CN105975799A - Method and system for calculating carbon emissions - Google Patents

Abstract

The invention discloses a method and system for calculating carbon emissions. A plurality of sets of preset carbon emission influence parameters and corresponding carbon emission data are normalized, and a plurality of sets of normalized data are obtained, wherein the carbon emission influence parameters comprise unit loads, the primary air rate, boiler frequency and the like; the normalized data is divided into a training sample and a testing sample; the BP neural network is trained based on the training sample and a preset nerve cell transfer function, and a network training model is obtained; the network training model is tested according to the testing sample, and whether the test result is within a preset range or not is acquired and judged; if not, the step of acquiring the network training model is executed again; if yes, the network training model is determined as a BP neural network calculating model, and the test result is anti-normalized; based on the BP neural network calculating model, calculation is carried out according to the current carbon emission influence parameters, and the corresponding anti-normalization carbon emission calculated value is obtained. Based on the method, the accuracy of the carbon emission calculating result is improved.

Description

A kind of carbon emission amount calculates method and system

Technical field

The present invention relates to field of measuring technique, particularly relate to a kind of carbon emission amount and calculate method and system.

Background technology

Global Greenhouse Effect aggravation in this year, the temperature rises, the natural calamity such as Melting Glaciers, Marine Storm Genesis, desertification Evil constantly manifests.According to international environmental protection tissue data, compared to 20 years former world temperature to rise 0.75 DEG C, its chief-criminal brings disaster upon The content of the carbon dioxide in head i.e. air rises year by year, wherein enters the carbon dioxide of environment with boiler smoke and accounts for total amount 30%.For with Fossil fuel for China of the main body of energy resource consumption, carbon emission amount computational problem has become as everybody altogether Same focus of attention.

Calculating about carbon emission amount is all to utilize empirical equation to estimate roughly at present, have ignored coal-fired quality, pot The factor impacts on carbon emission such as the efficiency of furnace, unit load, primary air ratio and excess air coefficient, cause calculating error the biggest.

From the foregoing, existing technical scheme cannot ensure the accurate of result of calculation when calculating carbon emission amount Degree.

Summary of the invention

In view of this, the present invention provides a kind of carbon emission amount to calculate method and system, exists solving existing technical scheme The problem that cannot ensure the accuracy of budget result when carbon emission amount is carried out budget.Technical scheme is as follows:

A kind of carbon emission amount computational methods, are applied to carbon emission amount and calculate system, including:

Obtain the carbon emission amount data of each self-generating under the many groups carbon emission amount affecting parameters preset, by described many group carbon rows High-volume affecting parameters and corresponding carbon emission amount data are normalized, and obtain organizing normalization data more；

Described many group normalization datas are divided into training sample and test sample；

Based on described training sample and default neural transferring function f (x), BP neutral net is trained, obtains network Training pattern；

According to described test sample, described network training model is tested, obtain and judge that whether test result is in advance If in scope；

If it is not, return perform described based on described training sample and default neural transferring function f (x) to BP neutral net It is trained, the step for of obtaining network training model；

If so, described network training model is defined as BP calculating model of neural networks, and returns counter for described test result One changes；

Based on described BP calculating model of neural networks, calculate according to current carbon emission amount affecting parameters, obtain corresponding The carbon emission amount value of calculation of renormalization.

Preferably, described based on described training sample and default neural transferring function f (x), BP neutral net is instructed Practice, obtain network training model, including:

Build carbon emission neuron computational mathematics modelWherein, ω_pqArrive for neuron p The weighted value of neuron q, x_pT () represents the input information from neuron p that t neuron q receives, f (x) is described pre- If neural transferring function；

Based on BP neutral net, respectively organize normalization carbon emission amount affecting parameters as institute using what described training sample comprised State the input layer vector X=(x of BP neutral net₁,x₂,…,x_i,…,x_n)^T, wherein, x_iIt it is one group of normalization carbon emission amount impact I-th in parameter n；

Based on each group normalization carbon emission comprised in described carbon emission neuron computational mathematics model, described training sample The input layer vector X=(x that amount affecting parameters is corresponding₁,x₂,…,x_i,…,x_n)^T, obtain the hidden layer vector Y of described BP neutral net =(y₁,y₂,…,y_j,…,y_m)^T, wherein, y_jIt is vectorial for the jth neuron in hidden layer m neuron,v_ijWeighted value for input layer i-th neuron to hidden layer jth neuron；

Based on described hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith described default neural transferring function f (x), meter The carbon emission amount of the output layer output calculating described BP neutral net corresponding to described each group of normalization carbon emission amount affecting parameters is pre- Calculation valueWherein, ω_jkFor hidden layer jth neuron to output layer kth neuron weighted value, k's Value is 1；

According to each group normalization carbon emission amount data comprised in described training sample, obtain described each group of normalization carbon row The actual carbon emission value b that high-volume affecting parameters is corresponding, and it is corresponding to calculate described each group of normalization carbon emission amount affecting parameters Output error e, e=1/2 (b-a)²；

Take the quadratic sum of output error e corresponding to described each group of normalization carbon emission amount affecting parameters as parameter error, And judge that described parameter error is whether in default error allowed band；

If so, described BP neutral net is defined as network training model；

If it is not, according to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described input layer vector X=(x₁,x₂,…,x_i,…,x_n)^T, described hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f X (), calculates input layer i-th neuron corresponding to described each group of normalization carbon emission amount affecting parameters neural to hidden layer jth Weighted value correction amount v of unit_ijWith hidden layer jth neuron to output layer kth neuron weighted value correction amount w_jk, and Described BP neutral net is modified, revised BP neutral net is defined as network training model.

