CN117408378A - NOx emission monitoring method of thermal generator set - Google Patents

Abstract

The invention discloses a method for monitoring NOx emission of a thermal power generating unit, which comprises the steps of acquiring original data of NOx emission of the thermal power generating unit, establishing a GM (1, N) gray prediction model for predicting the NOx emission, establishing an SVM support vector machine prediction model, and combining the gray prediction model with the support vector machine prediction model to obtain a combined optimization model; taking relevant parameters of the thermal generator set in a period of time, respectively predicting through a gray prediction model, a support vector machine prediction model and a combined optimization model, and verifying the prediction accuracy of the combined optimization model through data comparison; and predicting the future NOx emission of the thermal generator set by using the combined optimization model. The method is simple and feasible, the data acquisition is convenient, the data processing and prediction processes are simple, the prediction result is accurate, and the method is an effective method for predicting the NOx emission in the future thermal power industry.

Description

NOx emission monitoring method of thermal generator set

Technical Field

The invention belongs to the technical field of thermal generator sets, and particularly relates to a NOx emission monitoring method of a thermal generator set.

Background

NOx is a substance mixed in the flue gas generated by a thermal power generating unit, and is destroyed to a certain extent to the environment, and as NOx is mixed in the flue gas generated by the thermal power generating unit, the emission amount of the NOx cannot be directly monitored, namely the NOx is difficult to meter; the sensor is used for measuring the NOx content in the flue gas, so that the monitoring of certain nitrogen oxides can only be carried out singly, the monitoring of other nitrogen oxides still needs to be carried out in an estimation mode, and the NOx emission amount generated by the thermal power generating unit is generally predicted in the estimation mode at present.

In the prior art, for the prediction of the NOx emission amount of a certain area, there are mainly an emission factor method and a typical power plant method.

The emission factor method is to multiply the emission factor of the coal-fired NOx by the coal-fired quantity to obtain the emission quantity of NOx, and the coal-fired quantity in the current year is selected aiming at measuring the emission quantity of the NOx in the current year; aiming at predicting the NOx emission amount of the future year, selecting a predicted value of the fuel coal amount of the future year; in the implementation of the emission factor method, certain inaccuracy exists in the adoption of empirical data of the emission factor of the coal-fired NOx;

the typical power plant method is to estimate or predict the NOx emission amount of a typical thermal power plant on the basis of estimating or predicting the NOx emission amount of the thermal power industry in a measured area, firstly classifying thermal power units in the measured area, determining the NOx emission level of each typical unit, estimating the NOx emission amount of the thermal power industry in the measured area all the year according to the installed capacity of each unit in the current year, or calculating the NOx emission amount of the thermal power industry in the measured area according to the predicted value of the installed capacity of each unit in the future; because of the large variety and number of thermal power generating units in the detected area, the prediction process is complex, the workload is large, the time and economic cost are high, and the prediction method is not an ideal choice.

Disclosure of Invention

The invention provides a simple and feasible method for monitoring NOx emission of a thermal generator set, which is convenient for data acquisition, simple in data processing and prediction process and accurate in prediction result, and is an effective method for predicting the NOx emission in the future thermal power industry, in order to avoid the defects in the prior art

The invention adopts the following technical scheme for realizing the purpose:

the NOx emission monitoring method of the thermal generator set is characterized by comprising the following steps:

step 1: acquiring original data of NOx emission of a thermal power generating unit;

step 2: establishing a GM (1, N) gray prediction model for predicting NOx emission by using the raw data;

step 3: constructing an SVM support vector machine prediction model by using a sample training set;

step 4: combining the gray prediction model and a support vector machine prediction model to obtain a combined optimization model;

step 5: and predicting the future NOx emission of the thermal generator set by adopting the combined optimization model.

The NOx emission monitoring method of the thermal generator set is also characterized in that:

taking relevant parameters of the thermal generator set in a period of time, respectively predicting by using a gray prediction model, a support vector machine prediction model and a combined optimization model to obtain prediction data of each prediction model, comparing the prediction data of each prediction model, and verifying the prediction accuracy of the combined optimization model; the relevant parameters include: parameters of thermal power coal consumption, installed capacity, thermal power generation amount and NOx emission amount.

