CN109499364A - A kind of catalyst auxiliary design method based on digital mirror image - Google Patents
A kind of catalyst auxiliary design method based on digital mirror image Download PDFInfo
- Publication number
- CN109499364A CN109499364A CN201811444673.0A CN201811444673A CN109499364A CN 109499364 A CN109499364 A CN 109499364A CN 201811444673 A CN201811444673 A CN 201811444673A CN 109499364 A CN109499364 A CN 109499364A
- Authority
- CN
- China
- Prior art keywords
- catalyst
- layer
- input
- parameter
- channel
- 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.)
- Granted
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D53/00—Separation of gases or vapours; Recovering vapours of volatile solvents from gases; Chemical or biological purification of waste gases, e.g. engine exhaust gases, smoke, fumes, flue gases, aerosols
- B01D53/34—Chemical or biological purification of waste gases
- B01D53/74—General processes for purification of waste gases; Apparatus or devices specially adapted therefor
- B01D53/86—Catalytic processes
- B01D53/8621—Removing nitrogen compounds
- B01D53/8625—Nitrogen oxides
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D53/00—Separation of gases or vapours; Recovering vapours of volatile solvents from gases; Chemical or biological purification of waste gases, e.g. engine exhaust gases, smoke, fumes, flue gases, aerosols
- B01D53/34—Chemical or biological purification of waste gases
- B01D53/74—General processes for purification of waste gases; Apparatus or devices specially adapted therefor
- B01D53/86—Catalytic processes
- B01D53/88—Handling or mounting catalysts
Landscapes
- Engineering & Computer Science (AREA)
- Environmental & Geological Engineering (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Analytical Chemistry (AREA)
- General Chemical & Material Sciences (AREA)
- Oil, Petroleum & Natural Gas (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Exhaust Gas Treatment By Means Of Catalyst (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention develops a kind of catalyst auxiliary design method based on digital mirror image, belongs to coal-fired denitration technology field, includes the following steps: to establish catalyst one channel model, and carry out numerical simulation to one channel model using multiple groups different condition and analyze；According to large result obtained by numerical simulation, BP neural network model and database are established；Finally using the output parameter under the corresponding input condition of database lookup, corresponding flow field situation is obtained；The suction parameter condition of the given each one channel of first layer catalyst, and outlet parameter condition is obtained according to database lookup, and using upper layer outlet parameter as lower layer's suction parameter.The information of the problems such as the invention avoids the calculating cycle time of numerical simulation is long and database purchase excessively idealizes, realize the simulation to catalyst in flow field unevenness, the unequal situation of concentration field, with this come for Catalyst Design, running optimizatin, changing the outfit provides foundation, for Catalyst Design and change the outfit, SCR denitration system running optimizatin provides guidance.
Description
Technical field
The invention belongs to coal-fired denitration technology fields, and in particular to a kind of catalyst Computer Aided Design side based on digital mirror image
Method.
Background technique
Currently, China requires to be increasingly stringenter to Pollutant in Coal Burning Boiler emission limit, and most of coal-burning power plants are configured with
SCR denitration system is to meet coal-burning boiler NOxEmission request.It is limited by factors such as place, layouts, SCR denitration system changes
It is limited to make space, flue gas and SCR reducing agent are well-mixed apart from limited, thus enter the flow field of reactor, distribution of concentration
It is easy that there are deviations.
It in addition, the design of part SCR device comes into operation not in the progress of boiler design initial stage, but is to meet state in the later period
Family's environmental regulations require installation.This allows for most units after later period SCR denitration system installation and operation, and it is equal that there are flow fields
The problems such as even property is poor.
