CN115524976A - SCR system ammonia injection amount adjusting method considering boiler combustion state - Google Patents
SCR system ammonia injection amount adjusting method considering boiler combustion state Download PDFInfo
- Publication number
- CN115524976A CN115524976A CN202211328977.7A CN202211328977A CN115524976A CN 115524976 A CN115524976 A CN 115524976A CN 202211328977 A CN202211328977 A CN 202211328977A CN 115524976 A CN115524976 A CN 115524976A
- Authority
- CN
- China
- Prior art keywords
- scr
- value
- scr system
- ammonia injection
- nox emission
- 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
- QGZKDVFQNNGYKY-UHFFFAOYSA-N Ammonia Chemical compound N QGZKDVFQNNGYKY-UHFFFAOYSA-N 0.000 title claims abstract description 117
- 238000000034 method Methods 0.000 title claims abstract description 112
- 229910021529 ammonia Inorganic materials 0.000 title claims abstract description 58
- 238000002347 injection Methods 0.000 title claims abstract description 47
- 239000007924 injection Substances 0.000 title claims abstract description 47
- 238000002485 combustion reaction Methods 0.000 title claims abstract description 22
- 230000008569 process Effects 0.000 claims abstract description 82
- 238000005457 optimization Methods 0.000 claims abstract description 31
- 238000012937 correction Methods 0.000 claims abstract description 25
- 238000005070 sampling Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 20
- 238000012360 testing method Methods 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 10
- 239000000243 solution Substances 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000002068 genetic effect Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 238000005507 spraying Methods 0.000 claims description 5
- 230000007246 mechanism Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 claims description 2
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 19
- 239000003546 flue gas Substances 0.000 description 17
- MWUXSHHQAYIFBG-UHFFFAOYSA-N nitrogen oxide Inorganic materials O=[N] MWUXSHHQAYIFBG-UHFFFAOYSA-N 0.000 description 12
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 6
- 239000001301 oxygen Substances 0.000 description 6
- 229910052760 oxygen Inorganic materials 0.000 description 6
- 238000010531 catalytic reduction reaction Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 229910002091 carbon monoxide Inorganic materials 0.000 description 2
- 239000003245 coal Substances 0.000 description 2
- 239000012895 dilution Substances 0.000 description 2
- 238000010790 dilution Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- BULVZWIRKLYCBC-UHFFFAOYSA-N phorate Chemical compound CCOP(=S)(OCC)SCSCC BULVZWIRKLYCBC-UHFFFAOYSA-N 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000035484 reaction time Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Treating Waste Gases (AREA)
Abstract
The invention provides an SCR system ammonia injection amount adjusting method considering a boiler combustion state, which comprises the steps of firstly obtaining data representing a unit combustion state and process parameters in an SCR system, calculating delay time between each process parameter and SCR outlet NOx emission amount based on MIC, establishing an initial prediction model of the SCR outlet NOx emission amount based on ELM, obtaining an initial prediction value of the SCR outlet NOx emission amount at the current moment, establishing a dynamic error correction model based on ELM, obtaining an error correction value of the SCR outlet NOx emission amount at the current moment, overlapping prediction results of the two models, further establishing a dynamic prediction model of the SCR outlet NOx emission amount, considering that the SCR outlet NOx emission amount meets national emission standards, reducing ammonia injection cost of the SCR system and other targets, designing an intelligent ammonia injection amount adjusting method based on NSGA-II, further completing optimization and adjustment of the ammonia injection amount, reducing the NOx emission concentration at the SCR system while reducing the ammonia injection cost of a denitration system of a power plant, and having important significance for guiding operation of the power plant.
Description
Technical Field
The invention belongs to the technical field of thermal power generation, and particularly relates to an ammonia injection amount adjusting method of an SCR system considering a boiler combustion state.
Background
At present, new energy electric power such as wind energy, solar energy and the like is developed rapidly, and after the large-scale random and fluctuating new energy electric power is connected into a power grid, deep peak regulation requirements are generated, so that a coal-fired unit is required to operate flexibly. The deep peak regulation of the coal-fired unit means that the load of the unit fluctuates rapidly in a large range, the combustion state of the unit is influenced, the discharge amount of nitrogen oxides generated by combustion fluctuates violently, and the difficulty in accurately controlling the ammonia spraying amount of a Selective Catalytic Reduction (SCR) system is increased. The SCR denitration system is an important technical means for treating the flue gas of the power plant, realizes accurate dynamic modeling of the SCR denitration system and intelligent optimization and adjustment of ammonia injection amount, and has important significance for ultralow emission and economic operation of nitrogen oxides in the power plant.