Preferably, the span of described hidden layers numbers is the integer more than 0, specially 1.

Preferably, described output error e corresponding according to described each group of normalization carbon emission amount affecting parameters, described input Layer vector X=(x₁, x₂..., x_i..., x_n)^T, described hidden layer vector Y=(y₁, y₂..., y_j..., y_m)^TPass with default neuron Delivery function f (x), calculates input layer i-th neuron corresponding to described each group of normalization carbon emission amount affecting parameters to hidden layer jth Weighted value correction amount v of individual neuron_ijWith hidden layer jth neuron to output layer kth neuron weighted value correction amount w_jk, including:

According to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described input layer vector X= (x₁,x₂,…,x_i,…,x_n)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission amount impact ginseng The input layer i-th neuron of number correspondence is to weighted value correction amount v of hidden layer jth neuron_ij,Wherein,For corresponding defeated of described each group of normalization carbon emission amount affecting parameters Entering layer to hidden layer error signal, η is default learning rate, and the value of η is the constant between 0 to 1；

According to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described hidden layer vector Y=(y₁, y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission amount affecting parameters Corresponding hidden layer jth neuron is to output layer kth neuron weighted value correction amount w_jk,Wherein,For the hidden layer that described each group of normalization carbon emission amount affecting parameters is corresponding To output layer error signal.

Preferably, described default neural transferring function f (x) is function sigmoid, accordingly, and hidden layer modified weight amount Δw_jk=-η a (1-a) y_j, output layer modified weight amount Δ v_ij=-η (b-a) a (1-a) w_ijy_j(1-y_j)x_i。

A kind of carbon emission amount calculates system, including:

Normalization module, for obtaining the carbon emission amount number of each self-generating under default many groups carbon emission amount affecting parameters According to, described many group carbon emission amount affecting parameters and corresponding carbon emission amount data are normalized, obtain organizing normalizing more Change data；

Sample divides module, for described many group normalization datas are divided into training sample and test sample；

Network training model acquisition module, for based on described training sample and default neural transferring function f (x) to BP Neutral net is trained, and obtains network training model；

Obtain judge module, for described network training model being tested according to described test sample, obtain and sentence Whether disconnected test result is in preset range；If it is not, send the first execution signal to the first control module；If so, to the second control Molding block sends the second execution signal；

Described first control module, is used for receiving described first and performs signal, return described in performing based on described training sample BP neutral net is trained by this and default neural transferring function f (x), the step for of obtaining network training model；

Described second control module, is used for receiving described second and performs signal, described network training model is defined as BP Calculating model of neural networks, and by described test result renormalization；

Computing module, for based on described BP calculating model of neural networks, is carried out according to current carbon emission amount affecting parameters Calculate, obtain the carbon emission amount value of calculation of corresponding renormalization.

Preferably, described network training model acquisition module includes:

Carbon emission neuron computational mathematics model construction unit, is used for building carbon emission neuron computational mathematics modelWherein, ω_pqFor the weighted value of neuron p to neuron q, x_pT () represents t neuron q The input information from neuron p received, f (x) is described default neural transferring function；

Input layer acquiring unit, for based on BP neutral net, each group normalization carbon that will comprise in described training sample Discharge capacity affecting parameters is as the input layer vector X=(x of described BP neutral net₁,x₂,…,x_i,…,x_n)^T, wherein, x_iIt is one Group normalization carbon emission amount affecting parameters_nI-th in individual；

Hidden layer acquiring unit, for comprising based in described carbon emission neuron computational mathematics model, described training sample Input layer vector X=(x corresponding to each group normalization carbon emission amount affecting parameters₁,x₂,…,x_i,…,x_n)^T, obtain described BP The hidden layer vector Y=(y of neutral net₁,y₂,…,y_j,…,y_m)^T, wherein, y_jFor the jth neuron in hidden layer m neuron Vector,v_ijWeighted value for input layer i-th neuron to hidden layer jth neuron；

Output layer acquiring unit, for based on described hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith described default god Through unit's transmission function f (x), calculate the output of described BP neutral net corresponding to described each group of normalization carbon emission amount affecting parameters The carbon emission amount estimated value of layer outputWherein, ω_jkFor hidden layer jth neuron to output layer kth Individual neuron weighted value, the value of k is 1；

Output error computing unit, for respectively organizing normalization carbon emission amount data according to what described training sample comprised, Obtain the actual carbon emission value b that described each group of normalization carbon emission amount affecting parameters is corresponding, and calculate described each group of normalization Corresponding output error e of carbon emission amount affecting parameters, e=1/2 (b-a)²；

Parameter error obtains judging unit, for taking output corresponding to described each group of normalization carbon emission amount affecting parameters by mistake The quadratic sum of difference e is as parameter error, and judges that described parameter error is whether in default error allowed band；If so, to One control unit sends the 3rd execution signal；If it is not, send the 4th execution signal to the second control unit；

Described first control unit, is used for receiving the described 3rd and performs signal, described BP neutral net is defined as network Training pattern；

Described second control unit, is used for receiving the described 4th and performs signal, according to described each group of normalization carbon emission amount Output error e that affecting parameters is corresponding, described input layer vector X=(x₁,x₂,…,x_i,…,x_n)^T, described hidden layer vector Y= (y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission amount affecting parameters Corresponding input layer i-th neuron is to weighted value correction amount v of hidden layer jth neuron_ijArrive with hidden layer jth neuron Output layer kth neuron weighted value correction amount w_jk, and described BP neutral net is modified, revised BP is neural Network is defined as network training model.

