CN1763478A - Hot spot detection method for air preheater based on fuzzy kernel function support vector machine - Google Patents
Hot spot detection method for air preheater based on fuzzy kernel function support vector machine Download PDFInfo
- Publication number
- CN1763478A CN1763478A CN 200510096043 CN200510096043A CN1763478A CN 1763478 A CN1763478 A CN 1763478A CN 200510096043 CN200510096043 CN 200510096043 CN 200510096043 A CN200510096043 A CN 200510096043A CN 1763478 A CN1763478 A CN 1763478A
- Authority
- CN
- China
- Prior art keywords
- parameter
- vector machine
- support vector
- kernel function
- function
- 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
Images
Landscapes
- Feedback Control In General (AREA)
Abstract
The invention discloses a hot spot detecting method of power station boiler air preheater based on ambiguity nuclear function support vector machine, which is characterized by the following: finishing three different classification methods of air-conditioner fire condition through the support vector machine based on ambiguity nucleus function to support vector machine; selecting the measurable temperature value as the input variable of classification unit and fire pattern mark as output variable; dividing the data sample into two parts that two thirds as exercise and residual sample as detection; exercising three classification devices separately; estimating the classification device property through detection sample; computing the classification device ROC curve to compare the classification device property; adapting the adaptive parameter optimization method to select regularized parameter and nuclear parameter of support vector machine.
Description
Technical field
The invention belongs to industrial Automatic Measurement Technique field, relate to a kind of method of industrial process detection, especially relate to a kind of air pre-heater hot spot detection method based on fuzzy kernel function support vector machine.
Background technology
Station boiler air preheater is to utilize the smoke discharging residual heat of boiler to add the heat exchanger of hot-air.The excess of fuel that accumulates on the air preheater heating surface (comprising carbon black and oil droplet) heats up through oxidation, reaches the pyrophoricity accident that ignition temperature will cause air preheater itself.The air preheater burning has two conditions: 1. accumulating on the heating surface has fuel; Air in the air preheater, flue gas flow rate is low or inhomogeneous, causes the radiating condition variation.When the oil gas of burning or partially combusted fuel, particularly atomization badness not condenses upon on the heat transfer component of air preheater, when temperature was elevated to 350 ℃ of left and right sides, sediment can be dried and be lighted.The initial stage scope is less owing to catch fire, and is difficult to be found, and when component temperature continues to rise to 700 ℃, just is enough to cause steel heat accumulating element and whole air preheater on fire, thereby influences the normal operation of whole unit.This burning is called as the burning again (claiming secondary combustion again) of air preheater.Because burning again from taking place in air preheater, to causing fire is a slow evolution, and the oxidation reaction that it will experience carbon black and oil droplet is to spontaneous combustion, spreads to the adjacent corrugated plate case from the burning of a corrugated boxes, and constantly enlarging such process, this process often needs several hours.This shows, as long as can make accurate judgement in early days what burning again took place air preheater, and in time adopt an effective measure, the air preheater fire damage that causes of burning again can be controlled in the not half, therefore be necessary to adopt the sniffer that catches fire, take effective fire prevention measure ahead of time.Successively having developed the air preheater hot spot detection system at present both at home and abroad also comes into operation.These detection systems all adopt infrared sensor or thermopair as temperature element, the temperature value that measures is compared with the alarm threshold value of setting in advance, thereby judged whether that the condition of a fire takes place.But the experience that depends on and field working conditions that the setting of alarm threshold value is too much are failed to report and are reported by mistake easily.Also have simultaneously document with Application of Neural Network in the detection of fire signal, and obtained satisfied result.But in actual applications, the structure of neural network is too complicated, and the parameter that needs to estimate seems too much that with respect to less data sample cause resulting neural network model that data are crossed study, generalization ability is not enough, and detection accuracy is not high.The structure of neural network is difficult to select also to have limited its application in addition.
