CN112287598A - Fly ash carbon content prediction method based on particle swarm parameter optimization - Google Patents
Fly ash carbon content prediction method based on particle swarm parameter optimization Download PDFInfo
- Publication number
- CN112287598A CN112287598A CN202011041478.0A CN202011041478A CN112287598A CN 112287598 A CN112287598 A CN 112287598A CN 202011041478 A CN202011041478 A CN 202011041478A CN 112287598 A CN112287598 A CN 112287598A
- Authority
- CN
- China
- Prior art keywords
- fly ash
- carbon content
- particle swarm
- measurement model
- parameters
- 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
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 title claims abstract description 158
- 229910052799 carbon Inorganic materials 0.000 title claims abstract description 158
- 239000010881 fly ash Substances 0.000 title claims abstract description 155
- 239000002245 particle Substances 0.000 title claims abstract description 68
- 238000005457 optimization Methods 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000002485 combustion reaction Methods 0.000 claims abstract description 56
- 238000005259 measurement Methods 0.000 claims abstract description 56
- 238000012706 support-vector machine Methods 0.000 claims abstract description 44
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 41
- 238000012216 screening Methods 0.000 claims abstract description 10
- 238000004458 analytical method Methods 0.000 claims abstract description 6
- 239000003245 coal Substances 0.000 claims description 53
- 238000012549 training Methods 0.000 claims description 23
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 21
- 229910052760 oxygen Inorganic materials 0.000 claims description 21
- 239000001301 oxygen Substances 0.000 claims description 21
- 238000012360 testing method Methods 0.000 claims description 18
- 238000013507 mapping Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 239000013598 vector Substances 0.000 claims description 8
- 230000009977 dual effect Effects 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000012417 linear regression Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000000295 complement effect Effects 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 abstract description 4
- 238000010219 correlation analysis Methods 0.000 abstract description 2
- 230000007774 longterm Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 34
- 239000000843 powder Substances 0.000 description 14
- 239000000446 fuel Substances 0.000 description 11
- 239000000126 substance Substances 0.000 description 7
- 239000002956 ash Substances 0.000 description 4
- 239000002817 coal dust Substances 0.000 description 4
- 239000000571 coke Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 3
- 238000004939 coking Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 239000003546 flue gas Substances 0.000 description 3
- 239000007789 gas Substances 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005315 distribution function Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 239000005373 porous glass Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 239000000779 smoke Substances 0.000 description 2
- 210000004127 vitreous body Anatomy 0.000 description 2
- 239000004484 Briquette Substances 0.000 description 1
- 238000003723 Smelting Methods 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000007429 general method Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 230000017525 heat dissipation Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000001590 oxidative effect Effects 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 239000002893 slag Substances 0.000 description 1
- 239000004071 soot Substances 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
Abstract
The invention relates to a prediction method of fly ash carbon content based on particle swarm parameter optimization, according to the generation mechanism of fly ash carbon content and measurable operation parameters of a combustion system, influence factors influencing the fly ash carbon content are initially selected through mechanism analysis; obtaining long-term historical data of the influencing factors from a historical database, removing redundant related variables and variables with low correlation with the carbon content of the fly ash by means of mutual information correlation analysis, and screening out input variables of a soft measurement model of the carbon content of the fly ash; on the basis, a soft measurement model of the carbon content of the fly ash is established by adopting a support vector machine and a particle swarm optimization algorithm, so that the real-time prediction of the carbon content of the fly ash is realized. The invention can effectively guide the boiler to operate in time and improve the combustion efficiency of the boiler.
Description
Technical Field
The invention relates to the technical field of machine learning, in particular to a fly ash carbon content prediction method based on particle swarm parameter optimization.
Background
The carbon content of fly ash is one of more important operation economic indicators and technical indicators which influence the combustion efficiency of a boiler. However, due to the restriction of human factors and conditions such as the current industrial level, the method has the serious defects that the test result has time lag, the representativeness of the measurement result is low, and the like, and can not effectively guide the operation of the boiler in time and improve the combustion efficiency of the boiler.
The combustion efficiency of a boiler is closely related to its various heat losses, the main heat losses of the boiler include: heat loss caused by mechanical incomplete combustion, heat loss caused by overhigh exhaust gas temperature, heat loss caused by chemical incomplete combustion, great heat loss caused by ash and slag physics and heat dissipation loss caused by poor heat preservation effect. The heat loss is analyzed, and the most important heat loss of the boiler is mainly occupied by the following two terms, namely: the heat loss caused by the smoke exhaust loss and the incomplete combustion of the machine has many and complicated factors. Among them, fly ash combustible is an important component of the heat loss caused by mechanical incomplete combustion, and is not negligible. One important measure of the combustion efficiency of a boiler is the level of combustible fly ash. The combustible substance of the fly ash is higher, and the following points are approximately shown:
(1) the thermal efficiency of the boiler can be increased negatively along with the higher combustible substance of the fly ash;
(2) because the combustion is insufficient, reducing atmosphere can exist in the hearth, the water wall tube can be collided by carbon-containing carbon particles in an anoxic area, and the water wall tube is easy to corrode due to long-time collision. The coking can also cause the coking of a superheater, a water-cooled wall and the like, which is mainly caused by the fact that the smelting point of the coking is lower than that of an oxidizing atmosphere in the reducing gas existing for a long time;
(3) when the fly ash combustible substance is higher, the pulverized coal can not be completely burnt out, the flame center moves upwards at this time, the flue gas temperature of the outlet area of the hearth is higher, the high-temperature superheater and the reheater pipe are positioned in the area, the temperature of the high-temperature superheater and the reheater pipe is ultrahigh, the main steam temperature can be stably adjusted and affected, the operation is long, and the probability of sudden accidents such as pipe explosion can be increased. If the carbon content of the fly ash is high, the rear smoke well part of the boiler can be affected accordingly, and the abrasion caused by the overhigh temperature of the low-temperature section superheater is mainly reflected; the adverse effects are aggravated after long-time operation, and great challenges are brought to the safe operation of the boiler;
(4) the fly ash has high carbon content, can cause certain pollution to the environment, and the efficiency of electric dust removal is generally reduced.