The NOx emission monitoring method of the thermal generator set is also characterized in that:

in the step 1, the consumption of thermal power coal, the installed capacity and the thermal power generation amount are selected as factors influencing the NOx emission amount of the thermal power industry, and each factor is collected from daily data of a power plant, so that an original data sequence X based on the original data is obtained _i ⁽⁰⁾ As formula (1):

X _i ⁽⁰⁾ ＝{x _i ⁽⁰⁾ (1)，x _i ⁽⁰⁾ (2)，...x _i ⁽⁰⁾ (m)} (1)

in the formula (1):

i represents the i-th factor, i=1, 2,..n, n is the total amount of factors, each factor containing m state values;

in x _i ⁽⁰⁾ (1) A 1 st state value representing an i-th factor;

in x _i ⁽⁰⁾ (2) A 2 nd state value representing an i-th factor;

……

in x _i ⁽⁰⁾ (m) represents an mth state value of an ith factor.

The NOx emission monitoring method of the thermal generator set is also characterized in that:

in said step 2, for said original data sequence X _i ⁽⁰⁾ Accumulating once according to an accumulation formula to obtain a once iteration data sequence X _i ⁽¹⁾ As formula (2):

X _i ⁽¹⁾ ＝{x _i ⁽¹⁾ (1)，x _i ⁽¹⁾ (2)，...x _i ⁽¹⁾ (m)} (2)

in the formula (2):

x _i ⁽¹⁾ (1) Is x _i ⁽⁰⁾ (1) Is used for the first iteration data of the (a);

x _i ⁽¹⁾ (2) Is x _i ⁽⁰⁾ (2) Is used for the first iteration data of the (a);

……

x _i ⁽¹⁾ (m) is x _i ⁽⁰⁾ The one iteration data of (m);

using an iterative data sequence X _i ⁽¹⁾ Establishing a first-order differential equation as shown in formula (3):

wherein alpha, b ₁ ，b ₂ ，...b _n-1 The parameter sequence is the parameter sequence of the gray prediction model;

expressing formula (3) in vector form; comprising the following steps: obtaining a parameter B according to the formula (4), and obtaining a parameter y according to the formula (5):

obtaining grey prediction model parameters by the calculation of the formula (6) according to the least square method

Using the gray prediction model parametersThis builds a grey prediction model.

The NOx emission monitoring method of the thermal generator set is also characterized in that:

in said step 3, a sample training set (x _i ，y _i )，x _i For input space, y _i For output values, the sample training set (x _i ，y _i ) And constructing an SVM support vector machine prediction model, and obtaining an output set Y of the NOx emission according to the input set X of the NOx emission by utilizing the SVM support vector machine prediction model.

The NOx emission monitoring method of the thermal generator set is also characterized in that:

in the step 4, the gray prediction model and the support vector machine prediction model are combined to obtain a combined optimization model as shown in formula (7):

in the formula (7):

at f _x Representing a predicted value of the optimized combined prediction model;

at f ₁ Predicted values expressed as gray prediction models; at f ₂ Representing a predicted value of the support vector machine model;

e is as follows ₁ Representing a prediction error of the gray prediction model; e is as follows ₂ Representing a prediction error of the support vector machine model;

with D (e) ₁ ) Representation e ₁ Variance of the values, expressed as D (e ₂ ) Representation e ₂ Variance of values.

Compared with the prior art, the invention has the beneficial effects that:

on the basis of predicting by utilizing a gray model and a support vector machine model, a combined optimization prediction model is established, the advantages of two single prediction models are better combined, and the annual coal consumption, installed capacity and power generation of a thermal power unit are selected as influencing factors influencing the annual NOx emission of the thermal power unit by taking the past thermal power NOx emission as examples, so that the combined optimization model is verified, and a better prediction effect is obtained; compared with an emission factor method and a typical power plant method, the method is simple and feasible, convenient in data acquisition, simple in data processing and prediction process, accurate in prediction result, and an effective method for predicting the NOx emission in the future thermal power industry.

