CN108446799A - Waste pressure turbine device generated power forecasting method based on Elman neural networks - Google Patents
Waste pressure turbine device generated power forecasting method based on Elman neural networks Download PDFInfo
- Publication number
- CN108446799A CN108446799A CN201810198547.5A CN201810198547A CN108446799A CN 108446799 A CN108446799 A CN 108446799A CN 201810198547 A CN201810198547 A CN 201810198547A CN 108446799 A CN108446799 A CN 108446799A
- Authority
- CN
- China
- Prior art keywords
- layer
- model
- prediction
- neural networks
- error
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 36
- 239000002699 waste material Substances 0.000 title claims abstract description 14
- 238000013277 forecasting method Methods 0.000 title claims abstract description 6
- 238000012549 training Methods 0.000 claims abstract description 23
- 238000005070 sampling Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000010606 normalization Methods 0.000 claims abstract description 5
- 238000000034 method Methods 0.000 claims description 36
- 239000007789 gas Substances 0.000 claims description 31
- 230000006870 function Effects 0.000 claims description 12
- 238000011084 recovery Methods 0.000 claims description 9
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 claims description 6
- 229910002092 carbon dioxide Inorganic materials 0.000 claims description 6
- 238000010248 power generation Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 5
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical group [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 claims description 4
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 claims description 4
- 229910002091 carbon monoxide Inorganic materials 0.000 claims description 4
- 239000003034 coal gas Substances 0.000 claims description 4
- 229910052739 hydrogen Inorganic materials 0.000 claims description 4
- 239000001257 hydrogen Substances 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 239000001569 carbon dioxide Substances 0.000 claims description 3
- 238000009738 saturating Methods 0.000 claims 1
- 238000005457 optimization Methods 0.000 description 10
- 230000001419 dependent effect Effects 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 230000000306 recurrent effect Effects 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 3
- 229910000831 Steel Inorganic materials 0.000 description 2
- GWEVSGVZZGPLCZ-UHFFFAOYSA-N Titan oxide Chemical compound O=[Ti]=O GWEVSGVZZGPLCZ-UHFFFAOYSA-N 0.000 description 2
- 239000003245 coal Substances 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000002203 pretreatment Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 239000010959 steel Substances 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011109 contamination Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 239000004408 titanium dioxide Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Marketing (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Manufacturing & Machinery (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
Abstract
The waste pressure turbine device generated power forecasting method based on Elman neural networks that the invention discloses a kind of.Step 1：The data of Non uniform sampling are subjected to interpolation processing, form it into the equal sequence of time interval, the whole tendency and related coefficient of the every input variable of analysis, select six variables high with the generated energy degree of correlation as input variable, input variable is pre-processed into row interpolation, normalization using python and Matlab platforms, keeps sample frequency uniform；Step 2：Basic model of the Elman neural networks as prediction is chosen, using error back propagation learning algorithm（BPTT）And sliding window model is combined, it is used for the prediction of waste pressure turbine device generated output；Step 3：After model initialization, with normalized training sample training pattern, trained model is used for the prediction of waste pressure turbine device generated output.The present invention makes the big prediction of TRT generated outputs have higher precision, smaller mean square error, can be used on line predicting in real time.
Description
Technical field
The invention belongs to industrial process monitoring, modeling and simulation field, more particularly to a kind of Elman-BPTT neural networks
The method for predicting waste pressure turbine device generated energy.
Technical background
Blast furnace waste pressure retracting device is commonly used with recovery section energy in steel industry.Based on to blast furnace gas top pressure
The capability forecasting of recovery gas turbine power generator (Top Gas Pressure Recovery Turbine, abbreviation TRT) is optimization
Control the important foundation of its power generation.Accurate capability forecasting (model prediction in other words) is to improve real-time optimization (Real-time
Optimization, RTO) it is essential.Because RTO uses static models, when interfering, controlled system reaches steady again
State can just optimize, so optimization lag, and RTO Nonlinear Steady optimization cycle mistakes can be solved by being combined with capability forecasting
The long and too short inconsistency of circuit controlling cycle.Direct-on-line optimization method uses Nonlinear Model Predictive Control, limited
On-line optimization economic performance index in step-length, the above method require accurate nonlinear process model.