Based on the analysis, it is very important to accurately monitor the boiler carbon content signal in time. And direct and indirect factors influencing the fly ash carbon content of the coal-fired boiler are many, such as: the coal type condition for burning, the design, manufacture and installation level of the boiler, the operation level of the boiler and the like. The influence factors are numerous and the mutual relation is complex, the accurate expression by a simple formula is not feasible, and the current general method is to adopt a real furnace test method to measure the carbon content value of the fly ash and utilize reasonable air-coal ratio distribution to reduce the carbon content of the fly ash. The on-site real furnace test has the defects of limited operation test working conditions and the like, namely the operation working conditions of all load points cannot be included as much as possible. However, in the actual operation of the thermal power plant, the operation parameters are varied at many ends, and each of the parameters is different from the others and is measured for a long time, so that the method for obtaining reasonable operation is unreliable.
Disclosure of Invention
In order to solve the technical problems, the invention provides a fly ash carbon content prediction method based on particle swarm parameter optimization.
The technical scheme adopted by the invention for solving the technical problems is as follows: a fly ash carbon content prediction method based on particle swarm parameter optimization is constructed, and comprises the following steps:
extracting the historical data of the carbon content of the fly ash from a database, screening the influence parameters of the historical data corresponding to the carbon content of the fly ash, and dividing a training set and a test set;
establishing a fly ash carbon content measurement model by adopting a support vector machine and a particle swarm optimization algorithm, inputting the influence parameters corresponding to the fly ash carbon content historical data in a training set as input variables into the fly ash carbon content measurement model for training, and testing the training result through a test set after the training is finished;
and inputting the real-time influence parameters of the carbon content of the fly ash into the trained fly ash carbon content measurement model, and outputting the corresponding fly ash carbon content.
The method comprises the following steps of screening influence parameters of historical data corresponding to the carbon content of fly ash, wherein the step of screening the influence parameters comprises the following steps:
primarily selecting 28 combustion parameters as the influence parameters of the carbon content of the fly ash according to the combustion operation parameters of the boiler; wherein, the 28 combustion parameters are load, total air quantity, total coal quantity, total steam quantity, oxygen quantity, A side oxygen quantity, B side oxygen quantity, A side primary air pressure, B side primary air pressure, 2 air preheater outlet secondary air temperatures, 6 coal mill outlet temperatures, 2 hearth outlet temperatures and 9 secondary air door openness respectively;
carrying out data deduplication and abnormal value removal on the influence parameters, and deleting repeated data in the influence parameters and data with an output value exceeding a normal range caused by noise interference factors;
and extracting data characteristic quantity of the primarily selected influence parameters according to a mutual information theory, calculating the weight of each influence parameter, and selecting the influence parameters with the weight larger than a preset threshold value as input variables of the fly ash carbon content measurement model.
The ten influence parameters as the input variables of the fly ash carbon content measurement model are respectively as follows: secondary air temperature B, secondary air temperature A, total steam flow, B side oxygen quantity, A primary air quantity, total coal quantity, oxygen quantity, load and B primary air quantity.
In the step of establishing the measurement model of the carbon content in the fly ash by adopting the support vector machine and the particle swarm optimization algorithm, the measurement model of the carbon content in the fly ash of the support vector machine is established, the measurement model of the carbon content in the fly ash is optimized by the particle swarm optimization algorithm, and the establishment of the measurement model of the carbon content in the fly ash of the support vector machine specifically comprises the following steps:
the inner product formula of the kernel function of the support vector machine is expressed as formula (1):
Φ(xi)T·Φ(xj)＝K(xi,xj) (1)
wherein phi (x) is a non-mapping relation, K (x)i,xj) Is a kernel function of the support vector machine;
set a set of non-linear sample sets S { (x)i,yi) I is 1,2, … l, where x isi∈RnIn order to predict the factors, the method comprises the following steps,
yi∈Rnthen y ═ f (x) for the predicted objects have a typical nonlinear relationship, and the described problem is a nonlinear regression problem;
the criterion is a linear epsilon-insensitive loss function, as shown in equation (2):
determining generalization capability and number of SVM support vectors by epsilon, and defining loss as 0 if the error of a predicted value and an actually measured value is within the epsilon range;
after mapping the sample space to another high-dimensional space and implementing corresponding linear regression analysis, using reverse reduction to make the initial sample data complete regression operation, the functional relationship expressed by the process is shown as formula (3):
f(x)＝w·Φ(x)+b (3)
by adding relaxation factors xii＝(ξ1,…ξn) 1,2, …, n, expressing the constrained optimization objective function problem as equation (4):
wherein, w and b are weight vector and bias, c is punishment parameter keeping positive correlation with fitting degree;
like equation (5), introduce Lagrange's function,
lagrange factor
The extreme value of L should accord with Karush-Kuhn-Tucker condition (KKT condition), namely the constraint optimization problem of equality processed by Lagrange multiplier method is popularized to inequality, and the necessary conditions of the optimal solution include the constant equation, original feasibility, dual feasibility and complementary relaxation of Lagrange function, so as to reach a pair ai,
λi,
Maximization of (2);
the corresponding dual problem is converted, as shown in formula (6):
substituting w, b into f (x) w · Φ (x) + b to obtain a nonlinear SVM expression with a kernel function:
in the step of optimizing the fly ash carbon content measurement model by the particle swarm optimization algorithm, the radial basis kernel function parameter g and the penalty coefficient C in the SVM expression of the fly ash carbon content measurement model are adjusted and optimized by the particle swarm optimization algorithm.