Drawings

FIG. 1 is a flow chart of the method of the present invention.

Detailed Description

Referring to fig. 1, the NOx emission monitoring method of the thermal power generating unit in this embodiment includes the steps of:

step 1: acquiring original data of NOx emission of a thermal power generating unit;

step 2: establishing a GM (1, N) gray prediction model for predicting the NOx emission amount by using the original data;

step 3: constructing an SVM support vector machine prediction model by using a sample training set;

step 4: combining the gray prediction model and the support vector machine prediction model to obtain a combined optimization model;

step 5: and predicting the future NOx emission of the thermal generator set by adopting a combined optimization model.

In specific implementation, the corresponding technical measures also comprise:

obtaining relevant parameters of the thermal generator set within a period of time, respectively predicting by using a gray prediction model, a support vector machine prediction model and a combined optimization model, obtaining prediction data of each prediction model, comparing the prediction data of each prediction model, and verifying the prediction accuracy of the combined optimization model; the relevant parameters include: parameters of thermal power coal consumption, installed capacity, thermal power generation amount and NOx emission amount.

In step 1, the consumption of thermal power coal, the installed capacity and the thermal power generation amount are selected as factors influencing the NOx emission amount of the thermal power industry, wherein each factor is collected from daily data of a power plant, so that an original data sequence X based on the original data is obtained _i ⁽⁰⁾ As formula (1):

X _i ⁽⁰⁾ ＝{x _i ⁽⁰⁾ (1)，x _i ⁽⁰⁾ (2)，...x _i ⁽⁰⁾ (m)} (1)

in the formula (1):

i represents the i-th factor, i=1, 2,..n, n is the total amount of factors, each factor containing m state values;

in x _i ⁽⁰⁾ (1) A 1 st state value representing an i-th factor;

in x _i ⁽⁰⁾ (2) A 2 nd state value representing an i-th factor;

……

in x _i ⁽⁰⁾ (m) represents an mth state value of an ith factor.

In step 2, for the original data sequence X _i ⁽⁰⁾ Accumulating once according to an accumulation formula to obtain a once iteration data sequence X _i ⁽¹⁾ As formula (2):

X _i ⁽¹⁾ ＝{x _i ⁽¹⁾ (1)，x _i ⁽¹⁾ (2)，...x _i ⁽¹⁾ (m)} (2)

in the formula (2):

x _i ⁽¹⁾ (1) Is x _i ⁽⁰⁾ (1) Is used for the first iteration data of the (a);

x _i ⁽¹⁾ (2) Is x _i ⁽⁰⁾ (2) Is used for the first iteration data of the (a);

……

x _i ⁽¹⁾ (m) is x _i ⁽⁰⁾ The one iteration data of (m);

using an iterative data sequence X _i ⁽¹⁾ Establishing a first-order differential equation as shown in formula (3):

wherein alpha, b ₁ ，b ₂ ，...b _n-1 The parameter sequence is the parameter sequence of the gray prediction model;

expressing formula (3) in vector form; comprising the following steps: obtaining a parameter B according to the formula (4), and obtaining a parameter y according to the formula (5):

obtaining grey prediction model parameters by the calculation of the formula (6) according to the least square method

Using grey predictive model parametersThis builds a grey prediction model.

In step 3, a sample training set (x _i ，y _i )，x _i For input space, y _i For output values, a sample training set (x _i ，y _i ) And constructing an SVM support vector machine prediction model, and obtaining an output set Y of the NOx emission according to the input set X of the NOx emission by utilizing the SVM support vector machine prediction model.

In step 4, combining the gray prediction model and the support vector machine prediction model to obtain a combined optimization model as shown in formula (7):

in the formula (7):

at f _x Representing a predicted value of the optimized combined prediction model;

at f ₁ Predicted values expressed as gray prediction models; at f ₂ Representing a predicted value of the support vector machine model;

e is as follows ₁ Representing a prediction error of the gray prediction model; e is as follows ₂ Representing support vector machine modelIs a prediction error of (2);

with D (e) ₁ ) Representation e ₁ Variance of the values, expressed as D (e ₂ ) Representation e ₂ Variance of values.

Taking the weight of the predicted values of the gray predicted model and the support vector machine model in the predicted values of the optimized combined predicted model as omega ₁ And omega ₂ And has:

ω ₁ +ω ₂ ＝1 (8)

ω ₁ f ₁ +ω ₂ f ₂ ＝f _x (9)

since the gray prediction model and the support vector machine model are two independent models, cov (e ₁ ，e ₂ ) =0; then there are:

according to the weights represented by the formula (10) and the formula (11) of ω ₁ And omega ₂ Obtaining a combinatorial optimization model characterized by formula (7) according to formula (9);

since the expression (12) is established, it can be seen that the prediction accuracy of the combined model is higher than that of the gray prediction model and the support vector machine model.

D(e _z ) _min ≤min[D(e ₁ )，D(e ₁ )] (12)

In the formula (12):

e is as follows _z Representing a prediction error of the combined optimization model; with D (e) _z ) Representation e _z Variance of values.

Examples: in grey prediction, respectively using four sequences of x1, x2, x3 and x4 to represent NOx emission and four influencing factors; in support vector machine prediction, four influencing factors are taken as an input set of NOx emission, namely X= { X1, X2, X3, X4}, and the NOx emission is taken as an output set;

firstly, a GM (1, 1) model of each sample data is built according to the original sample data, then a GM (1, N) prediction model of the multi-factor influence of the NOx emission is built, and the NOx emission of a detected area is predicted according to the model.

Selecting annual thermal power coal consumption, installed capacity, thermal power generation amount and annual NOx emission of thermal power in the region 1998-2008 as shown in table 1

TABLE 1

GM (1, 1) model of thermal power NOx emissions:

GM (1, 1) model of coal consumption:

GM (1, 1) model of installed capacity:

GM (1, 1) model of thermal power generation:

the ash parameter sequence of the GM (1, 1) model of the thermal power NOx emission is as follows:

from this, the whitening first-order differential equation of the GM (1, 1) model of thermal power NOx emission can be found:

substituting the gray parameters into the whitened first-order differential equation of the model can obtain the time response function of the GM (1, 1) model:

in the middle of

And obtaining a predicted value of the thermal power NOx emission through accumulation reduction.

The predictions of the GM (1, 1) model and the support vector machine model are shown in table 2:

TABLE 2

According to the errors of the two models in table 2, the variance of the corresponding errors is calculated as follows:

D(e ₁ )＝9.3427，D(e ₂ )＝5.8456

the specific weights of the GM (1, N) model and the gray support vector machine model (SVM) obtained by the method in the combined optimization model are respectively as follows:

w ₂ ＝1-w ₁ ＝0.6151

according to the weight, a prediction formula of the optimized combined prediction model is obtained:

f _z ＝w ₁ f ₁ +w ₂ f ₂ ＝0.3849f ₁ +0.6151f ₂ ；

in combination, the thermal power NOx emission amount in 1999-2007 is predicted, and the result is shown in Table 2, so that the predicted result of the optimized combination prediction model is superior to that of the independent GM (1, N) model and SVM model.

The foregoing has shown and described the basic principles, principal features and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined in the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (6)

1. A NOx emission monitoring method of a thermal generator set is characterized by comprising the following steps:

step 1: acquiring original data of NOx emission of a thermal power generating unit;

step 2: establishing a GM (1, N) gray prediction model for predicting NOx emission by using the raw data;

step 3: constructing an SVM support vector machine prediction model by using a sample training set;

step 4: combining the gray prediction model and a support vector machine prediction model to obtain a combined optimization model;

step 5: and predicting the future NOx emission of the thermal generator set by adopting the combined optimization model.

2. The NOx emission monitoring method of a thermal power generation unit according to claim 1, characterized in that

Taking relevant parameters of the thermal generator set in a period of time, respectively predicting by using a gray prediction model, a support vector machine prediction model and a combined optimization model to obtain prediction data of each prediction model, comparing the prediction data of each prediction model, and verifying the prediction accuracy of the combined optimization model;

the relevant parameters include: parameters of thermal power coal consumption, installed capacity, thermal power generation amount and NOx emission amount.

3. The NOx emission monitoring method of a thermal power generation unit according to claim 1, characterized in that:

in the step 1, the consumption of thermal power coal, the installed capacity and the thermal power generation amount are selected as factors influencing the NOx emission amount of the thermal power industry, and each factor is collected from daily data of a power plant, so that an original data sequence X based on the original data is obtained _i ⁽⁰⁾ As formula (1):

X _i ⁽⁰⁾ ＝{x _i ⁽⁰⁾ (1),x _i ⁽⁰⁾ (2),...x _i ⁽⁰⁾ (m)} (1)

in the formula (1):

i represents the i-th factor, i=1, 2,..n, n is the total amount of factors, each factor containing m state values;

in x _i ⁽⁰⁾ (1) A 1 st state value representing an i-th factor;

in x _i ⁽⁰⁾ (2) A 2 nd state value representing an i-th factor;

……

in x _i ⁽⁰⁾ (m) represents an mth state value of an ith factor.

4. A method of monitoring NOx emissions from a thermal power plant according to claim 3, characterized by:

in said step 2, for said original data sequence X _i ⁽⁰⁾ Accumulating once according to an accumulation formula to obtain a once iteration data sequence X _i ⁽¹⁾ As formula (2):

X _i ⁽¹⁾ ＝{x _i ⁽¹⁾ (1)，x _i ⁽¹⁾ (2)，...x _i ⁽¹⁾ (m)} (2)

in the formula (2):

x _i ⁽¹⁾ (1) Is x _i ⁽⁰⁾ (1) Is used for the first iteration data of the (a);

x _i ⁽¹⁾ (2) Is x _i ⁽⁰⁾ (2) Is used for the first iteration data of the (a);

……

x _i ⁽¹⁾ (m) is x _i ⁽⁰⁾ The one iteration data of (m);

using an iterative data sequence X _i ⁽¹⁾ Establishing a first-order differential equation as shown in formula (3):

wherein alpha, b ₁ ,b ₂ ,...b _n-1 The parameter sequence is the parameter sequence of the gray prediction model;

expressing formula (3) in vector form; comprising the following steps: obtaining a parameter B according to the formula (4), and obtaining a parameter y according to the formula (5):

obtaining grey prediction model parameters by the calculation of the formula (6) according to the least square method

Using the gray prediction model parametersThis builds a grey prediction model.

5. The NOx emission monitoring method of a thermal power generation unit according to claim 1, characterized in that:

in said step 3, a sample is setTraining set (x) _i ,y _i )，x _i For input space, y _i For output values, the sample training set (x _i ,y _i ) And constructing an SVM support vector machine prediction model, and obtaining an output set Y of the NOx emission according to the input set X of the NOx emission by utilizing the SVM support vector machine prediction model.

6. The NOx emission monitoring method of a thermal power generation unit according to claim 1, characterized in that:

in the step 4, the gray prediction model and the support vector machine prediction model are combined to obtain a combined optimization model as shown in formula (7):

in the formula (7):

at f _x Representing a predicted value of the optimized combined prediction model;

at f ₁ Predicted values expressed as gray prediction models; at f ₂ Representing a predicted value of the support vector machine model;

e is as follows ₁ Representing a prediction error of the gray prediction model; e is as follows ₂ Representing a prediction error of the support vector machine model;

with D (e) ₁ ) Representation e ₁ Variance of the values, expressed as D (e ₂ ) Representation e ₂ Variance of values.