The SCR denitration reaction process is complex and is influenced by various factors such as unit load, flue gas temperature and the like, so that the SCR system has the characteristics of nonlinearity, strong disturbance and the like; and the SCR System has the characteristic of dynamic time delay due to the influence of factors such as measurement delay of a Continuous flue gas Emission Monitoring System (CEMS) and dynamic change of denitration reaction time. The above characteristics cause dynamic modeling of the SCR system and optimization of ammonia injection amount to be a challenging problem. In the existing research method aiming at the intelligent adjustment of the ammonia injection amount, an optimization set value of the ammonia injection amount under the actual operation working condition is mainly solved through a single-target optimization algorithm, only a single optimization target for reducing the NOx emission concentration at the outlet of an SCR system is generally considered, the ammonia injection amount consumption cost is not considered, and the method has certain limitation on guiding the economic operation of a denitration system.
Disclosure of Invention
Based on the problems, the invention provides an ammonia injection amount adjusting method of an SCR system considering the combustion state of a boiler, which comprises the following steps:
step 1: acquiring a process parameter data set D representing a unit combustion state and an SCR system running state; wherein D = { X t ,Y t }，
Y t ＝{y t }∈R t×1 ，t＝t 1 ,t 2 ,…,t N T represents the sampling time, t 1 To the initial sampling time, t N For terminating the sampling time, N is the number of modeling samples, i =1,2,3, \8230, m and m represent the number of process parameters, and X t For a set of m process parameters,
for the sample value of the ith process parameter at time t, y t At time t SCR system outlet NOx emission concentration;
the process parameters are divided into output parameters and input parameters, wherein the output parameters are NOx emission concentration at the outlet of the SCR system, and the input parameters comprise 3 process parameters representing the combustion state of the unit, such as unit load, total air volume and total coal volume, and 12 process parameters representing the operation state of the SCR system, such as ammonia injection amount, inlet flue gas temperature, outlet flue gas temperature, inlet flue gas pressure, outlet flue gas pressure, inlet oxygen concentration, outlet oxygen concentration, inlet nitrogen oxide concentration, inlet carbon monoxide concentration, flue gas flow, dilution fan current and ammonia escape concentration;
step 2: and carrying out normalization operation on the process parameter data set D by adopting a most-value normalization method to obtain a normalized modeling data set ND, wherein the most-value normalization formula is as follows:
wherein,
is the minimum value of the ith input parameter,
is the maximum value of the ith input parameter,
normalized values for the ith input parameter,
is the minimum value of the ith output parameter,
is the maximum value of the ith output parameter,
normalizing the value of the ith output parameter;
and 3, step 3: calculating delay time between each process parameter and the NOx emission concentration at the outlet of the SCR system by adopting a Maximum Information Coefficient (MIC), and reconstructing a modeling data set according to the calculated delay time; the method comprises the following steps:
step 3-1: determining a range of delay times (0, K) for each process parameter, wherein K is a maximum delay time;
step 3-2: for is to
Each process parameter in
In the delay time range (0, K)]Sequentially sampling from 1 moment before sampling time to k moments before sampling time to construct a MIC calculation data set
Wherein
Is the value of the ith process variable at the 1 st time instant before the sampling time instant t1,
the value of the ith process variable k times before the sampling time tN,
step 3-3: computing a dataset using a MIC algorithm
The column vectors and the output vector
The column vector with the maximum MIC value is selected as the corresponding process parameter
Delay time k of i (i＝1,2,…,m),k i I.e. the respective process parameters
Continuously repeating the process at the corresponding optimal delay time to obtain m optimal delay times;
step 3-4: each group of process parameters corresponding to the optimal delay time
And an output variable Y t Combining the data into a reconstructed data set xc, wherein a first column represents data of an output variable, and each of the remaining columns is a value of each process parameter at the optimal delay time;
in the formula,
for the 1 st process parameter at the corresponding delay time k 1 The value of (c) time of day,
for the mth process parameter at the corresponding delay time k m The value of time;
and 4, step 4: screening characteristic variables in the reconstructed data set xc by adopting an extreme Gradient Boosting (XGboost) algorithm to serve as modeling data SD; the concrete expression is as follows:
combining denitration reaction mechanism analysis, taking the concentration of NOx at the inlet of the SCR system and the ammonia injection amount as characteristic variables of modeling, determining the importance of the other variables by adopting a limit gradient lifting algorithm, and calculating and reconstructing a data set xc to remove the concentration of NOx at the inlet of the SCR systemImportance of the rest process parameters except the ammonia injection amount relative to the NOx emission concentration at the outlet of the SCR system in the first column in the reconstructed data set xc is set, an importance threshold value is set, the process parameters larger than the importance threshold value are selected as characteristic variables for modeling, and the data set after characteristic selection is SD = { S = { S = t ,Y t t＝t 1 ,…,t N }，