Two important indicators for influencing SCR denitration system operation are denitration efficiency and the escaping of ammonia rate.Influence SCR denitration performance
Not only have speed, ammonia nitrogen ratio, temperature, catalyst activity, it is also related with the flow field in SCR reactor, the distribution of concentration field.It is de-
Nitre Installed System Memory will have a direct impact on denitration efficiency and escape with ammonia the problems such as Flow Field Distribution is uneven, chaotic, ammonia concentration is unevenly distributed
Ease rate.As the inhomogeneities of various affecting parameters increases, the service life of catalyst can be reduced, and caused by the inhomogeneities in flow field
It influences bigger.When the inhomogeneities of flue gas flow field increases, it is possible that local velocity is excessive when flue gas enters SCR reactor
Or too small situation, and due to containing a large amount of flying dust in flue gas, if speed is higher when by catalyst duct, flying dust
Meeting abrasive catalyst surface, if speed is lower, flying dust is piled up in catalyst duct, influences gradually to add under the passage of time
Weight.
Core as catalyst as SCR denitration system, design can have very big shadow with type selecting to denitrating system performance
It rings.Currently, the main method of design catalyst is to carry out simulation meter by the way of numerical simulation when designing SCR denitration system
It calculates.Numerical simulation is increasingly used in the optimization design of SCR catalyst as a kind of efficient and convenient mode.It is this logical
Numerical simulation is crossed to provide the method for guidance for Catalyst Design, it is long there are calculating cycle the problems such as.By numerical simulation result,
Limitation setting is carried out to reactor inlet speed and reducing agent distribution when designing SCR system, so that outlet drain reaches existing
Capable standard.But for this method that adjustment inlet velocity and reducing agent distribution is set in advance, the numerical simulation calculation time is long,
It is not easy adjustment setting, the relatively complicated complexity of method, and is not easy to be operated at the scene.
Summary of the invention
Goal of the invention: of the existing technology in order to solve the problems, such as, the present invention provides a kind of catalysis based on digital mirror image
Agent auxiliary design method carries out more effective Catalyst Design, running optimizatin by this method and changes the outfit, to improve denitration effect
Rate reduces the escaping of ammonia rate.
Technical solution: in order to achieve the above-mentioned object of the invention, the present invention adopts the following technical scheme:
A kind of catalyst auxiliary design method based on digital mirror image, includes the following steps:
1) catalyst one channel model is established, and numerical simulation is carried out according to model；
2) using SCR catalyst suction parameter, outlet parameter as the training input of BP neural network, training output data；
BP neural network is trained using the trained input data, training output data；Selected part numerical simulation data is made
For test input, using the resulting BP neural network model prediction catalyst outlet parameter of training, established according to BP neural network
Associated databases；
3) catalyst is divided into several one channel regions by catalyst self structure, to simulate live Flow Field Distribution not
Different primary condition is arranged for each one channel region in equal situation；
4) according to the initial parameter in each duct, established database is inquired, obtains its outlet parameter, and three layers are urged
Agent entrance, outlet parameter coupling；
5) according to catalyst outlet parameter, the operation and maintenance of design selection and subsequent catalyst to catalyst change the outfit
Auxiliary information is provided.
Further, the step 1) specific steps are as follows: establish catalyst one channel model, and counted according to model
Value simulation, to obtain influence of the operating parameter to its denitration performance；It is given for model after establishing SCR catalyst one channel model
Different initial input parameters, analog rate range is 3-8m/s, ammonia nitrogen than range is 0.8-1.2, temperature range 270-360
DEG C, entrance NO concentration range be 50-500mg/m3, amount to 1250 samples；Numerical simulation is carried out according to different input parameters,
Obtain the escaping of ammonia rate, NO concentration at the outlet output parameter.