Preferably, described second control unit includes:

First computing unit, for output error e corresponding according to described each group of normalization carbon emission amount affecting parameters, institute State input layer vector X=(x₁, x₂..., x_i..., x_n)^TWith default neural transferring function f (x), calculate described each group of normalization Input layer i-th neuron corresponding to carbon emission amount affecting parameters is to weighted value correction amount v of hidden layer jth neuron_ij,Wherein,For corresponding defeated of described each group of normalization carbon emission amount affecting parameters Entering layer to hidden layer error signal, η is default learning rate, and the value of η is the constant between 0 to 1；

Second computing unit, for output error e corresponding according to described each group of normalization carbon emission amount affecting parameters, institute State hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate described each group of normalization carbon Hidden layer jth neuron corresponding to discharge capacity affecting parameters is to output layer kth neuron weighted value correction amount w_jk,Wherein,For hidden layer corresponding to described each group of normalization carbon emission amount affecting parameters to defeated Go out a layer error signal.

Comparing and prior art, what the present invention realized has the beneficial effect that

A kind of carbon emission amount the most provided by the present invention calculates method and system, and system is by obtaining the many groups carbon preset The carbon emission amount data of each self-generating under discharge capacity affecting parameters, will organize carbon emission amount affecting parameters and corresponding carbon emission amount more Data are normalized, and obtain organizing normalization data more；It is divided into training sample and test specimens by organizing normalization data more This；Based on training sample and default neural transferring function, BP neutral net is trained, obtains network training model；Foundation Network training model is tested by test sample, obtains and judges that test result is whether in preset range；Hold if it is not, return Row based on training sample and default neural transferring function f (x), BP neutral net is trained, obtain network training model this One step；If so, network training model is defined as BP calculating model of neural networks, and by test result renormalization；Based on BP calculating model of neural networks, calculates according to current carbon emission amount affecting parameters, obtains the carbon emission of corresponding renormalization Amount value of calculation.Based on above-mentioned carbon emission amount computational methods, improve the accuracy of result when carbon emission amount is calculated.

Accompanying drawing explanation

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing In having technology to describe, the required accompanying drawing used is briefly described, it should be apparent that, the accompanying drawing in describing below is only this Inventive embodiment, for those of ordinary skill in the art, on the premise of not paying creative work, it is also possible to according to The accompanying drawing provided obtains other accompanying drawing.

Fig. 1 is the flow chart of a kind of carbon emission amount computational methods disclosed in the embodiment of the present invention one；

Fig. 2 is the flow chart of a kind of carbon emission amount computational methods disclosed in the embodiment of the present invention two；

Fig. 3 is the flow chart of another kind of carbon emission amount computational methods disclosed in the embodiment of the present invention two；

Fig. 4 is the structural representation that a kind of carbon emission amount disclosed in the embodiment of the present invention three calculates system；

Fig. 5 is the structural representation that a kind of carbon emission amount disclosed in the embodiment of the present invention four calculates system；

Fig. 6 is the structural representation that another kind of carbon emission amount disclosed in the embodiment of the present invention four calculates system.

Detailed description of the invention

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete Describe, it is clear that described embodiment is only a part of embodiment of the present invention rather than whole embodiments wholely.Based on Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under not making creative work premise Embodiment, broadly falls into the scope of protection of the invention.

Embodiment one

A kind of carbon emission amount computational methods disclosed in the embodiment of the present invention, are applied to carbon emission amount and calculate system, flow chart As it is shown in figure 1, carbon emission amount computational methods include:

S101, obtains the carbon emission amount data of each self-generating under the many groups carbon emission amount affecting parameters preset, will organize carbon more Discharge capacity affecting parameters and corresponding carbon emission amount data are normalized, and obtain organizing normalization data more；

Performing during step S101, the carbon emission amount affecting parameters preset includes: unit load, primary air ratio, Fixed carbon content and volatile matter content and fineness of pulverized coal etc. in boiler efficiency, excess air coefficient, fuel, carried out multi-group data Normalized, is i.e. limited in [0,1].

S102, is divided into training sample and test sample by organizing normalization data more；

Performing during step S102, it is preferred that be training sample and survey by organizing normalization data random division more Sample is originally.

S103, is trained BP neutral net based on training sample and default neural transferring function f (x), obtains net Network training pattern；

Preferably, presetting neural transferring function f (x) is function sigmoid；

S104, tests network training model according to test sample, obtains and judges whether test result is being preset In the range of；

S105, performs to enter BP neutral net based on training sample and default neural transferring function f (x) if it is not, return Row training, the step for of obtaining network training model；

S106, is if so, defined as BP calculating model of neural networks by network training model, and by anti-for test result normalizing Change；

S107, based on BP calculating model of neural networks, calculates according to current carbon emission amount affecting parameters, obtains corresponding The carbon emission amount value of calculation of renormalization；

During performing step S107, based on fixed BP calculating model of neural networks, by current carbon emission amount Affecting parameters is normalized, and obtains normalization current carbon emission amount affecting parameters；By current for normalization carbon emission amount shadow Ring parameter vectorial as BP calculating model of neural networks input layer, obtain BP calculating model of neural networks output layer output result, And output result is carried out renormalization process, renormalization is exported result corresponding as current carbon emission amount affecting parameters Carbon emission amount value of calculation.