Statistical Learning Theory is (Statistical learning theory, SLT) a kind of theory of specializing in machine learning rule under the small sample situation of setting up by Vapnik, (Support vectormachines SVM) is a kind of new classification that grows up and the instrument of recurrence to support vector machine on this theoretical foundation.Support vector machine improves generalization ability by the structural risk minimization principle, preferably resolves practical problemss such as small sample, non-linear, high dimension, local minimum point.Document " based on the station boiler air preheater hot spot detection system research of least square method supporting vector machine " (Liu Han, Li Qi, Liu Ding, Liang Yanming, Song Nianlong: " based on the station boiler air preheater hot spot detection system research of least square method supporting vector machine ", Proceedings of the CSEE, 2005,25 (3); 147-152), using least square method supporting vector machine (LS-SVM) studies the air preheater hot spot detection system, but the precision of the least square method supporting vector machine that this method is used classification is relatively low, execution speed is slow, and the method for crosscheck, the more experience that depends on are adopted in the selection of parameter.
Summary of the invention
Defective or deficiency at above-mentioned existing air preheater hot spot detection technique existence, the purpose of this invention is to provide a kind of method that detects based on the air preheater hot spot of fuzzy kernel function support vector machine, this method has provided the method that the support vector machine classifier optimized parameter is adjusted; Adopt fuzzy kernel function to replace the kernel function of standard, simplified calculating, improved the efficient that engineering is used, for the application of support vector machine in the focus detection technique provides reliable basis and foundation.
To achieve these goals, the present invention takes following technical scheme:
A kind of air pre-heater hot spot detection method based on fuzzy kernel function support vector machine, it is characterized in that, this method is with 5 temperature focus values can surveying in air preheater input variable as sorter, the mark of condition of a fire type is as output variable, and input variable carried out normalized, input is finished by support vector machine with the mapping relations of output, and construct three two-value sorters, data sample is divided into two parts, 2/3 sample is as training, and remaining sample is used for testing, and trains three sorters respectively, and assessed the performance of sorter originally with test specimens, calculate the performance of the ROC curve of sorter with comparator-sorter.Using support vector machine to carry out the branch time-like, adopting the method for auto-adaptive parameter optimization to choose the regularization parameter and the nuclear parameter of support vector machine, and make the kernel function obfuscation.
Method of the present invention has been avoided calculated amount too big shortcoming in experience that depends on the deviser too much in the empirical method and the cross validation method.Parameter determination method of the present invention has strict theoretical foundation, and succinct, convenient, easy-to-use, is very suitable for practical application.Simultaneously, the present invention can significantly reduce calculated amount with the kernel function obfuscation when making practical application, is convenient to engineering and uses.
Description of drawings
Synoptic diagram expressed in the fuzzy language of Fig. 1 sigmoid kernel function;
The membership function of the fuzzy sigmoid kernel function of Fig. 2.
The present invention is described in further detail below in conjunction with drawings and Examples.
Embodiment
Present air preheater hot spot detecting method is existing some problems in varying degrees, is difficult to look after various aspects.The present invention has provided a kind of air preheater hot spot detecting method based on fuzzy kernel function support vector machine, can effectively overcome the defective that detection method had based on neural network of present more use.At the selection difficult problem of regularization parameter in the support vector machine and nuclear parameter, the present invention has directly provided concrete parameter determination method.
This method is with 5 temperature focus values can surveying in air preheater input variable as sorter, and the mark of condition of a fire type is as output variable, and input is finished by support vector machine with the mapping relations of output, comprises following content:
1) based on the condition of a fire sorter of fuzzy kernel function support vector machine
The model that belongs to black box one class based on the air preheater condition of a fire sorter of support vector machine, input is finished by support vector machine with the mapping relations of output, 5 temperature values can surveying in the air preheater are as the input variable of sorter, and the mark of condition of a fire type is as output variable.Condition of a fire situation in the air preheater is divided into three grades: " 1 " level takes place fire should be arranged; The corresponding fire alarm of " 2 " level, expression might will breaking out of fire; The corresponding no condition of a fire of " 3 " level is safest state.In the air preheater hot spot detection technique, need make correct classification to the condition of a fire of three kinds of different stages, in order to finish so many-valued classification problem, construct three two-value sorters, difference called after " 1 grade and 2 grades ", " 1 grade and 3 grades ", " 2 grades and 3 grades ", with known data sample these three sorters are trained up, these three sorters are made of the support vector machine based on fuzzy kernel function.The sorter that trains is as the arbiter of the condition of a fire, and system gathers 5 temperature points of air preheater in real time, and is classified in real time by sorter, thereby finish different condition of a fire situations is carried out accurate classification.
2) system of selection of sorter optimized parameter
Divide time-like carrying out support vector machine, two class parameters are arranged in the sorter: regularization parameter C and nuclear parameter α and β need adjust, the method of adjusting is as follows: these three parameters are carried out five equilibrium respectively in its span, and will wait the value behind the branch to make up three-dimensional data space, the parameter value in space of fetching data respectively carries out the training of support vector machine, chooses the parameter that the parameter of minimum study error is adjusted the most.Be implemented as follows:
A. at first determine the span of regularization parameter C and nuclear parameter α, β.Based on the object that the principle and the present technique of support vector machine are studied, the span of choosing C, α and β is respectively [0.01,100], [0.01,100] and [0.01,100];
B. make up parameter to { C
i, α
j, β
k: i=1 ..., m; J=1 ..., n; K=1 ..., h} is about to span m, n and k equal portions respectively, then constitutes the three-dimensional parameter space { C of m * n * h
i, α
j, β
k.
C. the parameter that generates is learnt being applied to support vector machine, and calculated the study error.Get minimum study error corresponding parameters to { C
i, α
j, β
k}
EminBe optimized parameter.
If d. error precision can not reach requirement, then with { C
i, α
j, β
k}
EminBe the center, dwindle parameter range, repeating step c constantly optimizes the parameter of support vector machine, until reaching desired error precision.
3) the fuzzy sigmoid kernel function of structure
Kernel function at present commonly used is nonlinear function, and these kernel functions can be brought very big computation burden in the training of support vector machine with when calculating.By utilizing fuzzy logic technology, with the kernel function obfuscation, to simplify calculating, thereby improved the efficient when engineering is used greatly, its principle is as follows:
The kernel function of selecting for use in support vector machine is the sigmoid kernel function, as the formula (1), calculates its obfuscation for simplifying:
K(x
i，x
j)＝tanh(α·x
i·x
j+β) (1)
In the formula: α is a constant, the smooth degree of expression sigmoid function, x
iX
jThe inner product of expression vector, α and β are the parameter of kernel function.
Sigmoid kernel function based on fuzzy logic is come out (1) exactly with the formal description of fuzzy language, shown in (2).Wherein α is a constant, has represented the smooth degree of sigmoid function, sees shown in the accompanying drawing 1.
According to (2) formula, sigmoid endorses to be defined as a series of membership functions.For the sake of simplicity, adopt leg-of-mutton membership function, as shown in Figure 2.Fuzzy subregion is respectively by low, medium, and high forms, and can certainly select the low by very, low, medium, high, very high constitutes, but can add the burden of computation like this.Because the sigmoid kernel function must be continuous, the value of therefore calculating the upper and lower boundary of triangular function is respectively (β ± 1)/α.Finally (2) can be written as the function of following α and β:
Wherein
M (x
i, x
j, α, β)=2 (x
iX
j+ beta/alpha)-α
2(x
iX
j+ beta/alpha) | x
iX
j+ beta/alpha | (4) are that the sigmoid kernel function can be described by (5):
K(x
i，x
j)＝(-1)μ
1(x
i·x
j)+ax
i·x
jμ
2(x
i·x
j)+(+1)μ
3(x
i·x
j) (5)
Sigmoid kernel function after the obfuscation has following advantage:
1. every bit all can be little in whole field of definition for function;
2. accompanying drawing 2 defined membership functions can simply be realized by hardware such as microprocessor, digital signal processors;
3. by the method for this obfuscation, select the membership function of varying number, differing complexity just can be similar to the kernel function of different nonlinear degrees.
A concrete example below is provided, so that technical scheme of the present invention is done further to understand:
Example: the experimental data of the 300MW unit of the 200MW unit of Shandong Province A power plant, the B of Guizhou Province power plant and the 300MW unit of the C of Henan Province power plant.The concrete operations step is:
Step 1: gather the input variable of 5 temperature points as requested as sorter;
Step 2: for improving counting yield, input variable is carried out normalized,
Property value x after the normalization
i∈ [0,1];
Step 3: select the kernel function of fuzzy sigmoid kernel function for use as support vector machine;
Step 4: utilize the self-adaptation setting method of regularization parameter and nuclear parameter to adjust;
Step 5: construct three two-value sorters: difference called after " 1 grade and 2 grades ", " 1 grade and 3 grades ", " 2 grades and 3 grades ", use the data that obtained respectively three sorters to be trained, wherein 2/3 sample is as training, and remaining sample is used for testing;
Step 6: finish the detection of air preheater hot spot and the judgement of the condition of a fire with setting up good sorter.
Three groups of samples and number that table 1 is gathered for the training support vector machine classifier, as shown in table 1.
Other training sample logarithm of the different condition of a fire levels of table 1
| | 1 | 2 grades | 3 grades | Amount to |
| A power plant | 13 | 14 | 12 | 39 |
| B power plant | 23 | 8 | 11 | 42 |
| C power plant | 17 | 24 | 8 | 49 |
| Amount to | 53 | 46 | 31 | 130 |
The training and testing result of the fuzzy sigmoid kernel function svm classifier device of table 2
| 2 grades of 1 grade of vs | 3 grades of 1 grade of vs | 3 grades of 2 grades of vs | |
| Number of training | 66 | 56 | 52 |
| The test specimens given figure | 33 | 28 | 25 |
| C | 31 | 27.1 | 14.9 |
| α | 1.4273 | 2.6961 | 0.9721 |
| β | 0.3709 | 2.1346 | 0.8790 |
| The training mistake is divided the sample number average | 5.5609 | 3.0103 | 4.1279 |
| The training mistake is divided the sample number standard deviation | 0.9915 | 1.0056 | 1.4167 |
| The accuracy rate of training | 91.78％ | 94.89％ | 90.79％ |
| The test mistake is divided the sample number average | 3.5671 | 2.3367 | 3.8957 |
| The test mistake is divided the sample number standard deviation | 1.6980 | 1.2756 | 0.7481 |
| Accuracy rating of tests | 89.98％ | 92.34％ | 84.78％ |
Regularization parameter C and adjusting of nuclear parameter α and β after the process parameter adaptive is adjusted the results are shown in Table 2,130 groups of samples altogether of being gathered have been carried out training and testing to three sorters respectively, result from test, the accuracy rate of this method is very high, can be applied to engineering reality fully.The result of the detection method of the present invention and the described least square method supporting vector machine of background technology is compared, adopted the data source identical with method described in the background technology.Table 3 is ROC (the Receiver Operating Characteristic) characteristic of two kinds of sorting techniques and the comparative result of operation efficiency, as can be seen, the AUC of the ROC characteristic of sorter of the present invention (Area under curve) value is wanted obviously greater than the sorter described in the background technology, show that sorter of the present invention is more accurate and efficient, and the computing execution speed is greatly improved.
The performance of two kinds of different sorters of table 3 relatively
| Sorter | Object of classification | AUC | Differentiate accuracy rate % | CPU average execution time (second) |
| | 2 grades of 1 grade of vs | 0.8507 | 89.71 | 2.1327 |
| 3 grades of 1 grade of vs | 0.8613 | 91.73 | 2.0103 | |
| 3 grades of 2 grades of vs | 0.7954 | 84.01 | 2.3458 | |
| The | 2 grades of 1 grade of vs | 0.8760 | 89.98 | 1.3829 |
| 3 grades of 1 grade of vs | 0.8937 | 92.34 | 1.4237 | |
| 3 grades of 2 grades of vs | 0.8263 | 84.78 | 1.4521 |
Claims (4)
1. air pre-heater hot spot detection method based on fuzzy kernel function support vector machine, it is characterized in that, this method is with 5 temperature focus values can surveying in air preheater input variable as sorter, the mark of condition of a fire type is as output variable, and input variable carried out normalized, input is finished by support vector machine with the mapping relations of output, and construct three two-value sorters, by finishing the different condition of a fire situation of sorter is carried out accurate classification based on fuzzy kernel function support vector machine, data sample is divided into two parts, 2/3 sample is as training, remaining sample is used for testing, train three sorters respectively, and assessed the performance of sorter originally with test specimens, and calculate the performance of the ROC curve of sorter with comparator-sorter, using support vector machine to carry out the branch time-like, adopt the method for auto-adaptive parameter optimization to choose the regularization parameter and the nuclear parameter of support vector machine, and make the kernel function obfuscation.
2. the method for claim 1 is characterized in that, this method specifically comprises following content:
1.1 condition of a fire sorter based on fuzzy kernel function support vector machine
Condition of a fire situation in the air preheater is divided into " 1 ", " 2 ", " 3 " level, and wherein the expression of " 1 " level has fire to take place; " 2 " level expression fire alarm promptly might breaking out of fire; The no condition of a fire of " 3 " level expression is safe condition;
According to above-mentioned classification situation, construct three two-value sorters, called after " 1 grade and 2 grades ", " 1 grade and 3 grades ", " 2 grades and 3 grades " carry out accurate classification by finishing based on fuzzy kernel function support vector machine to the different condition of a fire situation of sorter respectively;
1.2 the selection of sorter optimized parameter
Divide time-like carrying out support vector machine, adopt the method for auto-adaptive parameter optimization to choose the regularization parameter and the nuclear parameter of support vector machine, regularization parameter C and nuclear parameter α and β all need to adjust;
1.3 structure fuzzy kernel function
Selecting kernel function in support vector machine for use is the sigmoid kernel function, shown in (1) formula:
K(x
i，x
j)＝tanh(α·x
i·x
j+β) (1)
Calculate its obfuscation for simplifying:
Sigmoid kernel function based on fuzzy logic is exactly that standard sigmoid kernel function is come out with the formal description of fuzzy language, and shown in (2) formula, wherein a is a constant, has represented the smooth degree of sigmoid function;
According to (2) formula, sigmoid nuclear is defined as a series of membership functions, for the sake of simplicity, adopt leg-of-mutton membership function, fuzzy subregion is respectively by low, medium, high forms, because the sigmoid kernel function must be continuous, the value of therefore calculating the upper and lower boundary of triangular function is respectively (β ± 1)/α, and final (2) formula is written as the function of following α and β:
In the formula: a is a constant, the smooth degree of expression sigmoid function, x
iX
jThe inner product of expression vector, α and β are the parameter of kernel function.
Wherein
M (x
i, x
j, α, β)=2 (x
iX
j+ beta/alpha)-α
2(x
iX
j+ beta/alpha) | x
iX
j+ beta/alpha | (4) are that the sigmoid kernel function is described by following formula:
K(x
i，x
j)＝(-1)μ
1(x
i·x
j)+ax
i·x
jμ
2(x
i·x
j)+(+1)μ
3(x
i·x
j) (5)
Every bit all can be little in whole field of definition for function; Defined membership function is simply realized by microprocessor, digital signal processor hardware;
By the method for this obfuscation, select the membership function of varying number, differing complexity just can be similar to the kernel function of different nonlinear degrees.
3. the method for claim 1 is characterized in that, described input variable is carried out normalized and satisfied following formula:
Property value x after the normalization
i∈ [0,1].
4. method as claimed in claim 2 is characterized in that, described regularization parameter C and nuclear parameter α and β setting method are according to the following step:
1) span of at first definite regularization parameter C and nuclear parameter α, β, the span of choosing C, α and β is respectively [0.01,100], [0.01,100] and [0.01,100];
2) make up parameter to { C
i, α
j, β
k: i=1 ..., m; J=1 ..., n; K=1 ..., h} is about to span m, n and h equal portions respectively, then constitutes the three-dimensional parameter space { C of m * n * h
i, α
j, β
k;
3) with the parameter value of the data space that generates to being applied to support vector machine study, and calculate the study error, get minimum study error corresponding parameters to { C
i, α
j, β
k}
EminBe optimized parameter;
4) if error precision can not reach requirement, then with { C
i, α
j, β
k}
EminBe the center, dwindle parameter range, and repeating step 3., continue to optimize the parameter of support vector machine, until reaching desired error precision.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100960435A CN100360901C (en) | 2005-09-19 | 2005-09-19 | Hot spot detection method for air preheater based on fuzzy kernel function support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005100960435A CN100360901C (en) | 2005-09-19 | 2005-09-19 | Hot spot detection method for air preheater based on fuzzy kernel function support vector machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1763478A true CN1763478A (en) | 2006-04-26 |
| CN100360901C CN100360901C (en) | 2008-01-09 |
Family
ID=36747732
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2005100960435A Expired - Fee Related CN100360901C (en) | 2005-09-19 | 2005-09-19 | Hot spot detection method for air preheater based on fuzzy kernel function support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100360901C (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464964B (en) * | 2007-12-18 | 2011-04-06 | 同济大学 | Pattern recognition method capable of holding vectorial machine for equipment fault diagnosis |
| CN101540008B (en) * | 2009-04-24 | 2011-04-06 | 北京工业大学 | Analogy method based on activated sludge purification process of HPP cellular automaton model |
| CN103472867A (en) * | 2013-09-22 | 2013-12-25 | 浙江大学 | Pesticide production waste liquid incinerator temperature optimization system and method of support vector machine |
| CN108038510A (en) * | 2017-12-22 | 2018-05-15 | 湖南源信光电科技股份有限公司 | A kind of detection method based on doubtful flame region feature |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100708337B1 (en) * | 2003-06-27 | 2007-04-17 | 주식회사 케이티 | Apparatus and method for automatic video summarization using fuzzy one-class support vector machines |
| CN1327376C (en) * | 2004-04-08 | 2007-07-18 | 上海交通大学 | Soft measuring meter moduling method based on supporting vector machine |
| CN1585458A (en) * | 2004-05-27 | 2005-02-23 | 上海交通大学 | Method for positioning and extracting video frequency caption by supporting vector computer |
-
2005
- 2005-09-19 CN CNB2005100960435A patent/CN100360901C/en not_active Expired - Fee Related
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464964B (en) * | 2007-12-18 | 2011-04-06 | 同济大学 | Pattern recognition method capable of holding vectorial machine for equipment fault diagnosis |
| CN101540008B (en) * | 2009-04-24 | 2011-04-06 | 北京工业大学 | Analogy method based on activated sludge purification process of HPP cellular automaton model |
| CN103472867A (en) * | 2013-09-22 | 2013-12-25 | 浙江大学 | Pesticide production waste liquid incinerator temperature optimization system and method of support vector machine |
| CN103472867B (en) * | 2013-09-22 | 2015-11-04 | 浙江大学 | The optimizing temperature of pesticide production waste liquid incinerator system and method for support vector machine |
| CN108038510A (en) * | 2017-12-22 | 2018-05-15 | 湖南源信光电科技股份有限公司 | A kind of detection method based on doubtful flame region feature |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100360901C (en) | 2008-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | A novel optimal support vector machine ensemble model for NOX emissions prediction of a diesel engine | |
| CN110173710B (en) | Energy-saving type climate feedback dynamic regulation and control device and method for eliminating smoke plume | |
| CN111144609A (en) | Boiler exhaust emission prediction model establishing method, prediction method and device | |
| AU2019305721B2 (en) | Boiler coal saving control method | |
| CN110288048B (en) | Submarine pipeline risk assessment method of SVM directed acyclic graph | |
| WO2021159585A1 (en) | Dioxin emission concentration prediction method | |
| CN109508818B (en) | Online NOx prediction method based on LSSVM | |
| CN109670629A (en) | Coal-burning boiler thermal efficiency forecast method based on shot and long term Memory Neural Networks | |
| CN111754093A (en) | Fly ash carbon content prediction method based on coal quality prediction and PSO-SVM | |
| CN109376858B (en) | Method for predicting performance of condensing heat exchanger based on partial load rate | |
| CN113469230B (en) | Rotor system deep migration fault diagnosis method, system and medium | |
| CN1763478A (en) | Hot spot detection method for air preheater based on fuzzy kernel function support vector machine | |
| CN105468932A (en) | Heating efficiency online calculation method for boiler | |
| CN103676822A (en) | Control device of thermal power plant and control method thereof | |
| CN105116730A (en) | Hydrogen fuel engine ignition advance angle optimization system based on particle swarm fuzzy neural network, and optimization method thereof | |
| CN118009990B (en) | High-precision real-time forecasting method for tide level based on transducer model | |
| CN115510904A (en) | Boiler heating surface ash deposition monitoring method based on time sequence prediction | |
| Khilar et al. | Improving the efficiency of photovoltaic panels using machine learning approach | |
| CN115146537A (en) | Atmospheric pollutant emission estimation model construction method and system based on power consumption | |
| CN101825869A (en) | Method for identifying superheater model parameters based on data drive | |
| CN107742000A (en) | Boiler combustion oxygen content modeling method | |
| CN115437245A (en) | Boiler optimization control method based on combustion state prediction model | |
| CN103793746A (en) | Method for identifying parameters of coal-fired power plant boiler superheater model | |
| Kocer | Prediction of the higher heating values of biomass using machine learning methods based on proximate and ultimate analysis | |
| CN114611398B (en) | Soft measurement method for nitrogen oxides in urban solid waste incineration process based on brain-like modularized neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20080109 Termination date: 20100919 |