CN106056248A - User participation degree prediction method for interruptible load project - Google Patents

Abstract

The invention discloses a user participation degree prediction method for an interruptible load project. The user participation degree prediction method comprises the steps of: (1) analyzing user participation degree influencing factors in the interruptible load project based on a unified theory of acceptance and use of technology model, and establishing a user participation model of the interruptible load project; (2) carrying out quantification processing on the user participation degree influencing factors; (3) and establishing a user participation degree prediction model based on a support vector machine. The user participation degree prediction method provided by the invention establishes the user participation model of the interruptible load project, carries out quantification processing on influencing factors of the user participation model based on the user participation degree influencing factors of the interruptible load project, adopts the support vector machine for predicting user participation degrees, and provides a basis for efficient implementation of the interruptible load project.

Description

A kind of interruptible load project user participation Forecasting Methodology

Technical field

The present invention relates to a kind of interruptible load project user participation Forecasting Methodology, belong to Power System and its Automation Technology.

Background technology

In recent years, electricity consumption continues to increase rapidly, and electricity consumption peak-valley difference strengthens, and seasonal power is in short supply to happen occasionally.Need Seeking response technology is to solve this problem to provide many schemes flexibly, it is possible to use the demand response skill of relative inexpensiveness Art realizes load from user's side angle degree and cuts down, it is ensured that the equilibrium of supply and demand.

As a kind of important way of demand response, interruptible load can effectively alleviate peak times of power consumption power supply and demand lance Shield, has stronger realistic meaning and far-reaching significance.Tradition is often not related to electric power in the research of interruptible load project The participation of user, but in the interruptible load project implementation process, due to different user to economic benefit, rules and regulations and The acceptance of the publicity factors such as guiding is different, and the participation of interruptible load project is often existed bigger uncertain by user Property.Hence set up user's participation forecast model significant.

Summary of the invention

Goal of the invention: in order to overcome the deficiencies in the prior art, the present invention provides a kind of interruptible load project to use Family participation Forecasting Methodology, by the influence factor of power consumer participation in deep anatomy interruptible load project, sets up and uses Family participation model, and use support vector machine to build user's participation forecast model, thus realize in interruptible load project The calculating of user's participation, the high efficiency enforcement for interruptible load project provides basis.

Technical scheme: for achieving the above object, the technical solution used in the present invention is:

A kind of interruptible load project user participation Forecasting Methodology, comprises the steps:

(1) based on integrating information technology acceptance and using model (Unified Theory of Acceptance and Use of Technology Model, is called for short UTAUT), analyze the user's participation influent factor in interruptible load project, The user setting up interruptible load project participates in model；

(2) user's participation influent factor is carried out quantification treatment；

(3) based on support vector machine, user's participation forecast model is set up.

Concrete, in described step (1), user's participation influent factor includes expected utility PE, makes great efforts expectation EE and society Can affect SI, the user of foundation participates in model representation and is:

B=f (PE, EE, SI) (1)

Wherein: expected utility PE characterize user participate in the desired profit of interruptible load project (save the electricity charge or other Some favourable outcomes) degree, make great efforts expectation EE and characterize user and participate in the complexity of interruptible load project, social influence SI Characterizing user and consider to participate in social influence's degree suffered during interruptible load project, user's participation B characterizes user and responds secondary Number issues the ratio of demand response number of times with interruptible load project；Above-mentioned user participates in model and thinks in interruptible load project User's participation B mainly by expected utility PE, make great efforts expectation EE and tri-key factors of social influence SI are determined.

Concrete, in described step (2), user's participation influent factor is carried out quantification treatment, particularly as follows:

(2-1) quantification treatment of expected utility PE:

In interruptible load project, expected utility PE of user is closely bound up with the economic benefit of interruptible load project； We calculate user based on economic benefit and participate in the benefit-cost-ratio of interruptible load project, to obtain expected utility PE of user；

Specifying in demand response agreement, the electricity cut down in interruptible load project for user, user can The subsidy of the unit quantity of electricity obtained is r₁；Then user i participates in income B of interruptible load project_i,PEFor:

B_i,PE=x_i×r₁×m_i (2)

Wherein: x_iRepresent that user i participates in the electricity that interruptible load project is cut down, m every time_iRepresent user i participate in can in The number of times of disconnected load project；

The value added of the cost unit quantity of electricity cost that user participates in interruptible load project represents, i.e. user participate in can in During disconnected load project, the required cost increased of every response units electricity；If user i participates in interruptible load project Time, the required cost increased of every response units electricity is r_2,i, then user i participates in the cost of interruptible load project is C_i,PE:

C_i,PE=x_i×r_2,i×m_i (3)

Participate in income and the cost of interruptible load project based on user i, calculate user i and participate in interruptible load project Benefit-cost-ratio BCR_iFor:

${\mathrm{BCR}}_i=\frac{B_{i,PE}}{C_{i,PE}}=\frac{r_1}{r_{2,i}}---\left(4\right)$

Thus calculate expected utility PE of user i_i:

${\mathrm{PE}}_i=\left\{\begin{array}{cc}0& {\mathrm{BCR}}_i\&le;{\mathrm{BCR}}_{i,\min }\\ \frac{{\mathrm{BCR}}_i-{\mathrm{BCR}}_{i,\min }}{{\mathrm{BCR}}_{i,\max }-{\mathrm{BCR}}_{i,\min }}& {\mathrm{RCR}}_{i,\min }<{\mathrm{BCR}}_i<{\mathrm{BCR}}_{i,\max }\\ 1& {\mathrm{RCR}}_i\&GreaterEqual;{\mathrm{BCR}}_{i,\max }\end{array}\right.---\left(5\right)$

Wherein: BCR_i,minAnd BCR_i,maxIt is respectively benefit-cost-ratio BCR of user i_iLower threshold and upper limit threshold；

(2-2) quantification treatment of effort expectation EE:

First, will strive to expect that the influence factor of EE is defined as three: 1. user participates in interruptible load project Complexity, the complexity that 2. user makes a profit from interruptible load project, 3. user make appropriate arrangements produce complexity； Then, each influence factor refinement is split as three relevant issues, and each relevant issues are used Li Kete five subscale Investigate；Then, analytic hierarchy process (AHP) (AHP) is used to try to achieve weights and the weights of nine relevant issues of three influence factors； Finally, calculate user i and participate in the effort expectation score value EE of interruptible load project_i' it is:

${{\mathrm{EE}}_i}^{\&prime;}={\mathrm{\&omega;}}_{21}\underset{j=1}{\overset{3}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}+{\mathrm{\&omega;}}_{22}\underset{j=4}{\overset{6}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}+{\mathrm{\&omega;}}_{23}\underset{j=7}{\overset{9}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}---\left(6\right)$

Wherein: ω₂₁、ω₂₂And ω₂₃It is respectively influence factor's weights 1., 2. and 3., ω_3jFor the weights of relevant issues j, s_ijFor user i, Li Kete five subscale of relevant issues j is given a mark；J=1,2 ..., 9, ω₃₁、ω₃₂And ω₃₃The most corresponding shadow Ring factor three relevant issues 1., ω₃₄、ω₃₅And ω₃₆The most corresponding influence factor's three relevant issues 2., ω₃₇、ω₃₈ And ω₃₉The most corresponding influence factor's three relevant issues 3.；

Owing to each relevant issues maximum score value in Li Kete five subscale is 5 points, therefore to making great efforts expectation score value EE_i' be normalized, obtain user i and participate in the effort expectation EE of interruptible load project_iFor:

${\mathrm{EE}}_i=\frac{{{\mathrm{EE}}_i}^{\&prime;}}{5}---\left(7\right)$

(2-3)) the quantification treatment of social influence SI:

User on the participation of interruptible load project can by society and other people affected, therefore interruptible load project Enforcement body can be guided the social influence's degree expanding interruptible load project by publicity, so that more user participates in In interruptible load project；Therefore, with expenses on publicity as standard, calculate user i and participate in the society of interruptible load project Affect SI_iFor:

${\mathrm{SI}}_i=\left\{\begin{array}{cc}0& C_{SI}<C_{SI,i,\min }\\ \frac{C_{SI}-C_{SI,i,\min }}{C_{SI,i,\max }-C_{SI,i,\min }}& C_{SI,i,\min }\&le;C_{SI}\&le;C_{SI,i,\max }\\ 1& C_{SI}>C_{SI,i,\max }\end{array}\right.---\left(8\right)$

Wherein: C_SIFor expenses on publicity, C_SI,i,minAnd C_SI,i,maxLower threshold expenses on publicity responded for user i and Upper limit threshold.

Concrete, in described step (2-2), ω₂₁、ω₂₂、ω₂₃And ω_3jObtaining value method as follows:

The assessment indicator system setting up user i includes destination layer, rule layer and indicator layer, wherein: destination layer is first Layer, represents the effort expectation EE of user i_i, it is designated as A={A}；Rule layer is the second layer, represents and makes great efforts expectation EE_iThree impacts Factor, is designated as B={B₁,B₂,B₃}；Indicator layer is third layer, represents nine relevant issues of three influence factors, is designated as C= {C₁,C₂,C₃,C₄,C₅,C₆,C₇,C₈,C₉}；A is the last layer of B, and B is the last layer of C, and C is next layer of B, and B is next of A Layer；