Carbon emission amount computational methods disclosed in the embodiment of the present invention, by obtaining the many groups carbon emission amount affecting parameters preset Under the carbon emission amount data of each self-generating, carbon emission amount affecting parameters will be organized and corresponding carbon emission amount data are normalized more Process, obtain organizing normalization data more；It is divided into training sample and test sample by organizing normalization data more；Based on training sample With default neural transferring function, BP neutral net is trained, obtains network training model；According to test sample to network Training pattern is tested, and obtains and judges that test result is whether in preset range；Perform based on training sample if it is not, return With default neural transferring function f (x), BP neutral net is trained, the step for of obtaining network training model；If so, will Network training model is defined as BP calculating model of neural networks, and by test result renormalization；Based on BP neural computing Model, calculates according to current carbon emission amount affecting parameters, obtains the carbon emission amount value of calculation of corresponding renormalization.

Embodiment two

Based on carbon emission amount computational methods a kind of disclosed in the invention described above embodiment one, step as illustrated in FIG. 1 In S103, based on training sample and default neural transferring function f (x), BP neutral net is trained, obtains network training The process that specifically performs of model, as in figure 2 it is shown, comprise the steps:

S201, builds carbon emission neuron computational mathematics model Y_j；

This carbon emission neuron computational mathematics model Y in step s 201_jIt is specially formula (1)；

$Y_j=f\left({\mathrm{net}}_j\right)=f\[\underset{p=0}{\overset{n}{\Σ}}{\mathrm{\ω}}_{pq}{x}_p\left(t\right)\]---\left(1\right)$

Wherein, ω_pqFor the weighted value of neuron p to neuron q, x_p(t) represent t neuron q receive from god Through the input information of unit p, f (x) is for presetting neural transferring function；

S202, based on BP neutral net, using each group normalization carbon emission amount affecting parameters of comprising in training sample as The input layer vector X=(x of BP neutral net₁,x₂,…,x_i,…,x_n)^T, wherein, x_iIt it is one group of normalization carbon emission amount impact ginseng I-th in number n；

During performing step S202, using the many groups normalization carbon emission amount affecting parameters in training sample as defeated Entering layer vector to be simultaneously entered in BP neutral net, wherein, each group input layer vector is designated as X=(x₁,x₂,…,x_i,…,x_n)^T, can Choosing, make x for introducing threshold value to hidden neuron₀=0.

S203, based on each group normalization carbon emission amount comprised in carbon emission neuron computational mathematics model, training sample The input layer vector X=(x that affecting parameters is corresponding₁,x₂,…,x_i,…,x_n)^T, obtain the hidden layer vector Y=(y of BP neutral net₁, y₂,…,y_j,…,y_m)^T, wherein, y_jFor the jth neuron vector in hidden layer m neuron；

The jth neuron vector y in this hidden layer m neuron in step S203_jIt is specially formula (2)；

$y_j=f\left({\mathrm{net}}_j\right)=f\left(\underset{i=0}{\overset{n}{\Σ}}{v}_{ij}{x}_i\right)---\left(2\right)$

Wherein, v_ijWeighted value for input layer i-th neuron to hidden layer jth neuron；

During performing step S203, hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^T；Optionally, for giving output Layer neuron introduces threshold value and makes y₀=0；Preferably, the acquisition mode of hidden layer m neuron is trial and error procedure, chooses network training by mistake Difference and network training number of times combination hidden neuron number corresponding to optimum；Preferably, the quantity of hidden layer is 1, this be because of Error can be reduced for increasing hidden layers numbers, improve precision, but network also can be made to complicate simultaneously, thus increase network weight weight values Training time.

S204, based on hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate each The carbon emission amount estimated value a of the output layer output of the BP neutral net that group normalization carbon emission amount affecting parameters is corresponding；

In step S204, this carbon emission amount estimated value a is specially formula (3)；

$a=f\left({\mathrm{net}}_k\right)=f\left(\underset{j=0}{\overset{m}{\Σ}}{\mathrm{\ω}}_{jk}{y}_j\right)---\left(3\right);$

Wherein, ω_jkIt is 1 for hidden layer jth neuron to output layer kth neuron weighted value, the value of k；

S205, according to each group normalization carbon emission amount data comprised in training sample, obtains each group of normalization carbon emission Amount actual carbon emission value b corresponding to affecting parameters, and calculate each group of normalization carbon emission amount affecting parameters and export accordingly by mistake Difference e,

In step S205, this is respectively organized normalization carbon emission corresponding output error e of amount affecting parameters and is specially formula (4)；

E=1/2 (b-a)²(4)；

S206, takes the quadratic sum of each group of output error e corresponding to normalization carbon emission amount affecting parameters as parameter error, And judge that parameter error is whether in default error allowed band；

S207, is if so, defined as network training model by BP neutral net；

S208, if it is not, according to output error e corresponding to each group of normalization carbon emission amount affecting parameters, input layer vector X= (x₁,x₂,…,x_i,…,x_n)^T, hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), meter Calculate each group of input layer i-th neuron corresponding to the normalization carbon emission amount affecting parameters weighted value to hidden layer jth neuron Correction amount v_ijWith hidden layer jth neuron to output layer kth neuron weighted value correction amount w_jk, and to BP nerve net Network is modified, and revised BP neutral net is defined as network training model.

Carbon emission amount computational methods disclosed in the embodiment of the present invention, according to training sample and default neural transferring function f X BP neutral net is trained by (), the network training model needed for acquisition, ensure that meter when calculating carbon emission amount Calculate the accuracy of result.

Based on carbon emission amount computational methods a kind of disclosed in the invention described above embodiment two, step as illustrated in FIG. 2 In S208, according to output error e corresponding to each group of normalization carbon emission amount affecting parameters, input layer vector X=(x₁,x₂,…, x_i,…,x_n)^T, hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate each group of normalizing Change input layer i-th neuron corresponding to carbon emission amount affecting parameters weighted value correction amount v to hidden layer jth neuron_ij With hidden layer jth neuron to output layer kth neuron weighted value correction amount w_jkThe process that specifically performs, as it is shown on figure 3, Comprise the steps:

S301, according to output error e corresponding to each group of normalization carbon emission amount affecting parameters, input layer vector X=(x₁, x₂,…,x_i,…,x_n)^TWith default neural transferring function f (x), calculate each group of normalization carbon emission amount affecting parameters corresponding Input layer i-th neuron is to weighted value correction amount v of hidden layer jth neuron_ij；

In step S301, this input layer i-th neuron is to weighted value correction amount v of hidden layer jth neuron_ijTool Body is formula (5)；

${\mathrm{\Δv}}_{ij}=-\mathrm{\η}\frac{\∂e}{\∂w_{ij}}=-{\mathrm{\ηf}}^{\′}\left({\mathrm{net}}_j\right)x_i={\mathrm{\η\δ}}_j^0{x}_i---\left(5\right)$

Wherein,For input layer corresponding to each group of normalization carbon emission amount affecting parameters to hidden layer error signal, η is default Learning rate, the value of η is the constant between 0 to 1；

During performing step S301, it is preferred that presetting neural transferring function f (x) is function sigmoid, phase Answer, weighted value correction amount v of this input layer i-th neuron to hidden layer jth neuron_ij=-η (b-a) a (1-a) w_ijy_j(1-y_j)x_i。

S302, according to output error e corresponding to each group of normalization carbon emission amount affecting parameters, hidden layer vector Y=(y₁, y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate each group of normalization carbon emission amount affecting parameters corresponding Hidden layer jth neuron is to output layer kth neuron weighted value correction amount w_jk；

This hidden layer jth neuron is to output layer kth neuron weighted value correction amount w in step s 302_jkSpecifically For formula (6)；

${\mathrm{\Δw}}_{jk}=-\mathrm{\η}\frac{\∂e}{\∂w_{jk}}=-{\mathrm{\ηf}}^{\′}\left({\mathrm{net}}_k\right)y_j={\mathrm{\η\δ}}_i^0{y}_j---\left(6\right)$

Wherein,For hidden layer corresponding to each group of normalization carbon emission amount affecting parameters to output layer error signal；

During performing step S302, it is preferred that presetting neural transferring function f (x) is function sigmoid, phase Answering, hidden layer jth neuron is to output layer kth neuron weighted value correction amount w_jk=-η a (1-a) y_j。

Carbon emission amount computational methods disclosed in the embodiment of the present invention, choose test sample and enter the network training model obtained Row test, further ensures network training model close to required BP calculating model of neural networks, is carrying out carbon emission amount The accuracy of result of calculation is ensure that during calculating.

Embodiment three

The carbon emission amount computational methods provided based on each embodiment of the invention described above, the present embodiment three then correspondence discloses to be held The carbon emission amount of the above-mentioned carbon emission amount computational methods of row calculates system, and as shown in Figure 4, carbon emission amount calculates system to its structural representation System 400 includes:

Normalization module 401, for obtaining the carbon emission amount of each self-generating under default many groups carbon emission amount affecting parameters More data, will organize carbon emission amount affecting parameters and corresponding carbon emission amount data are normalized, obtain organizing normalization more Data；

Sample divides module 402, for being divided into training sample and test sample by organizing normalization data more；

Network training model acquisition module 403, for based on training sample and default neural transferring function f (x) to BP Neutral net is trained, and obtains network training model；

Obtain judge module 404, for network training model being tested according to test sample, obtain and judge test Whether result is in preset range；If it is not, send the first execution signal to the first control module；If so, to the second control module Send the second execution signal；

First control module 405, for receiving the first execution signal, returns and performs based on training sample and default neuron BP neutral net is trained by transmission function f (x), the step for of obtaining network training model；

Second control module 406, for receiving the second execution signal, is defined as BP neutral net meter by network training model Calculate model, and by test result renormalization；

Computing module 407, for based on BP calculating model of neural networks, counts according to current carbon emission amount affecting parameters Calculate, obtain the carbon emission amount value of calculation of corresponding renormalization.

Carbon emission amount disclosed in the embodiment of the present invention calculates system, including: normalization module, obtain the many groups carbon row preset The high-volume carbon emission amount data of each self-generating under affecting parameters, will organize carbon emission amount affecting parameters and corresponding carbon emission amount number more According to being normalized, obtain organizing normalization data more；Sample divides module, is divided into training sample by organizing normalization data more Basis and test sample；Network training model acquisition module, refreshing to BP based on training sample and default neural transferring function f (x) It is trained through network, obtains network training model；Obtain judge module, according to test sample, network training model is surveyed Examination, obtains and judges that test result is whether in preset range；If it is not, send the first execution signal to the first control module；If It is to send the second execution signal to the second control module；First control module, receive first execution signal, return perform based on BP neutral net is trained by training sample and default neural transferring function f (x), obtains this step of network training model Suddenly；Second control module, receives the second execution signal, and network training model is defined as BP calculating model of neural networks, and will Test result renormalization；Computing module, based on BP calculating model of neural networks, is carried out according to current carbon emission amount affecting parameters Calculate, obtain the carbon emission amount value of calculation of corresponding renormalization.Calculate system based on carbon emission amount disclosed above, carbon is being arranged The accuracy of result of calculation is ensure that when high-volume calculating.

Embodiment four

Calculating system in conjunction with carbon emission amount disclosed in above-described embodiment three, the present embodiment four also discloses a kind of carbon emission amount Calculating system, its structural representation as it is shown in figure 5,

Wherein, network training model acquisition module 403 includes:

Carbon emission neuron computational mathematics model construction unit 501, is used for building carbon emission neuron computational mathematics modelWherein, ω_pqFor the weighted value of neuron p to neuron q, x_pT () represents t neuron q The input information from neuron p received, f (x) is for presetting neural transferring function；

Input layer acquiring unit 502, for based on BP neutral net, each group normalization carbon row that will comprise in training sample High-volume affecting parameters is as the input layer vector X=(x of BP neutral net₁,x₂,…,x_i,…,x_n)^T, wherein, x_iIt it is one group of normalizing Change the i-th in carbon emission amount affecting parameters n；

Hidden layer acquiring unit 503, for based on each group comprised in carbon emission neuron computational mathematics model, training sample The input layer vector X=(x that normalization carbon emission amount affecting parameters is corresponding₁,x₂,…,x_i,…,x_n)^T, obtain BP neutral net Hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^T, wherein, y_jIt is vectorial for the jth neuron in hidden layer m neuron,v_ijWeighted value for input layer i-th neuron to hidden layer jth neuron；

Output layer acquiring unit 504, for based on hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TPass with default neuron Delivery function f (x), calculates the carbon emission of the output layer output of each group of BP neutral net corresponding to normalization carbon emission amount affecting parameters Amount estimated valueWherein, ω_jkFor hidden layer jth neuron to output layer kth neuron weight Value, the value of k is 1；

Output error computing unit 505, for according to each group normalization carbon emission amount data comprised in training sample, obtaining Take each group of actual carbon emission value corresponding to normalization carbon emission amount affecting parameters^b, and calculate each group of normalization carbon emission amount shadow Ring corresponding output error e of parameter, e=1/2 (b-a)²；

Parameter error obtains judging unit 506, for taking each group of output corresponding to normalization carbon emission amount affecting parameters by mistake The quadratic sum of difference e is as parameter error, and judges that parameter error is whether in default error allowed band；If so, to the first control Unit processed sends the 3rd execution signal；If it is not, send the 4th execution signal to the second control unit；

First control unit 507, for receiving the 3rd execution signal, is defined as network training model by BP neutral net；

Second control unit 508, for receiving the 4th execution signal, according to each group of normalization carbon emission amount affecting parameters pair Output error e answered, input layer vector X=(x₁, x₂..., x_i..., x_n)^T, hidden layer vector Y=(y₁, y₂..., y_j..., y_m)^T With default neural transferring function f (x), calculate each group of input layer i-th corresponding to normalization carbon emission amount affecting parameters neural Unit is to weighted value correction amount v of hidden layer jth neuron_ijWith hidden layer jth neuron to output layer kth neuron weight Value correction amount w_jk, and BP neutral net is modified, revised BP neutral net is defined as network training model.

Carbon emission amount disclosed in the embodiment of the present invention calculates system, chooses training sample and neural transferring function f (x) is right BP neutral net is trained, the network training model needed for acquisition, ensure that calculating knot when calculating carbon emission amount The accuracy of fruit.

Calculating system in conjunction with carbon emission amount disclosed in above-described embodiment three, the present embodiment four also discloses another kind of carbon emission Amount calculating system, its structural representation as shown in Figure 6,

Wherein, the second control unit 508 includes:

First computing unit 601, for output error e corresponding according to each group of normalization carbon emission amount affecting parameters, defeated Enter layer vector X=(x₁,x₂,…,x_i,…,x_n)^TWith default neural transferring function f (x), calculate each group of normalization carbon emission amount Input layer i-th neuron corresponding to affecting parameters is to weighted value correction amount v of hidden layer jth neuron_ij,Wherein,For the input layer that each group of normalization carbon emission amount affecting parameters is corresponding To hidden layer error signal, η is for presetting learning rate, and the value of η is the constant between 0 to 1；

Second computing unit 602, for output error e corresponding according to each group of normalization carbon emission amount affecting parameters, hidden Layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate each group of normalization carbon emission amount shadow Ring hidden layer jth neuron corresponding to parameter to output layer kth neuron weighted value correction amount w_jk,Wherein,For hidden layer corresponding to each group of normalization carbon emission amount affecting parameters to defeated Go out a layer error signal.

Carbon emission amount disclosed in the embodiment of the present invention calculates system, chooses test sample and enters the network training model obtained Row test, further ensures network training model close to required BP calculating model of neural networks, is carrying out carbon emission amount The accuracy of result of calculation is ensure that during calculating.

It should be noted that each embodiment in this specification all uses the mode gone forward one by one to describe, each embodiment weight Point explanation is all the difference with other embodiments, and between each embodiment, identical similar part sees mutually. For device disclosed in embodiment, owing to it corresponds to the method disclosed in Example, so describe is fairly simple, phase See method part in place of pass to illustrate.

Also, it should be noted in this article, the relational terms of such as first and second or the like is used merely to one Entity or operation separate with another entity or operating space, and not necessarily require or imply between these entities or operation There is relation or the order of any this reality.And, term " includes ", " comprising " or its any other variant are intended to contain Comprising of lid nonexcludability, so that include the key element that the process of a series of key element, method, article or equipment are intrinsic, Or also include the key element intrinsic for these processes, method, article or equipment.In the case of there is no more restriction, The key element limited by statement " including ... ", it is not excluded that including the process of described key element, method, article or equipment In there is also other identical element.

Described above to the disclosed embodiments, makes professional and technical personnel in the field be capable of or uses the present invention. Multiple amendment to these embodiments will be apparent from for those skilled in the art, as defined herein General Principle can realize without departing from the spirit or scope of the present invention in other embodiments.Therefore, the present invention It is not intended to be limited to the embodiments shown herein, and is to fit to and principles disclosed herein and features of novelty phase one The widest scope caused.

Claims (8)

1. carbon emission amount computational methods, it is characterised in that be applied to carbon emission amount and calculate system, including:

Obtain the carbon emission amount data of each self-generating under the many groups carbon emission amount affecting parameters preset, by described many group carbon emission amounts Affecting parameters and corresponding carbon emission amount data are normalized, and obtain organizing normalization data more；

Described many group normalization datas are divided into training sample and test sample；

Based on described training sample and default neural transferring function f (x), BP neutral net is trained, obtains network training Model；

According to described test sample, described network training model is tested, obtain and judge that whether test result is at default model In enclosing；

Based on described training sample and default neural transferring function f (x), BP neutral net is carried out described in execution if it is not, return Training, the step for of obtaining network training model；

If so, described network training model is defined as BP calculating model of neural networks, and by described test result renormalization；

Based on described BP calculating model of neural networks, calculate according to current carbon emission amount affecting parameters, obtain the most counter returning The one carbon emission amount value of calculation changed.

Method the most according to claim 1, it is characterised in that described transmit based on described training sample and default neuron BP neutral net is trained by function f (x), obtains network training model, including:

Build carbon emission neuron computational mathematics modelWherein, ω_pqFor neuron p to neural The weighted value of unit q, x_pT () represents the input information from neuron p that t neuron q receives, f (x) is described default god Function is transmitted through unit；

Based on BP neutral net, using each group normalization carbon emission amount affecting parameters of comprising in described training sample as described BP The input layer vector X=(x of neutral net₁,x₂,…,x_i,…,x_n)^T, wherein, x_iIt it is one group of normalization carbon emission amount affecting parameters I-th in n；

Based on each group normalization carbon emission amount shadow comprised in described carbon emission neuron computational mathematics model, described training sample Ring the input layer vector X=(x that parameter is corresponding₁,x₂,…,x_i,…,x_n)^T, obtain the hidden layer vector Y=of described BP neutral net (y₁,y₂,…,y_j,…,y_m)^T, wherein, y_jIt is vectorial for the jth neuron in hidden layer m neuron,v_ijWeighted value for input layer i-th neuron to hidden layer jth neuron；

Based on described hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith described default neural transferring function f (x), calculate institute State the carbon emission amount estimated value of the output layer output of each group of described BP neutral net corresponding to normalization carbon emission amount affecting parametersWherein, ω_jkFor hidden layer jth neuron to output layer kth neuron weighted value, the value of k It is 1；

According to each group normalization carbon emission amount data comprised in described training sample, obtain described each group of normalization carbon emission amount The actual carbon emission value b that affecting parameters is corresponding, and calculate described each group of normalization carbon emission amount affecting parameters and export accordingly Error e, e=1/2 (b-a)²；

Take the quadratic sum of output error e corresponding to described each group of normalization carbon emission amount affecting parameters as parameter error, and sentence Whether disconnected described parameter error is in default error allowed band；

If so, described BP neutral net is defined as network training model；

If it is not, according to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described input layer vector X= (x₁,x₂,…,x_i,…,x_n)^T, described hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f X (), calculates input layer i-th neuron corresponding to described each group of normalization carbon emission amount affecting parameters neural to hidden layer jth Weighted value correction amount v of unit_ijWith hidden layer jth neuron to output layer kth neuron weighted value correction amount w_jk, and Described BP neutral net is modified, revised BP neutral net is defined as network training model.

Method the most according to claim 2, it is characterised in that the span of described hidden layers numbers is the integer more than 0, It is specially 1.

Method the most according to claim 2, it is characterised in that described according to described each group of normalization carbon emission amount impact ginseng Output error e that number is corresponding, described input layer vector X=(x₁,x₂,…,x_i,…,x_n)^T, described hidden layer vector Y=(y₁, y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission amount affecting parameters pair The input layer i-th neuron answered is to weighted value correction amount v of hidden layer jth neuron_ijWith hidden layer jth neuron to defeated Go out layer kth neuron weighted value correction amount w_jk, including:

According to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described input layer vector X=(x₁, x₂,…,x_i,…,x_n)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission amount affecting parameters pair The input layer i-th neuron answered is to weighted value correction amount v of hidden layer jth neuron_ij,Wherein,For corresponding defeated of described each group of normalization carbon emission amount affecting parameters Entering layer to hidden layer error signal, η is default learning rate, and the value of η is the constant between 0 to 1；

According to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described hidden layer vector Y=(y₁, y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission amount affecting parameters Corresponding hidden layer jth neuron is to output layer kth neuron weighted value correction amount w_jk,Wherein,For the hidden layer that described each group of normalization carbon emission amount affecting parameters is corresponding To output layer error signal.

