CN115657466A - Boiler system of intelligent control ammonia input volume - Google Patents
Boiler system of intelligent control ammonia input volume Download PDFInfo
- Publication number
- CN115657466A CN115657466A CN202211089112.XA CN202211089112A CN115657466A CN 115657466 A CN115657466 A CN 115657466A CN 202211089112 A CN202211089112 A CN 202211089112A CN 115657466 A CN115657466 A CN 115657466A
- Authority
- CN
- China
- Prior art keywords
- concentration
- boiler
- neural network
- outlet
- ammonia
- 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
Images
Landscapes
- Treating Waste Gases (AREA)
Abstract
The invention discloses a boiler system for intelligently controlling the input amount of ammonia gas, which comprises: a furnace into which water is fed from an inlet pipe into a water-cooled wall; a coal-fired burner capable of heating the furnace; a pulverized coal pipe capable of supplying pulverized coal to the burner; the hearth is connected with a flue, and NO is arranged in the flue x The concentration on-line measuring instrument and the flue gas flowmeter have ammonia gas input device in the flue and do not need to be monitored x The concentration and the flue gas flow automatically control the input amount of the ammonia gas. The invention realizes the accurate control of ultralow emission of tail gas pollutants and the efficient operation of the SCR denitration system.
Description
Technical Field
The invention relates to the field of heat energy and power engineering, belongs to the technical field of intelligent control of coal-fired boilers, and particularly relates to a boiler system for intelligently controlling the input amount of ammonia gas.
Background
The traditional design of coal-fired boilers measures their individual performance in multi-configuration combustion by analyzing the combustion products of all burners as a whole, e.g. measuring oxygen and/or carbon dioxide in the flue gas channel.
This design basis is in fact erroneous, in that conventional designs do not provide for control or measurement of the fuel-air ratio of the individual burners at any particular load on the boiler; and in particular, does not include control measurements or calibrations of fuel-to-air ratio or load variations on the boiler, which is a common problem when the boiler is associated with a power generation system. In conventional designs, there is no separate fuel flow accurately metered to the individual combustors individually in this configuration. Such measurements are typically made within the main fuel supply system of the boiler as a whole.
The flue gas generated by burning coal in thermal power plant is atmosphere NO x One of the important sources of pollution causes the emission of nitrogen oxides to be more and more, and a large part of large-area haze in northern areas in recent years is caused by tail gas of a power plantIn order to respond to national policies and construct a green and harmonious society, industrial technologies are developed to reduce NO x The amount of discharge of (c). In response to national policy, to reach NO x Emission standard also practices thrift the running cost and promotes denitration efficiency in order to control SCR system and spout ammonia volume simultaneously, just constantly optimizes boiler combustion system and SCR deNOx systems, carries out intelligent predictive control.
The operation condition of the boiler may frequently change according to the actual condition, so that the flue gas parameters at the outlet of the hearth frequently change, and NO at the outlet of the hearth is caused x The concentration may be too high when entering the SCR reactor, affecting the subsequent denitration process, and also affecting the denitration efficiency. In addition, currently NO x The detection of boiler operation condition parameters of a boiler combustion system and furnace outlet flue gas parameter signals in an emission control system is lagged, and the reaction process of an SCR reactor is greatly delayed, so that the traditional PID control cannot obtain a satisfactory control effect on the SCR denitration process with complex reaction mechanism, nonlinear characteristics, large delay, multivariable coupling and the like, and the DMC-PID cascade controller is combined with a BP neural network to just overcome the difficulties.
Therefore, it is imperative to develop a boiler system for intelligently controlling the ammonia gas input amount, which has important significance for safe and economic operation of a boiler combustion system and an SCR denitration device, and the technical problem is solved by the scheme.
Disclosure of Invention
The invention designs a boiler system for intelligently controlling the input amount of ammonia gas, which is mainly used for NO in the SCR denitration process x The concentration prediction control system is optimally designed, so that the timely and accurate prediction control on the ammonia spraying amount in the SCR denitration system is realized, and the national NO is achieved with the optimal operation cost and denitration efficiency x Ultra-low emission standards; prediction of furnace outlet NO mainly through BP neural network model x Concentration (SCR reactor inlet NO) x Concentration), the flue gas parameters, the opening degree of an ammonia spraying valve and the NO at the outlet of the SCR reactor are mixed again x The measured concentration value is used as the input variable of the BP neural network, and the online mapping approaches to the NO at the outlet of the SCR reactor x Between predicted value of concentration and input variableAnd is related to SCR reactor outlet NO x The deviation feedback signal between the measured concentration value and the set value is jointly transmitted to the DMC prediction controller and the PID controller to jointly and accurately and quickly control the ammonia injection amount, and because the feedforward signal and the feedback signal in the invention are both related to the BP neural network model for prediction in advance, the system can deal with the complex change of the operation condition in advance, can quickly and timely obtain response, and avoids the reaction of the SCR denitration system and the lag of detection.