Further, the step 2) specific steps are as follows:
2.1) data acquire
Input that the numerical simulation result obtained in step 1) is modeled as BP neural network, output data；And it will be defeated
Enter data and utilizes formula xk=(xk-xmin)/(xmax-xmin) be normalized, wherein xmin、xmaxRespectively training input number
According to the minimum value of middle input variable, maximum value；2.2) according to formula l < n-1,L=log2N determines that network is hidden
The range of the l of number containing node layer, wherein n, m are input layer number, and a is the integer between 1~10；Hidden layer is obtained according to above formula
The range of number of nodes l gathers according to examination for several times in range obtain required node in hidden layer l later；
2.3) neural metwork training
According to connection weight, each layer threshold value between input data, each layer, the output of hidden layer and output layer is obtained；And
Output layer output data is compared with desired output, obtains training error；According to training error, gradient modification method is utilized
It is constantly reversed to update network weight and threshold value, until training terminates；If calculating error is less than the minimal error set in model, i.e.,
Training terminates；
2.4) neural network prediction
Using each SCR catalyst one channel entrance primary condition as test input data, obtained after being input to training
BP neural network model, prediction obtains denitration efficiency and the escaping of ammonia rate as test output.
Further, the step 3) the specific steps are be divided by its own design feature catalyst several
A one channel catalyst, for be arranged at each one channel area entry of first layer catalyst initial speed, ammonia nitrogen ratio, temperature,
Entrance NO concentration.
Further, the step 4) the specific steps are the entry condition that step 3) is set is input to data
In library, corresponding outlet parameter is obtained by searching；And it is corresponding using the output parameter of each one channel of first layer as the second layer
The input parameter of one channel, and continue to search the operation of database, and so on.
Further, the step 5) the specific steps are escape according to catalyst outlet parameter, such as denitration efficiency, ammonia
Ease rate, the operation and maintenance of design selection and subsequent catalyst to catalyst, changing the outfit provides auxiliary information, more efficiently, just
It is prompt；According to the above method, denitration efficiency caused by uneven, interior flow field is unevenly distributed for catalyst inlet Flow Field Distribution is low
Under, the escaping of ammonia rate problem, the historical data information by obtaining SCR catalyst flow field represents SCR denitration system actual motion
When catalyst the case where；It, can be appropriate in Catalyst Design type selecting according to historical data information and the database established
The distributed architecture of catalyst is adjusted, operation and maintenance are changed the outfit time, catalyst inactivation with when changing the outfit by database prediction.
Inventive principle: the design of catalyst is a more complicated problem, is related to various factors.Traditional catalyst
Design commonly relies on previous design experiences to establish mechanism or mathematical model.Recognized in this way according to the experience to reaction mechanism
The problems such as progress Catalyst Design, there are errors greatly, separate practical, its scope of application is caused to be very limited.The present invention couple
The single duct of catalyst carries out numerical simulation, and is incorporated into monolithic catalyst module, can adequately simulate field condition,
More closing to reality.Using neural network database, realizes the storage and update of catalyst related data, set for catalyst
Meter, running optimizatin, changing the outfit provides comprehensive, convenient, efficient data transmitting and application.Neural network is not direct reflection catalyst
Interaction between each component performance, but by being associated with layer by layer, the action function using non-linear implication is more objective
The coupled relation established between catalyst inlet, outlet parameter.Neural net model establishing is a black box process, is avoided a large amount of
The numerical simulation of time and cumbersome catalyst reaction mechanism, find out the relationship between input, output data automatically, are catalyst
Design, running optimizatin, changing the outfit provides mass data relationship.
The utility model has the advantages that compared with prior art, a kind of catalyst auxiliary design method based on digital mirror image of the invention,
The case where establishing associated databases according to the denitration reaction situation of one channel, make each duct closer to actual conditions, and with
Multiple one channel simulation flood catalyst modules couple monolithic catalyst with single-layer catalyst, can preferably reflect that SCR reacts
Device entrance flow field, distribution of concentration unevenness situation, make Catalyst Design more closing to reality；Utilize database purchase Numerical-Mode
Bulk information after quasi-, both avoided numerical simulation the calculating cycle time it is long the problems such as, in turn avoid the letter of database purchase
Breath excessively idealizes, realize to catalyst flow field is uneven, the simulation in the unequal situation of concentration field, is set with this for catalyst
Meter, running optimizatin, changing the outfit provides foundation, for Catalyst Design and change the outfit, SCR denitration system running optimizatin provides guidance.