Method the most according to claim 2, it is characterised in that described default neural transferring function f (x) is function Sigmoid, accordingly, hidden layer modified weight amount Δ w_jk=-η a (1-a) y_j, output layer modified weight amount Δ v_ij=-η (b-a) a (1-a)w_ijy_j(1-y_j)x_i。

6. a carbon emission amount calculates system, it is characterised in that including:

Normalization module, for obtaining the carbon emission amount data of each self-generating under default many groups carbon emission amount affecting parameters, will Described many group carbon emission amount affecting parameters and corresponding carbon emission amount data are normalized, and obtain organizing normalization number more According to；

Sample divides module, for described many group normalization datas are divided into training sample and test sample；

Network training model acquisition module, for neural to BP based on described training sample and default neural transferring function f (x) Network is trained, and obtains network training model；

Obtain judge module, for described network training model being tested according to described test sample, obtain and judge to survey Whether test result is in preset range；If it is not, send the first execution signal to the first control module；If so, mould is controlled to second Block sends the second execution signal；

Described first control module, is used for receiving described first and performs signal, return perform described based on described training sample and Preset neural transferring function f (x) BP neutral net is trained, the step for of obtaining network training model；

Described second control module, is used for receiving described second and performs signal, described network training model is defined as BP neural Network computing model, and by described test result renormalization；

Computing module, for based on described BP calculating model of neural networks, calculates according to current carbon emission amount affecting parameters, Obtain the carbon emission amount value of calculation of corresponding renormalization.

System the most according to claim 6, it is characterised in that described network training model acquisition module includes:

Carbon emission neuron computational mathematics model construction unit, is used for building carbon emission neuron computational mathematics modelWherein, ω_pqFor the weighted value of neuron p to neuron q, x_pT () represents t neuron q The input information from neuron p received, f (x) is described default neural transferring function；

Input layer acquiring unit, for based on BP neutral net, each group normalization carbon emission that will comprise in described training sample Amount affecting parameters is as the input layer vector X=(x of described BP neutral net₁,x₂,…,x_i,…,x_n)^T, wherein, x_iIt is one group to return One changes the i-th in carbon emission amount affecting parameters n；

Hidden layer acquiring unit, for each based on comprise in described carbon emission neuron computational mathematics model, described training sample The input layer vector X=(x that group normalization carbon emission amount affecting parameters is corresponding₁,x₂,…,x_i,…,x_n)^T, obtain described BP neural The hidden layer vector Y=(y of network₁,y₂,…,y_j,…,y_m)^T, wherein, y_jFor the jth neuron in hidden layer m neuron to Amount,v_ijWeighted value for input layer i-th neuron to hidden layer jth neuron；

Output layer acquiring unit, for based on described hidden layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith described default neuron Transmission function f (x), the output layer calculating described BP neutral net corresponding to described each group of normalization carbon emission amount affecting parameters is defeated The carbon emission amount estimated value gone outWherein, ω_jkFor hidden layer jth neuron to output layer kth god Through unit's weighted value, the value of k is 1；

Output error computing unit, for according to each group normalization carbon emission amount data comprised in described training sample, obtaining The actual carbon emission value b that described each group of normalization carbon emission amount affecting parameters is corresponding, and calculate described each group of normalization carbon row High-volume corresponding output error e of affecting parameters, e=1/2 (b-a)²；

Parameter error obtains judging unit, for taking output error e that described each group of normalization carbon emission amount affecting parameters is corresponding Quadratic sum as parameter error, and judge that described parameter error is whether in default error allowed band；If so, to the first control Unit processed sends the 3rd execution signal；If it is not, send the 4th execution signal to the second control unit；

Described first control unit, is used for receiving the described 3rd and performs signal, described BP neutral net is defined as network training Model；

Described second control unit, is used for receiving the described 4th and performs signal, according to described each group of normalization carbon emission amount impact Output error e corresponding to parameter, described input layer vector X=(x₁,x₂,…,x_i,…,x_n)^T, described hidden layer vector Y=(y₁, y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission amount affecting parameters pair The input layer i-th neuron answered is to weighted value correction amount v of hidden layer jth neuron_ijWith hidden layer jth neuron to defeated Go out layer kth neuron weighted value correction amount w_jk, and described BP neutral net is modified, by revised BP nerve net Network is defined as network training model.

System the most according to claim 7, it is characterised in that described second control unit includes:

First computing unit, for according to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described defeated Enter layer vector X=(x₁,x₂,…,x_i,…,x_n)^TWith default neural transferring function f (x), calculate described each group of normalization carbon row The input layer i-th neuron that high-volume affecting parameters is corresponding is to weighted value correction amount v of hidden layer jth neuron_ij,Wherein,For corresponding defeated of described each group of normalization carbon emission amount affecting parameters Entering layer to hidden layer error signal, η is default learning rate, and the value of η is the constant between 0 to 1；

Second computing unit, for according to output error e corresponding to described each group of normalization carbon emission amount affecting parameters, described hidden Layer vector Y=(y₁,y₂,…,y_j,…,y_m)^TWith default neural transferring function f (x), calculate described each group of normalization carbon emission Measure hidden layer jth neuron corresponding to affecting parameters to output layer kth neuron weighted value correction amount w_jk,Wherein,The hidden layer corresponding for described each group of normalization carbon emission amount affecting parameters arrives Output layer error signal.