S t Is a process parameter set after feature selection, wherein sm is less than or equal to m;
and 5: establishing a prediction model M of the outlet NOx emission concentration of the SCR system based on an Extreme Learning Machine (ELM) h The concrete expression is as follows: the prediction model M h Initial predictive model M including ELM-based SCR system outlet NOx emission concentration P And a dynamic error correction model M E Using the modeling data set SD after feature selection and adopting an initial prediction model M P Obtaining an initial predicted value y of the test sample p (t) then applying the error correction model M E Obtaining the error predicted value e of the current predicted sample p (t) superposing the prediction results of the two models to finally obtain the final prediction value y of the NOx emission concentration at the outlet of the SCR system h (t)；
The initial prediction model M P The construction process of (A) is as follows:
step 5-1-1: dividing modeling data into a training set and a test set;
step 5-1-2: the training of a weight matrix between an ELM hidden layer and an output layer is completed by utilizing a training set, and a trained model is called as an initial prediction model M P ；
Step 5-1-3: using an initial prediction model M P Predicting the test sample to obtain an initial predicted value y of the NOx emission concentration at the outlet of the SCR system p (t)；
The dynamic error correction model M E The construction process of (A) is as follows:
step 5-2-1: obtaining an initial prediction model M P Initial predictor y on training set p (t) And thus makes a difference with the actual measurement value to obtain an error data sequence e of the training set p (t)，e p (t) is as follows;
e p (t)＝y p (t)-y m (t) (4)
wherein, y p (t) is the initial model M at the current time t P Predicted value of (a), y m (t) is an actual measurement of the SCR system outlet NOx emission concentration at the current time t;
step 5-2-2: the historical errors e of the previous three moments p (t-1)、e p (t-2)、e p (t-3) and input S of the current initial prediction model t As input, the error e of the current time p (t) as output, training the ELM network to obtain an error correction model M E ；
Step 5-2-3: using error correction model M E Completing the prediction of the test set sample to obtain the error correction value of the test set sample;
finally, the initial model M is superimposed P Sum error correction model M E Finally obtaining a final predicted value y of the NOx emission concentration at the outlet of the SCR system h (t)；
Step 6: selecting ammonia spraying quantity Q as a controllable variable, establishing a multi-objective optimization function of the controllable variable, and solving the multi-objective optimization function by using a non-dominated sorting genetic algorithm NSGA-II with an elite strategy to obtain an optimal solution set meeting an optimization objective; the method comprises the following steps:
step 6-1: establishing a multi-objective optimization function as:
in the formula, f NOx (s) representation of the prediction model M h Deviation of output predicted NOx concentration from expected value, f price (Q) represents the ammonia injection consumption of the SCR system, and s is a prediction model M h Sm is the dimension of the input sample, Q is the ammonia injection quantity, s 1 ,s 2 ,…,s sm-1 For removing spray from input sampleProcess variables other than the amount of ammonia, E NOx Desired value, M, for the SCR system outlet NOx emission concentration h (s) is derived from a prediction model M h The obtained predicted value, M, of the NOx emission concentration at the outlet of the SCR system h (s) less than 50mg/Nm 3 ；
Step 6-2: and solving a multi-objective optimization function by using a non-dominated sorting genetic algorithm NSGA-II, solving a pareto frontier solution set of the ammonia injection consumption and the error between the NOx predicted value output by the SCR hybrid prediction model and an expected value, and obtaining an optimal solution set meeting the optimization objective. And operating personnel can select the optimal solution concentration result to control the ammonia spraying amount.
The beneficial effects of the invention are:
the invention provides an SCR system ammonia injection amount adjusting method considering the combustion state of a boiler, which estimates the delay time of each variable by using a maximum information coefficient, carries out data reconstruction and can make up for the defect of large lag of the original system; the extreme gradient lifting algorithm is adopted to screen the modeling variables of the reconstructed data set, so that the dynamic prediction of the NOx emission concentration at the outlet of the SCR system is realized, the limitation of the network structure of the traditional algorithm can be overcome, the deep level characteristics of the data can be extracted, and a prediction model and an error correction model can be simultaneously established, so that the method has the advantage of high prediction precision; the ammonia injection cost is reduced while the requirement of NOx emission concentration is met, the multi-objective optimization function is solved by using the non-dominated sorting genetic algorithm NSGA-II with the elite strategy, and the solution set of the ultra-low ammonia injection amount meeting the requirement of the NOx emission concentration is obtained.