A boiler system for intelligently controlling ammonia gas input, comprising: a furnace into which water is fed from an inlet duct into a water-cooled wall; a coal-fired burner capable of heating the furnace; a pulverized coal pipe capable of supplying pulverized coal to the burner; the hearth is connected with a flue, and NO is arranged in the flue x The concentration on-line measuring instrument and the flue gas flowmeter have ammonia gas input device in the flue and do not need to be monitored x The concentration and the flue gas flow automatically control the input amount of the ammonia gas.
With prediction of NO x The concentration is increased continuously, and the ammonia gas demand is increased nonlinearly.
With NO x The concentration is increased continuously, and the required amount of ammonia is increased more and more greatly.
The ammonia gas demand increases nonlinearly with the increasing flue gas amount.
The required amount of ammonia gas increases more and more along with the increase of the smoke gas amount.
The system also comprises a DMC predictive controller and a PID controller; collecting boiler operation data and carrying out correlation preprocessing, constructing a BP neural network intelligent prediction model, and predicting NO at a furnace outlet x Concentration;
DMC is used as a main controller, PID is used as a secondary controller to form a DMC-PID cascade feedback control structure, rolling optimization and feedback correction are carried out, prediction deviation compensation is carried out by combining a BP neural network model, and online mapping approaches to the outlet NO of an SCR denitration system x The complicated nonlinear relation between the predicted concentration value and each variable can directly realize accurate control on the ammonia injection amount of the SCR denitration system in advance according to different working conditions of a boiler combustion systemAnd (5) preparing.
At the outlet of the SCR denitration system, NO is installed x The concentration on-line measuring instrument can monitor and measure NO at the outlet of the SCR denitration system in real time x Concentration and transmitted as a feedback input signal to the BP neural network and DMC predictive controller.
And acquiring the operation data of the boiler combustion system and the operation data of the SCR denitration system at different time intervals under different working conditions, wherein the operation data need to include the data under different working conditions.
Establishing a BP neural network prediction model based on collected historical data of denitration operation of a boiler combustion system and an SCR reactor, setting the BP neural network prediction model into an input layer, a hidden layer and an output layer, and predicting NO at the inlet of the SCR denitration system in real time on line according to the established BP neural network x Concentration and NO at outlet of SCR denitration system x And (4) concentration.
According to historical data of denitration operation of a boiler combustion system and an SCR (selective catalytic reduction) reactor, determining an input variable with large correlation with a target function, establishing the number of neurons of a corresponding hidden layer and the number of layers of the hidden layer according to the number of the input variables, which is related to the accuracy of a BP (back propagation) neural network prediction result, and realizing the NO at the outlet of a hearth x Concentration prediction and SCR denitration system outlet position NO x The concentration prediction value and each variable have complex nonlinear relations.
PID controller predicts NO input by controller based on DMC x The concentration signal can calculate and control the ammonia injection amount required by the SCR denitration system in real time, so that the initial control of the opening of the ammonia injection valve is achieved, then prediction deviation compensation, rolling optimization and feedback correction are carried out by combining a BP neural network model, the opening of the ammonia injection valve is further controlled rapidly and accurately, and the delay lag disadvantage brought by a common control system can be avoided.
In order to facilitate further understanding of the technical scheme, in more detail, the technical scheme design idea of the invention is specifically realized as follows:
firstly, designing a whole predictive control system: accurate and timely prediction of furnace outlet NO by constructing BP neural network model x Concentration according to furnace outlet NO x In response to the predicted concentration value, the DMC-PID cascade controller preliminarily controls the ammonia injection amount required by the SCR reactor; then the neural network maps and approaches to the outlet NO of the SCR reactor on line x Nonlinear function relation between predicted concentration value and multiple variable parameters and SCR outlet position NO x The deviation feedback signal between the measured concentration value and the set value is transmitted to the DMC-PID controller together, the opening of the ammonia injection valve is accurately regulated and controlled again, and rolling optimization is carried out, so that the NO can be predicted in advance x The concentration and the opening degree of the ammonia injection valve are controlled to reach the ultralow emission standard, and the hysteresis of system reaction and detection is avoided.