The method for optimizing the fly ash carbon content measurement model by the particle swarm optimization algorithm comprises the following steps:
initializing particle swarm parameters; wherein, the parameters to be initialized at least comprise: inertia factors, population size, particle speed, particle position and maximum iteration number;
determining a fitness function; the fitness function is the mean square error of the fly ash carbon content value predicted by the fly ash carbon content measurement model and the fly ash carbon content value measured by the local carbon meter;
setting particle swarm parameter optimization termination conditions; setting iteration times, and stopping optimizing when the iteration times are reached;
and when the optimization is finished, combining the optimal parameters searched by the particle swarm algorithm in the fitness function meaning by the fly ash carbon content measurement model regressed by the support vector machine to monitor the fly ash carbon content numerical value.
Different from the prior art, the method for predicting the carbon content of the fly ash based on particle swarm parameter optimization preliminarily selects the influence factors influencing the carbon content of the fly ash through mechanism analysis according to the generation mechanism of the carbon content of the fly ash and the measurable operation parameters of a combustion system; obtaining long-term historical data of the influencing factors from a historical database, removing redundant related variables and variables with low correlation with the carbon content of the fly ash by means of mutual information correlation analysis, and screening out input variables of a soft measurement model of the carbon content of the fly ash; on the basis, a soft measurement model of the carbon content of the fly ash is established by adopting a support vector machine and a particle swarm optimization algorithm, so that the real-time prediction of the carbon content of the fly ash is realized. The invention can effectively guide the boiler to operate in time and improve the combustion efficiency of the boiler.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a schematic flow chart of a fly ash carbon content prediction method based on particle swarm parameter optimization according to the present invention.
FIG. 2 is a logic diagram of the fly ash carbon content measurement model establishment based on the particle swarm parameter optimization fly ash carbon content prediction method provided by the invention.
FIG. 3 is a schematic flow chart of a PSO-SVM algorithm of the fly ash carbon content prediction method based on particle swarm parameter optimization provided by the invention.
Detailed Description
For a more clear understanding of the technical features, objects and effects of the present invention, embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
Referring to fig. 1, the invention provides a fly ash carbon content prediction method based on particle swarm parameter optimization, comprising the following steps:
extracting the historical data of the carbon content of the fly ash from a database, screening the influence parameters of the historical data corresponding to the carbon content of the fly ash, and dividing a training set and a test set;
establishing a fly ash carbon content measurement model by adopting a support vector machine and a particle swarm optimization algorithm, inputting the influence parameters corresponding to the fly ash carbon content historical data in a training set as input variables into the fly ash carbon content measurement model for training, and testing the training result through a test set after the training is finished;
and inputting the real-time influence parameters of the carbon content of the fly ash into the trained fly ash carbon content measurement model, and outputting the corresponding fly ash carbon content.
The formation of fly ash in boiler furnace is a very complicated physical and chemical change, and the specific process is as follows:
firstly, the coal powder enters the boiler along with air, the low-gasification-temperature substance content is combusted firstly, volatile components are naturally escaped in the process, coal powder particles with a plurality of small holes on the surface are formed, the number of the small holes is increased sharply, and the coal powder particles become coke along with the continuous combustion. The coal powder contains organic matters and inorganic matters, wherein the organic matters are combustible, and the inorganic matters are left, so that after carbon components in the coal powder such as coke are combusted, the residual particles can continuously keep the original shape due to the existence of the inorganic matters, and a porous glass body is formed. However, combustion continues, and the porous glass body formed continues to melt and shrink, and the gaps between the pores are correspondingly reduced again, resulting in a higher density and a smaller soot particle radius, so that a dense, high-volume, dense glass body is finally formed. As can be seen from the above description, the formation of a dense vitreous body is a porous vitreous body, and porous carbon particles and coke are formed when the combustion of pulverized coal is particularly sufficient and insufficient. While coke is the particle with the highest carbon content in the fly ash.
The carbon content of fly ash is the product of boiler combustion, the coal briquette is ground into coal powder by a coal mill, the coal powder mixed with air is sent into a boiler hearth by a blower to be combusted, and whether the combustion process is sufficient or not is the main reason for the carbon content of fly ash. When the pulverized coal is mixed with primary air and enters the hearth for combustion, the pulverized coal must be enabled to be capable of being combusted at a fast and relatively stable speed, if the combustion speed is slow, the pulverized coal is blown away from the hearth before being completely combusted, and the flue gas discharged from the hearth contains a certain proportion of unburned and sufficient pulverized coal, namely the carbon content of fly ash is high. Therefore, every variable in the boiler combustion process will be a factor that affects the carbon content of fly ash. The influence of the main parameters was analyzed as follows:
(1) coal quality
The coal quality refers to the quality of fuel burned in a hearth, and main factors influencing the coal quality include ash content, moisture, volatile matters and the like. The ash content is the residue after coal combustion, and the higher the ash content is, the residue generated by combustion absorbs the heat generated by combustion, so that the coal powder combustion is slowed down, the coal powder cannot be fully combusted, and the carbon content of fly ash is increased; the moisture refers to free water and combined water contained in the fire coal, the higher the moisture is, the lower the heat generated by combustion is, the lower the temperature in the furnace is, and the carbon content of fly ash is increased; the volatile component means that mineral substances are decomposed into gas and liquid to overflow in the combustion process of the coal dust, the lower the volatile component is, the higher the required ignition temperature is, the more difficult the coal dust is to be completely and fully combusted, and the carbon content of fly ash is increased.