Use P={P₁,P₂,…,P_m,…,P_MCharacterize last layer, use Q={Q₁,Q₂,…,Q_n,…,Q_NCharacterize next Layer, the either element of last layer all elements to next layer have dominance relation, set up with element P_mAppointing for judgment criterion Anticipate two element Q_nBetween multilevel iudge matrix R_Pm；Multilevel iudge matrix R_PmIn element R_ijReflect for element P_m, unit Element Q_iRelative to element Q_jSignificance level, i=1,2 ..., N, j=1,2 ..., N；Multilevel iudge matrix R_PmIt it is a reciprocal square Battle array, R_ijThere is following character:

$R_{ij}>0;R_{ij}=\frac{1}{R_{ji}};R_{ii}=1---\left(9\right)$

With the judgment matrix of ground floor all elements as criterion:

[R_A1]=1 (10)

The judgment matrix solving second layer all elements is:

$R_B=\left[\begin{array}{ccc}{R}_{B11}& R_{B12}& R_{B13}\\ R_{B21}& R_{B22}& R_{B23}\\ R_{B31}& R_{B32}& R_{B33}\end{array}\right]---\left(11\right)$

The judgment matrix solving third layer all elements is:

$R_C=\left[\begin{array}{ccccccccc}{R}_{C11}& R_{C12}& R_{C13}& 0& 0& 0& 0& 0& 0\\ R_{C21}& R_{22}& R_{C23}& 0& 0& 0& 0& 0& 0\\ R_{C31}& R_{C32}& R_{C33}& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& R_{C44}& R_{C45}& R_{C46}& 0& 0& 0\\ 0& 0& 0& R_{C54}& R_{C55}& R_{C56}& 0& 0& 0\\ 0& 0& 0& R_{C64}& R_{C65}& R_{C66}& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& R_{C77}& R_{C78}& R_{C79}\\ 0& 0& 0& 0& 0& 0& R_{C87}& R_{C88}& R_{C89}\\ 0& 0& 0& 0& 0& 0& R_{C97}& R_{C98}& R_{C99}\end{array}\right]---\left(12\right)$

Based on multilevel iudge matrix R_BAnd R_C, the weight matrix W of each element of the second layer is obtained by solving formula (13)_B, pass through Solve formula (14) and obtain the weight matrix W of each element of third layer_C:

R_BW_B=λ_max,BW_B (13)

R_CW_C=λ_max,CW_C (14)

Try to achieve W_A、W_BAnd W_CAs follows:

W_A=[ω₁₁]=1 (15)

W_B=[ω₂₁ ω₂₂ ω₂₃]^T (16)

$W_C={\left[\begin{array}{ccccccccc}{\mathrm{\&omega;}}_{31}& {\mathrm{\&omega;}}_{32}& {\mathrm{\&omega;}}_{33}& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& {\mathrm{\&omega;}}_{34}& {\mathrm{\&omega;}}_{35}& {\mathrm{\&omega;}}_{36}& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& {\mathrm{\&omega;}}_{37}& {\mathrm{\&omega;}}_{38}& {\mathrm{\&omega;}}_{39}\end{array}\right]}^T---\left(17\right)$

Wherein: λ_max,BFor R_BEigenvalue of maximum, W_BIt is corresponding λ_max,BCharacteristic vector；λ_max,CFor R_CMaximum feature Value, W_CIt is corresponding λ_max,CCharacteristic vector.

Concrete, in described step (3), for user i, support vector machine has three input quantities, respectively expected utilities PE_i, make great efforts expectation EE_iWith social influence SI_i, support vector machine has an output, for user's participation B_i；Based on known Expected utility PE_i, make great efforts expectation EE_i, social influence SI_iWith user's participation B_iSupport vector machine is trained, the most available User participates in forecast model.

Beneficial effect: the interruptible load project user participation Forecasting Methodology that the present invention provides, going deep into anatomy can interrupt The influence factor of power consumer participation in load project, sets up user's participation model, and uses support vector machine to build use Family participation forecast model, thus realize the calculating of user's participation in interruptible load project, for interruptible load project High efficiency enforcement provides basis.

Accompanying drawing explanation

Fig. 1 is the real time process flow figure of the inventive method；

Fig. 2 is for making great efforts to expect the quantizing process of EE and Li Kete five subscale composition.

Detailed description of the invention

Below in conjunction with the accompanying drawings the present invention is further described.

Be illustrated in figure 1 a kind of interruptible load project user participation Forecasting Methodology, below whole implementation process is added To illustrate.

Step one: accept based on integrating information technology and use model, analyzing the user in interruptible load project and participate in Degree influent factor, the user setting up interruptible load project participates in model.

User's participation influent factor includes expected utility PE, makes great efforts expectation EE and social influence SI, user's ginseng of foundation With model representation it is:

B=f (PE, EE, SI) (1)

Wherein: expected utility PE characterize user participate in the desired profit of interruptible load project (save the electricity charge or other Some favourable outcomes) degree, make great efforts expectation EE and characterize user and participate in the complexity of interruptible load project, social influence SI Characterizing user and consider to participate in social influence's degree suffered during interruptible load project, user's participation B characterizes user and responds secondary Number issues the ratio of demand response number of times with interruptible load project；Above-mentioned user participates in model and thinks in interruptible load project User's participation B mainly by expected utility PE, make great efforts expectation EE and tri-key factors of social influence SI are determined.