The specific operation steps are as follows:
step1: based on historical operating data of a combustion system of a coal-fired thermal power generating unit, collecting multiple groups of boiler operating parameters, and establishing accurate prediction of NO at a hearth outlet x A BP neural network model of concentration;
description of the drawings: the BP neural network model comprises an input layer, a hidden layer and an output layer, and the working principle is as follows:
1. the design has more input variables, and for the sake of simplifying the principle, an input layer is supposed to have three variables (boiler parameters); the hidden layer is a layer and comprises two neurons; the output level is an objective function (furnace outlet NO) x Concentration);
2. firstly, a hidden layer connecting an input layer and an output layer is provided with an activation function for processing data, and a sigmoid function is generally adopted:
3. input data of the input layer:
wherein x 1 x 2 x 3 Respectively referring to the operating parameters of the boiler, the BP neural network can simulate x 1 x 2 x 3 And (3) weight matrixes corresponding to two neurons of the hidden layer:
hidden layerCorresponding to the output layer weight matrix: w is a 0 ＝[w 7 w 8 ]Hidden layer neuron threshold matrix:
output layer threshold: b is a mixture of 0 ；
4. According to the above formula, the method comprises the following steps,
substituting the Z value into the sigmoid function:
the value after the hidden layer activation is derived:
final objective function prediction (furnace outlet NO) x Concentration): y = w 0 ·h+b 0 ＝w 7 h 1 +w 8 h 2 +b 0 ；
The sigmoid function is an activation function of neurons in the hidden layer, and each parameter in the input layer corresponds to a threshold value and a weight value for each neuron in the hidden layer, as shown in the formula
And
then inputting variables and corresponding threshold values and weight values to calculate the independent variable Z value of the sigmoid function, substituting the Z value into the f (Z) function, and inputting layer variables
Data after activation by the hidden layer activation function is
Hidden layer Each neuron to target function y(concentration of NOx at furnace outlet) also has corresponding threshold and weight, as shown in the above formula [ w ] 7 w 8 ]And b 0 Calculating the value of the target function y (the furnace outlet NO) by using the data h after activating the function and the threshold value and the weight value x Concentration), i.e. the above formula y = w 0 ·h+b 0 ＝w 7 h 1 +w 8 h 2 +b 0 (ii) a Through the calculation step, the target function is predicted by an input layer, a hidden layer and an output layer, and the target function is mainly obtained through the calculation of an activation function according to the threshold and the weight value associated between each layer of data and the next layer of data.
5. The steps are the whole principle and formula of BP neural network prediction, the predicted value is the result predicted by the BP neural network, wherein the weight and the threshold value between each layer of data are also the values of the BP neural network when the simulation error is minimum, and the formulas are not required to be written in practical situations because each boiler operation parameter and the NO at the outlet of a furnace chamber x The predicted value influence relationship of the concentration can be converted into weight and threshold value provided by BP neural network, and in addition, how boiler operation parameters influence the NO at the outlet of the furnace x The concentration, which are important influencing factors, can be proved by correlation calculation in a later step.
Carrying out simulation training prediction on the currently acquired unit operation data under different working conditions by using a BP neural network model, and predicting the furnace chamber outlet NO under the current boiler operation working condition x And (4) concentration.
step2: step1 furnace outlet NO x The predicted value of the concentration is the NO at the inlet of the SCR reactor x The concentration, according to the concentration of the nitrogen oxide at the moment, the ammonia injection amount required by the system operation under the current working condition is calculated, the signal of the required ammonia injection amount is transmitted to a DMC-PID controller, then an ammonia injection valve required by the SCR system is preliminarily controlled, and NO at the outlet of the SCR reactor is measured x The true value of (d).
step3: online approximation of SCR reactor outlet NO by BP neural network model x Predicted concentration value, flue gas parameter and NO at furnace outlet x Predicted concentration value, ammonia spraying valve opening degree and SCR (selective catalytic reduction) reactor outlet NO x A complex non-linear relationship between the measured values of (a).
Description of the drawings: the BP neural network here functions as in step1, i.e. to couple the SCR reactor outlet NO x As a prediction of the target function, wherein these important influencing factors and the SCR reactor outlet NO are reflected x The BP neural network gives the weight and the threshold value between each layer of data according to the nonlinear relation between measured values, and the specific principle is as described in step 1.
Transmitting the signal and a deviation feedback signal to a DMC-PID cascade controller, and re-regulating an ammonia injection valve based on the signal DMC-PID cascade controller to enable the outlet NO of the SCR reactor x The concentration meets the emission standard.
So far, the whole process design steps are completely realized, and the coal-fired flue gas NO is regulated and controlled based on systematic intelligent prediction and the whole process x The emission control system is basically realized, the ammonia injection amount in the SCR reactor is controlled by combining a feedforward signal and a deviation feedback signal together, so that NO in the flue gas is reduced x The emission reaches the national standard, even can reach ultralow emission, and simultaneously can realize the economic operation of each equipment.
Regarding the establishment of the BP neural network prediction model in step1, the normal operation condition of the actual power plant unit and the selection condition of the boiler operation parameters are combined.