Detailed description of the invention
Fig. 1 is SCR catalyst one channel and integral module schematic diagram；
Fig. 2 is SCR catalyst module subregion schematic diagram；
Fig. 3 is neural net model establishing block diagram；
Fig. 4 is primary condition schematic diagram；
Fig. 5 is schematic diagram of calculation flow.
Specific embodiment
The present invention is further described with specific implementation example with reference to the accompanying drawing.
A kind of catalyst auxiliary design method based on digital mirror image, includes the following steps:
1) catalyst one channel model is established, and numerical simulation is carried out according to model, to obtain operating parameter to its denitration
The influence of performance；After establishing SCR catalyst one channel model, different initial input parameters, analog rate model are given for model
Enclosing for 3-8m/s, ammonia nitrogen than range is 0.8-1.2, temperature range is 270-360 DEG C, entrance NO concentration range is 50-500mg/
m3, amount to 1250 samples；Numerical simulation is carried out according to different input parameters, it is defeated to obtain the escaping of ammonia rate, NO concentration at the outlet etc.
Parameter out；
2) using SCR catalyst suction parameter, outlet parameter as the training input of BP neural network, training output data；
BP neural network is trained using the trained input data, training output data；Selected part numerical simulation data is made
For test input, using the resulting BP neural network model prediction catalyst outlet parameter of training, established according to BP neural network
Associated databases；
3) catalyst is divided into several one channel regions by catalyst self structure, to simulate live Flow Field Distribution not
Different primary condition, such as speed, ammonia nitrogen ratio, temperature, entrance NO concentration is arranged for each one channel region in situations such as equal；
4) according to the initial parameter in each duct, established database is inquired, obtains its outlet parameter, and three layers are urged
Agent entrance, outlet parameter coupling；
5) according to catalyst outlet parameter, such as denitration efficiency, the escaping of ammonia rate, design selection to catalyst and subsequent
The operation and maintenance of catalyst, changing the outfit provides auxiliary information, more efficiently, convenient；According to the above method, for catalyst inlet stream
The problems such as field distribution is uneven, interior flow field is unevenly distributed low caused denitration efficiency, the escaping of ammonia rate can be urged by obtaining SCR
The historical data information in agent flow field is come when representing SCR denitration system actual motion the case where catalyst.Believed according to historical data
Breath and the database established can run dimension in Catalyst Design type selecting with the distributed architecture etc. of appropriate adjustment catalyst
Change the outfit time, catalyst inactivation etc. can also be predicted by database when protecting and changing the outfit.
Step 2) specific steps are as follows:
2.1) data acquire
Input that the numerical simulation result obtained in step 1) is modeled as BP neural network, output data.And it will be defeated
Enter data and utilizes formula xk=(xk-xmin)/(xmax-xmin) be normalized, wherein xmin、xmaxRespectively training input number
According to the minimum value of middle input variable, maximum value；
2.2) according to l < n-1,L=log2N formula determines network node in hidden layer；Wherein n, m are
Input layer number, a are the integer between 1~10；The approximate range that node in hidden layer l is obtained according to above formula, later in model
Interior basis is enclosed to try to gather for several times to obtain best node in hidden layer；
2.3) neural metwork training
According to connection weight, each layer threshold value between input data, each layer, the output of hidden layer and output layer is obtained, and
Output layer output data is compared with desired output, obtains training error；According to training error, gradient modification method is utilized
It is constantly reversed to update network weight and threshold value, until training terminates；If calculating error is less than the minimal error set in model, i.e.,
Training terminates；
2.4) neural network prediction
Using each SCR catalyst one channel entrance primary condition as test input data, obtained after being input to training
BP neural network model, prediction obtains denitration efficiency and the escaping of ammonia rate as test output.