Step 2: user's participation influent factor is carried out quantification treatment.

(2-1) quantification treatment of expected utility PE:

In interruptible load project, expected utility PE of user is closely bound up with the economic benefit of interruptible load project； We calculate user based on economic benefit and participate in the benefit-cost-ratio of interruptible load project, to obtain expected utility PE of user.

Specifying in demand response agreement, the electricity cut down in interruptible load project for user, user can The subsidy of the unit quantity of electricity obtained is r₁；Then user i participates in income B of interruptible load project_i,PEFor:

B_i,PE=x_i×r₁×m_i (2)

Wherein: x_iRepresent that user i participates in the electricity that interruptible load project is cut down, m every time_iRepresent user i participate in can in The number of times of disconnected load project；

The value added of the cost unit quantity of electricity cost that user participates in interruptible load project represents, i.e. user participate in can in During disconnected load project, the required cost increased of every response units electricity；If user i participates in interruptible load project Time, the required cost increased of every response units electricity is r_2,i, then user i participates in the cost of interruptible load project is C_i,PE:

C_i,PE=x_i×r_2,i×m_i (3)

Participate in income and the cost of interruptible load project based on user i, calculate user i and participate in interruptible load project Benefit-cost-ratio BCR_iFor:

${\mathrm{BCR}}_i=\frac{B_{i,PE}}{C_{i,PE}}=\frac{r_1}{r_{2,i}}---\left(4\right)$

Thus calculate expected utility PE of user i_i:

${\mathrm{PE}}_i=\left\{\begin{array}{cc}0& {\mathrm{BCR}}_i\&le;{\mathrm{BCR}}_{i,\min }\\ \frac{{\mathrm{BCR}}_i-{\mathrm{BCR}}_{i,\min }}{{\mathrm{BCR}}_{i,\max }-{\mathrm{BCR}}_{i,\min }}& {\mathrm{RCR}}_{i,\min }<{\mathrm{BCR}}_i<{\mathrm{BCR}}_{i,\max }\\ 1& {\mathrm{RCR}}_i\&GreaterEqual;{\mathrm{BCR}}_{i,\max }\end{array}\right.---\left(5\right)$

Wherein: BCR_i,minAnd BCR_i,maxIt is respectively benefit-cost-ratio BCR of user i_iLower threshold and upper limit threshold.

(2-2) quantification treatment of effort expectation EE:

First, will strive to expect that the influence factor of EE is defined as three: 1. user participates in interruptible load project Complexity, the complexity that 2. user makes a profit from interruptible load project, 3. user make appropriate arrangements produce complexity； Then, each influence factor refinement is split as three relevant issues, and each relevant issues are used Li Kete five subscale Investigate；Then, analytic hierarchy process (AHP) (AHP) is used to try to achieve weights and the weights of nine relevant issues of three influence factors； Finally, calculate user i and participate in the effort expectation score value EE of interruptible load project_i' it is:

${{\mathrm{EE}}_i}^{\&prime;}={\mathrm{\&omega;}}_{21}\underset{j=1}{\overset{3}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}+{\mathrm{\&omega;}}_{22}\underset{j=4}{\overset{6}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}+{\mathrm{\&omega;}}_{23}\underset{j=7}{\overset{9}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}---\left(6\right)$

Wherein: ω₂₁、ω₂₂And ω₂₃It is respectively influence factor's weights 1., 2. and 3., ω_3jFor the weights of relevant issues j, s_ijFor user i, Li Kete five subscale of relevant issues j is given a mark；J=1,2 ..., 9, ω₃₁、ω₃₂And ω₃₃The most corresponding shadow Ring factor three relevant issues 1., ω₃₄、ω₃₅And ω₃₆The most corresponding influence factor's three relevant issues 2., ω₃₇、ω₃₈ And ω₃₉The most corresponding influence factor's three relevant issues 3..

