CN107356710A - A kind of waste incineration dioxin in flue gas class concentration prediction method and system - Google Patents

A kind of waste incineration dioxin in flue gas class concentration prediction method and system Download PDF

Info

Publication number
CN107356710A
CN107356710A CN201710544474.6A CN201710544474A CN107356710A CN 107356710 A CN107356710 A CN 107356710A CN 201710544474 A CN201710544474 A CN 201710544474A CN 107356710 A CN107356710 A CN 107356710A
Authority
CN
China
Prior art keywords
concentration
flue gas
dioxin
waste incineration
characteristic vector
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201710544474.6A
Other languages
Chinese (zh)
Inventor
卢加伟
肖晓东
海景
朱锋
雷鸣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China Institute of Environmental Science of Ministry of Ecology and Environment
Original Assignee
South China Institute of Environmental Science of Ministry of Ecology and Environment
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by South China Institute of Environmental Science of Ministry of Ecology and Environment filed Critical South China Institute of Environmental Science of Ministry of Ecology and Environment
Priority to CN201710544474.6A priority Critical patent/CN107356710A/en
Publication of CN107356710A publication Critical patent/CN107356710A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
    • G01N33/0004—Gaseous mixtures, e.g. polluted air
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines

Abstract

The invention discloses a kind of waste incineration dioxin in flue gas class concentration prediction method and system.Methods described includes：(1) duty parameter and Conventional pollution concentration during acquisition waste incinerator stable operation, as characteristic vector；(2) characteristic vector is inputted into supporting vector machine model, carries out regression forecasting；(3) according to Support vector regression prediction result, waste incineration dioxin in flue gas class concentration is demarcated.The system includes：Characteristic vector acquisition device, regression forecasting device and concentration calibration device.Method real-time provided by the invention is good, cost is low；System provided by the invention, without increasing detection device, realize low cost monitoring waste incineration dioxin in flue gas class concentration in real time.