Step 3) specific steps are as follows:
Catalyst is divided into several one channel catalyst by its own design feature, is each list of first layer catalyst
Initial speed, ammonia nitrogen ratio, temperature, entrance NO concentration etc. are set at the area entry of duct；
Step 4) specific steps are as follows:
The entry condition that step 3) is set is input in database, obtains corresponding outlet parameter by searching；And
The input parameter of one channel is corresponded to using the output parameter of each one channel of first layer as the second layer, and continues to search data
The operation in library, and so on.
Fig. 1 shows SCR one channel model and catalyst whole relation schematic diagram, and integral module includes three layers of catalysis
Agent.SCR catalyst reactor is placed using module, and the number of plies depends on required catalyst reaction surface area.Typical cloth
Mode is set as 2~3 layers of catalyst layer of arrangement.Though integral module studies the flow field that can preferably simulate catalyst in actual motion
Situation, but due in catalyst micropore structure and distribution it is sufficiently complex, micro unit should be directed to the research of its response situation
Carry out.One channel denitration reaction is an important factor for measuring entire catalyst denitration ability.Catalyst is integrally pressed to its structure point
For several one channels, the different entry condition of each one channel can be very good the feelings of simulated flow pattern, distribution of concentration unevenness
Condition.
Shown in Fig. 2, actual SCR catalyst one channel is reduced to be a square duct simplified model, flue gas is by cigarette
After gas entrance enters channel, flue gas completes denitrification process along airflow direction, and denitrification process is by flowing, heat and mass and denitration
What reaction collectively constituted.Wherein mass transport process includes that reaction gas caused by flowing with gas transports, and because concentration gradient is drawn
The reaction gas diffusion risen.Work as NH3When passing through catalyst layer with smoke mixture, reaction process are as follows: in duct, reacted constituent
(NO、NH3) from mainstream smoke catalyst duct wall surface is diffused to, and then in micro- pore diffusion of catalyst wall surface.It adsorbs later
In on catalyst, and chemically reacted on a catalyst.The N of generation2And H2O desorption from catalyst surface.After desorption
Reaction product, to external diffusion, is then spread from catalyst duct wall surface to mainstream smoke in the micropore of catalyst wall surface.I.e.
Complete a microreaction.Reaction more than constantly carrying out in catalyst duct is to realize denitrification process.Finally gone out by flue gas
Mouth outflow, and enter the corresponding channel of next layer of catalyst.
Catalyst module subregion schematic diagram as shown in Figure 2, if every layer of catalyst is divided into according to catalyst self structure
Dry block one channel region, each one channel area size are identical.Each one channel is arranged with different initial flow-fields, concentration field
Condition, i.e. analog SCR reactor inlet flue gas flow field, the situation of concentration distribution unevenness.One of one channel is carried out more
The numerical simulation of group primary condition, obtains the initial data of neural network model.
Fig. 3 is the topology diagram of the BP neural network model of present example.The BP neural network that the present invention constructs point
It is 3 layers, including input layer, hidden layer, output layer.Its main thought is: input learning sample, using back-propagation algorithm to net
The weight and deviation of network carry out adjusting training repeatedly, and the vector and Mean Vector for making output are close to when network is defeated
Training is completed when the error sum of squares of layer is less than specified error out, saves the weight and deviation of network.BP algorithm utilizes output
Error afterwards estimates the error of the direct preceding conducting shell of output layer, then with the error of this estimation error more preceding layer, such one
As soon as the anti-pass of layer layer is gone down, the estimation error of every other each layer is obtained.Its learning process is to be transmitted by the forward direction of signal
To the back transfer of error.Using the numerical simulation result of Fig. 2 catalyst one channel model established, BP neural network is carried out
Modeling.By forward-propagating and back-propagation process, constantly reduction network error, optimize network.
BP neural network modeling mainly includes acquisition phase, training stage, forecast period, optimizing phase.Acquisition phase is
Numerical simulation is obtained into a large amount of result data as the trained data of neural net model establishing.Due to the tool between input data
There are magnitude differences, needs that input data is normalized.The normalized result of k-th of data in input variable
Are as follows:
xk=(xk-xmin)/(xmax-xmin)
Wherein xminFor the minimum value of input variable in training input data, xmaxFor maximum value.