Owing to each relevant issues maximum score value in Li Kete five subscale is 5 points, therefore to making great efforts expectation score value EE_i' be normalized, obtain user i and participate in the effort expectation EE of interruptible load project_iFor:

${\mathrm{EE}}_i=\frac{{{\mathrm{EE}}_i}^{\&prime;}}{5}---\left(7\right)$

In this step, ω₂₁、ω₂₂、ω₂₃And ω_3jObtaining value method as follows:

The assessment indicator system setting up user i includes destination layer, rule layer and indicator layer, wherein: destination layer is first Layer, represents the effort expectation EE of user i_i, it is designated as A={A}；Rule layer is the second layer, represents and makes great efforts expectation EE_iThree impacts Factor, is designated as B={B₁,B₂,B₃}；Indicator layer is third layer, represents nine relevant issues of three influence factors, is designated as C= {C₁,C₂,C₃,C₄,C₅,C₆,C₇,C₈,C₉}；A is the last layer of B, and B is the last layer of C, and C is next layer of B, and B is next of A Layer；

Use P={P₁,P₂,…,P_m,…,P_MCharacterize last layer, use Q={Q₁,Q₂,…,Q_n,…,Q_NCharacterize next Layer, the either element of last layer all elements to next layer have dominance relation, set up with element P_mAppointing for judgment criterion Anticipate two element Q_nBetween multilevel iudge matrix R_Pm；Multilevel iudge matrix R_PmIn element R_ijReflect for element P_m, unit Element Q_iRelative to element Q_jSignificance level, i=1,2 ..., N, j=1,2 ..., N；Multilevel iudge matrix R_PmIt it is a reciprocal square Battle array, R_ijThere is following character:

$R_{ij}>0;R_{ij}=\frac{1}{R_{ji}};R_{ii}=1---\left(8\right)$

9 grades of scaling laws of employing AHP are to the element assignment of judgment matrix, as shown in table 1:

The value of table 1 judgment matrix element

With the judgment matrix of ground floor all elements as criterion:

[R_A1]=1 (9)

The judgment matrix solving second layer all elements is:

$R_B=\left[\begin{array}{ccc}{R}_{B11}& R_{B12}& R_{B13}\\ R_{B21}& R_{B22}& R_{B23}\\ R_{B31}& R_{B32}& R_{B33}\end{array}\right]---\left(10\right)$

The judgment matrix solving third layer all elements is:

$R_C=\left[\begin{array}{ccccccccc}{R}_{C11}& R_{C12}& R_{C13}& 0& 0& 0& 0& 0& 0\\ R_{C21}& R_{22}& R_{C23}& 0& 0& 0& 0& 0& 0\\ R_{C31}& R_{C32}& R_{C33}& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& R_{C44}& R_{C45}& R_{C46}& 0& 0& 0\\ 0& 0& 0& R_{C54}& R_{C55}& R_{C56}& 0& 0& 0\\ 0& 0& 0& R_{C64}& R_{C65}& R_{C66}& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& R_{C77}& R_{C78}& R_{C79}\\ 0& 0& 0& 0& 0& 0& R_{C87}& R_{C88}& R_{C89}\\ 0& 0& 0& 0& 0& 0& R_{C97}& R_{C98}& R_{C99}\end{array}\right]---\left(11\right)$

Based on multilevel iudge matrix R_BAnd R_C, the weight matrix W of each element of the second layer is obtained by solving formula (12)_B, pass through Solve formula (13) and obtain the weight matrix W of each element of third layer_C:

R_BW_B=λ_max,BW_B (12)

R_CW_C=λ_max,CW_C (13)

Try to achieve W_A、W_BAnd W_CAs follows:

W_A=[ω₁₁]=1 (14)

W_B=[ω₂₁ ω₂₂ ω₂₃]^T (15)

$W_C={\left[\begin{array}{ccccccccc}{\mathrm{\&omega;}}_{31}& {\mathrm{\&omega;}}_{32}& {\mathrm{\&omega;}}_{33}& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& {\mathrm{\&omega;}}_{34}& {\mathrm{\&omega;}}_{35}& {\mathrm{\&omega;}}_{36}& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& {\mathrm{\&omega;}}_{37}& {\mathrm{\&omega;}}_{38}& {\mathrm{\&omega;}}_{39}\end{array}\right]}^T---\left(16\right)$

Wherein: λ_max,BFor R_BEigenvalue of maximum, W_BIt is corresponding λ_max,BCharacteristic vector；λ_max,CFor R_CMaximum feature Value, W_CIt is corresponding λ_max,CCharacteristic vector.

(2-3) quantification treatment of social influence SI:

User on the participation of interruptible load project can by society and other people affected, therefore interruptible load project Enforcement body can be guided the social influence's degree expanding interruptible load project by publicity, so that more user participates in In interruptible load project；Therefore, with expenses on publicity as standard, calculate user i and participate in the society of interruptible load project Affect SI_iFor:

${\mathrm{SI}}_i=\left\{\begin{array}{cc}0& C_{SI}<C_{SI,i,\min }\\ \frac{C_{SI}-C_{SI,i,\min }}{C_{SI,i,\max }-C_{SI,i,\min }}& C_{SI,i,\min }\&le;C_{SI}\&le;C_{SI,i,\max }\\ 1& C_{SI}>C_{SI,i,\max }\end{array}\right.---\left(17\right)$

Wherein: C_SIFor expenses on publicity, C_SI,i,minAnd C_SI,i,maxLower threshold expenses on publicity responded for user i and Upper limit threshold.