At present to the research of similar industrial process, it is broadly divided into two big modes.
First, traditional mechanisms model, on the basis of analyzing TRT system working mechanisms, the dynamic mathematical modulo of each unit is established
Type derives TRT systems in conjunction with each unit model based on blast furnace gas generating process empirical equation and material balance relationship
Generated energy mathematical model.But since blast furnace ironmaking process is complicated, it is difficult to establish accurate model by mechanism, actual value pole can
It can predict to differ larger with modeling.
Second is that data-driven models.Mainly statistical method and machine learning method are used.Wherein, statistical method
Include mainly：Partial Least Squares, principal component analysis, auto-regressive analysis, regression analysis.And the method for machine learning has nerve net
Network model and support vector machines etc..In the case where there is mass data that can carry out analyzing or training neural network, prediction is expected
Obtain preferable effect.
But the shortcomings that data-driven modeling also has oneself, due to certain features of big data in process industrial, traditional number
Some problems can be all encountered according to modeling method.First, available data modeling method focuses mostly in the modeling to rule sampling data
With analysis, the data of irregular sampling can not be modeled and be analyzed.This requires the data to irregular sampling to carry out in advance
Processing, but this partial routine can lost part data.Second, Data Modeling Method is mainly the static number for extracting space at present
It analyzes according to statistics, does not consider temporal dynamic property, operating condition characteristic information cannot be extracted from history dynamic big data, and it is industrial
Model needs dynamic analysis.
Third, available data modeling method need free of contamination data or to pretreated data modeling.Otherwise, abnormal
Point brings very big influence back to model parameter, even results in model mismatch.
Invention content
For overcome the deficiencies in the prior art, the present invention provides a kind of waste pressure turbine device power generation based on Elman models
Power forecasting method.
A kind of waste pressure turbine device generated power forecasting method based on Elman neural networks,
Step 1：The data of Non uniform sampling are subjected to interpolation processing, form it into the equal sequence of time interval, are analyzed
The whole tendency and related coefficient of every input variable have selected six variables high with the generated energy degree of correlation to become as input
Amount, the input variable are carbon monoxide content, titanium dioxide in gas flow, gas inlet pressure, gas inlet temperature, coal gas
Carbon content, hydrogen content pre-process input variable into row interpolation, normalization using python and Matlab platforms, make sampling
Frequency-flat；
Step 2：Basic model of the Elman neural networks as prediction is chosen, using error back propagation learning algorithm
(BPTT) and sliding window model is combined, is used for the prediction of waste pressure turbine device generated output；
Step 3：After model initialization, with normalized training sample training pattern, trained model is used for remaining
Press the prediction of turbine installation generated output.
Step 1 related coefficient is as follows：
Gas flow, unit m3/ h, related coefficient 0.9294；
Gas inlet pressure, unit kPa, related coefficient -0.0397；
Gas inlet temperature, unit DEG C, related coefficient 0.0352；
CO volume contents, 0.1278%；
CO2Volume content, 0.1935%；
For related coefficient between 0 to 1, bigger expression correlation is stronger；
Method for normalizing described in step 1 is as follows：
Result is normalized between 0 to 1.
The structure of Elman neural networks described in step 2 is as follows：
ELman neural networks are constituted by four layers, are input layer respectively, hidden layer, are accepted layer and output layer：
1) by input layer to hidden layer：
X (k)=f (w1xc(k)+w2(u(k-1)))；
2) by hidden layer to output layer：
Y (k)=g (w3x(k))；
3) layer storage information in a short time is accepted：
xc(k)=x (k-1)；
4) learn target function：Using sum of squared errors function, work as E<Terminate to train when 0.0001
5) weight renewing method：
w1To accept layer to the connection weight of hidden layer, w2For the connection weight of input layer to hidden layer, w3Extremely for hidden layer
The connection weight of output layer, three weights are required for by the adjustment that iterates, and error is prediction and reality in training process
Difference, η represents each feedback adjustment degree, is set as 0.001, by constantly feeding back update weights, finally makes study target function value
It meets the requirements, reaches the training completed to neural network
w1(n)=w1(n-1)-η·2·error·w3·xc,
w2(n）=w2(n-1)-η·2·error·w3U,
w3(n)=w3(n-1)-η·2·error·w3·xc。
Sliding window model principle described in step 2 is as follows：
Sliding window model is established on a kind of hypothesis, i.e., current output depends on current input, and inputs defeated
Mapping ruler between going out is obtained by historical data, according to this it is assumed that presetting a certain amount of training set sample, then
It is continuously updated sample data and gives up earliest data point, with the sliding of window, T-S fuzzy neural networks constantly update it
Structural parameters simultaneously provide newest predicted value.
Model described in step 3 is suitable for the time-varying of blast furnace coal gas residual pressure recovery turbine power generation, dynamic, non-linear, strong used
The characteristic of property.
The present invention makes the big prediction of TRT generated outputs have higher precision, smaller mean square error, can be used for pre- in real time on line
It surveys.
Description of the drawings
Present invention will be further explained below with reference to the attached drawings and examples.
Fig. 1 is the structural schematic diagram of Elman neural networks；
Fig. 2 is the schematic diagram of sliding window.
Specific implementation mode
During blast furnace ironmaking, chemical reaction is slower, and the current working of a furnace and history working of a furnace correlation is strong, shows as one
A dynamic time series, it is desirable that neural network dynamic remembers historical information.Therefore recurrent neural network is used
(Recurrent Neural Network, RNN) carries out TRT generated outputs and is predicted.In the branch of recurrent neural network,
The characteristics of being worked according to TRT, we have selected the Elman neural network sensitive to historical data, it can be regarded as a tool
The recurrent neural network for having local mnemon and local feedback link, in its hidden layer, output is by accepting prolonging for layer
Late with storage, it is linked to its input certainly, makes it that there is sensibility to the data of historic state, and this internal feedback also enhances
The ability of Elman network processes multidate informations.
At present in industrial data excavation, Elman neural networks have prodigious potentiality in terms of analysis time sequence.This hair
The bright prediction that this neural network model is used for TRT generated outputs, achieves preferable effect, to needing to analyze dynamic time
For the industrial occasions of sequence, it should which there is higher application value.
The present invention proposes a kind of waste pressure turbine device generated energy prediction technique based on Elman models, and this method includes
Following steps：
Step 1：Whole tendency, related coefficient in conjunction with the every dependent variable of analysis and consideration of related data, it is determined that
6 with the higher variable of the generated energy degree of correlation as input variable (gas flow, gas inlet pressure, gas inlet temperature, coal
Carbon monoxide content, carbon dioxide content, hydrogen content in gas), input variable is carried out using python and Matlab platforms
The pretreatments such as interpolation, normalization keep sample frequency uniform.
Step 2：Basic model (as shown in Figure 1) of the Elman neural networks as prediction is chosen, is reversely passed using error
It broadcasts learning algorithm (BPTT) and combines sliding window model (as shown in Figure 2), be used for the prediction of TRT device generated outputs.
Step 3：After model initialization, with normalized training sample training pattern, trained model is used for
The prediction of TRT device generated outputs.
1, the data of Non uniform sampling are carried out interpolation processing by step 1, form it into the equal sequence of time interval, are passed through
The analysis of related data and whole tendency, the related coefficient of every dependent variable are crossed, select 6 has with TRT device generated outputs
The variable of pass.Related coefficient is as shown in the table：
| Variable name | Unit | Related coefficient | |
| X1 | Gas flow | m3/h | 0.9294 |
| X2 | Gas inlet pressure | kPa | -0.0397 |
| X3 | Gas inlet temperature | ℃ | 0.0352 |
| X4 | CO contents | % | 0.1278 |
| X5 | CO2Content | % | 0.1935 |
| X6 | H2Content | % | 0.1645 |
Method for normalizing described in step 1 is as follows：
Result is normalized between 0 to 1.
2, the structure of the Elman neural networks described in step 2 is as follows：
ELman neural networks are constituted by four layers, are input layer respectively, hidden layer, are accepted layer and output layer.
(1) by input layer to hidden layer：
X (k)=f (w1xc(k)+w2(u(k-1)))
(2) by hidden layer to output layer：
Y (k)=g (w3x(k))
(3) layer storage information in a short time is accepted：
xc(k)=x (k-1)
(4) learn target function：Using sum of squared errors function, work as E<Terminate to train when 0.0001
(5) weight renewing method：
w1To accept layer to the connection weight of hidden layer, w3For the connection weight of input layer to hidden layer, w3Extremely for hidden layer
The connection weight of output layer.In this method, three weights are required for by the adjustment that iterates, and error is pre- in training process
It surveys and practical difference, η represents each feedback adjustment degree, be set as 0.001, by constantly feeding back update weights, finally make study
Target function value is met the requirements, and reaches the training completed to neural network
w1(n)=w1(n-1)-η·2·error·w3·xc
w2(n)=w2(n-1)-η·2·error·w3·u
w3(n)=w3(n-1)-η·2·error·w3·xc。
3, the sliding window model principle described in step 2 is as follows：
Sliding window model is built upon in a kind of hypothesis, i.e., current output depends on current input, and inputs defeated
Mapping ruler between going out can be obtained by historical data.According to this it is assumed that we preset a certain amount of training set
Then sample is continuously updated sample data and gives up earliest data point.With the sliding of window, T-S fuzzy neural networks
Its structural parameters can be constantly updated and provide newest predicted value.
Model is suitable for the time-varying of blast furnace coal gas residual pressure recovery turbine power generation, the characteristic of dynamic, non-linear, strong inertia.
Embodiment
Blast furnace waste pressure retracting device is commonly used with recovery section energy in steel industry.Based on to blast furnace gas top pressure
The capability forecasting of recovery gas turbine power generator (Top Gas Pressure Recovery Turbine, abbreviation TRT) is optimization
Control the important foundation of its power generation.Accurate capability forecasting (model prediction in other words) is to improve real-time optimization (Real-time
Optimization, RTO) it is essential.
We verify the accuracy of the model of proposition by the data in April, 2016 to the June of research Liu Gang.In the following, I
Implementation steps are explained in detail in conjunction with detailed process：
Step 1：Whole tendency, related coefficient in conjunction with the every dependent variable of analysis and consideration of related data, it is determined that
6 with the higher variable of the generated energy degree of correlation as input variable (gas flow, gas inlet pressure, gas inlet temperature, coal
Carbon monoxide content, carbon dioxide content, hydrogen content in gas), input variable is carried out using python and Matlab platforms
The pretreatments such as interpolation, normalization keep sample frequency uniform.
Step 2：Basic model of the Elman neural networks as prediction is chosen, using error back propagation learning algorithm
(BPTT) and sliding window model is combined, is used for the prediction of TRT device generated outputs.
Step 3：After model initialization, with normalized training sample training pattern, trained model is used for
The prediction of TRT device generated outputs.
1, the data of Non uniform sampling are carried out interpolation processing by step 1, form it into the equal sequence of time interval, are passed through
The analysis of related data and whole tendency, the related coefficient of every dependent variable are crossed, select 6 has with TRT device generated outputs
The variable of pass.Method for normalizing described in step 1 is as follows：
Result is normalized between 0 to 1.
2, the structure of the Elman neural networks described in step 2 is as follows：
ELman neural networks are constituted by four layers, are input layer respectively, hidden layer, are accepted layer and output layer.
(1) by input layer to hidden layer：
X (k)=f (w1xc(k)+w2(u(k-1)))
(2) by hidden layer to output layer：
Y (k)=g (w3x(k))
(3) layer storage information in a short time is accepted：
xc(k)=x (k-1)
(4) learn target function：Using sum of squared errors function, work as E<Terminate to train when 0.0001
(5) weight renewing method：
w1To accept layer to the connection weight of hidden layer, w2For the connection weight of input layer to hidden layer, w3Extremely for hidden layer
The connection weight of output layer.
In this method, three weights are required for by the adjustment that iterates, and error is prediction and reality in training process
Difference, η represents each feedback adjustment degree, is set as 0.001, by constantly feeding back update weights, finally makes study target function value
It meets the requirements, reaches the training completed to neural network
w1(n)=w1(n-1)-η·2·error·w3·xc
w2(n)=w2(n-1)-η·2·error·w3·u
w3(n)=w3(n-1)-η·2·error·w3·xc。
3, the sliding window model principle described in step 2 is as follows：
Sliding window model is built upon in a kind of hypothesis, i.e., current output depends on current input, and inputs defeated
Mapping ruler between going out can be obtained by historical data.According to this it is assumed that we preset a certain amount of training set
Then sample is continuously updated sample data and gives up earliest data point.With the sliding of window, T-S fuzzy neural networks
Its structural parameters can be constantly updated and provide newest predicted value.
The data at 500 time points are predicted, under this premeasuring, the results of Elman algorithms is relatively stablized, substantially without
Fluctuation.We verify the precision of model prediction with prediction hit rate ratio and mean square error mse two indices：
In the actual production process, prediction error is less than 0.1 and can meet the requirements.It is proposed that the hit rate of model reach
To 93.2%, mean square error 0.0018.With very high precision, it is entirely capable of meeting the needs of actual production.
Above-described embodiment is used for illustrating the present invention, rather than limits the invention, the present invention spirit and
In scope of the claims, to any modifications and changes that the present invention makes, protection scope of the present invention is belonged to.
Claims (5)
1. a kind of waste pressure turbine device generated power forecasting method based on Elman neural networks, which is characterized in that
Step 1：The data of Non uniform sampling are subjected to interpolation processing, form it into the equal sequence of time interval, analysis is every
The whole tendency and related coefficient of input variable have selected six variables high with the generated energy degree of correlation as input variable,
The input variable is carbon monoxide content, carbon dioxide in gas flow, gas inlet pressure, gas inlet temperature, coal gas
Content, hydrogen content pre-process input variable into row interpolation, normalization using python and Matlab platforms, make sampling frequency
Rate is uniform；
Step 2：Basic model of the Elman neural networks as prediction is chosen, using error back propagation learning algorithm (BPTT)
And sliding window model is combined, it is used for the prediction of waste pressure turbine device generated output；
Step 3：It is with normalized training sample training pattern, trained model is saturating for overbottom pressure after model initialization
The prediction of leveling device generated output.
2. according to the method described in claim 1, it is characterized in that, step 1 related coefficient is as follows：
Gas flow, unit m3/ h, related coefficient 0.9294；
Gas inlet pressure, unit kPa, related coefficient -0.0397；
Gas inlet temperature, unit DEG C, related coefficient 0.0352；
CO volume contents, 0.1278%；
CO2Volume content, 0.1935%；
For related coefficient between 0 to 1, bigger expression correlation is stronger；
Method for normalizing described in step 1 is as follows：
Result is normalized between 0 to 1.
3. according to the method described in claim 1, it is characterized in that, the structure of the Elman neural networks described in step 2 is as follows：
ELman neural networks are constituted by four layers, are input layer respectively, hidden layer, are accepted layer and output layer：
1) by input layer to hidden layer：
X (k)=f (w1xc(k)+w2(u(k-1)))；
2) by hidden layer to output layer：
Y (k)=g (w3x(k))；
3) layer storage information in a short time is accepted：
xc(k)=x (k-1)；
4) learn target function：Using sum of squared errors function, work as E<Terminate to train when 0.0001
5) weight renewing method：
w1To accept layer to the connection weight of hidden layer, w2For the connection weight of input layer to hidden layer, w3For hidden layer to output
The connection weight of layer, for three weights all by the adjustment that iterates, error is that prediction and practical difference, η are represented in training process
Each feedback adjustment degree, is set as 0.001, by constantly feeding back update weights, finally meets the requirements study target function value,
Reach the training completed to neural network
w1(n)=w1(n-1)-η·2·error·w3·xc,
w2(n)=w2(n-1)-η·2·error·w3U,
w3(n)=w3(n-1)-η·2·error·w3·xc。
4. according to the method described in claim 1, it is characterized in that, the sliding window model principle described in step 2 is as follows：
Sliding window model is established on a kind of hypothesis, i.e., current output depends on current input, and input and output it
Between mapping ruler obtained by historical data, according to this it is assumed that presetting a certain amount of training set sample, then constantly
Ground updates sample data and gives up earliest data point, and with the sliding of window, T-S fuzzy neural networks constantly update its structure
Parameter simultaneously provides newest predicted value.
5. according to the method described in claim 1, it is characterized in that, the model described in step 3 is returned suitable for blast-furnace top gas recovery
Receive time-varying, the characteristic of dynamic, non-linear, strong inertia of turbine power generation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810198547.5A CN108446799B (en) | 2018-03-12 | 2018-03-12 | Residual pressure turbine device power generation power prediction method based on Elman neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810198547.5A CN108446799B (en) | 2018-03-12 | 2018-03-12 | Residual pressure turbine device power generation power prediction method based on Elman neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108446799A true CN108446799A (en) | 2018-08-24 |
| CN108446799B CN108446799B (en) | 2021-08-03 |
Family
ID=63194059
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810198547.5A Active CN108446799B (en) | 2018-03-12 | 2018-03-12 | Residual pressure turbine device power generation power prediction method based on Elman neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108446799B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542882A (en) * | 2023-07-06 | 2023-08-04 | 浙江大学 | Photovoltaic power generation smoothing method, system and storage medium |
| CN118070003A (en) * | 2024-04-19 | 2024-05-24 | 中国西安卫星测控中心 | Spacecraft telemetry data interpolation method based on neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102994672A (en) * | 2012-11-30 | 2013-03-27 | 武汉钢铁(集团)公司 | Automatic control method for top pressure of TRT (blast furnace top gas recovery turbine unit) system |
| CN104268638A (en) * | 2014-09-11 | 2015-01-07 | 广州市香港科大霍英东研究院 | Photovoltaic power generation system power predicting method of elman-based neural network |
| CN106709197A (en) * | 2016-12-31 | 2017-05-24 | 浙江大学 | Molten iron silicon content predicting method based on slide window T-S fuzzy neural network model |
-
2018
- 2018-03-12 CN CN201810198547.5A patent/CN108446799B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102994672A (en) * | 2012-11-30 | 2013-03-27 | 武汉钢铁(集团)公司 | Automatic control method for top pressure of TRT (blast furnace top gas recovery turbine unit) system |
| CN104268638A (en) * | 2014-09-11 | 2015-01-07 | 广州市香港科大霍英东研究院 | Photovoltaic power generation system power predicting method of elman-based neural network |
| CN106709197A (en) * | 2016-12-31 | 2017-05-24 | 浙江大学 | Molten iron silicon content predicting method based on slide window T-S fuzzy neural network model |
Non-Patent Citations (4)
| Title |
|---|
| CHUNJIE YANG ETAL.: "Neural Networks Based Predictive Control for TRT", 《2005 INTERNATIONAL CONFERENCE ON NEURAL NETWORKS AND BRAIN》 * |
| JING LIU ETAL.: "Gas/Liquid Two-Phase Flow Regime Recognition by Combining the Features of Shannon"s Entropy with the Improved Elman Network", 《2009 FIFTH INTERNATIONAL CONFERENCE ON NATURAL COMPUTATION》 * |
| 张笑天 等: "《MATLAB 7.x基础教程》", 30 April 2008 * |
| 石磊: "并网光伏电站发电功率预测方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116542882A (en) * | 2023-07-06 | 2023-08-04 | 浙江大学 | Photovoltaic power generation smoothing method, system and storage medium |
| CN116542882B (en) * | 2023-07-06 | 2023-09-19 | 浙江大学 | Photovoltaic power generation smoothing method, system and storage medium |
| CN118070003A (en) * | 2024-04-19 | 2024-05-24 | 中国西安卫星测控中心 | Spacecraft telemetry data interpolation method based on neural network |
| CN118070003B (en) * | 2024-04-19 | 2024-07-23 | 中国西安卫星测控中心 | Spacecraft telemetry data interpolation method based on neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108446799B (en) | 2021-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103730006B (en) | A kind of combination forecasting method of Short-Term Traffic Flow | |
| Cheng et al. | ThermalNet: A deep reinforcement learning-based combustion optimization system for coal-fired boiler | |
| Lindemann et al. | Anomaly detection and prediction in discrete manufacturing based on cooperative LSTM networks | |
| CN103711523B (en) | Based on the gas density real-time predicting method of local decomposition-Evolutionary Neural Network | |
| Liu et al. | Data-driven based model for flow prediction of steam system in steel industry | |
| CN111695666A (en) | Wind power ultra-short term conditional probability prediction method based on deep learning | |
| Kedong et al. | Improved grey prediction model based on exponential grey action quantity | |
| CN108197764A (en) | Predict the method and its equipment of electric power enterprise comprehensive energy consumption | |
| CN106779384A (en) | A kind of long-term interval prediction method of steel and iron industry blast furnace gas based on Information Granularity optimum allocation | |
| Lugaresi et al. | Real-time validation of digital models for manufacturing systems: A novel signal-processing-based approach | |
| CN108446799A (en) | Waste pressure turbine device generated power forecasting method based on Elman neural networks | |
| CN116305985A (en) | Local intelligent ventilation method based on multi-sensor data fusion | |
| CN117235595A (en) | LSTM-Attention-based sewage treatment dosing prediction method and system | |
| CN107706938A (en) | A kind of wind power waving interval analysis method returned based on quantile | |
| CN110033126A (en) | Shot and long term memory network prediction technique based on attention mechanism and logistic regression | |
| Yang et al. | A novel exhaust gas temperature prediction method of hot blast stove | |
| CN109270917B (en) | Intelligent power plant steam turbine bearing-oriented closed-loop control system fault degradation state prediction method | |
| CN109376957A (en) | A kind of prediction technique of thermal power plant's load | |
| Li et al. | An optimal method for prediction and adjustment on gasholder level and self-provided power plant gas supply in steel works | |
| Zheng et al. | Power generation forecast of top gas recovery turbine unit based on Elman model | |
| Koprinkova-Hristova et al. | Recurrent neural networks for predictive maintenance of mill fan systems | |
| Qiang et al. | The combination prediction of BTP in sintering process based on Bayesian framework and LS-SVM | |
| Li et al. | Performance prediction of a production line with variability based on grey model artificial neural network | |
| Jia et al. | Data-Driven Tube-based Model Predictive Control of an Industrial Thickener | |
| Meng et al. | Prediction of Silicon Content of Hot Metal in Blast Furnace Based on Optuna-GBDT |
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 |