Further, according to a large number of references and the correlation among calculation parameters, the correlation between the input variable and the output variable is selected to be more than 0.5, and the correlation is as the name implies: the data association and the mutual influence rule refer to the association and mutual influence rule among the data, some of the data association and mutual influence rule are promotion relations, the data association is called positive association, the numerical value is larger than 0, the larger the numerical value is, the larger the association among the data is, the stronger the mutual influence is, and the data association and mutual influence rule can be called as main factors, namely important influence factors; some of the data are in inhibition relation, called negative correlation, the numerical value is smaller than 0, the smaller the numerical value is, the tighter the relation among the data is, and the stronger the influence is; in the operation of a common boiler plant, when the correlation is about 0.5, the data can be proved to be closely related and have stronger mutual influence. The output variable being the furnace outlet NO as defined above x Concentration (inlet NO of SCR reactor) x Concentration).
Further, the selected boiler operation parameters are input variables: parameters such as boiler load, primary air quantity, secondary air quantity, over-fire air quantity, oxygen quantity, hearth temperature and the like are used as neural networks to predict NO at hearth outlet x Input variables for concentration.
Further, a BP neural network model is established according to the selected input parameters, and the hidden layer of the neural network is designed into a double layer in consideration of the model prediction performance due to more input variables.
Further, the selected input variable data is preprocessed, each group of abnormal data including data under extreme conditions is removed, and then the same data is deleted, for example, multiple groups of data under the same conditions or similar conditions.
Further, the BP neural network model predicts the NO of the outlet of the furnace chamber x Concentration is the inlet NO of the SCR reactor x Concentration according to NO x And calculating the corresponding ammonia injection amount according to the concentration, and controlling the opening of an ammonia injection valve by a DMC-PID cascade controller according to the ammonia injection amount required by the SCR reactor, thereby preliminarily and accurately controlling the ammonia injection amount in the SCR reactor.
Further, the neural network is used for mapping and approaching the outlet NO of the SCR reactor on line x The nonlinear relation between the predicted concentration value and the variable parameter is transmitted to a DMC prediction controller together with a deviation feedback signal, and the ammonia injection amount required by the system is adjusted and corrected in time to meet the requirement of the NO at the outlet of the SCR denitration system x The concentration set value is required to meet the national NO x And (4) emission requirement standards. When the predictive control system detects the working condition change of a boiler combustion system and an SCR reactor, the system can adaptively adjust DMC-PID parameters, thereby realizing optimal control.
The invention has the following advantages:
(1) Can be in real time according to boiler operating parameter under the different operating modes, the required ammonia injection amount of direct prediction value control SCR deNOx systems, can make the system under normal economic operation condition, regulate and control in advance, according to source direct control result, the delay lag shortcoming that general control system goes the regulation and control process to bring according to the result has been avoided, the control problem that detection system delay and the big delay characteristic of SCR deNOx systems brought has also been overcome simultaneously, reserve more reaction time for control system like this, reduce the control overshoot, thereby the fault-tolerant rate of lift system. The ammonia escape is reduced on the premise of saving the ammonia spraying amount, and the accurate control of the ammonia spraying amount is realized, so that the national emission standard is met, and the economic operation of the system is also met.
(2) The control system is a highly automatic control system, and the operation of workers is greatly reduced, so that the error of manual operation is reduced, and the ammonia spraying amount of the SCR denitration system is controlled more accurately.
(3) The deviation between the concentration of nitrogen oxides at the outlet of the SCR denitration system and a set value can be reduced to be within an acceptable range, the required ammonia injection amount is corrected according to different deviation signals, and NO is achieved under different operation conditions of the system x And (4) stable discharge.
(4) Coal-fired flue gas NO regulated and controlled in whole process x The system response rate of the emission control system is high, the investment is low, the modification period is short, and the denitration efficiency is high.
Drawings
FIG. 1 is a basic structure of a single-layer hidden layer BP neural network.
Fig. 2 is a basic structure of a two-layer hidden layer BP neural network.
FIG. 3 is a flow chart of the control of the amount of ammonia required by the system.
FIG. 4 is a diagram of neural network prediction of furnace outlet NO of the present invention x Concentration flow chart.
FIG. 5 shows the SCR denitration system NO of the present invention x And (4) an overall flow chart of emission prediction control.
Detailed Description
In order that the present invention may be more readily and clearly understood, the following detailed description of the present invention is given with reference to specific embodiments thereof, and the accompanying drawings are included to illustrate and not limit the scope of the present invention, and various equivalent modifications thereof will fall within the scope of the appended claims after reading the present invention.
First, it should be noted that:
the boiler combustion system is the combination of equipment and corresponding smoke, wind and coal (coal powder) pipelines required for fully combusting fuel in a boiler hearth and discharging smoke generated by combustion into the atmosphere, and the operation condition is directly related to the NO at the outlet of the hearth x And (4) concentration. The SCR denitration system is a complex control system, the reaction is influenced by factors such as catalyst activity, flue gas temperature, flue gas flow rate and oxygen concentration, a mathematical model established according to a mechanism method can be operated with a good effect when the working condition is unchanged, but the mathematical model can be greatly changed when the working condition is changed. Therefore, in the face of ultra-low emission control of nitrogen oxides, predictive control is constantly optimized for both boiler combustion systems and SCR denitration systems.
The single-layer hidden layer BP neural network structure in FIG. 1 comprises an input layer, a hidden layer and an output layer, wherein the output layer is a target function to be predicted, and the hidden layer neuron number empirical formula:
h is the number of hidden layer neurons, m is the number of input layer nodes, i.e. the dimension of the input variable, n is the number of output variables, i.e. the number of objective functions, and c is a number between 0 and 10.
Fig. 2 is a model of a double-layer hidden layer BP neural network, and since the number of input variables is increased, the number of hidden layers is increased in order to improve the accuracy of the model, the design adopts the double-layer hidden layer, and other design rules are consistent with those of a single-layer hidden layer BP neural network.
FIG. 3 shows the calculation of ammonia amount based on flue gas parameters and NOx concentration at the inlet of the SCR reactor, and the signal is transmitted to the PID controller to control the opening of the ammonia injection valve.
FIG. 4 is a flowchart of a BP neural network for predicting the NOx concentration at the outlet of a furnace, which comprises the following steps: collecting data, selecting parameters, preprocessing the data, constructing a BP neural network structure, and performing data simulation training and prediction.
Preferably, the content of the data preprocessing is as follows: firstly, calculating the boiler operation parameters and NO at the outlet of a furnace chamber by using a Correl function or a Pearson function through Excel software x The correlation among the concentrations is obtained, boiler operation parameters with the correlation larger than 0.5 are selected as main factors and also used as input variables of a BP neural network model; and then processing the data of each group of input variables, for example, eliminating the data which do not conform to the actual conditions under extreme working conditions, wherein the data contain the data under each working condition as much as possible, and finally establishing a BP neural network structure.
FIG. 5 shows SCR denitration system NO x An emission prediction control overall flow chart, prediction comprises predicting a furnace outlet NO by a BP neural network x Concentration (SCR reactor inlet NO) x Concentration), and SCR reactor outlet NO x Concentration prediction, SCR reactor outlet NO x Predicted concentration value and SCR reactor outlet NO x And the deviation signals of the concentration measured value and the set value are transmitted to the DMC prediction controller together, and are received by the PID controller to carry out optimization control.
The implementation case is based on the specific implementation of a SCR denitration system of a front-wall and rear-wall opposed boiler of a certain 300MW coal-fired thermal power generating unit, and the system intelligent prediction and overall process regulation and control based on the coal-fired flue gas NO x The emission control system is designed and researched, and the specific implementation process is as follows:
the invention mainly provides a coal-fired flue gas NOx emission control method based on systematic intelligent prediction regulation and whole-process regulation, and mainly relates to intelligent prediction control of flue gas NO by combining BP neural network model prediction and DMC-PID control unit x And (4) discharging.
According to the invention, a large amount of historical boiler operation data is collected, wherein a large amount of documents, boiler operation principles, denitration professional knowledge and industrial experience are referred to select certain boiler operation parameters, including primary air volume, secondary air volume, burnout air volume, oxygen volume, coal mill combination mode, boiler load, hearth temperature and other boiler operation parameters.
Setting the boiler operation parameters as input variables and the furnace outlet NO according to the selected parameters x Setting the concentration as an output variable, performing correlation calculation on the input variable and the output variable, taking the input variable with the correlation larger than 0.5, and then constructing a double-layer implicit variable according to the obtained input variable and the output variableA layer neural network model.
It should be noted that, when the simulation data is simulated by using the neural network, the data is normalized and denormalized, and the data is normalized to the range of [ -1,1], so that the error caused by the numerical problem can be reduced.
Normalization formula: y = (y) max -y min )*(x-x min )/(x max -x min )+y min (ii) a Wherein y is max Is 1,y min Is 1,x max And x min For each row or column of each variable in the collected data, y is the normalized data.
Predicting NO at the outlet of the furnace chamber by using the constructed neural network model according to the boiler operation parameters under different working conditions x And (4) concentration. Then according to NO x Calculating the ammonia amount of the reducing agent liquid required by the SCR reactor by concentration, transmitting the required ammonia amount signal to a DMC-PID control unit, preliminarily and rapidly controlling the ammonia injection amount in the SCR reactor, and simultaneously measuring the NO at the outlet of the SCR reactor x And (4) concentration.
The required ammonia injection amount is based on the SCR reactor inlet NO x Concentration calculation using predicted furnace outlet NO x The concentration is multiplied by the smoke gas amount, and then multiplied by the ammonia nitrogen molar ratio to calculate the required ammonia injection amount. Wherein the flue gas amount refers to the actually measured flue gas amount of the boiler, and the unit is Nm 3 H is used as the reference value. The ammonia nitrogen molar ratio refers to the concentration ratio of ammonia gas and nitrogen oxide in denitration, is a common concept in denitration, and is generally 0.7-0.9. The formula for calculating the "ammonia amount demand" is therefore: ammonia demand = nitrogen oxides at the inlet of the SCR reactor x flue gas x ammonia nitrogen molar ratio.
As a further improvement, with predicting NO x The concentration is increased continuously, and the ammonia gas demand is increased nonlinearly; the required amount of ammonia gas increases nonlinearly with the increasing amount of flue gas.
Further improvement, with NO x The concentration is increased continuously, and the increase of the ammonia gas demand is increased more and more; along with the continuous increase of the smoke quantity, the ammonia quantity demand quantity is increasedIs increasingly large.
Preferably, the ammonia gas amount demand control method is as follows:
the controller stores standard data ammonia gas demand R and forecasts NO x Concentration C and flue gas amount V, are predicted NO x And when the concentration C and the flue gas volume V are reached, the ammonia gas volume demand R meets the requirement.
The standard data represents data satisfying a certain exhaust gas condition. For example, NO in the exhaust gas can be satisfied within a certain range x The requirement of concentration.
Then NO x When the concentration is changed into c and the flue gas volume v, the ammonia gas volume demand r meets the following requirements:
(v*c)/(V*C)＝a*Ln((r-R standard of merit )/(R-R Standard of merit ) + b, a, b are parameters, satisfying the following formula:
(r-R standard of merit )/(R-R Standard of merit )<1，1.041<a<1.044，1.0<b<1.005；
(r-R Standard of merit )/(R-R Standard of merit )＝1，b＝1；
(r-R Standard of merit )/(R-R Standard of merit )>1，1.045<a<1.052；0.991<b<1；
Preferably, (R-R) Standard of merit )/(R-R Standard of merit )<1，a＝1.042,b＝1.003。
Preferably, (R-R) Standard of merit )/(R-R Standard of merit )>1，a＝1.048,b＝0.997。
Preferably, (R-R) Standard of merit )/(R-R Standard of merit )<1, with (R-R) Standard of merit )/(R-R Standard of reference ) Increasing, a getting larger and b getting smaller.
Preferably, (R-R) Standard of merit )/(R-R Standard of merit )>1, as R/R increases, a becomes larger and b becomes smaller. The design is that the smoke emission can meet the requirement as soon as possible.
Wherein R is Standard of merit NO in exhaust flue gas for satisfying the smoke exhaust requirement of normal operation of boiler x The concentration value may preferably be NO satisfying the requirements x Upper limit of concentration.
At the upper partThe following conditions need to be satisfied in the formula of the mode: 0.88<＝(r-R Standard of merit )/(R-R Standard of merit )<＝1.13；
Preferably, the controller stores a plurality of sets of standard data.
Preferably, when a plurality of sets of standard data are satisfied ((1-V/V) 2 +(1－c/C) 2 ) A set v and c with the smallest value of (a); of course, a first set of v and c satisfying the requirement may be selected, or a set of v and c satisfying the condition may be randomly selected.
Through the design of foretell intelligent control's operation mode, compare with linear input ammonia quantity, can so that the quantity of the ammonia of more accurate calculation input for boiler exhaust emission can be faster better satisfy the emission requirement, further improved the system intellectuality, reach energy-concerving and environment-protective requirement.
Then, the NO at the outlet of the SCR reactor is approximated according to the online mapping of the neural network model x Nonlinear relation between predicted concentration value and variable parameter, and its relation with SCR reactor outlet NO x And the deviation signals of the concentration are transmitted to the DMC-PID control unit together, the ammonia injection amount required by the SCR system is reversely controlled and adjusted, and the ammonia injection amount in the SCR system is accurately regulated and controlled in time.
The feedback signal is transmitted to a DMC-PID cascade controller, wherein the parameters of a control unit need to be set according to the working conditions on site to obtain the optimal parameters, and the method comprises the following steps:
(1) And during initial debugging, obtaining characteristic parameters of the system under different working conditions according to field test, establishing an input and output predictive control system model of the SCR denitration system, and determining PID parameters through the model.
(2) And in the operation process, acquiring boiler operation data and SCR reactor working data in real time, correcting the model, and then obtaining the optimal PID parameter under the current working condition by adopting an optimization algorithm based on the corrected model so as to realize self-adaptive PID control.
By combining the control system of the DMC-PID cascade controller through the neural network, national NO can be achieved x And the emission standard is ultralow, so that the operation cost is reduced, and the denitration efficiency is improved. Because of the fact thatThe invention relates to prediction and control at the same time, and selects a DMC-PID cascade controller, the system response rate is faster in the practical process, and the ammonia control amount at the rear end can be directly reflected according to different working conditions of a front-end boiler combustion system, so that the reaction is more timely, and in addition, NO at the SCR outlet is x The concentration can be controlled in a lower range and the deviation from the set value can be smaller, so that NO can be reduced x The emission concentration is lower, and the accuracy of a control system is improved.
In order to ensure that the algorithm can be applied in the field and the control algorithm is kept secret, the implementation of the scheme is based on the optimized control NO x A drainage platform implementation.
Note:
the applicant of the invention refers to a large number of documents of applying a BP neural network to a prediction function of a thermal power plant, and finds that when the BP neural network faces the relationship between random sudden change of unit operation conditions and uncertain nonlinearity in an actual system, the BP neural network has good nonlinear approximation capability and stronger robustness, the prediction performance is more stable, and Dynamic Matrix Control (DMC) has the defect of difficulty in timely prediction control at the moment, so that prediction deviation is generated. Therefore, the applicant utilizes the BP neural network to carry out online compensation on DMC prediction error, the BP neural network and the DMC prediction controller are connected in series to work, and the BP neural network can approach to the outlet NO of the SCR denitration system online x The concentration predicted value and each variable are in a complex nonlinear relationship, so that the prediction error is reduced to the minimum, and the ammonia injection amount in the SCR denitration system is accurately controlled in time.
According to the scheme, the operation data collection is carried out by selecting the operation data containing all the working conditions as far as possible within a period of time of equipment working, manual collection is not needed, the error of manual collection is reduced, and the number of collection points or the collection range can be increased as far as possible in the large data collection. The diversity of the data is beneficial to the fitting prediction performance of the BP neural network model, so that the error of the target function is smaller, and the prediction result of the model is more stable.
In addition, regarding the establishment of the BP neural network model, the establishment of the hidden layer is mainly involved, generally, the number of layers of the hidden layer is one or two, and the hidden layer is determined according to the dimension of an input variable or the comparison of prediction results;
the determination formula of the number of the neurons in the hidden layer of the BP neural network is
h is the number of neurons, m is the number of nodes of the input layer, i.e. the dimension of the input variable, n is the number of output variables, i.e. the number of objective functions, and c is a number between 0 and 10. The number of the neurons of the hidden layers of different models is different, and the number needs to be determined according to comparison of prediction fitting effects of actual conditions.
Regarding the determination of DMC-PID cascade controller parameters, the initial parameters of the control unit need to be obtained by debugging according to the operation data of a plurality of groups of boiler systems and denitration systems, then the test is carried out according to the field operation data, the model is corrected, and then the optimal DMC-PID parameters under the current working condition are obtained by adopting an optimization algorithm based on the corrected model, thereby realizing the self-adaptive DMC-PID control.
The above-mentioned specific implementation method is only described according to the technical solution of the present invention, and is taken as an example, but the protection scope of the present invention is not limited thereto, and a person skilled in the art may modify the technical solution of the present invention or substitute the same, but the present invention shall be included in the protection scope of the present invention, and the protection scope of the present invention shall be subject to the claims.
Claims (9)
1. A boiler system for intelligently controlling ammonia gas input, comprising: a furnace into which water is fed from an inlet duct into a water-cooled wall; a coal-fired burner capable of heating the furnace; a pulverized coal pipe capable of supplying pulverized coal to the burner; the hearth is connected with a flue, and NO is arranged in the flue x The concentration on-line measuring instrument and the flue gas flowmeter have ammonia gas input device in the flue and do not need to be monitored x The concentration and the flue gas flow automatically control the input amount of the ammonia gas.
2. The boiler system of claim 1, whereinIn that, with predicting NO x The concentration is increased continuously, and the ammonia gas demand is increased nonlinearly.
3. The boiler system of claim 2, wherein with NO, there is provided x The concentration is increased continuously, and the required amount of ammonia is increased more and more greatly.
4. The boiler system according to claim 1, wherein the ammonia gas demand increases non-linearly with increasing flue gas volume.
5. The boiler system according to claim 1, wherein the required amount of ammonia gas increases more and more with increasing amount of flue gas.