(2) Fineness of coal powder
The coal blocks are ground into coal powder by a coal mill, and the fineness of the coal powder is also an important factor influencing whether the combustion is sufficient or not. The finer the coal powder is, the larger the contact area of the coal powder and the flame is, the faster the combustion speed is, the faster the volatile components and the moisture in the fuel are separated out, the full implementation of the combustion process is facilitated, and the carbon content of the fly ash is reduced.
(3) Primary air quantity
The concentration of the primary air directly influences the concentration of the pulverized coal blown into a hearth by the air blower, so that the carbon content of the fly ash is also influenced significantly. The larger the primary air quantity is, the lower the pulverized coal concentration is, the unstable combustion state of the boiler can be caused, the relatively lower combustion temperature in the boiler is not beneficial to the separation of volatile components in the fuel, the combustion efficiency is reduced, and the carbon content of fly ash is increased; however, if the primary air quantity is too small, the concentration of the pulverized coal is large, and the pulverized coal concentration can cause the primary air and the secondary air to support the pulverized coal unstably, so that the combustion process is unstable. Therefore, an appropriate primary air volume is a necessary condition for ensuring stable combustion.
(4) Opening degree of secondary air door
The secondary air is combustion-supporting air and is sent out by a blower, and the primary function of the secondary air is to supply enough oxygen for fuel combustion. The opening degree of the secondary air door influences the ratio of fuel to air quantity in the boiler, the secondary air quantity is increased, so that the oxygen quantity required by combustion is sufficient, and the carbon content of fly ash is reduced when the combustion is sufficient; meanwhile, the temperature of secondary air is lower than that of flame, and the temperature of a hearth is reduced due to the fact that the secondary air is mixed into the furnace greatly, so that insufficient combustion is caused, and the carbon content of fly ash is increased. Therefore, the opening degree of the secondary air door is not larger, and the opening degree is better, and the opening degree needs to be properly changed according to the combustion process, so that the combustion efficiency of the boiler can be improved.
(5) Amount of oxygen
The necessary condition for complete combustion of fuel is proper air-fuel ratio, when the fuel is combusted, sufficient oxygen can not be obtained, the fuel can not be combusted fully, and the carbon content of fly ash can be increased; when the oxygen amount sent into the hearth is sufficient, the fuel is fully combusted, unburned fuel carried in the flue gas is less, and the carbon content of fly ash is reduced.
(6) Boiler load
The boiler load is used as one of main indexes influencing the combustion efficiency, and the research on the combustion condition of the pulverized coal under different loads has important significance. The load rise and fall is in positive correlation with the coal blowing-in quantity, the blown-in coal dust is less, the natural boiler load is low, the low load is that the temperature in a hearth is lower, the combustion working condition is relatively poor, and the coal dust is not sufficiently combusted, so that the carbon content of fly ash is increased. When the boiler load is high, the amount of coal required for the boiler is larger, which for coal-fired boilers is generally expressed by the main steam flow on both sides of the boiler. Under the condition of keeping other influencing factors unchanged, the carbon content of the fly ash is inversely proportional to the boiler load, and the carbon content of the fly ash is larger when the boiler load is small. The main reason for this is to control the load of the boiler and to adjust the load variation by starting and stopping the mill and adjusting the coal supply of the coal feeder. As boiler load increases, the calorific value of the pulverized coal increases, and therefore the carbon content of the pulverized coal decreases.
According to the combustion operation parameters of the boiler, initially selecting 28 combustion parameters as the influence factors of the carbon content of the fly ash, wherein the influence factors comprise: load, total steam quantity, opening degree of all secondary air doors, all primary air quantity, coal feeding quantity of a coal feeder, primary air temperature, secondary air temperature, oxygen quantity and the like.
The method comprises the following steps of screening influence parameters of historical data corresponding to the carbon content of fly ash, wherein the step of screening the influence parameters comprises the following steps:
primarily selecting 28 combustion parameters as the influence parameters of the carbon content of the fly ash according to the combustion operation parameters of the boiler;
carrying out data deduplication and abnormal value removal on the influence parameters, and deleting repeated data in the influence parameters and data with an output value exceeding a normal range caused by noise interference factors;
data deduplication, also known as deduplication, finds and deletes duplicate data, keeping only unique data units. Data reconstruction is considered at the same time of deletion, namely, although part of the content of the file is deleted, the whole content of the file is reconstructed when needed, and index information between the file and the unique data unit needs to be reserved. The abnormal value of the data is removed, and the purpose is to remove the data of which the output value exceeds the normal range due to noise interference and other factors.
In order to eliminate the noise interference and improve the smoothness of the curve, the sampled data needs to be smoothed. The general data filtering method adopts a five-value average method. In addition, filtering can also be performed by a hysteresis loop.
And extracting data characteristic quantity of the primarily selected influence parameters according to a mutual information theory, calculating the weight of each influence parameter, and selecting the influence parameters with the weight larger than a preset threshold value as input variables of the fly ash carbon content measurement model.