Description

A kind of waste incineration dioxin in flue gas class concentration prediction method and system
Technical field
It is dense more particularly, to a kind of waste incineration dioxin in flue gas class the invention belongs to technical field of waste treatment Spend Forecasting Methodology and system.
Background technology
In order to tackle the refuse pollution problem being on the rise, alleviate Pressure on Energy, waste incineration and generating electricity conduct can be realized Garbage harmless, minimizing, the processing mode of recycling are by global common concern and application.Each incineration plant prevention and cure of pollution Horizontal uneven, neighbour caused by pollutant discharge beyond standards keeps away particularly thorny, needs badly so that monitoring in real time is means for a long time, to burning Factory's prevention and cure of pollution ability carries out dynamic evaluation prediction and Classification Management.
Dioxin content caused by waste incineration is low but harm is big, is the main inducing that incineration plant neighbour keeps away contradiction, It is the difficult point of environmental monitoring and management.Currently, waste incineration dioxin be difficult to as duty parameter as Conventional pollution it is real Existing on-line continuous monitoring, standard method are tested and analyzed using high resolution spectral estimation combination method, cost High, cycle length, poor real.
The content of the invention
For the disadvantages described above or Improvement requirement of prior art, the invention provides a kind of waste incineration dioxin in flue gas Class concentration prediction method and system, its object is to the duty parameter easily obtained by monitoring in real time and Conventional pollution concentration, Predict, real-time calibration rubbish dioxin in incineration smoke class concentration thus solve existing rubbish and burn using Support vector regression Dioxin in flue gas class concentration monitor cost is high, the cycle is long, the technical problem of poor real.
To achieve the above object, according to one aspect of the present invention, there is provided a kind of waste incineration dioxin in flue gas class Concentration prediction method, comprises the following steps：
(1) duty parameter and Conventional pollution concentration during acquisition waste incinerator stable operation, as characteristic vector；
(2) characteristic vector obtained in step (1) is inputted into supporting vector machine model, carries out regression forecasting；
(3) it is dense according to Support vector regression prediction result in step (2), demarcation waste incineration dioxin in flue gas class Degree.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its described duty parameter include：Burner hearth Interior incineration temperature；The Conventional pollution concentration includes：HCl concentration.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its described duty parameter also include：Pot Outlet of still cigarette temperature and flue gas flow；The Conventional pollution concentration also includes： SO2Concentration and particle concentration.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its step (2) described SVMs Model kernel function is：D rank multinomials kernel function, Radial basis kernel function, sigmoid kernel functions.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its step (2) described SVMs Model uses ε insensitive loss functions support vector regression (ε-SVR) instrument.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its insensitive damage of step (2) ε The insensitive loss coefficient ε values of function support vector regression (ε-SVR) instrument of mistake are between 0.01~0.1.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its insensitive damage of step (2) ε The insensitive loss coefficient ε values of function support vector regression (ε-SVR) instrument of mistake are 0.04.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its step (2) include：
The characteristic vector obtained in step (1) is standardized pretreatment so that characteristic vector value after processing [0, 1] between.Specifically, standardized value can calculate according to equation below：
Wherein, xsFor the standardized value of the characteristic vector value, xiFor the value of the characteristic vector, xminFor the feature to Measure the minimum value in training sample, xmaxFor maximum of the characteristic vector in training sample.
Preferably, the waste incineration dioxin in flue gas class concentration prediction method, its step (3) described dioxin are dense Spend Wei dioxins total concentration and/or total toxic equivalent.
It is another aspect of this invention to provide that providing a kind of waste incineration dioxin in flue gas class concentration prediction system, wrap Include：
Characteristic vector acquisition device, the duty parameter and Conventional pollution during for obtaining waste incinerator stable operation are dense Degree, as characteristic vector, submits to regression forecasting device；
The regression forecasting device, is stored with housebroken supporting vector machine model, is filled for being obtained according to characteristic vector The characteristic vector of submission is put, calculates regression forecasting result, and submit to concentration calibration device；
The supporting vector machine model is ε insensitive loss functions support vector regression (ε-SVR)；ε value is 0.01 Between~0.1, preferred ε=0.04；Its kernel function is d rank multinomials kernel function, Radial basis kernel function or sigmoid kernel functions；
The concentration calibration device, for the regression forecasting result submitted according to regression forecasting device, demarcate waste incineration Dioxin in flue gas class concentration；The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.
In general, by the contemplated above technical scheme of the present invention compared with prior art, it can obtain down and show Beneficial effect：
Waste incineration dioxin in flue gas class concentration prediction method and system provided by the invention are by selecting supporting vector Machine characterized by duty parameter and Conventional pollution concentration, carries out machine learning regression forecasting, so as in real time as mathematical modeling Detect waste incineration dioxin in flue gas class concentration.Method real-time provided by the invention is good, and cost is low.System provided by the invention System, without increasing detection device, you can realize low cost monitoring waste incineration dioxin in flue gas class concentration in real time.For rubbish Processing, there is important application value and researching value.
Brief description of the drawings
Fig. 1 is waste incineration dioxin in flue gas class concentration prediction method prediction result contrast provided in an embodiment of the present invention Figure；Wherein Fig. 1 a Wei dioxin total concentrations prediction result contrasts；The total toxic equivalent prediction result contrast of Fig. 1 b Wei dioxins；
Fig. 2 is that the embodiment of the present invention 1 to 3 predicts absolute percent error map；Fig. 2 a Wei dioxin total concentrations are pre- Survey result absolute percent error map；The total toxic equivalent of Fig. 2 b Wei dioxins surveys the distribution of result absolute percent error Figure；
Fig. 3 is the characteristic vector measure of merit result figure that the present invention uses.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and It is not used in the restriction present invention.As long as in addition, technical characteristic involved in each embodiment of invention described below Conflict can is not formed each other to be mutually combined.
Waste incineration dioxin in flue gas class concentration prediction method provided by the invention, comprises the following steps：
(1) obtain waste incinerator stable operation when to duty parameter and Conventional pollution concentration, as characteristic vector；
The duty parameter, including：Incineration temperature in burner hearth, preferably also includes：Boiler export cigarette temperature and flue gas stream Amount；The Conventional pollution concentration, including：HCl concentration, preferably also includes：SO2Concentration and particle concentration.
For duty parameter, on the one hand, incineration temperature and oxygen concentration influence dioxin production in the burner hearth of burning zone It is raw, such as in 650~900 DEG C of sections of after-burning section, the goodness of fit R of temperature Yu dioxin content linear regressions2Up to 0.80；Separately On the one hand, the stable incineration plant of Incineration circumstances discharges dioxins essentially from low temperature heterogeneous catalytic reaction, with boiler export Cigarette kelvin relation is close, linear regression fit goodness R2Up to 0.83, also influenceed by deduster temperature, when deduster temperature is 180~ At 270 DEG C, linear regression fit goodness R2Up to 0.85 , dioxins can carry out de novo formation between 200~400 DEG C (de novo) reacts, therefore shortens residence time of the flue gas between this temperature range, carries out flue gas chilling, can effectively reduce Dioxin generates.
Because Conventional pollution can realize on-line continuous monitoring, it is more desirable to realize bioxin using Conventional pollution data The real-time monitoring of class.Research finds dioxins content and dust concentration, HCl concentration and SO in flue gas2Concentration exist compared with Strong correlation.But compared with duty parameter or chloride precursor, this correlation is on the weak side.
Input variable can excessively cause model excessively complicated, especially into irrelevant variable or multicollinear variable Cause the reduction of model generalization ability.According to document to flue gas dioxin class and the correlation of Incineration circumstances and Conventional pollution point Analysis, the present invention choose incineration temperature, boiler export cigarette temperature, flue gas flow, SO in burner hearth2Concentration, HCl concentration and particle concentration Deng 6 variables as characteristic vector.