Drawings
FIG. 1 is a flow chart of a method for adjusting ammonia injection amount of an SCR system in consideration of combustion state of a boiler according to the present invention.
FIG. 2 is a flow chart of a method for establishing a predictive model of SCR system outlet NOx emission concentration in accordance with the present invention.
FIG. 3 is a predicted value of SCR system outlet NOx emission concentration on a test set of the present invention.
FIG. 4 is a graph comparing ammonia injection consumption and outlet NOx emission concentration before and after optimization adjustments according to the present invention.
Detailed Description
The invention is further described with reference to the following figures and specific examples.
In order to accurately optimize and adjust the ammonia injection amount in the SCR system, the invention provides a multi-target optimization and adjustment method for the ammonia injection amount of the SCR system, which takes the combustion state of a boiler into consideration. The method comprises the following steps: acquiring data representing a unit combustion state and process parameters in an SCR (selective catalytic reduction) system through a plant-level monitoring information system (SIS) of a thermal power plant; calculating a delay time between each process parameter and an SCR system outlet NOx emission concentration based on a Maximum Information Coefficient (MIC); establishing an initial prediction model of the outlet NOx emission concentration of the SCR system based on the extreme learning machine to obtain an initial prediction value of the outlet NOx emission concentration of the SCR system at the current moment, establishing a dynamic error correction model based on the ELM to obtain an error correction value of the outlet NOx emission concentration of the SCR system at the current moment, and overlapping the prediction results of the two models to further establish a dynamic prediction model of the outlet NOx emission concentration of the SCR system; considering the aims of meeting the national emission standard of the outlet NOx emission concentration of the SCR system, reducing the ammonia injection cost of the SCR system and the like, the intelligent ammonia injection amount adjusting method is designed based on the non-dominated sorting genetic algorithm (NSGA-II) with the elite strategy, and then the optimization and adjustment of the ammonia injection amount are completed.
As shown in fig. 1, an SCR system ammonia injection amount adjustment method considering a combustion state of a boiler includes:
step 1: acquiring a process parameter data set D representing a unit combustion state and an SCR system running state through a plant-level monitoring information system; wherein D = { X t ,Y t }，
Y t ＝{y t }∈R，t＝t 1 ,t 2 ,…,t N T represents the sampling time, t 1 Is the initial sampling time, t N For terminating the sampling time, N is the number of modeling samples, i =1,2,3, \8230;, m, m represent the number of process parameters, and X represents the number of process parameters t Is a set of m-dimensional process parameters,
for the sample value of the ith process parameter at time t, y t The concentration of NOx emitted at the outlet of the SCR system at the moment t;
in this embodiment, a data set containing 16 parameters is collected, wherein the output parameter is the NOx emission concentration at the outlet of the SCR system, and the output parameter includes 3 process parameters representing the combustion state of the unit, such as unit load, total air volume, and total coal volume, and 12 process parameters representing the operation state of the SCR system, such as ammonia injection amount, inlet flue gas temperature, outlet flue gas temperature, inlet flue gas pressure, outlet flue gas pressure, inlet oxygen concentration, outlet oxygen concentration, inlet nitrogen oxide concentration, inlet carbon monoxide concentration, flue gas flow, dilution fan current, and ammonia slip concentration. 9210 sets of data were collected at 10s intervals, starting time 2016, month 4, day 2, 25.
Step 2: and carrying out normalization operation on the process parameter data set D by adopting a most-value normalization method to obtain a normalized modeling data set ND, wherein the most-value normalization formula is as follows:
wherein,
is the minimum value of the ith input parameter,
is the maximum value of the ith input parameter,
normalized values for the ith input parameter,
is the minimum value of the ith output parameter,
is the maximum value of the ith output parameter,
normalizing the value of the ith output parameter;
and step 3: calculating delay time between each process parameter and the NOx emission concentration at the outlet of the SCR system by adopting a Maximum Information Coefficient (MIC), and reconstructing a modeling data set according to the calculated delay time; the method comprises the following steps:
step 3-1: determining the maximum delay time range (0, K) of each process parameter according to the operation experience of the field unit]Wherein K is the maximum delay time; suppose the delay time of each process parameter is d 1 ,d 2 ,…,d m In which d is i For the ith process parameter
The delay time of (d); and setting the maximum delay time range of the process parameters to be 600s according to the actual operation condition and the related operation experience of the selected unit. Suppose that the delay times of the 15 process parameters are d 1 ,d 2 ,…,d 15 . Since the sampling interval is 10s, the maximum delay time k =60 for each process parameter.