Step 3: based on support vector machine, sets up user's participation forecast model.

For user i, support vector machine has three input quantities, respectively expected utilities PE_i, make great efforts expectation EE_iWith society's shadow Ring SI_i, support vector machine has an output, for user's participation B_i；Based on known expected utility PE_i, make great efforts expectation EE_i、 Social influence SI_iWith user's participation B_iSupport vector machine is trained, i.e. can get user and participate in forecast model.

Prior art as, the training process of support vector machine is as follows:

For sample set (x₁,y₁),(x₂,y₂),…,(x_l,y_l), first look for a input space (x₁,x₂,…,x_l) arrive Output space (y₁,y₂,…,y_l) nonlinear mapping Φ, then the data of sample set are mapped to higher dimensional space F, use following line Property function carries out linear regression to sample set, it may be assumed that

F (x)=w Φ (x)+b, Φ: Rⁿ→F,w∈F (18)

Wherein: b is threshold value.Introduce structure risk function, be shown below:

$R_{reg}=\frac{1}{2}\left|\right|w|{|}^2+\underset{i=1}{\overset{l}{\&Sigma;}}\mathrm{\&epsiv;}\left(f\right(x_i)-y_i)---\left(19\right)$

Wherein: | | w | | is described function, l represents that the number of sample, ε () are loss function, is defined as follows formula:

Complexity for control function should make linear regression function the most smooth, though Euler's scope | | w | | of w²Minimum, And consider to introduce relaxation factor beyond the error of fitting of precisionThe solution making following formula exists:

$\begin{array}{c}\min \frac{1}{2}\left|\right|w|{|}^2+C\underset{i}{\overset{l}{\&Sigma;}}({\mathrm{\&zeta;}}_i+{\mathrm{\&zeta;}}_i^{*})\\ \begin{array}{cc}s.t.& y_i-w\&CenterDot;\mathrm{\&Phi;}\left(x_i\right)-b\&le;\mathrm{\&epsiv;}+{\mathrm{\&zeta;}}_i\\ & w\&CenterDot;\mathrm{\&Phi;}\left(x_i\right)+b-y_i\&le;\mathrm{\&epsiv;}+{\mathrm{\&zeta;}}_i^{*}\\ & {\mathrm{\&zeta;}}_i,{\mathrm{\&zeta;}}_i^{*}\&GreaterEqual;0\end{array}\end{array}---\left(21\right)$

Utilize Lagrangian and the principle of duality, can be with primal-dual optimization problem:

$\begin{array}{c}\min \frac{1}{2}\underset{i,j=1}{\overset{l}{\&Sigma;}}({\mathrm{\&alpha;}}_i-{\mathrm{\&alpha;}}_i^{*})({\mathrm{\&alpha;}}_j-{\mathrm{\&alpha;}}_j^{*})\left(\mathrm{\&Phi;}\right(x_i)\&CenterDot;\mathrm{\&Phi;}(x_i\left)\right)-\underset{i=1}{\overset{l}{\&Sigma;}}{\mathrm{\&alpha;}}_i(\mathrm{\&epsiv;}-y_i)-\underset{i=1}{\overset{l}{\&Sigma;}}{\mathrm{\&alpha;}}_i^{*}(\mathrm{\&epsiv;}+y_i)\\ \begin{array}{cc}s.t.& \underset{i=1}{\overset{l}{\&Sigma;}}({\mathrm{\&alpha;}}_i-{\mathrm{\&alpha;}}_i^{*})=0\\ & {\mathrm{\&alpha;}}_i,{\mathrm{\&alpha;}}_i^{*}\&Element;\&lsqb;0,C\&rsqb;\end{array}\end{array}---\left(22\right)$

Wherein: α_i,For Lagrange multiplier.Solve this quadratic programming problem and can try to achieve the value of α, try to achieve simultaneouslyThus can try to achieve linear regression function:

$f\left(x\right)=\underset{i=1}{\overset{l}{\&Sigma;}}({\mathrm{\&alpha;}}_i-{\mathrm{\&alpha;}}_i^{*})K(x,x_i)+b---\left(23\right)$

Wherein: K (x, x_i)=Φ (x_i) Φ (x), K (x, x_i) it is kernel function.The kernel function selecting multi-form can generate Different support vector machine, conventional kernel function has: RBF, polynomial function, perceptron (Sigmoid) function, line Property function etc..

The above is only the preferred embodiment of the present invention, it should be pointed out that: for the ordinary skill people of the art For Yuan, under the premise without departing from the principles of the invention, it is also possible to make some improvements and modifications, these improvements and modifications also should It is considered as protection scope of the present invention.

Claims (5)

1. an interruptible load project user participation Forecasting Methodology, it is characterised in that: comprise the steps:

(1) based on integrating information technology acceptance and using model, analyzing the user's participation impact in interruptible load project will Element, the user setting up interruptible load project participates in model；

(2) user's participation influent factor is carried out quantification treatment；

(3) based on support vector machine, user's participation forecast model is set up.