It is preferred that obtain as follows：
The waste incinerator being equipped with for municipal solid waste incinerator, fume sample collection is carried out, sampled point is respectively provided at de- After acid tower, before activated carbon injection and sack cleaner, and after sack cleaner, each point gathers 2~3 times.In order to ensure number According to matching, every CIU before and after gas cleaning simultaneously carry out fume sample collection.Adopted under different working conditions Collect fume sample, each sampling process continues 4 hours, respectively stable conditions time 2 h, flue gas dioxin class sample collection 2 hours.During sample collection, using modes such as manual monitoring, portable instrument monitoring or on-line monitoring equipment monitorings, to work (incineration temperature, boiler export cigarette temperature, cloth bag export cigarette temperature, flue gas flow, O to condition parameter in burner hearth2With H2O content) it is dirty with routine Contaminate thing concentration (CO2、SO2, NOx, HCl, CO, particulate matter, HF and NH3) monitoring is synchronized, collect within every 10 minutes 1 monitoring number According to totally 12 times, end product takes its average value.
(2) characteristic vector obtained in step (1) is inputted into supporting vector machine model, carries out regression forecasting；
The supporting vector machine model, for training sample set { xi,yi},xi∈Rn,yi∈ R (i=1,2 ..., l), obtain Regression function f (x)=wx+b parameter w, b；Optimization aim is：
Wherein,For slack variable, C is penalty factor, and Controlling model exceeds the punishment degree of error, C for sample The bigger fitting degree to training sample is higher.
It is primal-dual optimization problem that it is equations turned, which to introduce Lagrange, as follows：
Wherein,For Lagrange multipliers, and
Using the necessary and sufficient condition (KKT conditions) of optimization, goal regression estimation function can be obtained：
Due to pertaining only to the inner product operation between training sample, the complexity of calculating depends on sample number and is not dependent on The dimension in space, so as to effectively handle higher-dimension problem.For non-linear training sample, the thought of SVMs is by non- Linear Mapping Φ, training sample input data x is mapped in higher dimensional space, so as to which nonlinear function estimation problem be converted into Linear function estimation problem in higher dimensional space, it is specific as follows：
F (x)=w Φ (x)+b
With above similar, the goal regression estimation function under the conditions of this can be obtained：
Wherein, K (xiX)=Φ (xi)·Φ(xj) be SVMs kernel function, be that SVMs overcomes higher-dimension Number calculates and solved nonlinear key factor.Conventional kernel function has d rank multinomials kernel function, Radial basis kernel function (Radial Basis Function, RBF), sigmoid kernel functions, functional form difference are as follows：
D rank multinomial kernel functions：K(x,xi)=[γ (xxi)+1]d
RBF kernel functions：K(x,xi)=exp (- γ | | x-xi||2)
Sigmoid kernel functions：K(x,xi)=tanh [γ (xxi)+c]
The parameter of Polynomial kernel function is relatively more, and when its exponent number d is higher, computation complexity can greatly increase； RBF kernel functional parameters are few, have preferable performance for large sample and small sample, therefore most widely used；Using sigmoid core letters Several SVMs is accomplished that the multilayer perceptron comprising a hidden layer and in the absence of the local minimum point in neutral net Problem, but it only meets that some specified conditions could meet that (any positive semi-definite function can serve as Mercer conditions Kernel function), thus it is somewhat limited in the application.
It is preferred that ε insensitive loss functions support vector regression (ε-SVR) instrument of using.The selection pair of SVMs parameter Prediction result has a major impact, for ε insensitive loss functions support vector regression (ε-SVR) instrument, it is thus necessary to determine that parameter There is the parameter γ of insensitive loss coefficient ε, penalty factor and kernel function, the selection of these parameters controls the study of model Ability and generalization ability are, it is necessary to consider influencing each other for 3 factors, it is determined that stable, quick parameter.
ε of the present invention value is between 0.01~0.1, preferred ε=0.04, and it is best now to model effect.Utilize K- simultaneously Folding cross validation combination grid data service determines the parameter γ and penalty factor of kernel function.Because sample number is less, use 10- rolls over and 5- folding cross validations, and parameter γ and C hunting zone are all [2-10, 210], step-size in search is 0.4 (i.e. 2-10, 2-9.6, 2-9.2...), parameter search time and fitting performance are all relatively good.
It is preferred that the characteristic vector obtained in step (1) is standardized pretreatment so that the characteristic vector value after processing Between [0,1].Specifically, standardized value can calculate according to equation below：
Wherein, xsFor the standardized value of the characteristic vector value, xiFor the value of the characteristic vector, xminFor the feature to Measure the minimum value in training sample, xmaxFor maximum of the characteristic vector in training sample.
(3) it is dense according to Support vector regression prediction result in step (2), demarcation waste incineration dioxin in flue gas class Degree.
The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.
Waste incineration dioxin in flue gas class concentration prediction system provided by the invention, including：
Characteristic vector acquisition device, during for obtaining waste incinerator stable operation to duty parameter and Conventional pollution Concentration, as characteristic vector, submit to regression forecasting device；
The regression forecasting device, is stored with housebroken supporting vector machine model, is filled for being obtained according to characteristic vector The characteristic vector of submission is put, calculates regression forecasting result, and submit to concentration calibration device；
The supporting vector machine model is ε insensitive loss functions support vector regression (ε-SVR)；ε value is 0.01 Between~0.1, preferred ε=0.04；Its kernel function is d rank multinomials kernel function, Radial basis kernel function or sigmoid kernel functions.
The concentration calibration device, for the regression forecasting result submitted according to regression forecasting device, demarcate waste incineration Dioxin in flue gas class concentration；The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.
Compared with directly detecting, the present invention makes full use of existing monitoring conditions and data, by analyzing dioxin content With duty parameter and the dependency relation of Conventional pollution concentration, regressive prediction model is established, is carried out with indirect Dui dioxins real When monitor, can meet that dioxin becomes more meticulous the requirement of supervision with minimum cost.
Waste incineration produces the complicated mechanism of dioxin, and influence factor is more, and regression forecasting needs to establish multivariate model. But it is often few for modeling training dioxin data, multiple regression forecasting is met with Small Sample Database problem, and it is traditional Statistical method be that infinitely great progressive theory is tended to based on number of samples, it is difficult to reach the requirement of prediction.And machine learning In method, decision tree method is larger by artificial subjective impact, though artificial neural network method is based on empiric risk most using relatively extensively Based on smallization principle, local optimum is easily absorbed in, learning state is crossed and causes precision of prediction to decline, and support vector regression (Support Vector Regression, SVR) based on Statistical Learning Theory and structural risk minimization principle, effectively Evade above-mentioned risk, good prediction effect can be obtained.
It is embodiment below：
Support vector regression of the embodiment of the present invention based on 3 kinds of different kernel functions establishes dioxin concentration of emission and toxicity Equivalent (I-TEQ) regressive prediction model, and with multiple linear regression (Multiple Linear Regression, MLR) from plan Conjunction ability and the aspect of precision of prediction two are compared, to the forecast model learnt and generalization ability is strong, so as to indirectly real The real-time monitoring of existing waste incineration dioxin in flue gas class.
The embodiment of the present invention uses goodness of fit R2Evaluate the learning ability that method is provided in embodiment；Rolled over and handed over using 10- The mode of fork checking provides the generalization ability of method to evaluate in embodiment, evaluation index includes mean absolute error (MAE), put down Equal absolute percent error (MAPE) and maximum absolute percent error (MaxAPE).
Model data obtains in accordance with the following methods in following examples：
The same step of characteristic vector parameter acquiring (1) of training sample；Dioxin total concentration and total toxic equivalent are according to such as Lower method obtains：Each group sample dioxins are tested and analyzed, obtain total concentration and total toxic equivalent.Dioxin Total concentration is that total toxic equivalent of the concentration sum , dioxins of 17 kinds of poisonous congeners is according to the dense of 17 kinds of poisonous congeners Degree and《Consumer waste incineration contamination control standard》The toxic equivalency factor (I-TEF) that (GB 18485-2014) is provided is added Toxic equivalent sum obtained from power calculates.
Embodiment 1
A kind of waste incineration dioxin in flue gas class concentration prediction method, comprises the following steps：
(1) obtain waste incinerator stable operation when to duty parameter and Conventional pollution concentration, as characteristic vector；
The duty parameter, including：Incineration temperature, boiler export cigarette temperature and flue gas flow in burner hearth；It is described conventional dirty Thing concentration is contaminated, including：SO2Concentration, HCl concentration and particle concentration.
Obtain as follows：
The waste incinerator being equipped with for municipal solid waste incinerator, fume sample collection is carried out, is respectively provided at using sampling point After extracting tower, before activated carbon injection and sack cleaner, and after sack cleaner, each point gathers 2~3 times.In order to ensure The matching of data, every CIU carry out fume sample collection simultaneously before and after gas cleaning.Under different working conditions Fume sample is gathered, each sampling process continues 4 hours, and respectively stable conditions time 2 h , dioxins sample uses sample 2 hours.During sample collection, using modes such as manual monitoring, portable instrument monitoring or on-line monitoring equipment monitorings, to work (incineration temperature, boiler export cigarette temperature, cloth bag export cigarette temperature, flue gas flow, O to condition parameter in burner hearth2With H2O content) it is dirty with routine Contaminate thing concentration (CO2、SO2, NOx, HCl, CO, particulate matter, HF and NH3) monitoring is synchronized, collect within every 10 minutes 1 monitoring number According to totally 12 times, end product takes its average value.
(2) characteristic vector obtained in step (1) is inputted into supporting vector machine model, carries out regression forecasting；
The supporting vector machine model, for training sample set { xi,yi},xi∈Rn,yi∈ R (i=1,2 ..., l), obtain Regression function f (x)=wx+b parameter w, b；Goal regression estimation function is：
Wherein, K (xiX)=Φ (xi)·Φ(xj) be SVMs kernel function, it is specific to use：
D rank multinomial kernel functions：K(x,xi)=[γ (xxi)+1]d
Using ε insensitive loss functions support vector regression (ε-SVR) instrument.Insensitive loss coefficient ε is 0.04, exponent number The parameter γ values of d=1, penalty factor, and kernel function are shown in Table 1.
The characteristic vector obtained in step (1) is standardized pretreatment so that characteristic vector value after processing [0, 1] between.Specifically, standardized value can calculate according to equation below：
Wherein, xsFor the standardized value of the characteristic vector value, xiFor the value of the characteristic vector, xminFor the feature to Measure the minimum value in training sample, xmaxFor maximum of the characteristic vector in training sample.
Table 1d rank multinomial kernel functional parameter values
(3) it is dense according to Support vector regression prediction result in step (2), demarcation waste incineration dioxin in flue gas class Degree.
The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.As a result it is as shown in table 1.
Embodiment 2
A kind of waste incineration dioxin in flue gas class concentration prediction method, comprises the following steps：
(1) obtain waste incinerator stable operation when to duty parameter and Conventional pollution concentration, as characteristic vector；
The duty parameter, including：Incineration temperature, boiler export cigarette temperature and flue gas flow in burner hearth；It is described conventional dirty Thing concentration is contaminated, including：SO2Concentration, HCl concentration and particle concentration.
Obtain as follows：
The waste incinerator being equipped with for municipal solid waste incinerator, fume sample collection is carried out, is respectively provided at using sampling point After extracting tower, before activated carbon injection and sack cleaner, and after sack cleaner, each point gathers 2~3 times.In order to ensure The matching of data, every CIU carry out fume sample collection simultaneously before and after gas cleaning.Under different working conditions Fume sample is gathered, continues 4 hours using sample process every time, respectively stable conditions time 2 h , dioxins sample uses Sample 2 hours.It is right using modes such as manual monitoring, portable instrument monitoring or on-line monitoring equipment monitorings during sample collection (incineration temperature, boiler export cigarette temperature, cloth bag export cigarette temperature, flue gas flow, O to duty parameter in burner hearth2With H2O content) and it is conventional Pollutant concentration (CO2、SO2, NOx, HCl, CO, particulate matter, HF and NH3) monitoring is synchronized, collect within every 10 minutes 1 monitoring Data, totally 12 times, end product takes its average value.
(2) characteristic vector obtained in step (1) is inputted into supporting vector machine model, carries out regression forecasting；
The supporting vector machine model, for training sample set { xi,yi},xi∈Rn,yi∈ R (i=1,2 ..., l), obtain Regression function f (x)=wx+b parameter w, b；Optimization aim is：
Wherein, K (xiX)=Φ (xi)·Φ(xj) be SVMs kernel function, it is specific to use：
RBF kernel functions：K(x,xi)=exp (- γ | | x-xi||2)
Using ε insensitive loss functions support vector regression (ε-SVR) instrument.Insensitive loss coefficient ε is 0.04, punishment The parameter γ values of factor C and kernel function are shown in Table 2.
Table 2RBF kernel functional parameter values
The characteristic vector obtained in step (1) is standardized pretreatment so that characteristic vector value after processing [0, 1] between.Specifically, standardized value can calculate according to equation below：
Wherein, xsFor the standardized value of the characteristic vector value, xiFor the value of the characteristic vector, xminFor the feature to Measure the minimum value in training sample, xmaxFor maximum of the characteristic vector in training sample.
(3) it is dense according to Support vector regression prediction result in step (2), demarcation waste incineration dioxin in flue gas class Degree.
The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.As a result it is as shown in table 1.
Embodiment 3
A kind of waste incineration dioxin in flue gas class concentration prediction method, comprises the following steps：
(1) obtain waste incinerator stable operation when to duty parameter and Conventional pollution concentration, as characteristic vector；
The duty parameter, including：Incineration temperature, boiler export cigarette temperature and flue gas flow in burner hearth；It is described conventional dirty Thing concentration is contaminated, including：SO2Concentration, HCl concentration and particle concentration.
Obtain as follows：
The waste incinerator being equipped with for municipal solid waste incinerator, fume sample collection is carried out, is respectively provided at using sampling point After extracting tower, before activated carbon injection and sack cleaner, and after sack cleaner, each point gathers 2~3 times.In order to ensure The matching of data, every CIU carry out fume sample collection simultaneously before and after gas cleaning.Under different working conditions Fume sample is gathered, continues 4 hours using sample process every time, respectively stable conditions time 2 h , dioxins sample uses Sample 2 hours.It is right using modes such as manual monitoring, portable instrument monitoring or on-line monitoring equipment monitorings during sample collection (incineration temperature, boiler export cigarette temperature, cloth bag export cigarette temperature, flue gas flow, O to duty parameter in burner hearth2With H2O content) and it is conventional Pollutant concentration (CO2、SO2, NOx, HCl, CO, particulate matter, HF and NH3) monitoring is synchronized, collect within every 10 minutes 1 monitoring Data, totally 12 times, end product takes its average value.
(2) characteristic vector obtained in step (1) is inputted into supporting vector machine model, carries out regression forecasting；
The supporting vector machine model, for training sample set { xi,yi},xi∈Rn,yi∈ R (i=1,2 ..., l), obtain Regression function f (x)=wx+b parameter w, b；Optimization aim is：
Wherein, K (xiX)=Φ (xi)·Φ(xj) be SVMs kernel function, it is specific to use：
Sigmoid kernel functions：K(x,xi)=tanh [γ (xxi)+c]
Using ε insensitive loss functions support vector regression (ε-SVR) instrument.Insensitive loss coefficient ε is 0.04, punishment Factor C and the parameter γ of kernel function are shown in Table 3.
Table 3sigmoid kernel functional parameter values
The characteristic vector obtained in step (1) is standardized pretreatment so that characteristic vector value after processing [0, 1] between.Specifically, standardized value can calculate according to equation below：
Wherein, xsFor the standardized value of the characteristic vector value, xiFor the value of the characteristic vector, xminFor the feature to Measure the minimum value in training sample, xmaxFor maximum of the characteristic vector in training sample.
(3) it is dense according to Support vector regression prediction result in step (2), demarcation waste incineration dioxin in flue gas class Degree.
The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.As a result it is as shown in table 1.
Comparative example：
A kind of waste incineration dioxin in flue gas class concentration prediction method, comprises the following steps：
(1) obtain waste incinerator stable operation when to duty parameter and Conventional pollution concentration, as characteristic vector；
The duty parameter, including：Incineration temperature, boiler export cigarette temperature and flue gas flow in burner hearth；It is described conventional dirty Thing concentration is contaminated, including：SO2Concentration, HCl concentration and particle concentration.
Obtain as follows：