It is first initialized when BP neural network training.Node in hidden layer influences very big, root to the precision of prediction of network
The range of node in hidden layer l is determined according to following formula:
l<n-1
L=log2n；
Wherein n is input layer number, and m is output layer number of nodes, and a is the integer between 1~10.It is obtained according to above formula hidden
The range of the l of number containing node layer gathers according to examination for several times in range obtain best node in hidden layer l later.
Determine input layer, hidden layer, output layer number of nodes after start carry out BP neural network training.According to input
Connection weight, each layer threshold value between data, each layer obtain the output of hidden layer and output layer.And by output layer output data
It is compared with desired output, obtains training error.According to training error, network is constantly reversely updated using gradient modification method
Weight and threshold value, until training terminates.If calculating error is less than the minimal error set in model, i.e. training terminates.
The primary condition for constantly changing SCR catalyst one channel entrance is right using each primary condition as test input data
After test input data is normalized, it is input to obtained BP neural network model after training, can be obtained pair
The output data that should be tested.
The primary condition schematic diagram of present example is as shown in Figure 4.For monolithic catalyst module, catalyst is pressed as schemed
Subregion shown in 2, and primary condition setting in each one channel region is as shown in Figure 4.To be arranged just at each one channel area entry
Speed, ammonia nitrogen ratio, temperature, NO concentration of beginning etc..In reaction, heat release and the thermal loss outwardly transmitted are smaller, can ignore
Its influence to gas temperature, it is believed that denitrification process carries out at a constant temperature.
Calculation process is as shown in Figure 5.The primary condition of first layer catalyst is set according to Fig.4, is given known defeated
Enter, and database is searched according to specified criteria, to give output parameter by database.Existing research shows the first floor at present
The denitration degree of irregularity of catalyst is affected to second layer catalyst inlet ammonia nitrogen than uneven distribution, and then will affect
Two layers of catalyst denitration effect.Thus example of the invention carries out overall applicability to three layers of catalyst.And by overlying catalyst
Entry condition of the exit result as lower catalyst agent, i.e., output parameter is passed into next layer of catalyst.Recycle lower layer
The entry condition of catalyst searches database and obtains output parameter, and so on.
In conclusion be applied to it is of the present invention it is a kind of based on the Catalyst Design method of digital mirror image, it can be achieved that urging
Agent design, changes the outfit at running optimizatin, introduces BP neural network and carries out numerical simulation data analysis, obtains catalyst information data
Library.The calculating cycle time for greatly shortening numerical simulation during Catalyst Design, so that Catalyst Design, running optimizatin
It is more efficiently convenient with changing the outfit, to improve SCR denitration system denitration efficiency, reduce the escaping of ammonia rate.
Method of the invention is described in detail with specific implementation method above, and gives corresponding implement in fact
Example.Certainly, besides these examples, the present invention can also have other embodiments, all to be formed using equivalent substitution or equivalent transformation
Technical solution, all fall within invention which is intended to be protected.
Claims (6)
1. a kind of catalyst auxiliary design method based on digital mirror image, characterized by the following steps:
1) catalyst one channel model is established, and numerical simulation is carried out according to model；
2) using SCR catalyst suction parameter, outlet parameter as the training input of BP neural network, training output data；It utilizes
The trained input data, training output data are trained BP neural network；Selected part numerical simulation data is as survey
Examination input is established corresponding using the resulting BP neural network model prediction catalyst outlet parameter of training according to BP neural network
Database；
3) catalyst is divided into several one channel regions by catalyst self structure, to simulate live Flow Field Distribution unevenness
Different primary condition is arranged for each one channel region in situation；
4) according to the initial parameter in each duct, established database is inquired, obtains its outlet parameter, and by three layers of catalyst
Entrance, outlet parameter coupling；
5) according to catalyst outlet parameter, the operation and maintenance of design selection and subsequent catalyst to catalyst, change the outfit offer
Auxiliary information.