Mutual information is a basic concept in information theory, and is generally used to describe statistical correlation between two systems, or how much information is contained in one system and in the other system. The total number of the influence factors of the fly ash carbon content model is 28, wherein main factors influencing the fly ash carbon content are extracted from the 28 influence factors, namely the information content occupied by each influence factor contained in the fly ash carbon content model is determined, the data characteristic content is extracted by adopting a mutual information method, and the weight of each influence factor in the fly ash carbon content model is respectively found.
In probability theory, two random variables, a and B, have marginal probability distributions pa (a) and pb (B), and their joint probability distributions pAB (a, B). When pAB (a, B) ═ pa (a) · pb (B), a and B are independent of each other. Mutual information I (A, B) the degree of dependence of A and B was obtained by calculating the difference between pAB (a, B) and pA (a) pB (B). Formally, setting X as one of the influencing parameters and Y as the carbon content of fly ash; mutual information of X and Y can be defined as:
where p (X, Y) is the joint probability distribution function of X and Y, and p (X) and p (Y) are the edge probability distribution functions of two random variables X and Y, respectively.
In the process of feature selection, we generally perform mutual information expression through entropy. The entropy refers to the uncertainty of one system, the system entropy of two discrete random variables X and Y, the conditional entropy of the two systems and the joint entropy of the two systems are calculated respectively, and the calculation result of mutual information can be obtained through derivation. The derivation process is as follows:
system entropy of discrete random variable X:
system entropy of discrete random variable Y:
joint entropy of both systems:
conditional entropy for both systems:
from the above calculated system entropy values, the mutual information can be obtained as:
I(A，B)＝H(A)+H(B)-H(A，B)
the mutual information values of the 28 influencing parameters and the carbon content of the fly ash are shown in the table 1:
TABLE 1 mutual information values
And setting a threshold value of the mutual information value, and selecting the influence parameter corresponding to the mutual information value larger than the set threshold value as an input variable of the fly ash carbon content measurement model.
The ten influence parameters as the input variables of the fly ash carbon content measurement model are respectively as follows: secondary air temperature B, secondary air temperature A, total steam flow, B side oxygen quantity, A primary air quantity, total coal quantity, oxygen quantity, load and B primary air quantity.
The method adopts a Support Vector Machine (SVM) in combination with Particle Swarm Optimization (PSO) to identify a fly ash carbon content measurement model, divides collected field operation data into training data and test data, establishes a fly ash carbon content measurement model aiming at the training data, completes the identification of non-linear model parameters with multiple input and single output of fly ash carbon content by utilizing a particle swarm optimization algorithm, and finally uses the test data for the identified model to test the generalization capability of the model, proves that the fly ash carbon content soft measurement model established based on the PSO-SVR can accurately reflect the carbon content in the fly ash, and the basic block diagram of the soft measurement model establishment is shown in figure 2.
The support vector machine is used as a brand-new machine learning algorithm, is an implementation of the SRM principle, and can obtain a model with the minimum structural risk, so that the maximum popularization capability of the model is ensured. The SVM algorithm was originally developed from the study of pattern recognition problems and then generalized to the regression estimation problem. The SVM essentially solves a convex optimization problem, theoretically, a global optimal solution can be obtained, and thus, a local extremum problem does not exist. The SVM is suitable for the problem of small samples on the basis of a statistical learning theory, and because the SVM adopts a structure risk minimization principle, the best compromise can be found between the complexity and the popularization capability of a learning machine, so that the problem of 'over-learning' is effectively avoided. The SVM introduces a kernel function, high-dimensional inner product operation in a Hilbert space is avoided, algorithm complexity is independent of the dimension of the Hilbert space, and the problem of dimension disaster is avoided. Due to these excellent properties, SVMs gradually replace the dominance of neural networks in the fields of pattern recognition, regression estimation, and the like.
Constructing a proper kernel K (x) according to the expansion theorem of the Mercer kernel function aiming at the nonlinear sample seti,xj)＝Φ(xi)T·Φ(xj) And a nonlinear mapping relation phi (x), mapping the sample space to another space with higher dimensionality, and solving a new linear problem in the new dimensionality space.
Referring to the Mercer theorem, there are three more commonly used kernel functions:
(1) linear kernel function: k (x)i,xj)＝xi T·xj
(2) Polynomial kernel function: k (x)i,xj)＝(xi T·xj+1)d
(3) Radial basis kernel function: k (x)i,xj)＝exp(-g.||xi-xj||2)
The basic principle of the kernel function is to map the input space data to another high-dimensional space by using a nonlinear mapping relationship, thereby improving the separation characteristic of the sample space and increasing the operation degree through inner products.
In the step of establishing the measurement model of the carbon content in the fly ash by adopting the support vector machine and the particle swarm optimization algorithm, establishing the measurement model of the carbon content in the fly ash of the support vector machine, optimizing the measurement model of the carbon content in the fly ash by the particle swarm optimization algorithm, and specifically comprising the following steps:
the inner product formula of the kernel function of the support vector machine is expressed as formula (1):
Φ(xi)T·Φ(xj)＝K(xi,xj) (1)
wherein phi (x) is a non-mapping relation, K (x)i,xj) Is a kernel function of the support vector machine;
set a set of non-linear sample sets S { (x)i,yi),i＝1,2… l, where x isi∈RnIn order to predict the factors, the method comprises the following steps,
yi∈Rnthen y ═ f (x) for the predicted objects have a typical nonlinear relationship, and the described problem is a nonlinear regression problem;
the criterion is a linear epsilon-insensitive loss function, as shown in equation (2):
determining generalization capability and number of SVM support vectors by epsilon, and defining loss as 0 if the error of a predicted value and an actually measured value is within the epsilon range;
after mapping the sample space to another high-dimensional space and implementing corresponding linear regression analysis, using reverse reduction to make the initial sample data complete regression operation, the functional relationship expressed by the process is shown as formula (3):
f(x)＝w·Φ(x)+b (3)
by adding relaxation factors xii＝(ξ1,…ξn) 1,2, …, n, expressing the constrained optimization objective function problem as equation (4):
wherein, w and b are weight vector and bias, c is punishment parameter keeping positive correlation with fitting degree;
like equation (5), introduce Lagrange's function,
lagrange factor
The extreme value of L should accord with KKT condition to reach ai,
λi,
Maximization of (2);
the corresponding dual problem is converted, as shown in formula (6):
substituting w, b into f (x) w · Φ (x) + b to obtain a nonlinear SVM expression with a kernel function:
in the step of optimizing the fly ash carbon content measurement model by the particle swarm optimization algorithm, the radial basis kernel function parameter g and the penalty coefficient C in the SVM expression of the fly ash carbon content measurement model are adjusted and optimized by the particle swarm optimization algorithm.