Step 3-2: to pair
Each process parameter of
In the delay time range (0, K)]Sequentially sampling from 1 moment before sampling time to k moments before sampling time, and constructing MIC calculation data set
Wherein,
is the value of the ith process variable 1 time prior to the sampling time t1,
the value of the ith process variable k times before the sampling time tN,
k =60,n =9210, yielding the variables
Wherein each column vector represents a range of delay instants (0, 60)]And sequentially delaying the reconstruction data sequences at different time instants.
Step 3-3: analyzing and reconstructing data matrix by adopting maximum information coefficient MIC
Middle row vector and output vector
The column vector with the maximum MIC value is selected as the corresponding process parameter
Delay time k of i (i =1,2, \8230;, m), the delay instant k for which the column vector corresponds i Is the process parameter
At the optimum delay time
Process parameter
Is optimum delay time
T is a sampling interval; sequentially calculating the delay time of all process parameters according to the method, and completing the solution of the delay time of each process parameter and the reconstruction of the whole modeling data set;
the specific calculation formula of the MIC is as follows:
wherein I (x, y) is a mutual information value between variables x and y, p (x) and p (y) are respectively edge probability distribution of the variables x and y, p (x, y) is joint probability distribution of the variables x and y, and B (n) is 0.6 th power of sample data volume n;
step 3-4: according to the optimum delay time of each variable
Selecting data corresponding to each variable optimal delay time as initial data and output data to rearrange, wherein the reconstructed data set is xc, the first column represents the output data, and the rest columns are process parameters reconstructed according to the optimal delay time;
in the formula,
for the 1 st process parameter at the corresponding delay time k 1 The value of (c) time of day,
for the mth process parameter at the corresponding delay time k m A value of time;
the calculated delay times for the various process parameters are shown in table 1.
TABLE 1 delay time Table for Process parameters
And 4, step 4: screening characteristic variables in the reconstructed data set xc by adopting an extreme Gradient Boosting (XGboost) algorithm to serve as modeling data; the concrete expression is as follows:
determining the importance of each process parameter in a reconstructed data set relative to the NOx emission concentration at the outlet of an SCR system by adopting a limit gradient lifting algorithm, selecting the process parameter greater than the importance threshold as a characteristic variable for modeling by setting the importance threshold, and simultaneously combining denitration reaction mechanism analysis, taking the NOx at the inlet of the SCR and the ammonia injection amount as the characteristic variables for modeling, wherein the data set after characteristic selection is SD = { S = t ,Y t |t＝t 1 ,…,t N }，
S t The feature variable set with the dimension sm after feature selection is obtained, wherein sm is less than or equal to m;
in the present embodiment, the importance threshold is set to 0.2. And calculating the importance of each process parameter by using the XGboost, normalizing the calculation result, and selecting the process parameter with the importance value of more than 0.2 after normalization as the input parameter of the prediction model, wherein the calculation result of the XGboost is shown in a table 2. The denitration reaction is considered as that inlet NOx and ammonia gas are subjected to selective catalytic reduction reaction in an SCR system reactor, so that the inlet NOx concentration is used as an input parameter of a prediction model. Meanwhile, the variables of the ammonia addition amount, the outlet flue gas temperature, the inlet flue gas temperature, the outlet oxygen concentration, the unit load, the total air volume and the like with the normalized importance greater than 0.2 are used as input parameters of the prediction model. Feature-selected input data set S t Has a characteristic dimension sm of 7.
Table 2 table of importance calculation results based on XGBoost
And 5: establishing a prediction model M of NOx emission concentration at an outlet of an SCR system based on Extreme Learning Machine (ELM) h The method comprises the following steps: initial prediction model M for establishing SCR system outlet NOx emission concentration based on extreme learning machine P And a dynamic error correction model M E Using the modeling data set after feature selection, using an initial prediction model M P Obtaining an initial predicted value y of the test sample p (t) then applying the error correction model M E Obtaining the error predicted value e of the current predicted sample p (t) superposing the prediction results of the two models, and finally establishing a mixed prediction model M of the NOx emission concentration at the outlet of the SCR system h Obtaining the final predicted value y of the NOx emission concentration at the outlet of the SCR system h (t)；
The initial prediction model M P The construction process of (2) is as follows:
step 5-1-1: dividing modeling data into a training set and a test set; the front 7500 groups of data were selected as training set and the remaining 1710 groups of data were selected as test set.