6. The boiler system of claim 1, further comprising a DMC predictive controller and a PID controller; collecting boiler operation data and carrying out correlation preprocessing, constructing a BP neural network intelligent prediction model, and predicting NO at a furnace outlet x Concentration;
DMC is used as a main controller, PID is used as a secondary controller to form a DMC-PID cascade feedback control structure, rolling optimization and feedback correction are carried out, prediction deviation compensation is carried out by combining a BP neural network model, and online mapping approaches to the outlet NO of an SCR denitration system x The complicated nonlinear relation between the predicted concentration value and each variable can directly realize accurate control on the ammonia injection amount of the SCR denitration system in advance according to different working conditions of a boiler combustion system.
7. The boiler system according to claim 6, wherein NO is installed at an outlet of the SCR denitration system x The concentration on-line measuring instrument can monitor and measure NO at the outlet of the SCR denitration system in real time x Concentration and transmitted as a feedback input signal to the BP neural network and DMC predictive controller.
8. The boiler system of claim 7, wherein the operating data of the boiler combustion system and the operating data of the SCR denitration system are collected at different time intervals under different operating conditions, and the data under different operating conditions are required to be included.
9. The boiler system according to claim 8, wherein a BP neural network prediction model is constructed based on collected historical data of denitration operation of the boiler combustion system and the SCR reactor, the BP neural network prediction model is set as an input layer, a hidden layer and an output layer, and NO at an inlet of the SCR denitration system is predicted online in real time according to the constructed BP neural network x Concentration and NO at outlet of SCR denitration system x And (4) concentration.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211089112.XA CN115657466A (en) | 2022-09-07 | 2022-09-07 | Boiler system of intelligent control ammonia input volume |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211089112.XA CN115657466A (en) | 2022-09-07 | 2022-09-07 | Boiler system of intelligent control ammonia input volume |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115657466A true CN115657466A (en) | 2023-01-31 |
Family
ID=84983554
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211089112.XA Pending CN115657466A (en) | 2022-09-07 | 2022-09-07 | Boiler system of intelligent control ammonia input volume |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115657466A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116185087A (en) * | 2023-05-04 | 2023-05-30 | 科大智能物联技术股份有限公司 | Closed-loop deamination control system based on machine learning |
-
2022
- 2022-09-07 CN CN202211089112.XA patent/CN115657466A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116185087A (en) * | 2023-05-04 | 2023-05-30 | 科大智能物联技术股份有限公司 | Closed-loop deamination control system based on machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104826492B (en) | Improvement method for selective catalytic reduction flue gas denitrification and ammonia injection control system | |
| CN105629738B (en) | SCR flue gas denitrification systems control method and equipment | |
| CN106681381A (en) | SCR denitration system ammonia spraying quantity optimal control system and method based on intelligent feedforward signals | |
| CN107561941A (en) | A kind of full working scope qualified discharge control method of fired power generating unit denitrating system | |
| CN115685743A (en) | Intelligent control coal-fired boiler and intelligent prediction regulation and control flue gas emission method thereof | |
| CN112580250A (en) | Thermal power generating unit denitration system based on deep learning and optimization control method | |
| US8135653B2 (en) | Power plant control device which uses a model, a learning signal, a correction signal, and a manipulation signal | |
| CN103759277B (en) | Coal-fired power station boiler intelligent ash blowing closed loop control method, device and system | |
| CN107544288A (en) | A kind of denitration optimal control method and system | |
| CN110263452B (en) | Flue gas time distribution characteristic analysis method and system in flue and denitration system | |
| CN112859780B (en) | Thermal power plant intelligent combustion control method based on cloud data and cloud computing | |
| CN112628712A (en) | Secondary air closed-loop optimization control system based on air door resistance coefficient | |
| CN115145152A (en) | Boiler combustion and denitration process collaborative optimization control method | |
| CN115111601A (en) | Multi-target boiler combustion optimization control method with embedded algorithm fusion under variable loads | |
| Liu et al. | Anticipatory NH3 injection control for SCR system based on the prediction of the inlet NOx concentration | |
| CN115657466A (en) | Boiler system of intelligent control ammonia input volume | |
| CN111841276A (en) | SNCR (selective non-catalytic reduction) denitration control method and device for circulating fluidized bed unit and storage medium | |
| CN108488831B (en) | Boiler combustion control system and method | |
| CN114675607A (en) | Fired coal heat value prediction and correction control method combining flame temperature | |
| CN111401652A (en) | Boiler optimization method and system based on CO online detection | |
| CN113418207B (en) | Power station hearth combustion monitoring and diagnosing device and method | |
| WO2022252414A1 (en) | Coal-air synchronous dynamic coordinated control method for coal-fired unit | |
| JP2005506540A (en) | Nitrogen oxide software sensor at the exhaust stack outlet of a heating furnace using air containing high concentration oxygen or pure oxygen as oxidant | |
| JP7374590B2 (en) | KPI improvement support system and KPI improvement support method | |
| CN207478283U (en) | A kind of fired power generating unit denitration real-time control apparatus |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |