LightGBM parameter prediction model after data dimension reduction, and the average value of the residual errors of the early warning parameters in the actual operation state is calculated as follows:
after actual operation data of the coal mill are collected, selecting multiple groups of actual operation data of other parameters with highest relevance to the early warning parameters to form an early warning data set, inputting the early warning data set into the optimized LightGBM parameter prediction model to obtain actual early warning parameter predicted values in an actual operation state, and calculating by using a sliding window method to obtain early warning parameter residual error average values in the actual operation state.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211403260.4A CN115879369A (en) | 2022-11-10 | 2022-11-10 | Coal mill fault early warning method based on optimized LightGBM algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211403260.4A CN115879369A (en) | 2022-11-10 | 2022-11-10 | Coal mill fault early warning method based on optimized LightGBM algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115879369A true CN115879369A (en) | 2023-03-31 |
Family
ID=85759573
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211403260.4A Pending CN115879369A (en) | 2022-11-10 | 2022-11-10 | Coal mill fault early warning method based on optimized LightGBM algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115879369A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116773234A (en) * | 2023-05-05 | 2023-09-19 | 华电莱州发电有限公司 | Coal pulverizer fault monitoring system |
| CN117972616A (en) * | 2024-03-28 | 2024-05-03 | 江西江投能源技术研究有限公司 | Pumped storage generator set safety state monitoring and diagnosing method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112598148A (en) * | 2020-09-30 | 2021-04-02 | 新天绿色能源股份有限公司 | Fan variable pitch motor temperature fault early warning method based on collaborative expression and LightGBM algorithm |
| CN113792762A (en) * | 2021-08-24 | 2021-12-14 | 华南理工大学 | Water chilling unit fault diagnosis method, system and medium based on Bayesian optimization LightGBM |
| CN114093503A (en) * | 2021-11-09 | 2022-02-25 | 北京石油化工学院 | Mortality prediction method and system based on LightGBM optimization |
| CN114721263A (en) * | 2022-03-16 | 2022-07-08 | 中国中材国际工程股份有限公司 | Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm |
-
2022
- 2022-11-10 CN CN202211403260.4A patent/CN115879369A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112598148A (en) * | 2020-09-30 | 2021-04-02 | 新天绿色能源股份有限公司 | Fan variable pitch motor temperature fault early warning method based on collaborative expression and LightGBM algorithm |
| CN113792762A (en) * | 2021-08-24 | 2021-12-14 | 华南理工大学 | Water chilling unit fault diagnosis method, system and medium based on Bayesian optimization LightGBM |
| CN114093503A (en) * | 2021-11-09 | 2022-02-25 | 北京石油化工学院 | Mortality prediction method and system based on LightGBM optimization |
| CN114721263A (en) * | 2022-03-16 | 2022-07-08 | 中国中材国际工程股份有限公司 | Intelligent regulation and control method for cement decomposing furnace based on machine learning and intelligent optimization algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 蒋敏 等；: "基于PSO优化VMD算法的轴承振动信号重构及故障诊断", 《机械设计与研究》, vol. 38, no. 05, 31 October 2022 (2022-10-31) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116773234A (en) * | 2023-05-05 | 2023-09-19 | 华电莱州发电有限公司 | Coal pulverizer fault monitoring system |
| CN117972616A (en) * | 2024-03-28 | 2024-05-03 | 江西江投能源技术研究有限公司 | Pumped storage generator set safety state monitoring and diagnosing method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Al-qaness et al. | Boosted ANFIS model using augmented marine predator algorithm with mutation operators for wind power forecasting | |
| CN109816221B (en) | Project risk decision method, apparatus, computer device and storage medium | |
| CN115879369A (en) | Coal mill fault early warning method based on optimized LightGBM algorithm | |
| CN111353631A (en) | Thermal power plant condenser vacuum degree prediction method based on multilayer LSTM | |
| CN111625516A (en) | Method and device for detecting data state, computer equipment and storage medium | |
| CN111461463A (en) | Short-term load prediction method, system and equipment based on TCN-BP | |
| CN104036328A (en) | Self-adaptive wind power prediction system and prediction method | |
| CN113325721A (en) | Model-free adaptive control method and system for industrial system | |
| CN115456287A (en) | Long-and-short-term memory network-based multi-element load prediction method for comprehensive energy system | |
| CN115759336A (en) | Prediction method and storage medium for short-term power load prediction | |
| CN114266201B (en) | Self-attention elevator trapping prediction method based on deep learning | |
| CN111598328A (en) | Power load prediction method considering epidemic situation events | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN114239396A (en) | Fan gear box state prediction method and system | |
| CN113128666A (en) | Mo-S-LSTMs model-based time series multi-step prediction method | |
| CN116861256A (en) | Furnace temperature prediction method, system, equipment and medium for solid waste incineration process | |
| CN112183872A (en) | Blast furnace gas generation amount prediction method combining generation of countermeasure network and neural network | |
| CN112149976A (en) | Power grid accurate investment project decision method based on artificial intelligence | |
| CN117113086A (en) | Energy storage unit load prediction method, system, electronic equipment and medium | |
| CN116205508A (en) | Distributed photovoltaic power generation abnormality diagnosis method and system | |
| CN114741952A (en) | Short-term load prediction method based on long-term and short-term memory network | |
| CN115619563A (en) | Stock price analysis method based on neural network | |
| CN115759455A (en) | Load probability density prediction method based on time sequence Gaussian mixture density network | |
| CN113947332B (en) | Underground engineering comprehensive security capability assessment method and system | |
| CN115526376A (en) | Multi-feature fusion generation countermeasure network ultra-short-term wind power prediction method |
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 |