The waste incinerator being equipped with for municipal solid waste incinerator, fume sample collection is carried out, is respectively provided at using sampling point After extracting tower, before activated carbon injection and sack cleaner, and after sack cleaner, each point gathers 2~3 times.In order to ensure The matching of data, every CIU carry out fume sample collection simultaneously before and after gas cleaning.Under different working conditions Fume sample is gathered, continues 4 hours using sample process every time, respectively stable conditions time 2 h , dioxins sample uses Sample 2 hours.It is right using modes such as manual monitoring, portable instrument monitoring or on-line monitoring equipment monitorings during sample collection (incineration temperature, boiler export cigarette temperature, cloth bag export cigarette temperature, flue gas flow, O to duty parameter in burner hearth2With H2O content) and it is conventional Pollutant concentration (CO2、SO2, NOx, HCl, CO, particulate matter, HF and NH3) monitoring is synchronized, collect within every 10 minutes 1 monitoring Data, totally 12 times, end product takes its average value.
(2) characteristic vector obtained in step (1) is inputted into MLR machine models, carries out regression forecasting；
(3) it is dense according to Support vector regression prediction result in step (2), demarcation waste incineration dioxin in flue gas class Degree.
The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.As a result it is as shown in table 4.
From the point of view of table 4, both are fine for the learning ability of sample, R2Average value is all higher than 0.95, illustrates sample sheet Preferable linear relationship be present in body.But SVR models are better than MLR, especially Dui bis- Evil for the generalization ability of Small Sample Database The prediction error of English class total concentration is smaller, and the absolute percent error and average absolute percentage of more than 70% test sample are missed Difference is all within 10%, but largest percentage error is larger, and the maximum of 3 seed nucleus function pair dioxin total concentrations prediction is definitely Percentage error appears at No. 2 samples.Big error sample proportion of the MLR prediction result absolute percent errors more than 20% compared with Reach 60% in height, the especially prediction result of the total toxic equivalent of Dui dioxins.
The prediction effect of 4 embodiment of table 1 to 3
The comparative analysis of embodiment 1 to 3：
Found during modeling and forecasting, the exponent number of the d rank multinomial kernel functions of embodiment 1 have when taking 1 preferably study with it is extensive Ability, and take 2~6 acquired results very poor to this sample data set learning ability, it is low to data sensitivity, therefore the present invention is preferably Polynomial kernel function exponent number takes 1.When training sample is 9 groups, the model learning ability of 3 kinds of kernel functions is all preferable, R2Average value More than 0.96.Embodiment 1 predicts that mean error is minimum for dioxin total concentration, and the Dui dioxins of embodiment 3 are total Concentration and total toxic equivalent entirety generalization ability are most strong, and ratio of the absolute percent error less than 10% reaches respectively in prediction result To 80% and 60%, embodiment 1 and embodiment 2 are above, but its largest percentage error is maximum.The Dui bioxin of embodiment 2 The forecast model generalization ability of the total toxic equivalent of class shows preferable harmony, a little higher than implementation of mean absolute percentage error Example 3, but its maximum absolute percent error is 22.38%, less than embodiment 1 and embodiment 3.
For the quality that further relatively 3 kinds of different kernel functions are predicted for small sample problem, the present invention examines rate training sample The estimated performance of difference kernel function during this reduction.By the way of 5 folding cross validations, every 8 groups of data as training set, remaining 2 Group data are respectively adopted 3 kinds of kernel functions and are modeled calculating, carry out 90 cross validations altogether and calculate as test set.Such as the institute of table 5 Show, table 5 has counted the R of each group experimental result2Average value, R2Maximum, R2Minimum value, mean absolute error, average absolute percentage The confidential interval of 95% time ratio error, maximum absolute percent error and confidence level mean absolute percentage error.Fig. 2 is real The kernel function of example 1 to 3 drag prediction absolute percent error map is applied, is that interval is counted with 10%.
The prediction result of 5 embodiment of table 1 to 3 counts
From table 3 it is observed that when training sample is reduced to 8 groups, embodiment 1 is with the ability of embodiment 2 significantly better than real Apply example 3, the part training sample learning ability of the latter's Dui dioxins total concentration and total toxic equivalent is poor, R2Minimum value is distinguished For 0.83 and 0.65, and the above two R2Minimum value is substantially all more than 0.90.Embodiment 1 is total for dioxin to embodiment 3 The generalization ability of concentration prediction is close, and substantially excellent with embodiment 2 for the prediction of the total toxic equivalent of dioxin, embodiment 1 In embodiment 3.Embodiment 1 and the performance of the generalization ability of embodiment 2 are close, but the former mean absolute percentage error can restrain To smaller section, slightly better than the latter.
Characteristic vector impact analysis：
In order to analyze influence degree size of 6 characteristic variables to model, 1 change is removed successively on the basis of embodiment 2 Amount, is modeled, the method and parameter of modeling are consistent with original model using other 5 variables as characteristic variable.By original mould Type is represented with M0, will remove incineration temperature in burner hearth, boiler export cigarette temperature, flue gas flow, SO successively2Concentration, HCl concentration and The model of grain thing concentration is represented with M1~M6 respectively.Referred to using the mean absolute percentage error of model prediction result as evaluation Mark, as a result as shown in Figure 3.
Found out by Fig. 3 result, influence of the incineration temperature to model is maximum in HCl concentration and burner hearth, right after removal variable The prediction mean absolute percentage error of dioxin total concentration and total toxic equivalent is significantly increased, wherein it is dense to remove HCl More than 25% is added after degree (M5), removes in burner hearth and adds 10% or so after incineration temperature (M1)；Variable pot is removed respectively Outlet of still cigarette temperature (M2), flue gas flow (M3), SO2Under the prediction error of concentration (M4) , dioxins total concentration afterwards has slightly Drop, and the prediction error of total toxic equivalent another 2 groups of essentially unchangedization in addition to M3 rises about 8%；It is dense to remove variable particulate matter (M6) is spent afterwards on dioxin total toxic equivalent prediction error also substantially without influence, but total concentration prediction error adds 9%.
Understand that 3 incineration temperature, particle concentration variable Dui dioxin total concentrations are pre- in HCl concentration, burner hearth through analysis Survey model to have a great influence, 3 incineration temperature, flue gas flow total toxic equivalents of variable Dui dioxins are pre- in HCl concentration, burner hearth Model is surveyed to have a great influence.Follow-up study can deepen the research to characteristic variable, optimized variable combination.
The embodiment of the present invention is based on 10 groups of actual monitoring data, by the way of K- rolls over cross validation, compared for being based on 3 kinds The support vector regression of kernel function and the estimated performance of multiple linear regression, the results showed that support vector regression is in generalization ability side Face has obvious advantage, and the absolute percent error of most prediction results is within 10%.But no matter which kind of model, The generalization ability of the total toxic equivalent of Yu Ce dioxins is below Yu Ce dioxin total concentrations.
Number of training has a great influence to the generalization ability of support vector regression, when number of training is 9, uses The embodiment 3 of sigmoid kernel functions shows overall less relative error, but when number of training reduces to 8, uses 1 rank The embodiment 2 of polynomial embodiment 1 and RBF kernel functions obtains smaller phase than the number of embodiment 3 using sigmoid core letters To error.
Influence of the variable to model is analyzed by way of model prediction error change size after rejecting certain variable successively, As a result HCl concentration is found, incineration temperature, flue gas flow and particle concentration have a great influence to model predictive error in burner hearth.
Herein for garbage burning factory flue gas dioxin class total concentration and the Small Sample Database modeling and forecasting of total toxic equivalent, General performance goes out preferable performance, helps to realize the indirect monitoring in real time of garbage burning factory flue gas dioxin class concentration, related Method is available for other small sample problems encountered in environmental monitoring to refer to.But present document relates to sample number it is less, model Use also need to collect more Monitoring Datas further to optimize, while can deepen research model characteristic variable combination To obtain more preferable performance.
As it will be easily appreciated by one skilled in the art that the foregoing is merely illustrative of the preferred embodiments of the present invention, not to The limitation present invention, all any modification, equivalent and improvement made within the spirit and principles of the invention etc., all should be included Within protection scope of the present invention.

Claims (10)

  1. A kind of 1. waste incineration dioxin in flue gas class concentration prediction method, it is characterised in that comprise the following steps：
    (1) duty parameter and Conventional pollution concentration during acquisition waste incinerator stable operation, as characteristic vector；
    (2) characteristic vector obtained in step (1) is inputted into supporting vector machine model, carries out regression forecasting；
    (3) according to Support vector regression prediction result in step (2), waste incineration dioxin in flue gas class concentration is demarcated.
  2. 2. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 1, it is characterised in that the operating mode Parameter includes：Incineration temperature in burner hearth；The Conventional pollution concentration includes：HCl concentration.
  3. 3. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 2, it is characterised in that the operating mode Parameter also includes：Boiler export cigarette temperature and flue gas flow；The Conventional pollution concentration also includes：SO2Concentration and Grain thing concentration.
  4. 4. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 1, it is characterised in that step (2) The supporting vector machine model kernel function is：D rank multinomials kernel function, Radial basis kernel function, sigmoid kernel functions.
  5. 5. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 1, it is characterised in that step (2) The supporting vector machine model uses ε insensitive loss function support vector regression instruments.
  6. 6. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 5, it is characterised in that step (2) The insensitive loss coefficient ε values of ε insensitive loss function support vector regression (ε-SVR) instrument 0.01~0.1 it Between.
  7. 7. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 6, it is characterised in that step (2) The insensitive loss coefficient ε values of ε insensitive loss function support vector regression (ε-SVR) instrument are 0.04.
  8. 8. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 6, it is characterised in that step (2) Including：
    The characteristic vector obtained in step (1) is standardized pretreatment so that characteristic vector value after processing [0,1] it Between.Specifically, standardized value can calculate according to equation below：
    <mrow> <msub> <mi>x</mi> <mi>s</mi> </msub> <mo>=</mo> <mfrac> <mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </mrow> <mrow> <msub> <mi>x</mi> <mi>max</mi> </msub> <mo>-</mo> <msub> <mi>x</mi> <mrow> <mi>m</mi> <mi>i</mi> <mi>n</mi> </mrow> </msub> </mrow> </mfrac> </mrow>
    Wherein, xsFor the standardized value of the characteristic vector value, xiFor the value of the characteristic vector, xminExist for the characteristic vector Minimum value in training sample, xmaxFor maximum of the characteristic vector in training sample.
  9. 9. waste incineration dioxin in flue gas class concentration prediction method as claimed in claim 1, it is characterised in that step (3) The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.
  10. A kind of 10. waste incineration dioxin in flue gas class concentration prediction system, it is characterised in that including：
    Characteristic vector acquisition device, duty parameter and Conventional pollution concentration during for obtaining waste incinerator stable operation, As characteristic vector, regression forecasting device is submitted to；
    The regression forecasting device, is stored with housebroken supporting vector machine model, for being carried according to characteristic vector acquisition device The characteristic vector of friendship, regression forecasting result is calculated, and submit to concentration calibration device；
    The supporting vector machine model is ε insensitive loss function support vector regressions；ε value is excellent between 0.01~0.1 Select ε=0.04；Its kernel function is d rank multinomials kernel function, Radial basis kernel function or sigmoid kernel functions；
    The concentration calibration device, for the regression forecasting result submitted according to regression forecasting device, demarcate flue gas of refuse burning Zhong dioxin concentration；The dioxin concentration Wei dioxins total concentration and/or total toxic equivalent.
CN201710544474.6A 2017-07-06 2017-07-06 A kind of waste incineration dioxin in flue gas class concentration prediction method and system Pending CN107356710A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710544474.6A CN107356710A (en) 2017-07-06 2017-07-06 A kind of waste incineration dioxin in flue gas class concentration prediction method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710544474.6A CN107356710A (en) 2017-07-06 2017-07-06 A kind of waste incineration dioxin in flue gas class concentration prediction method and system

Publications (1)

Publication Number Publication Date
CN107356710A true CN107356710A (en) 2017-11-17

Family

ID=60293288

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710544474.6A Pending CN107356710A (en) 2017-07-06 2017-07-06 A kind of waste incineration dioxin in flue gas class concentration prediction method and system

Country Status (1)

Country Link
CN (1) CN107356710A (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109613130A (en) * 2018-11-22 2019-04-12 环境保护部华南环境科学研究所 Transfer dissolving method and liquid relief component based on bioanalysis detection dioxin sample
CN110081440A (en) * 2019-05-05 2019-08-02 绿色动力环保集团股份有限公司 The on-line prediction method of dioxin and early warning and control system in consumer waste incineration
CN110135057A (en) * 2019-05-14 2019-08-16 北京工业大学 Solid waste burning process dioxin concentration flexible measurement method based on multilayer feature selection
JP2020067980A (en) * 2018-10-26 2020-04-30 富士通株式会社 Prediction program, prediction method, and prediction device
CN111308966A (en) * 2019-11-05 2020-06-19 光大环保技术研究院（南京）有限公司 Emergency emission supervision system, analysis system and user terminal equipment for waste incineration power generation
CN111882230A (en) * 2020-07-31 2020-11-03 浙江大学 Neural network-based dioxin emission online monitoring method
CN111882130A (en) * 2020-07-30 2020-11-03 浙江大学 Online dioxin emission prediction method based on generation path clustering and Box-Cox transformation
CN112102890A (en) * 2020-07-28 2020-12-18 浙江中控技术股份有限公司 MCS synthesis device reaction temperature prediction method based on machine learning model
CN112884203A (en) * 2021-01-22 2021-06-01 浙江大学 Online prediction method for dioxin emission based on Bayesian support vector regression
CN113094986A (en) * 2021-04-01 2021-07-09 瀚蓝绿电固废处理(佛山)有限公司 Method for constructing prediction model of pollutant emission in flue gas of waste incinerator and application
CN113804838A (en) * 2021-09-28 2021-12-17 浙江富春江环保科技研究有限公司 Boundary alarm method for dioxin online detection
CN114216128A (en) * 2021-12-06 2022-03-22 生态环境部华南环境科学研究所 Risk prevention and control method and system for dioxins in household garbage incineration plant
CN116153425A (en) * 2023-02-24 2023-05-23 北京和利时工业软件有限公司 Ammonia-carbon ratio soft measurement method, device, equipment and medium

