CN104239714A - User power consumption demand control method with dispatching control in intelligent power grid - Google Patents

Abstract

The invention relates to a user power consumption demand control method with dispatching control in an intelligent power grid. The method comprises the following steps of (1) in the intelligent power grid, solving the problem that the overall cost of the system including user dispatching and user dissatisfaction is minimized while the condition of reducing certain power deficit is met, and modeling the problem; (2) as the optimizing problem is related to the joint optimizing of continuous variable and discrete variable, splitting the problem into a bottom problem P1 and a top problem P2, so as to conveniently solve the problem; (3) utilizing the quick splitting method to solve the bottom problem P2; (4) on the basis of solution of bottom problem, utilizing a Gibbs sampling method with temperature varying process to solve the top problem P2, so as to obtain the final problem solution. The user power consumption demand control method with dispatching control in the intelligent power grid has the advantage that the purpose of minimizing the overall cost including user dispatching and user dissatisfaction is realized.

Description

With user's power consumption requirements control method of scheduling controlling in a kind of intelligent grid

Technical field

The present invention relates to intelligent grid field, with user's power consumption requirements control method of scheduling controlling in especially a kind of intelligent grid.

Background technology

Intelligent grid is a kind of electric power transmission network of robotization, and it has electric energy and information bidirectional flow feature, and the superiority of concentrated-distributed calculating simultaneously and real-time Communication for Power is in one.It is mutual that the core of intelligent grid is in electrical network between electricity provider and user, and this is mutual comprising information and electric energy, and these are realized by corresponding demand control procedure alternately.Demand modeling refers to the change due to electricity price, and the change of power consumption made by power consumer thereupon.Demand modeling is widely used, and a series ofly fact proved that it plays an important role in reduction peak load, mitigation network congestion, economize energy, minimizing gas discharging etc.In intelligent grid, demand modeling is usually adopted to reach the object of cutting down electric power deficit, maintaining the stabilization of power grids.The correlative study that current needs control is all often the precondition all participating in demand modeling based on users all in electrical network.But in a practical situation, we have to consider inevitably to produce extra cost when carrying out demand modeling to the power consumption of user.Therefore, how research adopts certain scheduling controlling to be significantly by controlling to cut down electric power deficit as much as possible to the power consumption requirements of user in electrical network simultaneously.

Summary of the invention

All participate in method that power consumption requirements controls in order to overcome in existing research to dispatch users all in electrical network and produce the deficiency of additional cost, the invention provides a kind of while satisfied reduction electric power deficit condition, ensure that scheduling process comprises user scheduling cost and user's dissatisfaction minimizing in the total cost of interior system, and algorithm complex is low, there is the user's power consumption requirements control method with scheduling controlling in good constringent intelligent grid.

The technical scheme that the technical matters that the present invention solves adopts is:

With user's power consumption requirements control method of scheduling controlling in intelligent grid, described control method comprises the following steps:

(1) in intelligent grid, control for cutting down the power consumption requirements of electric power deficit to user, the target reached meanwhile is needed to be minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction, so user's power consumption requirements optimization problem is described as following problem:

min?θ·∑ _iρ _i·a _i+(1-θ)·∑ _iD _i(x _i)

s.t.∑ _ia _i·(d _i-x _i)≥∑ _id _i-M

$d_i>x_i\≥x_{i,\min },\∀i$

Each parameter is defined as follows:

I: user i;

θ: weight coefficient, 0 < θ < 1, θ is larger, represents that, in designed target, minimizing of user scheduling cost is more important, and minimizing of user's dissatisfaction is then relatively inessential, and θ is less, then on the contrary;

ρ _i: dispatched users i participates in the scheduling cost caused by power consumption requirements control, and power consumer i is more important, ρ _ilarger;

D _i: the electrical energy demands that user i proposes in advance;

A _i: whether user i is scheduled participates in the control signal of power consumption requirements control, works as a _irepresent when=1 that user i is scheduled to participate in power consumption requirements and control, now for user i, it must cut down self power consumption, so just has x _{i, min}≤ x _i≤ d _i, wherein x _{i, min}it is the necessary minimal consumption electric energy of user i; Work as a _irepresent when=0 that user i is not scheduled and participate in power consumption requirements control, so x _i=d _i, namely the power consumption of user i is exactly the electrical energy demands that it proposes in advance;