Interruptible load project user participation Forecasting Methodology the most according to claim 1, it is characterised in that: described step (1), in, user's participation influent factor includes expected utility PE, makes great efforts expectation EE and social influence SI, and the user of foundation participates in Model representation is:

B=f (PE, EE, SI) (1)

Wherein: expected utility PE characterizes user and participates in the degree of the desired profit of interruptible load project, expectation EE table is made great efforts Requisition family participates in the complexity of interruptible load project, when social influence SI sign user considers to participate in interruptible load project Suffered social influence's degree, user's participation B characterizes user's response times and issues demand response with interruptible load project The ratio of number.

Interruptible load project user participation Forecasting Methodology the most according to claim 1, it is characterised in that: described step (2), in, user's participation influent factor is carried out quantification treatment, particularly as follows:

(2-1) quantification treatment of expected utility PE:

Specifying in demand response agreement, the electricity cut down in interruptible load project for user, user is obtained in that The subsidy of unit quantity of electricity be r₁；If user i participates in interruptible load project, the required one-tenth increased of every response units electricity This is r_2,i；Then user i participates in benefit-cost-ratio BCR of interruptible load project_iFor:

${\mathrm{BCR}}_i=\frac{B_{i,PE}}{C_{i,PE}}=\frac{r_1}{r_{2,i}}---\left(2\right)$

Thus calculate expected utility PE of user i_i:

${\mathrm{PE}}_i=\left\{\begin{array}{cc}0& {\mathrm{BCR}}_i\&le;{\mathrm{BCR}}_{i,\min }\\ \frac{{\mathrm{BCR}}_i-{\mathrm{BCR}}_{i,\min }}{{\mathrm{BCR}}_{i,\max }-{\mathrm{BCR}}_{i,\min }}& {\mathrm{BCR}}_{i,\min }<{\mathrm{BCR}}_i<{\mathrm{BCR}}_{i,\max }\\ 1& {\mathrm{BCR}}_i\&GreaterEqual;{\mathrm{BCR}}_{i,\max }\end{array}\right.---\left(3\right)$

Wherein: BCR_i,minAnd BCR_i,maxIt is respectively benefit-cost-ratio BCR of user i_iLower threshold and upper limit threshold；

(2-2) quantification treatment of effort expectation EE:

First, will strive to expect that the influence factor of EE is defined as three: 1. user participates in the difficulty or ease of interruptible load project Degree, the complexity that 2. user makes a profit from interruptible load project, 3. user make appropriate arrangements produce complexity；So After, each influence factor refinement is split as three relevant issues, and uses Li Kete five subscale to enter each relevant issues Row investigation；Then, analytic hierarchy process (AHP) is used to try to achieve weights and the weights of nine relevant issues of three influence factors；Finally, meter Calculate user i and participate in the effort expectation score value EE of interruptible load project_i' it is:

${{\mathrm{EE}}_i}^{\&prime;}={\mathrm{\&omega;}}_{21}\underset{j=1}{\overset{3}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}+{\mathrm{\&omega;}}_{22}\underset{j=4}{\overset{6}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}+{\mathrm{\&omega;}}_{23}\underset{j=7}{\overset{9}{\&Sigma;}}{s}_{ij}\&CenterDot;{\mathrm{\&omega;}}_{3j}---\left(4\right)$

Wherein: ω₂₁、ω₂₂And ω₂₃It is respectively influence factor's weights 1., 2. and 3., ω_3jFor the weights of relevant issues j, s_ijFor Li Kete five subscale of relevant issues j is given a mark by user i；J=1,2 ..., 9, ω₃₁、ω₃₂And ω₃₃Respectively correspondence affect because of Element three relevant issues 1., ω₃₄、ω₃₅And ω₃₆The most corresponding influence factor's three relevant issues 2., ω₃₇、ω₃₈With ω₃₉The most corresponding influence factor's three relevant issues 3.；

To making great efforts expectation score value EE_i' be normalized, obtain user i and participate in the effort expectation EE of interruptible load project_i For:

${\mathrm{EE}}_i=\frac{{{\mathrm{EE}}_i}^{\&prime;}}{5}---\left(5\right)$

(2-3)) the quantification treatment of social influence SI:

Calculate user i and participate in the social influence SI of interruptible load project_iFor:

${\mathrm{SI}}_i=\left\{\begin{array}{cc}0& C_{SI}<C_{SI,i,\min }\\ \frac{C_{SI}-C_{SI,i,\min }}{C_{SI,i,\max }-C_{SI,i,\min }}& C_{SI,i,\min }\&le;C_{SI}\&le;C_{SI,i,\max }\\ 1& C_{SI}>C_{SI,i,\max }\end{array}\right.---\left(6\right)$

Wherein: C_SIFor expenses on publicity, C_SI,i,minAnd C_SI,i,maxLower threshold expenses on publicity responded for user i and the upper limit Threshold value.