The kernel function of the support vector regression model selects a radial basis kernel function K (x)i,xj)＝exp(-g.||xi-xj||2) And the radial basis kernel function parameter g, the penalty coefficient C, the insensitive loss function epsilon and the final allowable error e during model training are training parameters of the SVM model. The accuracy of the SVM model is largely determined by these four parameters, and their determination is therefore very important for modeling the support vector machine. The two parameters of epsilon and e are artificially controlled, and the influence on the prediction capability of the model is very small and can be almost ignored due to the requirement degree of a modeler on the precision of the model; and the C and g parameters directly influence the calculation process of modeling and the properties of the model, and have great influence on the prediction accuracy and the generalization capability of the model.
In order to obtain a satisfactory set of parameters C and g, the invention optimizes the model parameters by using a global optimization algorithm. In engineering, various optimization algorithms can be applied, such as a random search method, a genetic algorithm, a particle swarm algorithm and the like. Literature research shows that the particle swarm optimization is good in finding the optimal solution efficiency. Therefore, the method selects the particle swarm optimization algorithm to carry out parameter optimization on the fly ash carbon content soft measurement model.
In order to try an automatic modeling idea combining a support vector machine and an optimization algorithm, MATLAB programming is adopted to simulate the model, and the main purpose is to realize the optimization of support vector machine model parameters C and g by utilizing the optimization function of a particle swarm algorithm, so that the model has good prediction capability. The SVM model parameters are obtained by optimizing the training data parameters through a particle swarm algorithm instead of relying on the experience of a user, and the method for selecting the parameters and the traditional support vector regression method are combined to form a support vector machine modeling method based on the particle swarm. The algorithm flow is shown in fig. 3.
The method for optimizing the fly ash carbon content measurement model by the particle swarm optimization algorithm comprises the following steps:
initializing particle swarm parameters; wherein, the parameters to be initialized at least comprise: inertia factors, population size, particle speed, particle position and maximum iteration number;
determining a fitness function; the fitness function is the mean square error of the fly ash carbon content value predicted by the fly ash carbon content measurement model and the fly ash carbon content value measured by the local carbon meter;
setting particle swarm parameter optimization termination conditions; setting iteration times, and stopping optimizing when the iteration times are reached;
and when the optimization is finished, combining the optimal parameters searched by the particle swarm algorithm in the fitness function meaning by the fly ash carbon content measurement model regressed by the support vector machine to monitor the fly ash carbon content numerical value.
And carrying out soft measurement modeling on the carbon content of the fly ash by applying a PSO-SVM algorithm. The model input variables selected from the foregoing are: the temperature of secondary air at the B side, the temperature of secondary air at the A side, the total steam flow, the oxygen quantity at the B side, the oxygen quantity at the A side, the primary air quantity at the A side, the total coal quantity, the oxygen quantity, the load and the primary air quantity at the B side. And establishing a fly ash carbon content measurement model aiming at the operation data of the power plant, and taking the operation data of the power plant SIS system in about two months and the hand-measured value of the fly ash carbon content at the corresponding time point. The first 70% of the data were selected as training data and the last 30% as test data.
In order to measure the training and testing effects of the neural network model, Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) which are representative in the control theory are used as evaluation indexes.
Wherein is yiThe actual output of the power supply is,
is the predicted output of the model, and n is the number of samples.
The evaluation result shows that the model fitting degree is good, the accuracy is high, and the accurate model can be established and accurately estimated for the carbon content of the boiler fly ash.
The carbon content of the fly ash is closely related to the boiler efficiency and is a basis for judging the quality of boiler combustion, however, the carbon content characteristic of the fly ash is influenced by various factors, the relationship is complex, and the accurate measurement of the carbon content of the fly ash is difficult. The project adopts the particle swarm optimization algorithm to identify the training data to obtain the learning parameters, effectively reduces the influence of human factors and the problem of uncertainty of precision in the modeling process, and has the advantages of high precision, good generalization capability and the like. The fly ash carbon content is modeled by adopting a PSO-SVM algorithm, and test data is input into the identified model, so that the fly ash carbon content model based on the PSO-SVM algorithm is verified to have good generalization capability, and the carbon content in the fly ash can be accurately reflected.
While the present invention has been described with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, which are illustrative and not restrictive, and it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (6)
1. A fly ash carbon content prediction method based on particle swarm parameter optimization is characterized by comprising the following steps:
extracting the historical data of the carbon content of the fly ash from a database, screening the influence parameters of the historical data corresponding to the carbon content of the fly ash, and dividing a training set and a test set;
establishing a fly ash carbon content measurement model by adopting a support vector machine and a particle swarm optimization algorithm, inputting the influence parameters corresponding to the fly ash carbon content historical data in a training set as input variables into the fly ash carbon content measurement model for training, and testing the training result through a test set after the training is finished;
and inputting the real-time influence parameters of the carbon content of the fly ash into the trained fly ash carbon content measurement model, and outputting the corresponding fly ash carbon content.