Step 5-1-2: the method comprises the steps of completing training of a weight matrix between an ELM hidden layer and an ELM output layer by utilizing a training set, setting the number of neurons of the hidden layer to be 25, randomly initializing the weight matrix between the input layer and the hidden layer and the bias of the neurons of the hidden layer, solving the output weight matrix between the hidden layer and the output layer through the training set, and completing the training process of the ELM. The trained model is called an initial prediction model M P ；
Step 5-1-3: using an initial prediction model M P Predicting 1710 groups of test samples to obtain an initial predicted value y of the NOx emission concentration at the outlet of the SCR system p (t)；
Initial prediction model M for SCR system outlet NOx emission concentration P Using ELM to mine M p Predicting regularity of error, and establishing dynamic error correction model M based on ELM E The construction process of (2) is as follows:
step 5-2-1: obtaining an initial prediction model M P The initial predicted value on the training set is subtracted from the actual measured value to obtain the error data sequence e of the training set p (t)，e p (t) is as follows;
e p (t)＝y p (t)-y m (t) (4)
wherein, y p (t) is the initial model M at the current time t P Predicted value of (a), y m (t) is an actual measurement of the SCR system outlet NOx emission concentration at the current time t;
step 5-2-2: historical errors e (t-1), e (t-2) and e (t-3) of the previous three moments and the input S of the current initial prediction model t Taking the error e (t) of the current moment as an input and taking the error e (t) as an output, training the ELM network to obtain an error correction model M E ；
Step 5-2-3: using error correction model M E Completing the prediction of the test set sample to obtain the error correction value of the test set sample;
step 5-2-4: the predicted value of the concentration of NOx emitted from the outlet of the SCR system on the test set is shown in FIG. 2, and an initial prediction model M is adopted P Obtaining an initial predicted value y of the test sample p (t) then applying the error correction model M E Obtaining the error predicted value e of the current predicted sample p (t), superposing the prediction results of the two models, and finally establishing a mixed prediction model M of the NOx emission concentration at the outlet of the SCR system h Obtaining a final predicted value y of the NOx emission concentration at the outlet of the SCR system h (t), the specific calculation formula is as follows:
y h (t)＝y p (t)+e p (t)
step 6: in order to reduce the NOx emission concentration at the outlet of the SCR system and reduce the ammonia injection consumption cost, and considering the controllability of relevant operation parameters in the SCR system, selecting the ammonia injection quantity Q as a controllable variable, establishing a multi-objective optimization function of the controllable variable, and solving the multi-objective optimization function by using a non-dominated sorting genetic algorithm NSGA-II with an elite strategy to obtain an optimal solution set meeting the optimization objective; the method comprises the following steps:
step 6-1: establishing a multi-objective optimization function as:
in the formula (f) NOx (s)、f price (Q) are respectively two constructed optimization objective functions which are respectively referred to as a minimum prediction model M h Error of output NOx predicted value from expected value, minimizing ammonia injection consumption in SCR system, s is mixed prediction model M h Sm is the dimension of the input sample, Q is the ammonia injection quantity, s 1 ,s 2 ,…,s sm-1 For non-adjustable process variables other than the quantity of injected ammonia in the input sample, E NOx Desired value, M, for the SCR system outlet NOx emission concentration h (s) is a mixture of prediction models M h The obtained actual value of the NOx emission concentration at the outlet of the SCR system meets M h (s) less than 50mg/Nm 3 ；
The dimension sm of the input sample s after feature selection is 7, wherein Q is the ammonia injection quantity and s 1 ,s 2 ,…,s sm-1 The process variables input into the sample except the ammonia spraying amount, namely, the variables such as outlet flue gas temperature, inlet flue gas temperature, outlet oxygen concentration, unit load, total air volume and the like. Considering model prediction error and actual operation condition of unit SCR system, in order to reduce NOx emission concentration at outlet NOx Set to 30mg/Nm 3 。
Step 6-2: solving a multi-objective optimization function by using a non-dominated sorting genetic algorithm NSGA-II with an elite strategy to obtain an optimal solution set meeting an optimization objective;
the determined objective function is solved by adopting an NSGA-II algorithm to obtain a pareto optimal solution set meeting an optimization objective, further the optimized ammonia injection amount and the SCR system outlet NOx emission concentration after the ammonia injection amount is optimized and adjusted are determined, the prediction result is shown in figure 3, 10 groups of samples are selected for optimization, and the optimization result is shown in figure 4.
Claims (6)
1. An ammonia injection amount adjusting method of an SCR system considering a combustion state of a boiler, comprising:
step 1: acquiring a process parameter data set D representing a unit combustion state and an SCR system running state;
and 2, step: carrying out normalization operation on the modeling data set D by adopting a most-valued normalization method to obtain a normalized modeling data set ND;
and step 3: calculating delay time between each process parameter and the NOx emission amount at the outlet of the SCR by adopting a maximum information coefficient, and reconstructing a modeling data set according to the calculated delay time;
and 4, step 4: screening the reconstructed characteristic variables by adopting a limit gradient lifting algorithm to serve as modeling data;
and 5: establishing a prediction model M of the SCR system outlet NOx emission concentration based on an extreme learning machine h Obtaining the final predicted value y of the NOx discharge amount at the SCR outlet h (t)；
And 6: selecting the ammonia spraying amount Q as a controllable variable, establishing a multi-objective optimization function of the controllable variable, and solving the multi-objective optimization function by using a non-dominated sorting genetic algorithm NSGA-II with an elite strategy to obtain an optimal solution set meeting the optimization objective.