Interruptible load project user participation Forecasting Methodology the most according to claim 3, it is characterised in that: described step (2-2) in, ω₂₁、ω₂₂、ω₂₃And ω_3jObtaining value method as follows:

The assessment indicator system setting up user i includes destination layer, rule layer and indicator layer, wherein: destination layer is ground floor, generation The effort expectation EE of table user i_i, it is designated as A={A}；Rule layer is the second layer, represents and makes great efforts expectation EE_iThree influence factors, It is designated as B={B₁,B₂,B₃}；Indicator layer is third layer, represents nine relevant issues of three influence factors, is designated as C={C₁,C₂, C₃,C₄,C₅,C₆,C₇,C₈,C₉}；A is the last layer of B, and B is the last layer of C, and C is next layer of B, and B is next layer of A；

Use P={P₁,P₂,…,P_m,…,P_MCharacterize last layer, use Q={Q₁,Q₂,…,Q_n,…,Q_NCharacterize next layer, on The either element of one layer all elements to next layer have dominance relation, set up with element P_mAny two for judgment criterion Element Q_nBetween multilevel iudge matrix R_Pm；Multilevel iudge matrix R_PmIn element R_ijReflect for element P_m, element Q_iPhase For element Q_jSignificance level, i=1,2 ..., N, j=1,2 ..., N；Multilevel iudge matrix R_PmIt is a reciprocal matrix, R_ij There is following character:

$R_{ij}>0;R_{ij}=\frac{1}{R_{ji}};R_{ii}=1---\left(7\right)$

With the judgment matrix of ground floor all elements as criterion:

[R_A1]=1 (8)

The judgment matrix solving second layer all elements is:

$R_B=\left[\begin{array}{ccc}{R}_{B11}& R_{B12}& R_{B13}\\ R_{B21}& R_{B22}& R_{B23}\\ R_{B31}& R_{B32}& R_{B33}\end{array}\right]---\left(9\right)$

The judgment matrix solving third layer all elements is:

$R_C=\left[\begin{array}{ccccccccc}{R}_{C11}& R_{C12}& R_{C13}& 0& 0& 0& 0& 0& 0\\ R_{C21}& R_{C22}& R_{C23}& 0& 0& 0& 0& 0& 0\\ R_{C31}& R_{C32}& R_{C33}& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& R_{C44}& R_{C45}& R_{C46}& 0& 0& 0\\ 0& 0& 0& R_{C54}& R_{C55}& R_{C56}& 0& 0& 0\\ 0& 0& 0& R_{C64}& R_{C65}& R_{C66}& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& R_{C77}& R_{C78}& R_{C79}\\ 0& 0& 0& 0& 0& 0& R_{C87}& R_{C88}& R_{C89}\\ 0& 0& 0& 0& 0& 0& R_{C97}& R_{C98}& R_{C99}\end{array}\right]---\left(10\right)$

Based on multilevel iudge matrix R_BAnd R_C, the weight matrix W of each element of the second layer is obtained by solving formula (11)_B, by solving Formula (12) obtains the weight matrix W of each element of third layer_C:

R_BW_B=λ_max,BW_B (11)

R_CW_C=λ_max,CW_C (12)

Try to achieve W_A、W_BAnd W_CAs follows:

W_A=[ω₁₁]=1 (13)

W_B=[ω₂₁ ω₂₂ ω₂₃]^T (14)

$W_C={\left[\begin{array}{ccccccccc}{\mathrm{\&omega;}}_{31}& {\mathrm{\&omega;}}_{32}& {\mathrm{\&omega;}}_{33}& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& {\mathrm{\&omega;}}_{34}& {\mathrm{\&omega;}}_{35}& {\mathrm{\&omega;}}_{36}& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& {\mathrm{\&omega;}}_{37}& {\mathrm{\&omega;}}_{38}& {\mathrm{\&omega;}}_{39}\end{array}\right]}^T---\left(15\right)$

Wherein: λ_max,BFor R_BEigenvalue of maximum, W_BIt is corresponding λ_max,BCharacteristic vector；λ_max,CFor R_CEigenvalue of maximum, W_CIt is Corresponding λ_max,CCharacteristic vector.

Interruptible load project user participation Forecasting Methodology the most according to claim 1, it is characterised in that: described step (3) in, for user i, support vector machine has three input quantities, respectively expected utilities PE_i, make great efforts expectation EE_iAnd social influence SI_i, support vector machine has an output, for user's participation B_i；Based on known expected utility PE_i, make great efforts expectation EE_i, society SI can be affected_iWith user's participation B_iSupport vector machine is trained, i.e. can get user and participate in forecast model.