2. The method for predicting fly ash carbon content based on particle swarm parameter optimization according to claim 1, wherein the step of screening the influence parameters of the historical data corresponding to fly ash carbon content comprises:
primarily selecting 28 combustion parameters as the influence parameters of the carbon content of the fly ash according to the combustion operation parameters of the boiler; wherein, the 28 combustion parameters are load, total air quantity, total coal quantity, total steam quantity, oxygen quantity, A side oxygen quantity, B side oxygen quantity, A side primary air pressure, B side primary air pressure, 2 air preheater outlet secondary air temperatures, 6 coal mill outlet temperatures, 2 hearth outlet temperatures and 9 secondary air door openness respectively;
carrying out data deduplication and abnormal value removal on the influence parameters, and deleting repeated data in the influence parameters and data with an output value exceeding a normal range caused by noise interference factors;
and extracting data characteristic quantity of the primarily selected influence parameters according to a mutual information theory, calculating the weight of each influence parameter, and selecting the influence parameters with the weight larger than a preset threshold value as input variables of the fly ash carbon content measurement model.
3. The particle swarm parameter optimization-based fly ash carbon content prediction method according to claim 2, wherein the ten influence parameters as the input variables of the fly ash carbon content measurement model are respectively: secondary air temperature B, secondary air temperature A, total steam flow, B side oxygen quantity, A primary air quantity, total coal quantity, oxygen quantity, load and B primary air quantity.
4. The method for predicting carbon content in fly ash based on particle swarm parameter optimization according to claim 1, wherein in the step of establishing the measurement model of carbon content in fly ash by using a support vector machine and a particle swarm optimization algorithm, the measurement model of carbon content in fly ash is established by using the support vector machine, the measurement model of carbon content in fly ash is optimized by using the particle swarm optimization algorithm, and the establishment of the measurement model of carbon content in fly ash by using the support vector machine specifically comprises the steps of:
the inner product formula of the kernel function of the support vector machine is expressed as formula (1):
Φ(xi)T·Φ(xj)＝K(xi,xj) (1)
wherein phi (x) is a non-mapping relation, K (x)i,xj) Is a kernel function of the support vector machine;
set a set of non-linear sample sets S { (x)i,yi) I is 1,2, … l, where x isi∈RnIn order to predict the factors, the method comprises the following steps,
yi∈Rnthen y ═ f (x) for the predicted objects have a typical nonlinear relationship, and the described problem is a nonlinear regression problem;
the criterion is a linear epsilon-insensitive loss function, as shown in equation (2):
determining generalization capability and number of SVM support vectors by epsilon, and defining loss as 0 if the error of a predicted value and an actually measured value is within the epsilon range;
after mapping the sample space to another high-dimensional space and implementing corresponding linear regression analysis, using reverse reduction to make the initial sample data complete regression operation, the functional relationship expressed by the process is shown as formula (3):
f(x)＝w·Φ(x)+b (3)
by adding relaxation factors xii＝(ξ1,…ξn) 1,2, …, n, expressing the constrained optimization objective function problem as equation (4):
wherein, w and b are weight vector and bias, c is punishment parameter keeping positive correlation with fitting degree;
like equation (5), introduce Lagrange's function,
lagrange factor ai,
The extreme value of L should accord with Karush-Kuhn-Tucker condition (KKT condition), namely the constraint optimization problem of equality processed by Lagrange multiplier method is popularized to inequality, and the necessary conditions of the optimal solution include the constant equation, original feasibility, dual feasibility and complementary relaxation of Lagrange function, so as to reach a pair ai,
λi,
Maximization of (2);
the corresponding dual problem is converted, as shown in formula (6):
substituting w, b into f (x) w · Φ (x) + b to obtain a nonlinear SVM expression with a kernel function:
5. the particle swarm parameter optimization-based fly ash carbon content prediction method according to claim 4, wherein in the step of optimizing the fly ash carbon content measurement model through the particle swarm optimization algorithm, the radial basis kernel function parameter g and the penalty coefficient C in the SVM expression of the fly ash carbon content measurement model are adjusted and optimized through the particle swarm optimization algorithm.