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
李大中 等: "垃圾焚烧发电污染物排放过程建模与优化", 《可再生能源》 *
杨俊燕 等: "ε不敏感损失函数支持向量机分类性能研究", 《西安交通大学学报》 *

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2020067980A (en) * 2018-10-26 2020-04-30 富士通株式会社 Prediction program, prediction method, and prediction device
JP7215077B2 (en) 2018-10-26 2023-01-31 富士通株式会社 Prediction program, prediction method and prediction device
CN109613130A (en) * 2018-11-22 2019-04-12 环境保护部华南环境科学研究所 Transfer dissolving method and liquid relief component based on bioanalysis detection dioxin sample
CN110081440A (en) * 2019-05-05 2019-08-02 绿色动力环保集团股份有限公司 The on-line prediction method of dioxin and early warning and control system in consumer waste incineration
CN110081440B (en) * 2019-05-05 2020-06-09 绿色动力环保集团股份有限公司 Online prediction method and early warning and control system for dioxin in household garbage incineration
WO2020228215A1 (en) * 2019-05-14 2020-11-19 北京工业大学 Soft measurement method for dioxin emission concentration in solid waste incineration process based on multi-layer feature selection
CN110135057A (en) * 2019-05-14 2019-08-16 北京工业大学 Solid waste burning process dioxin concentration flexible measurement method based on multilayer feature selection
CN111308966A (en) * 2019-11-05 2020-06-19 光大环保技术研究院（南京）有限公司 Emergency emission supervision system, analysis system and user terminal equipment for waste incineration power generation
CN112102890A (en) * 2020-07-28 2020-12-18 浙江中控技术股份有限公司 MCS synthesis device reaction temperature prediction method based on machine learning model
CN112102890B (en) * 2020-07-28 2023-12-15 浙江中控技术股份有限公司 MCS synthesis device reaction temperature prediction method based on machine learning model
CN111882130B (en) * 2020-07-30 2022-01-11 浙江大学 Online dioxin emission prediction method based on generation path clustering and Box-Cox transformation
CN111882130A (en) * 2020-07-30 2020-11-03 浙江大学 Online dioxin emission prediction method based on generation path clustering and Box-Cox transformation
CN111882230B (en) * 2020-07-31 2021-10-01 浙江大学 Neural network-based dioxin emission online monitoring method
CN111882230A (en) * 2020-07-31 2020-11-03 浙江大学 Neural network-based dioxin emission online monitoring method
CN112884203A (en) * 2021-01-22 2021-06-01 浙江大学 Online prediction method for dioxin emission based on Bayesian support vector regression
CN113094986A (en) * 2021-04-01 2021-07-09 瀚蓝绿电固废处理(佛山)有限公司 Method for constructing prediction model of pollutant emission in flue gas of waste incinerator and application
CN113804838B (en) * 2021-09-28 2022-07-26 浙江富春江环保科技研究有限公司 Boundary alarm method for dioxin online detection
CN113804838A (en) * 2021-09-28 2021-12-17 浙江富春江环保科技研究有限公司 Boundary alarm method for dioxin online detection
CN114216128A (en) * 2021-12-06 2022-03-22 生态环境部华南环境科学研究所 Risk prevention and control method and system for dioxins in household garbage incineration plant
CN116153425A (en) * 2023-02-24 2023-05-23 北京和利时工业软件有限公司 Ammonia-carbon ratio soft measurement method, device, equipment and medium
CN116153425B (en) * 2023-02-24 2024-04-02 北京和利时工业软件有限公司 Ammonia-carbon ratio soft measurement method, device, equipment and medium

Similar Documents

Publication Publication Date Title
CN107356710A (en) A kind of waste incineration dioxin in flue gas class concentration prediction method and system
CN108549792B (en) Soft measurement method for dioxin emission concentration in solid waste incineration process based on latent structure mapping algorithm
Jiang et al. An innovative hybrid air pollution early-warning system based on pollutants forecasting and Extenics evaluation
CN111461355B (en) Dioxin emission concentration migration learning prediction method based on random forest
CN109960873B (en) Soft measurement method for dioxin emission concentration in urban solid waste incineration process
CN110135057B (en) Soft measurement method for dioxin emission concentration in solid waste incineration process based on multilayer characteristic selection
CN107944173B (en) Dioxin soft measurement system based on selective integrated least square support vector machine
WO2021159585A1 (en) Dioxin emission concentration prediction method
Li et al. Heavy metals in submicronic particulate matter (PM1) from a Chinese metropolitan city predicted by machine learning models
Sun et al. Prediction of oxygen content using weighted PCA and improved LSTM network in MSWI process
Elbaz et al. Spatiotemporal air quality forecasting and health risk assessment over smart city of NEOM
CN112364562A (en) Cooperative control method and system for flue gas environmental protection island
Qin et al. A fuzzy composting process model
Xia et al. Dioxin emission modeling using feature selection and simplified DFR with residual error fitting for the grate-based MSWI process
CN1327376C (en) Soft measuring meter moduling method based on supporting vector machine
Zain et al. Development of a neural network Predictive Emission Monitoring System for flue gas measurement
CN112241800A (en) Method for calculating VOCs pollutant emission of coke oven
CN115017980A (en) Prediction method for heavy metal migration in waste incineration process based on random forest algorithm
JP3668405B2 (en) Waste incinerator control method and apparatus
Liu et al. Time series prediction of the chemical components of PM2. 5 based on a deep learning model
Wu et al. Design and analysis of an data-driven intelligent model for persistent organic pollutants in the internet of things environments
Saini et al. Environmental Scrutinizing System based on Soft Computing Technique
Tang et al. Dioxin Emission Concentration Soft Measuring Method Based on Selective Ensemble Least Square Support Vector Machine Algorithm
Xia et al. Dioxin emission concentration forecasting model for MSWI process with random forest-based transfer learning
CN117054616B (en) Remote middle station atmosphere pollution tracing system based on artificial intelligence

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20171117

RJ01 Rejection of invention patent application after publication