D _i(x _i): the dissatisfaction of user i, when user i is after scheduling, x _iwith d _iwhen being more or less the same, D _i(x _i) can be very little, this represents that user exists dissatisfaction hardly when user i only needs to cut down very low amount power consumption time; When user i is after scheduling, x _iclose to x _{i, min}time, D _i(x _i) a very large value can be reached, this represent when user i need go to cut down a large amount of power consumption time, user will to electric energy scheduling very be unsatisfied with; Setting dissatisfaction D _i(x _i) be one at [x _{i, min}, d _i) convex function of upper monotone decreasing;

M: the maximum electricity that electricity provider can provide;

(2) in the optimization problem of above-mentioned (1), the problems referred to above are split as a continuous variable optimization problem and a Discrete Variables Optimization, correspond to it an a bottom problem P1 and top layer problem P2 respectively; Bottom problem P1 solves, and when determining certain user scheduling scheme, namely when known scheduling signals A, descends the user dissatisfaction of change caused by user's power consumption is cut down most, wherein A=(a ₁, a ₂..., a _i-1, a _i); Top layer problem P2 solves, according to the result of bottom problem P1, and the scheme of further optimizing user scheduling, thus minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction;

When given scheduling signals A, solving of bottom problem P1 could start; Bottom problem P1 solution procedure is: after receiving the scheduling signals A sent from top layer problem P2, first the power consumer be scheduled is formed set omega, Ω={ a _i| a _i=1}, if so Ω set is infeasible, needs top layer problem P2 again to choose A, otherwise Ω set is feasible; Gather for feasible Ω, the optimization problem P1 of bottom problem is described as:

P1：Ψ({a _i} _i∈Ω)＝min∑ _i∈ΩD _i(x _i)

$s.t.\underset{i\&Element;\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}{a}_i\&CenterDot;(d_i-x_i)\&GreaterEqual;\underset{i}{\mathrm{\&Sigma;}}{d}_i-M$

$d_i>x_i\&GreaterEqual;x_{i,\min },\&ForAll;i\&Element;\mathrm{\&Omega;}$

In addition, the optimization problem P2 of top layer problem is described as:

P2：min?C(A)＝θ·∑ _iρ _i·a _i+(1-θ)·Ψ({a _i} _i∈Ω)

In problem P2, parameter is defined as follows:

C (A): total cost of system in given scheduling signals A situation;

(3) adopt fast to dividing an algorithm to solve bottom problem P1, concrete steps are as follows:

Step 3.1: after receiving the scheduling signals A sent from top layer problem P2, judge set omega, Ω={ a _i| a _iwhether=1} is feasible, if so Ω set is infeasible, jumps to step 3.6, otherwise carry out step 3.2;

Step 3.2: the bound determining λ, ${\mathrm{\&lambda;}}_{\min }=0,{\mathrm{\&lambda;}}_{\max }=\underset{i\&Element;\mathrm{\&Omega;}}{\max }\left({-D}_i^{\&prime;}\left(x_{i,\min }\right)\right);$

Step 3.3: calculate $\mathrm{\&lambda;}=\frac{{\mathrm{\&lambda;}}_{\min }+{\mathrm{\&lambda;}}_{\max }}{2};$

Step 3.4: calculate if the value obtained in interval [-η, η], then jumps to step 3.5; If the value obtained is greater than η, by λ _min=λ, if the value obtained is less than-η, by λ _max=λ, then returns step 3.3;

Step 3.5: computing formula and obtain result;

Step 3.6: terminate;

(4) in step (3), complete solving of bottom problem P1, based on the result of bottom problem P1, utilize the Gibbs sampling method with alternating temperature process to solve top layer problem P2, concrete steps are as follows:

Step 4.1: initialization A, γ ₀, k;

Step 4.2: start gibbs sampler iteration;

Step 4.3: utilize formula upgrade current temperature;

Step 4.4: the random integer M produced in [1, I];

Step 4.5: utilize bottom problem solving algorithm in step (3), calculates c (A _m, b _m);

Step 4.6: utilize formula $\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=\frac{1}{{1+e}^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}},$ Calculate π (A _m, b _m);

Step 4.7: the random several index produced between [0,1];

Step 4.8: judge index < π (A _m, b _m), a if so, then in A _m=b _motherwise, in A

Step 4.9: the need of renewal temperature, if so, then change k, and return step 4.3;

Step 4.10: judge whether iteration completes, if not, then return step 4.2;

Step 4.11: terminate;

Further, in described step 4.6, the expression formula of the probability distribution function of gibbs sampler is:

$\mathrm{\&pi;}\left(A\right)=\frac{1}{Z}{e}^{-\mathrm{\&gamma;C}\left(A\right)}$

In above formula, each parameter is defined as follows:

A: one group of vector, Λ represents the set of institute's directed quantity, A ∈ Λ;

γ a: controling parameters of gibbs sampler, γ > 0, is referred to as to accept the factor by γ, which control the speed that gibbs sampler reaches stable π (A), and physical significance is the accommodation degree to variation result;

C (A): objective function;

Z: ∑ _{a ' ∈ Λ}e ^{-γ C (A ')}, for the normalized of probability;

According to top layer problem P2 and gibbs sampler probability distribution function formula obtain:

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{{\mathrm{\&Sigma;}}_{\&ForAll;b_m^{*}\&Element;B_m}{e}^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m^{*})}}$

In above formula, each parameter is defined as follows:

B _m: all possible b _m, b _m∈ B _m;

π (A _m, b _m): at known A _mand B _mwhen, use b _mgo to replace a _mprobability;

Formula in this problem $\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{{\mathrm{\&Sigma;}}_{\&ForAll;b_m^{*}\&Element;B_m}{e}^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m^{*})}}$ Be expressed as:

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=1-\mathrm{\&pi;}(A_{\backslash m},b_m)$

$\begin{array}{c}\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}\\ =\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}}\end{array}$

In addition, order dynamically accepts factor gamma and is:

$\mathrm{\&gamma;}=\frac{\log (2+k)}{{\mathrm{\&gamma;}}_0}$

In above formula, each parameter is defined as follows:

K: variable one by one, when each temperature renewal process, increases the value of k;

γ ₀: an initialized constant of needs;

According to formula $\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}}$ And formula $\mathrm{\&gamma;}=\frac{\log (2+k)}{{\mathrm{\&gamma;}}_0};$

Described step 4.8 comprises the steps:

Step 4.8.1: when k mono-timing, γ ₀larger, γ is less, so no matter when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)>0$ And time constant or work as $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)<0$ Time constant, have trend towards 1, π (A _m, b _m) to trend towards 0.5, γ less, gibbs sampler is by Stochastic choice with b _m;

Step 4.8.2: when k mono-timing, γ ₀less, γ is larger, the first situation: when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)>0$ And time constant, more level off to 0, π (A _m, b _m) more close to 1, in the case, gibbs is just with larger probability b _mgo to replace a _m; The second situation: when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)<0$ Time constant, larger, π (A _m, b _m) more close to 0, in the case, gibbs sampler is just used with larger probability go to replace a _m;

Step 4.8.3: as k → ∞, so γ is larger, the operation in step 4.8.2.

Technical conceive of the present invention is: first carry out power consumption requirements to dispatched users in intelligent grid to it and control to reach this problem of object of cutting down electric power deficit and be described, then problem is converted into mathematical optimization problem and carries out analysis modeling.This simulated target is under the requirement of satisfied reduction electric power deficit, reduces and comprises the total cost of system of user scheduling cost and user's dissatisfaction, proposes user's power consumption requirements control method with scheduling controlling for solving this model.The method utilizes gibbs sampler to carry out user and chooses, core concept is by one group of random state, according to the target preset (maximize or minimize objective function), by building its probability distribution function, carry out successive ignition, obtain one group stable and meet the state of target.In addition in order to reach the effect of convergence better, utilizing alternating temperature mechanism, dynamically will increase and accept factor gamma, the probability making the method accept the result of difference diminishes gradually, after repeatedly running, reach the effect of global convergence, thus the scheme that is resolved more rapidly and effectively.

The invention has the beneficial effects as follows, while satisfied reduction electric power deficit condition, ensure that scheduling process comprises user scheduling cost and user's dissatisfaction minimizing in the total cost of interior system, and algorithm complex in the present invention is low, has good convergence.For user, do not need all users all to participate in power consumption requirements and control just can cut down energy consumption, the impact brought due to dispatching office also drops to minimum.

Accompanying drawing explanation

Fig. 1 is the structured flowchart of top layer problem and bottom problem in the present invention.

Fig. 2 is the FB(flow block) of bottom problem algorithm.

Fig. 3 is the FB(flow block) of top layer problem algorithm.

Embodiment

Below in conjunction with accompanying drawing, the present invention will be further described.

With reference to Fig. 1 ~ Fig. 3, with user's power consumption requirements control method of scheduling controlling in a kind of intelligent grid, carrying out the method can under the requirement of the certain electric power deficit of satisfied reduction, ensure to minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction, thus obtain a preferably energy Reduced measure.The present invention is based on and comprise a bottom problem and a top layer problem (as shown in Figure 1).Bottom problem P1 solves, and when determining certain user scheduling scheme, namely when known scheduling signals A, descends the user dissatisfaction of change caused by user's power consumption is cut down most, wherein A=(a ₁, a ₂..., a _i-1, a _i).Top layer problem P2 solves, according to the result of bottom problem P1, and the scheme of further optimizing user scheduling, thus minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction.Algorithm complex in this method is low, has good convergence.Propose, with the user's power consumption requirements control method with scheduling controlling that to minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction be target, to said method comprising the steps of for the situation of cutting down certain electric power deficit in intelligent grid:

(1) in intelligent grid, control for cutting down the power consumption requirements of electric power deficit to user, the target reached meanwhile is needed to be minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction, so this optimization problem can be described as following problem:

min?θ·∑ _iρ _i·a _i+(1-θ)·∑ _iD _i(x _i)

s.t.∑ _ia _i·(d _i-x _i)≥∑ _id _i-M

$d_i>x_i\&GreaterEqual;x_{i,\min },\&ForAll;i$

Each parameter is defined as follows:

I: user i;

θ: weight coefficient, 0 < θ < 1, θ is larger, represents that, in designed target, minimizing of user scheduling cost is more important, and minimizing of user's dissatisfaction is then relatively inessential, and θ is less, then on the contrary;

ρ _i: dispatched users i participates in the scheduling cost caused by power consumption requirements control, and power consumer i is more important, ρ _ilarger;

D _i: the electrical energy demands that user i proposes in advance;

A _i: whether user i is scheduled participates in the control signal of power consumption requirements control, works as a _irepresent when=1 that user i is scheduled to participate in power consumption requirements and control, now for user i, it must cut down self power consumption, so just has x _{i, min}≤ x _i≤ d _i, wherein x _{i, min}it is the necessary minimal consumption electric energy of user i; Work as a _irepresent when=0 that user i is not scheduled and participate in power consumption requirements control, so x _i=d _i, namely the power consumption of user i is exactly the electrical energy demands that it proposes in advance;

D _i(x _i): the dissatisfaction of user i, when user i is after scheduling, x _iwith d _iwhen being more or less the same, D _i(x _i) can be very little, this represents that user exists dissatisfaction hardly when user i only needs to cut down very low amount power consumption time.When user i is after scheduling, x _iclose to x _{i, min}time, D _i(x _i) a very large value can be reached, this represent when user i need go to cut down a large amount of power consumption time, user will to electric energy scheduling very be unsatisfied with; Setting dissatisfaction D _i(x _i) be one at [x _{i, min}, d _i) convex function of upper monotone decreasing;

M: the maximum electricity that electricity provider can provide;

(2) in the optimization problem of above-mentioned (1), owing to relating to the problem of continuous variable and discrete variable combined optimization, so direct solution is whole, the case is extremely complicated, conveniently solve, the problems referred to above are split as a continuous variable optimization problem and a Discrete Variables Optimization, correspond to it an a bottom problem P1 and top layer problem P2 respectively.Bottom problem P1 solves, and when determining certain user scheduling scheme, namely when known scheduling signals A, descends the user dissatisfaction of change caused by user's power consumption is cut down most, wherein A=(a ₁, a ₂..., a _i-1, a _i).Top layer problem P2 solves, according to the result of bottom problem P1, and the scheme of further optimizing user scheduling, thus minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction.

When given scheduling signals A, solving of bottom problem P1 could start; Bottom problem P1 solution procedure is: after receiving the scheduling signals A sent from top layer problem P2, first the power consumer be scheduled is formed set omega, Ω={ a _i| a _i=1}, if so Ω set is infeasible, needs top layer problem P2 again to choose A, otherwise Ω set is feasible; Gather for feasible Ω, the optimization problem P1 of bottom problem can be described as:

P1：Ψ({a _i} _i∈Ω)＝min∑ _i∈ΩD _i(x _i)

$s.t.\underset{i\&Element;\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}{a}_i\&CenterDot;(d_i-x_i)\&GreaterEqual;\underset{i}{\mathrm{\&Sigma;}}{d}_i-M$

$d_i>x_i\&GreaterEqual;x_{i,\min },\&ForAll;i\&Element;\mathrm{\&Omega;}$

In addition, the optimization problem P2 of top layer problem can be described as:

P2：min?C(A)＝θ·∑ _iρ _i·a _i+(1-θ)·Ψ({a _i} _i∈Ω)

In problem P2, parameter is defined as follows:

C (A): total cost of system in given scheduling signals A situation;

(3) adopt fast to dividing an algorithm to solve bottom problem P1, concrete steps are as follows:

Step 3.1: after receiving the scheduling signals A sent from top layer problem P2, judge set omega, Ω={ a _i| a _iwhether=1} is feasible, if so Ω set is infeasible, jumps to step 3.6, otherwise carry out step 3.2;

Step 3.2: the bound determining λ, ${\mathrm{\&lambda;}}_{\min }=0,{\mathrm{\&lambda;}}_{\max }=\underset{i\&Element;\mathrm{\&Omega;}}{\max }(-D_i^{\&prime;}\left(x_{i,\min }\right));$

Step 3.3: calculate $\mathrm{\&lambda;}=\frac{{\mathrm{\&lambda;}}_{\min }+{\mathrm{\&lambda;}}_{\max }}{2};$

Step 3.4: calculate if the value obtained in interval [-η, η], then jumps to step 3.5.If the value obtained is greater than η, by λ _min=λ, if the value obtained is less than-η, by λ _max=λ, then returns step 3.3;

Step 3.5: computing formula and obtain result;

Step 3.6: terminate;

(4) in step (3), complete solving of bottom problem P1, based on the result of bottom problem P1, utilize the Gibbs sampling method with alternating temperature process to solve top layer problem P2, concrete steps are as follows:

Step 4.1: initialization A, γ ₀, k;

Step 4.2: start gibbs sampler iteration;

Step 4.3: utilize formula upgrade current temperature;

Step 4.4: the random integer M produced in [1, I];

Step 4.5: utilize bottom problem solving algorithm in step (3), calculates c (A _m, b _m);

Step 4.6: utilize formula $\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}}=\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}},$ Calculate π (A _m, b _m);

Step 4.7: the random several index produced between [0,1];

Step 4.8: judge index < π (A _m, b _m), a if so, then in A _m=b _motherwise, in A

Step 4.9: the need of renewal temperature, if so, then change k, and return step 4.3;

Step 4.10: judge whether iteration completes, if not, then return step 4.2;

Step 4.11: terminate;

Further, in described step 4.6, the expression formula of the probability distribution function of gibbs sampler is:

$\mathrm{\&pi;}\left(A\right)=\frac{1}{Z}{e}^{-\mathrm{\&gamma;C}\left(A\right)}$

In above formula, each parameter is defined as follows:

A: one group of vector, Λ represents the set of institute's directed quantity, A ∈ Λ;

γ a: controling parameters of gibbs sampler, γ > 0, is referred to as to accept the factor by γ, which control the speed that gibbs sampler reaches stable π (A), and physical significance is the accommodation degree to variation result;

C (A): objective function;

Z: ∑ _{a ' ∈ Λ}e ^{-γ C (A ')}, for the normalized of probability;

According to top layer problem P2 and gibbs sampler probability distribution function formula obtain:

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{{\mathrm{\&Sigma;}}_{\&ForAll;b_m^{*}\&Element;B_m}{e}^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m^{*})}}$

In above formula, each parameter is defined as follows:

B _m: all possible b _m, b _m∈ B _m;

π (A _m, b _m): at known A _mand B _mwhen, use b _mgo to replace a _mprobability;

Formula in this problem $\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{{\mathrm{\&Sigma;}}_{\&ForAll;b_m^{*}\&Element;B_m}{e}^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m^{*})}}$ Be expressed as:

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=1-\mathrm{\&pi;}(A_{\backslash m},b_m)$

$\begin{array}{c}\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}\\ =\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}}\end{array}$

In addition, order dynamically accepts factor gamma and is:

$\mathrm{\&gamma;}=\frac{\log (2+k)}{{\mathrm{\&gamma;}}_0}$

In above formula, each parameter is defined as follows:

K: variable one by one, when each temperature renewal process, increases the value of k;

γ ₀: an initialized constant of needs;

According to formula

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}}$ And formula $\mathrm{\&gamma;}=\frac{\log (2+k)}{{\mathrm{\&gamma;}}_0};$

Described step 4.8 comprises the steps:

Step 4.8.1: when k mono-timing, γ ₀larger, γ is less, so no matter when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)>0$ And time constant or work as $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)<0$ Time constant, have trend towards 1, π (A _m, b _m) to trend towards 0.5, γ less, gibbs sampler is by Stochastic choice with b _m;

Step 4.8.2: when k mono-timing, γ ₀less, γ is larger, the first situation: when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)>0$ And time constant, more level off to 0, π (A _m, b _m) more close to 1, in the case, gibbs is just with larger probability b _mgo to replace a _m; The second situation: when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)<0$ Time constant, larger, π (A _m, b _m) more close to 0, in the case, gibbs sampler is just used with larger probability go to replace a _m;

Step 4.8.3: as k → ∞, so γ is larger, the operation in step 4.8.2.

Claims (2)

1. in intelligent grid with user's power consumption requirements control method of scheduling controlling, it is characterized in that: described control method comprises the following steps:

(1) in intelligent grid, control for cutting down the power consumption requirements of electric power deficit to user, the target reached meanwhile is needed to be minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction, so user's power consumption requirements optimization problem is described as following problem:

minθ·∑ _iρ _i·α _i+(1-θ)·∑ _iD _i(x _i)

s.t.∑ _iα _i·(d _i-x _i)≥∑ _id _i-M

$d_i>x_i\&GreaterEqual;x_{i,\min },\&ForAll;i$

Each parameter is defined as follows:

I: user i;

θ: weight coefficient, 0 < θ < 1, θ is larger, represents that, in designed target, minimizing of user scheduling cost is more important, and minimizing of user's dissatisfaction is then relatively inessential, and θ is less, then on the contrary;

ρ _i: dispatched users i participates in the scheduling cost caused by power consumption requirements control, and power consumer i is more important, ρ _ilarger;

D _i: the electrical energy demands that user i proposes in advance;

α _i: whether user i is scheduled participates in the control signal of power consumption requirements control, works as α _irepresent when=1 that user i is scheduled to participate in power consumption requirements and control, now for user i, it must cut down self power consumption, so just has x _{i, min}≤ x _i≤ d _i, wherein x _{i, min}it is the necessary minimal consumption electric energy of user i; Work as α _irepresent when=0 that user i is not scheduled and participate in power consumption requirements control, so x _i=d _i, namely the power consumption of user i is exactly the electrical energy demands that it proposes in advance;

D _i(x _i): the dissatisfaction of user i, when user i is after scheduling, x _iwith d _iwhen being more or less the same, D _i(x _i) can be very little, this represents that user exists dissatisfaction hardly when user i only needs to cut down very low amount power consumption time; When user i is after scheduling, x _iclose to x _{i, min}time, D _i(x _i) a very large value can be reached, this represent when user i need go to cut down a large amount of power consumption time, user will to electric energy scheduling very be unsatisfied with; Setting dissatisfaction D _i(x _i) be one at [x _{i, min}, d _i) convex function of upper monotone decreasing;

M: the maximum electricity that electricity provider can provide;

(2) in the optimization problem of above-mentioned (1), the problems referred to above are split as a continuous variable optimization problem and a Discrete Variables Optimization, correspond to it an a bottom problem P1 and top layer problem P2 respectively; Bottom problem P1 solves, and when determining certain user scheduling scheme, namely when known scheduling signals A, descends the user dissatisfaction of change caused by user's power consumption is cut down most, wherein A=(a ₁, a ₂..., a _i-1, a _i); Top layer problem P2 solves, according to the result of bottom problem P1, and the scheme of further optimizing user scheduling, thus minimize the total cost of the system comprising user scheduling cost and user's dissatisfaction;

When given scheduling signals A, solving of bottom problem P1 could start; Bottom problem P1 solution procedure is: after receiving the scheduling signals A sent from top layer problem P2, first the power consumer be scheduled is formed set omega, Ω={ a _i| a _i=1}, if $\underset{i\&Element;\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}{a}_i\&CenterDot;(d_i-x_{i,\min })<\underset{i}{\mathrm{\&Sigma;}}{d}_i-M,$ So Ω set is infeasible, needs top layer problem P2 again to choose A, otherwise Ω set is feasible; Gather for feasible Ω, the optimization problem P1 of bottom problem is described as:

P1：Ψ({α _i} _i∈Ω)＝min∑ _i∈ΩD _i(x _i)

$s.t.\underset{i\&Element;\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}{a}_i\&CenterDot;(d_i-x_i)\&GreaterEqual;\underset{i}{\mathrm{\&Sigma;}}{d}_i-M$

$d_i>x_i\&GreaterEqual;x_{i,\min },\&ForAll;i\&Element;\mathrm{\&Omega;}$

In addition, the optimization problem P2 of top layer problem is described as:

P2：min?C(A)＝θ·∑ _iρ _i·α _i+(1-θ)·Ψ({α _i} _i∈Ω)

In problem P2, parameter is defined as follows:

C (A): total cost of system in given scheduling signals A situation;

(3) adopt fast to dividing an algorithm to solve bottom problem P1, concrete steps are as follows:

Step 3.1: after receiving the scheduling signals A sent from top layer problem P2, judge set omega, Ω={ a _i| a _iwhether=1} is feasible, if $\underset{i\&Element;\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}{a}_i\&CenterDot;(d_i-x_{i,\min })<\underset{i}{\mathrm{\&Sigma;}}{d}_i-M,$ So Ω set is infeasible, jumps to step 3.6, otherwise carry out step 3.2;

Step 3.2: the bound determining λ, λ _min=0, ${\mathrm{\&lambda;}}_{\max }=\underset{i\&Element;\mathrm{\&Omega;}}{\max }(-D_i^{\&prime;}\left(x_{i,\min }\right));$

Step 3.3: calculate $\mathrm{\&lambda;}=\frac{{\mathrm{\&lambda;}}_{\min }+{\mathrm{\&lambda;}}_{\max }}{2};$

Step 3.4: calculate $\underset{i}{\mathrm{\&Sigma;}}{d}_i-M-\underset{i\&Element;\mathrm{\&Omega;}}{\mathrm{\&Sigma;}}(d_i-\left({\left(D_i^{\&prime;}\right)}^{-1}(-\mathrm{\&lambda;})\right)),$ If the value obtained in interval [-η, η], then jumps to step 3.5; If the value obtained is greater than η, by λ _min=λ, if the value obtained is less than-η, by λ _max=λ, then returns step 3.3;

Step 3.5: computing formula and obtain result;

Step 3.6: terminate;

(4) in step (3), complete solving of bottom problem P1, based on the result of bottom problem P1, utilize the Gibbs sampling method with alternating temperature process to solve top layer problem P2, concrete steps are as follows:

Step 4.1: initialization A, γ ₀, k;

Step 4.2: start gibbs sampler iteration;

Step 4.3: utilize formula upgrade current temperature;

Step 4.4: the random integer M produced in [1, I];

Step 4.5: utilize bottom problem solving algorithm in step (3), calculates c (A _m, b _m);

Step 4.6: utilize formula $\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}},$ Calculate π (A _m, b _m);

Step 4.7: the random several index produced between [0,1];

Step 4.8: judge index < π (A _m, b _m), a if so, then in A _m=b _motherwise, in A $a_m={\stackrel{\&OverBar;}{b}}_m;$

Step 4.9: the need of renewal temperature, if so, then change k, and return step 4.3;

Step 4.10: judge whether iteration completes, if not, then return step 4.2;

Step 4.11: terminate.

2. in intelligent grid as claimed in claim 1 with user's power consumption requirements control method of scheduling controlling, it is characterized in that: in described step 4.6, the expression formula of the probability distribution function of gibbs sampler is:

$\mathrm{\&pi;}\left(A\right)=\frac{1}{Z}{e}^{-\mathrm{\&gamma;C}\left(A\right)}$

In above formula, each parameter is defined as follows:

A: one group of vector, Λ represents the set of institute's directed quantity, A ∈ Λ;

γ a: controling parameters of gibbs sampler, γ > 0, is referred to as to accept the factor by γ, which control the speed that gibbs sampler reaches stable π (A), and physical significance is the accommodation degree to variation result;

C (A): objective function;

Z: ∑ _{a ' ∈ Λ}e ^{-γ C (A ')}, for the normalized of probability;

According to top layer problem P2 and gibbs sampler probability distribution function formula obtain:

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{{\mathrm{\&Sigma;}}_{\&ForAll;b_m^{*}\&Element;B_m}{e}^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m^{*})}}$

In above formula, each parameter is defined as follows:

B _m: all possible b _m, b _m∈ B _m;

π (A _m, b _m): at known A _mand B _mwhen, use b _mgo to replace a _mprobability;

Formula in this problem $\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{{\mathrm{\&Sigma;}}_{\&ForAll;b_m^{*}\&Element;B_m}{e}^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m^{*})}}$ Be expressed as:

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=1-\mathrm{\&pi;}(A_{\backslash m},b_m)$

$\begin{array}{c}\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}\\ =\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}}\end{array}$

In addition, order dynamically accepts factor gamma and is:

$\mathrm{\&gamma;}=\frac{\log (2+k)}{{\mathrm{\&gamma;}}_0}$

In above formula, each parameter is defined as follows:

K: variable one by one, when each temperature renewal process, increases the value of k;

γ ₀: an initialized constant of needs;

According to formula

$\mathrm{\&pi;}(A_{\backslash m},b_m)=\frac{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}}{e^{-\mathrm{\&gamma;C}(A_{\backslash m},b_m)}+e^{-\mathrm{\&gamma;C}(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)}}=\frac{1}{1+e^{-\mathrm{\&gamma;}(C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m))}}$ And formula $\mathrm{\&gamma;}=\frac{\log (2+k)}{{\mathrm{\&gamma;}}_0};$

Described step 4.8 comprises the steps:

Step 4.8.1: when k mono-timing, γ ₀larger, γ is less, so no matter when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)>0$ And time constant or work as $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)<0$ Time constant, have trend towards 1, π (A _m, b _m) to trend towards 0.5, γ less, gibbs sampler is by Stochastic choice with b _m;

Step 4.8.2: when k mono-timing, γ ₀less, γ is larger, the first situation: when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)>0$ And time constant, more level off to 0, π (A _m, b _m) more close to 1, in the case, gibbs is just with larger probability b _mgo to replace a _m; The second situation: when $C(A_{\backslash m},{\stackrel{\&OverBar;}{b}}_m)-C(A_{\backslash m},b_m)<0$ Time constant, larger, π (A _m, b _m) more close to 0, in the case, gibbs sampler is just used with larger probability go to replace a _m;

Step 4.8.3: as k → ∞, so γ is larger, the operation in step 4.8.2.