6. The particle swarm parameter optimization-based fly ash carbon content prediction method according to claim 5, wherein the step of optimizing the fly ash carbon content measurement model through a particle swarm optimization algorithm comprises:
initializing particle swarm parameters; wherein, the parameters to be initialized at least comprise: inertia factors, population size, particle speed, particle position and maximum iteration number;
determining a fitness function; the fitness function is the mean square error of the fly ash carbon content value predicted by the fly ash carbon content measurement model and the fly ash carbon content value measured by the local carbon meter;
setting particle swarm parameter optimization termination conditions; setting iteration times, and stopping optimizing when the iteration times are reached;
and when the optimization is finished, combining the optimal parameters searched by the particle swarm algorithm in the fitness function meaning by the fly ash carbon content measurement model regressed by the support vector machine to monitor the fly ash carbon content numerical value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011041478.0A CN112287598A (en) | 2020-09-28 | 2020-09-28 | Fly ash carbon content prediction method based on particle swarm parameter optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011041478.0A CN112287598A (en) | 2020-09-28 | 2020-09-28 | Fly ash carbon content prediction method based on particle swarm parameter optimization |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112287598A true CN112287598A (en) | 2021-01-29 |
Family
ID=74421587
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011041478.0A Pending CN112287598A (en) | 2020-09-28 | 2020-09-28 | Fly ash carbon content prediction method based on particle swarm parameter optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112287598A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113358811A (en) * | 2021-05-26 | 2021-09-07 | 中电华创电力技术研究有限公司 | Soft measurement method and device for fly ash carbon content based on firing method calibration |
| CN113836729A (en) * | 2021-09-27 | 2021-12-24 | 国能龙源环保有限公司 | Method for reducing content of combustible materials in fly ash of boiler of thermal power plant |
| CN114637212A (en) * | 2022-05-17 | 2022-06-17 | 浙江浩普智能科技有限公司 | Oxygen content prediction control method and system |
| CN117724431A (en) * | 2024-02-07 | 2024-03-19 | 长沙矿冶研究院有限责任公司 | Control method and control system for granularity of reaction materials in reaction kettle |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102778538A (en) * | 2012-07-06 | 2012-11-14 | 广东电网公司电力科学研究院 | Soft measuring method based on improved SVM (Support Vector Machine) for measuring boiler unburned carbon content in fly ash |
| CN106649919A (en) * | 2016-09-13 | 2017-05-10 | 上海电气电站环保工程有限公司 | Method and system for predicting carbon content of fly ash in coal-fired power plant boiler |
-
2020
- 2020-09-28 CN CN202011041478.0A patent/CN112287598A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102778538A (en) * | 2012-07-06 | 2012-11-14 | 广东电网公司电力科学研究院 | Soft measuring method based on improved SVM (Support Vector Machine) for measuring boiler unburned carbon content in fly ash |
| CN106649919A (en) * | 2016-09-13 | 2017-05-10 | 上海电气电站环保工程有限公司 | Method and system for predicting carbon content of fly ash in coal-fired power plant boiler |
Non-Patent Citations (2)
| Title |
|---|
| 王建峰 等: ""基于特征选择的飞灰含碳量影响因素分析"", 《仪器仪表用户》 * |
| 王涛 等: ""基于改进粒子群算法优化的飞灰含碳量建模研究"", 《黑龙江电力》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113358811A (en) * | 2021-05-26 | 2021-09-07 | 中电华创电力技术研究有限公司 | Soft measurement method and device for fly ash carbon content based on firing method calibration |
| CN113836729A (en) * | 2021-09-27 | 2021-12-24 | 国能龙源环保有限公司 | Method for reducing content of combustible materials in fly ash of boiler of thermal power plant |
| CN114637212A (en) * | 2022-05-17 | 2022-06-17 | 浙江浩普智能科技有限公司 | Oxygen content prediction control method and system |
| CN117724431A (en) * | 2024-02-07 | 2024-03-19 | 长沙矿冶研究院有限责任公司 | Control method and control system for granularity of reaction materials in reaction kettle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112287598A (en) | Fly ash carbon content prediction method based on particle swarm parameter optimization | |
| Shi et al. | Combustion optimization of ultra supercritical boiler based on artificial intelligence | |
| CN107016455B (en) | Prediction system and method for oxygen content of flue gas at hearth outlet of circulating fluidized bed household garbage incineration boiler | |
| Álvarez et al. | CFD modelling of oxy-coal combustion in an entrained flow reactor | |
| CN111754093B (en) | Fly ash carbon content prediction method based on coal quality prediction and PSO-SVM | |
| Hong et al. | Modeling for the bed temperature 2D-interval prediction of CFB boilers based on long-short term memory network | |
| CN108549792A (en) | A kind of solid waste burning process dioxin emission concentration flexible measurement method based on latent structure mapping algorithm | |
| WO2021159585A1 (en) | Dioxin emission concentration prediction method | |
| CN111780127B (en) | Garbage incinerator combustion management system | |
| CN109002068A (en) | A kind of temperature optimization control method of quirk | |
| Qiao | A novel online modeling for NOx generation prediction in coal-fired boiler | |
| Ye et al. | Multi-objective modeling of boiler combustion based on feature fusion and Bayesian optimization | |
| CN111401652A (en) | Boiler optimization method and system based on CO online detection | |
| Chen et al. | Modelling the furnace temperature field of a municipal solid waste incinerator using the numerical simulation and the deep forest regression algorithm | |
| Więckowski et al. | Numerical investigation of temperature distribution in the furnace of a coal fired grate boiler in part load conditions | |
| CN116401948A (en) | Online prediction method and system for generating amount of power station boiler ash based on LSTM | |
| Saptoro et al. | Prediction of coal hydrogen content for combustion control in power utility using neural network approach | |
| Tan et al. | Use of artificial intelligence techniques for optimisation of co-combustion of coal with biomass | |
| Wang et al. | Application research based on GA-FWA in prediction of sintering burning through point | |
| Sujatha et al. | Soft sensor with shape descriptors for flame quality prediction based on lstm regression | |
| Ndizihiwe et al. | Review of Stoichiometric Technique for Better Performance of the Boiler | |
| Manke | A Performance Comparison of Different Back Propagation Neural Networks for Nitrogen Oxides Emission Prediction in Thermal Power Plant | |
| Yan et al. | Improved LS-SVM Boiler Combustion Model Based on Affinity Propagation | |
| Tang et al. | Artificial neural networks model for predicting oxygen content in flue gas of power plant | |
| Zhu et al. | Mechanism-enhanced data-driven method for the joint optimization of boiler combustion and selective catalytic reduction systems considering gas temperature deviations |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210129 |