2. The method for adjusting the ammonia injection amount of the SCR system in consideration of the combustion state of the boiler as recited in claim 1, wherein the formula of the most-value normalization in the step 2 is as follows:
wherein,
is the minimum value of the ith process parameter,
is the maximum value of the ith process parameter,
the sample value at time t for the ith process parameter,
normalized values for the ith process parameter.
3. The method of claim 1, wherein step 3 comprises:
step 3-1: determining a maximum delay time range [0, K ] of each process parameter, wherein K is the maximum delay time;
step 3-2: in the delay time range (0, K) for each process parameter]Sequentially sampling from 1 moment before sampling time to k moments before sampling time to construct a calculation data set of a maximum information coefficient MIC
Step 3-3: computing a dataset using a MIC algorithm
In each column vector and output vector Y t Selecting the column vector with the maximum MIC value as the optimal delay time of the process parameters, and calculating the optimal delay time of the m process parameters;
step 3-4: corresponding process parameters and output variables Y of the m optimal delay moments t The combinations are a reconstructed data set xc, where the first column represents the elements of the output variable and the remaining columns each represent the value of the respective process parameter at the optimal delay time.
4. The method for adjusting the ammonia injection amount of the SCR system in consideration of the combustion state of the boiler as recited in claim 1, wherein the step 4 is specifically expressed as:
determining the importance of each process parameter relative to the NOx discharge amount at the SCR outlet in a reconstruction data set by adopting a limit gradient lifting algorithm, selecting the process parameter larger than the importance threshold as a characteristic variable for modeling by setting the importance threshold, simultaneously combining denitration reaction mechanism analysis, taking the NOx at the SCR inlet and the ammonia injection amount as the characteristic variables for modeling, and taking the data set after characteristic selection as SD = { S = { S } t ,Y t |t＝t 1 ,…,t N }，
S t The feature variable set with the dimension sm after feature selection is obtained, wherein sm is less than or equal to m.
5. The method as claimed in claim 1, wherein the step 5 comprises: initial prediction model M of SCR system outlet NOx emission concentration based on ELM P And a dynamic error correction model M E Using the modeling data set after feature selection, and adopting an initial prediction model M P Obtaining an initial predicted value y of the test sample p (t) then applying an error correction model M E Obtaining the error predicted value e of the current predicted sample p (t) superposing the prediction results of the two models to finally obtain the final prediction value y of the NOx emission concentration at the outlet of the SCR system h (t)；
The initial prediction model M P The construction process of (2) is as follows:
step 5-1-1: dividing modeling data into a training set and a test set;
step 5-1-2: training a weight matrix between a hidden layer and an output layer of the extreme learning machine by using a training set, wherein the trained model is called as an initial prediction model M P ；
Step 5-1-3: using an initial prediction model M P Predicting the test sample to obtain an initial predicted value y of the NOx emission amount at the outlet of the SCR p (t)；
The dynamic error correction model M E The construction process of (2) is as follows:
step 5-2-1: obtaining an initial prediction model M P Initial predictor y on training set p (t) subtracting the actual measurement value to obtain an error data sequence e of the training set p (t)；
Step 5-2-2: historical errors e (t-1), e (t-2) and e (t-3) of the previous three moments and the input S of the current initial prediction model t As input, the error e (t) of the current time is used as output, and the extreme learning machine network is trained to obtain an error correction model M E ；
Step 5-2-3: using error correction model M E And completing the prediction of the test set sample to obtain the error correction value of the test set sample.
6. The method for adjusting the ammonia injection amount of the SCR system in consideration of the combustion state of the boiler as recited in claim 1, wherein the multi-objective optimization function in step 6 is expressed as:
min f NOx (s)＝M h (s)-E NOx
min f price (Q)＝Q
s＝[Q,s 1 ,s 2 ,…,s sm-1 ]
s.t.M h (s)<50
in the formula, f NOx (s)、f price (Q) are respectively two constructed optimization objective functions which are respectively referred to as a minimized SCR mixed prediction model M h Error of output NOx predicted value from expected value, minimizing ammonia injection consumption in SCR system, s is mixed prediction model M h Sm is the dimension of the input sample, Q is the ammonia injection quantity, s 1 ,s 2 ,…,s sm-1 For non-adjustable process variables other than the quantity of injected ammonia in the input sample, E NOx Desired value for SCR system outlet NOx emission, M h (s) is a mixture of prediction models M h And obtaining an actual value of the NOx discharge amount at the outlet of the SCR.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211328977.7A CN115524976A (en) | 2022-10-27 | 2022-10-27 | SCR system ammonia injection amount adjusting method considering boiler combustion state |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211328977.7A CN115524976A (en) | 2022-10-27 | 2022-10-27 | SCR system ammonia injection amount adjusting method considering boiler combustion state |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115524976A true CN115524976A (en) | 2022-12-27 |
Family
ID=84702995
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211328977.7A Pending CN115524976A (en) | 2022-10-27 | 2022-10-27 | SCR system ammonia injection amount adjusting method considering boiler combustion state |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115524976A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116151438A (en) * | 2023-01-10 | 2023-05-23 | 南京工业大学 | Method and system for predicting emission concentration of pollutants in process industry |
| CN116832614A (en) * | 2023-05-24 | 2023-10-03 | 华能国际电力股份有限公司上海石洞口第二电厂 | Ammonia spraying amount control method and system for SCR denitration system |
-
2022
- 2022-10-27 CN CN202211328977.7A patent/CN115524976A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116151438A (en) * | 2023-01-10 | 2023-05-23 | 南京工业大学 | Method and system for predicting emission concentration of pollutants in process industry |
| CN116832614A (en) * | 2023-05-24 | 2023-10-03 | 华能国际电力股份有限公司上海石洞口第二电厂 | Ammonia spraying amount control method and system for SCR denitration system |
| CN116832614B (en) * | 2023-05-24 | 2024-05-24 | 华能国际电力股份有限公司上海石洞口第二电厂 | Ammonia spraying amount control method and system for SCR denitration system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115524976A (en) | SCR system ammonia injection amount adjusting method considering boiler combustion state | |
| CN104826492B (en) | Improvement method for selective catalytic reduction flue gas denitrification and ammonia injection control system | |
| CN112418284B (en) | Control method and system of SCR denitration system of all-condition power station | |
| CN107526292B (en) | A method of the regulation ammonia spraying amount based on inlet NOx concentration prediction | |
| CN104638644B (en) | Acquiring method for dynamic random optimal power flow of power system for wind-containing power field | |
| CN110263395A (en) | The power plant's denitration running optimizatin method and system analyzed based on numerical simulation and data | |
| CN109062053A (en) | A kind of denitration spray ammonia control method based on multivariate calibration | |
| CN113433911B (en) | Accurate control system and method for ammonia spraying of denitration device based on accurate concentration prediction | |
| CN109190848A (en) | A kind of SCR system NO based on Time-delay PredictionxConcentration of emission prediction technique | |
| CN114225662B (en) | Hysteresis model-based flue gas desulfurization and denitrification optimal control method | |
| CN111552175B (en) | Overall optimization scheduling and rapid variable load control method for supercritical coal-fired power plant-carbon capture system after chemical adsorption combustion | |
| CN110975597B (en) | Neural network hybrid optimization method for cement denitration | |
| CN112836884A (en) | Copula-DBiLSTM-based method for accurately predicting multi-element load of comprehensive energy system | |
| CN109670625A (en) | NOx emission concentration prediction method based on Unscented kalman filtering least square method supporting vector machine | |
| CN111897373A (en) | Model prediction-based ammonia injection flow adjusting method for SCR denitration device | |
| CN107918368A (en) | The dynamic prediction method and equipment of iron and steel enterprise's coal gas yield and consumption | |
| CN106991507A (en) | A kind of SCR inlet NOx concentration on-line prediction method and device | |
| CN110598929A (en) | Wind power nonparametric probability interval ultrashort term prediction method | |
| CN110737198B (en) | Large-scale coal-fired power plant CO based on BP neural network 2 Capture system prediction control method | |
| CN117190173B (en) | Optimal control method and control system for flue gas recirculation and boiler coupling system | |
| CN113962140A (en) | Method for optimizing steam turbine valve flow characteristic function based on GA-LSTM | |
| CN114740713B (en) | Multi-objective optimization control method for wet flue gas desulfurization process | |
| CN105955350A (en) | Fractional order prediction function control method for optimizing heating furnace temperature through genetic algorithm | |
| CN107038489B (en) | Multi-target unit combination optimization method based on improved NBI method | |
| CN113836819B (en) | Bed temperature prediction method based on time sequence attention |
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 |