2. a kind of catalyst auxiliary design method based on digital mirror image according to claim 1, it is characterised in that: described
Step 1) specific steps are as follows: after establishing SCR catalyst one channel model, give different initial input parameters, mould for model
Quasi- velocity interval is 3-8m/s, ammonia nitrogen than range is 0.8-1.2, temperature range is 270-360 DEG C, entrance NO concentration range is
50-500mg/m3, amount to 1250 samples；Numerical simulation is carried out according to different input parameters, obtains the escaping of ammonia rate, outlet NO
Concentration output parameter.
3. a kind of catalyst auxiliary design method based on digital mirror image according to claim 1, it is characterised in that: described
Step 2) specific steps are as follows:
2.1) data acquire
Input that the numerical simulation result obtained in step 1) is modeled as BP neural network, output data；And number will be inputted
According to utilization formula xk=(xk-xmin)/(xmax-xmin) be normalized, wherein xmin、xmaxRespectively train in input data
Minimum value, the maximum value of input variable；
2.2) according to formula l < n-1,L=log2N determines the range of network node in hidden layer l, wherein n,
M is input layer number, and a is the integer between 1~10；The range that node in hidden layer l is obtained according to above formula, later in range
Interior basis tries to gather for several times node in hidden layer l needed for acquisition；
2.3) neural metwork training
According to connection weight, each layer threshold value between input data, each layer, the output of hidden layer and output layer is obtained；And it will be defeated
Layer output data is compared with desired output out, obtains training error；It is continuous using gradient modification method according to training error
It is reversed to update network weight and threshold value, until training terminates；If calculating error is less than the minimal error set in model, that is, train
Terminate；
2.4) neural network prediction
Using each SCR catalyst one channel entrance primary condition as test input data, the BP obtained after being input to training
Neural network model, prediction obtain denitration efficiency and the escaping of ammonia rate as test output.
4. a kind of catalyst auxiliary design method based on digital mirror image according to claim 1, it is characterised in that: described
Step 3) the specific steps are, catalyst is divided into several one channel catalyst by its own design feature, be first
Initial speed, ammonia nitrogen ratio, temperature, entrance NO concentration are set at the layer each one channel area entry of catalyst.
5. a kind of catalyst auxiliary design method based on digital mirror image according to claim 4, it is characterised in that: described
Step 4) the specific steps are, the entry condition that step 3) is set is input in database, through lookup corresponded to
Outlet parameter；And the input parameter of one channel is corresponded to using the output parameter of each one channel of first layer as the second layer, and after
The continuous operation for carrying out searching database, and so on.
6. a kind of catalyst auxiliary design method based on digital mirror image according to claim 1, it is characterised in that: described
Step 5) the specific steps are, by obtain SCR catalyst flow field historical data information come represent SCR denitration system reality
When operation the case where catalyst；According to historical data information and the database established, adjusted in Catalyst Design type selecting
The distributed architecture of catalyst, operation and maintenance are changed the outfit time, catalyst inactivation with when changing the outfit by database prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811444673.0A CN109499364B (en) | 2018-11-29 | 2018-11-29 | Catalyst aided design method based on digital mirror image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811444673.0A CN109499364B (en) | 2018-11-29 | 2018-11-29 | Catalyst aided design method based on digital mirror image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109499364A true CN109499364A (en) | 2019-03-22 |
| CN109499364B CN109499364B (en) | 2021-05-28 |
Family
ID=65751234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811444673.0A Active CN109499364B (en) | 2018-11-29 | 2018-11-29 | Catalyst aided design method based on digital mirror image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109499364B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298142A (en) * | 2019-07-19 | 2019-10-01 | 国电环境保护研究院有限公司 | A kind of denitrating catalyst operational characteristic index prediction technique based on measured data |
| CN112403261A (en) * | 2020-11-20 | 2021-02-26 | 中煤能源研究院有限责任公司 | Method for digitally constructing SCR reactor of flue gas denitration core device |
| CN112461995A (en) * | 2020-11-03 | 2021-03-09 | 西安热工研究院有限公司 | Method for predicting ammonia escape of thermal power plant |
| CN114880943A (en) * | 2022-05-24 | 2022-08-09 | 安及义实业（上海）有限公司 | Bioreactor design method and system based on database |
-
2018
- 2018-11-29 CN CN201811444673.0A patent/CN109499364B/en active Active
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298142A (en) * | 2019-07-19 | 2019-10-01 | 国电环境保护研究院有限公司 | A kind of denitrating catalyst operational characteristic index prediction technique based on measured data |
| CN110298142B (en) * | 2019-07-19 | 2023-01-03 | 国电环境保护研究院有限公司 | Denitration catalyst process characteristic index prediction method based on measured data |
| CN112461995A (en) * | 2020-11-03 | 2021-03-09 | 西安热工研究院有限公司 | Method for predicting ammonia escape of thermal power plant |
| CN112403261A (en) * | 2020-11-20 | 2021-02-26 | 中煤能源研究院有限责任公司 | Method for digitally constructing SCR reactor of flue gas denitration core device |
| CN114880943A (en) * | 2022-05-24 | 2022-08-09 | 安及义实业（上海）有限公司 | Bioreactor design method and system based on database |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109499364B (en) | 2021-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109499364A (en) | A kind of catalyst auxiliary design method based on digital mirror image | |
| CN107243257B (en) | It is suitble to the intelligence spray ammonia control system of full load | |
| CN102693451B (en) | Method for predicting ammonia process flue gas desulphurization efficiency based on multiple parameters | |
| CN106599586B (en) | SCR neural network based intelligently sprays ammonia optimization method and device | |
| CN110263395A (en) | The power plant's denitration running optimizatin method and system analyzed based on numerical simulation and data | |
| CN105700504B (en) | SCR system autocontrol method based on spray ammonia sensitive valve locking | |
| CN101281698A (en) | Simulation platform and test method for flow field arrangement structure of flue gas denitration apparatus | |
| CN106870955A (en) | Serve the pipe network monitoring point optimization placement method of water supply network node water requirement inverting | |
| Maile et al. | Formalizing approximations, assumptions, and simplifications to document limitations in building energy performance simulation | |
| CN108760592B (en) | Fly ash carbon content online measurement method based on BP neural network | |
| US20240054262A1 (en) | Carbon emission flow calculation method and apparatus for regional integrated energy system | |
| CN113689917A (en) | Visual ammonia injection optimization method and device based on SCR outlet NOx concentration | |
| CN107831651A (en) | Coal dust station boiler SCR and SNCR denitration system cooperative control method | |
| CN204903485U (en) | Two reacting furnace activity detection devices of SCR denitration catalyst | |
| CN102842066A (en) | Modeling method for combustion optimization of biomass furnace | |
| CN110414089A (en) | The simulated prediction method of vehicle PEMS discharge based on Engine Universal Characteristics | |
| CN116189791A (en) | Ammonia spraying optimization method, device and storage medium for ammonia nitrogen double control of SCR system | |
| WO2024021943A1 (en) | Ammonia injection optimization method and apparatus based on scr reaction kinetic model, and medium | |
| CN111489605A (en) | Ammonia spraying optimization control simulation system based on Simulink and WinCC | |
| CN112613237B (en) | CFB unit NOx emission concentration prediction method based on LSTM | |
| CN109766666A (en) | Boiler smoke based on low nitrogen burning and SNCR-SCR Collaborative Control discharges NOxConcentration prediction method | |
| CN105498530A (en) | Method for instructing ammonia spray adjustment through introducing tracer gases in SCR system simulation | |
| CN112426856B (en) | Flue gas desulfurization flow field simulation method, system and device | |
| CN112365065A (en) | WFGD self-adaptive online optimization scheduling method | |
| CN107401784A (en) | Air conditioning system with variable and design method based on genetic algorithm |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |