CN103679284A - Accommodated wind power accessed fixed interval rolling scheduling method - Google Patents

Abstract

The invention relates to an accommodated wind power accessed fixed interval rolling scheduling method comprising the following steps: 1) an intra-day rolling scheduling model of an accommodated wind power accessed power grid is established on the basis of a day-ahead plan; 2) constrained conditions of an objective function of the rolling scheduling model are established; 3) an output of power of a wind power set optimizes solutions of sub-problems in a problem; and 4) the output of power of the wind power set solves the main problem. Compared with methods in the prior art, the method has advantages that the output of power of the wind power set can be planned in a short period of time so that real-time performance of control of the wind power set can be met.

Description

A kind of fixed interval rolling scheduling method of the wind-powered electricity generation access of dissolving

Technical field

The present invention relates to a kind of wind-powered electricity generation forecasting techniques, especially relate to a kind of fixed interval rolling scheduling method of the wind-powered electricity generation access of dissolving.

Background technology

The large-scale wind power of dissolving access is significant to Operation of Electric Systems.Reason has two aspects: one, and that wind energy has is uncontrollable, randomness and the feature such as intermittent; Its two, large-scale wind power access electric system produces many-sided impact to management and running, as meritorious/reactive power flow, voltage, system stability, the quality of power supply etc.

In order to improve the access capability of electrical network to wind-powered electricity generation, wind-powered electricity generation prediction is an element task, but the accuracy of wind-powered electricity generation prediction is difficult to guarantee.Along with the growth of predicted time, predicated error can increase gradually, so the online rolling amendment of generation schedule just seems most important.Because rolling scheduling is a multi-period optimization problems, its calculated amount is very large, and therefore, its online application exists very large challenge.

Current scholar both domestic and external controls research to the operation of wind energy turbine set and mainly concentrates on the aspects such as the voltage of wind power generating set and idle control, but relatively less in the research of the online meritorious scheduling aspect of grid control centre.

ZHANG Boming and WU Wenchuan Design of a multi-time scale coordinated active power dispatching system for accommodating large scale wind power penetration (Automation of Electric Power Systems.35 (1). (2011), have provided the design framework of the Multiple Time Scales coherent system of the large-scale wind power of dissolving in pp.1-6).

SHEN Wei and WU Wenchuan article An on-line rolling dispatch method and model for accommodating large-scale wind power (Automation of Electric Power Systems.35 (22). (2011), the online rolling scheduling strategy of the large-scale wind power of dissolving proposing pp.136) has provided solution with the problem of exerting oneself that model is the conventional unit of on line refreshable, the impact on operation of power networks after can solving well wind-powered electricity generation and accessing.

Yet there are two defects in said method.

First variation district recurrence Problem.As shown in Figure 1, first, in variation district recursive process, every quantity through 1 stage subregion will increase by 1 or 2 (this depends on the position of the optimal locus of decision making on last stage), and the quadratic function analytic expression of each subregion is all different, make solution procedure more sophisticated.

Secondly, subregion is numerous will cause each decision point to change on a large scale in, so optimal strategy is acute variation, thereby each unit general adjust and exert oneself with the bound of climbing for a long time.Solving also of the optimal locus of decision making will lose due meaning.

Finally, can prove, can not occur the interlaced situation of axis of symmetry of adjacent two interval quadratic functions, its limiting case is that axis of symmetry overlaps, and this means that the mode of subregion can further be simplified.

It two is backward induction method problems, and the schematic diagram of backward induction method as shown in Figure 2.As seen from the figure, reverse decision point and the actual motion point that pushes back the starting stage that first period tries to achieve may be far apart, therefore causes unit with climbing bound, to chase for a long time the problem of decision point.

Summary of the invention

Object of the present invention is exactly that a kind of fixed interval rolling scheduling method of the wind-powered electricity generation access of dissolving is provided in order to overcome the defect of above-mentioned prior art existence.

Object of the present invention can be achieved through the following technical solutions:

A fixed interval rolling scheduling method for the wind-powered electricity generation of dissolving access, is characterized in that, comprises the following steps:

1) set up the in a few days rolling scheduling model of the wind-powered electricity generation access electrical network of dissolving based on plan a few days ago;

2) set up the bound for objective function of rolling scheduling model;

3) the solving of subproblem in unit output optimization problem;

4) unit output primal problem solves.

The in a few days rolling scheduling model of the dissolve wind-powered electricity generation access electrical network of described foundation based on plan is a few days ago specially:

Take 15min as 1 period, within 1 day, have 96 periods, each period is corresponding one by one with each stage, take T period as the 1st stage, and t period is corresponding to T-t+1 stage;

The objective function of the optimizing decision that unit solves is:

$\min {\mathrm{\&theta;}}_i\left(p_{i,t}\right)=\{\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{a_{i,t}}^{\&prime;}{p}_{i,\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">t</figure-callout>}}^2+\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{b_{i,t}}^{\&prime;}{p}_{i,t}|p_{i,t}\&Element;D_e\}$

P wherein _{i, t}for upgrading in the works the conventional unit of i platform at plan value of exerting oneself of t period, D _econstraint set, a _{i, t}' and b _{i, t}' be respectively the conventional unit of i platform from t ₀period is to quadratic term and the Monomial coefficient of total generating expense of t period;

B _{i, t}' by following formula, determined:

b _i，t′＝b _i-w _t

W _ifor Lagrange multiplier vector corresponding to equality constraint, b _iit is the Monomial coefficient of the conventional unit of i platform.

The described bound for objective function of setting up rolling scheduling model is specially:

21) unit output bound constraint

p _i，t，min≤p _i，t≤p _i，t，max

P wherein _{i, t, min}and p _{i, t, max}be respectively the upper and lower bound that i platform unit was exerted oneself in the t period;

22) unit climbing rate constraint

p _i，t-1-Δp _i，t，dn≤p _i，t≤p _i，t-1+Δp _i，t，up

Δ p wherein _{i, t, dn}with Δ p _{i, t, up}be respectively the maximal value of exerting oneself and rising from power of falling that i platform unit allows from the t-1 period to the t period;

23) balancing the load constraint

$\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{p}_{i,t}={\stackrel{\&OverBar;}{D}}_t-{\stackrel{\&OverBar;}{W}}_t$

Wherein

with

be respectively the residue system loading predicted value of period and the wind-powered electricity generation predicted value of exerting oneself.

In described unit output optimization problem, solving of subproblem is specially:

First calculate the optimizing decision of the first period, now objective function is

$\min {\mathrm{\&theta;}}_i\left(p_{i,t}\right)=\{\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{a_{i,t}}^{\&prime;}{p}_{i,\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">t</figure-callout>}}^2+\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{b_{i,t}}^{\&prime;}{p}_{i,t}|p_{i,t}\&Element;D_e\}$

Solve the first stage, can obtain

$p_{i,1,s}=-\frac{{b_{i,1}}^{\&prime;}}{2{a_{i,1}}^{\&prime;}}$

P _{i, t, s}for the extreme value of quadratic function, this value obtains at axis of symmetry place, considers the constraint of unit bound simultaneously, and the optimizing decision in the 1st stage is

p _i,1,o＝min(max(p _i,1,min′,p _i,1,s),p _i,1,max′)

${p_{i,1,\min }}^{\&prime;}=\max (p_{i,1,\min },p_{i,t_0}-\mathrm{\&Delta;}{p}_{i,1,\mathrm{up}})$

${p_{i,1,\max }}^{\&prime;}=\min (p_{i,1,\max },p_{i,t_0}+\mathrm{\&Delta;}{p}_{i,1,\mathrm{dn}})$

P wherein _{i, 1, o}be the optimizing decision value in the 1st stage, p _{i, 1, min}' and p _{i, 1, max}' consider the unit decision-making interval of the 1st period after the constraint of lower limit in fact for unit;

Decision-making interval and optimizing decision value p in the known t stage _{i, t, min}', p _{i, t, max}', p _{i, t, o}basis on, the decision-making interval that can obtain the t+1 stage is [p _{i, t+1, min}', p _{i, t+1, max}'];

p _{i，t+1，min}′＝max(p _{i，t+1，min}，p _i，t，min′-Δp _i，t+1，dn)

p _{i，t+1，max}′＝min(p _{i，t+1，max}，p _i，t，max′+Δp _i，t+1，up)

Δ p _{i, t+1, dn}with Δ p _{i, t+1, up}being respectively unit i falls and exerts oneself and maximum emersion power in the maximum of t+1 period, p _{i, t+1, max}and p _{i, t+1, min}be respectively unit i in the bound of exerting oneself of t+1 period, unit i is at the optimizing decision p of t period _{i, t, o}the interval at place is

calculate all subregions in t+1 stage:

$[\underset{\&OverBar;}{p_{i,t}}-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}},p_{i,t,o}-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}}]$

[p _i,t,o-Δp _{i,t+1,dn,pi,t,o}+Δp _i,t+1,up]

$[p_{i,t,o}+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}},\stackrel{\&OverBar;}{p_{i,t}}+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}}]$

The form that is quadratic function in each interval, its quadratic term coefficient and Monomial coefficient are respectively:

[B wherein ₂, B ₃], [B ₃, B ₄], [B ₄, B ₅] represent that correction is interval, a _{i, t}and b _{i, t}for revising front unit i at quadratic term, the Monomial coefficient of the generating expense of t period, a _{i, t}' and b _{i, t}' for to revise rear unit i at quadratic term, the Monomial coefficient of the generating expense of t period;

Try to achieve after all optimal strategies of N unit subproblem, direction and compensation are revised in circulation, after final convergence, by the interval comparison of Piecewise Quadratic Functions, can obtain the optimizing decision p in t+1 stage _{i, t+1, o}piecewise function, be the optimal strategy of required unit

$f_i\left(p\right)=\left\{\begin{array}{c}{f_i}^{t+1}\left(p\right)+{f_i}^t(p+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}}),p\&Element;[{\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">B</figure-callout>}}_2,B_3]\\ {f_i}^{t+1}\left(p\right)+{f_i}^t\left(p_{i,t,o}\right),p\&Element;[{\mathrm{<figure-callout\; id="3"\; label="B"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">B</figure-callout>}}_3,B_4]\\ {f_i}^{t+1}\left(p\right)+{f_i}^t(p-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}}),p\&Element;[B_4,B_5]\end{array}\right.$

Wherein 3 interval generating expenses of correspondence of i unit are the piecewise functions about unit output p.

Described solving of unit output primal problem is specially:

Subproblem, obtained optimization solution

prerequisite under, primal problem need to be determined the correction direction dw of variable w ^(j)with correction step-length λ ^(j);

Revise step-length λ ^(j)along with the increase of iterations j, diminish gradually, therefore revise step-length and be taken as:

${\mathrm{\&lambda;}}^{\left(j\right)}=\frac{1}{\mathrm{Aj}+B}$

In formula, A and B are rule of thumb taken as A=1, B=4, thus can revise multiplier w:

w ^(j+1)＝w ^(j)+λ ^(j)dw ^(j)

。

Compared with prior art, the present invention has the following advantages:

1,, according to the feature of unit climbing restriction and protruding optimization, the improvement dynamic optimization algorithm that has proposed the fixed interval recursion of forward of subproblem accesses the demand in line computation to meet wind energy.

2, forward recursive has been avoided the acute variation of optimal strategy, prevents that each unit from adjusting and exert oneself with climbing bound for a long time.

3, constant subregion has been avoided the increase along with recursion number of times, the problem that the interval quantity of demand solution accumulates.

The feature 4, simultaneously with rapidity, compares and can in a short period of time exerting oneself of wind-powered electricity generation unit be planned with the intelligent algorithm such as neural network, thereby meets the requirement of real-time that unit is controlled.

Accompanying drawing explanation

Fig. 1 is that traditional rolling scheduling subproblem solves schematic diagram;

Fig. 2 is optimizing decision and optimal strategy schematic diagram;

Fig. 3 is forward recursive schematic diagram;

Fig. 4 is reverse recursion schematic diagram;

Fig. 5 is constant subregion forward recursive schematic diagram.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

Embodiment

The present invention is by the in a few days rolling scheduling model setting up the wind-powered electricity generation access electrical network of dissolving based on plan a few days ago, and expansion prediction obtains on the basis of current load system predicted value and wind-powered electricity generation gross capability predicted value by a few days rolling, solve the subproblem in unit output optimization problem, obtain the value of exerting oneself of unit.Thereby reach the compromise of set optimization and counting yield, obtain more practicable decision value for the adjustment of exerting oneself of unit simultaneously.

The complexity realizing for reducing algorithm, the present invention has ignored the factor less on the impact of result, as section tidal current security constraint and the electricity contract constraint of machine group day, only consider the factor larger on the impact of result, as the constraint of unit output bound, the constraint of unit climbing rate and balancing the load constraint, take 15min as a period is optimized, within 1 day, have 96 periods.

Detailed process of the present invention is as follows:

1) set up the in a few days rolling scheduling model of the wind-powered electricity generation access electrical network of dissolving based on plan a few days ago;

2) set up the bound for objective function of rolling scheduling model;

3) the solving of subproblem in unit output optimization problem;

4) unit output primal problem solves.

The in a few days rolling scheduling model of the dissolve wind-powered electricity generation access electrical network of described foundation based on plan is a few days ago specially:

Take 15min as 1 period, within 1 day, have 96 periods, each period is corresponding one by one with each stage, take T period as the 1st stage, and t period is corresponding to T-t+1 stage;

The objective function of the optimizing decision that unit solves is:

$\min {\mathrm{\&theta;}}_i\left(p_{i,t}\right)=\{\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{a_{i,t}}^{\&prime;}{p}_{i,\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">t</figure-callout>}}^2+\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{b_{i,t}}^{\&prime;}{p}_{i,t}|p_{i,t}\&Element;D_e\}$

P wherein _{i, t}for upgrading in the works the conventional unit of i platform at plan value of exerting oneself of t period, D _econstraint set, a _{i, t}' and b _{i, t}' be respectively the conventional unit of i platform from t ₀period is to quadratic term and the Monomial coefficient of total generating expense of t period;

B _{i, t}' by following formula, determined:

b _i，t′＝b _t-w _t

W _ifor the bright H multiplier vector of glug corresponding to equality constraint.

The described bound for objective function of setting up rolling scheduling model is specially:

21) unit output bound constraint

p _i，t，min≤p _i，t≤p _i，t，max

P wherein _{i, t, min}and p _{i, t, max}be respectively the upper and lower bound that i platform unit was exerted oneself in the t period;

22) unit climbing rate constraint

p _i，t-1-Δp _i，t，dn≤p _i，t≤p _i，t-1+Δp _i，t，up

Δ p wherein _{i, t, dn}with Δ p _{i, t, up}be respectively the maximal value of exerting oneself and rising from power of falling that i platform unit allows from the t-1 period to the t period;

23) balancing the load constraint

$\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{p}_{i,j}={\stackrel{\&OverBar;}{D}}_t-{\stackrel{\&OverBar;}{W}}_t$

Wherein

with

be respectively the residue system loading predicted value of period and the wind-powered electricity generation predicted value of exerting oneself.

Subproblem has obvious multi-period characteristic, is specifically divided into 2 steps.

1) draw each stage optimizing decision of unit output

As shown in Figure 3, filled circles is wherein optimizing decision, if this unit is at the value of the exerting oneself p in t period (T-t+1 stage) _{i, t, o}can guarantee to obtain optimum from t period T period objective function.For t ₀+ 1 period, demand solution

be total to T-t ₀the optimizing decision in individual stage.

2) on the basis of optimizing decision, provide optimal strategy

As shown in Figure 2, the path that open circles wherein forms is optimal strategy.Take the optimal locus of decision making as target, take climbing rate as restriction, generate to guarantee when unit obtains optimum solution the path at day part.

As shown in Figure 4, traditional subproblem solves the reverse recursion mode that adopts.The present invention is directed to the solution procedure (as shown in Figure 5) that subproblem adopts forward recursive, generating algorithm is described in detail in embodiment, in solution procedure, using constant subregion successively, can be in the immovable situation of number of partitions, the value of exerting oneself of conventional unit is refreshed, in calculated performance with on to the frequent degree of the adjusting of unit, reach better effect.

Solving of described subproblem is specially:

First calculate the optimizing decision of the first period, now objective function is

$\min {\mathrm{\&theta;}}_i\left(p_{i,t}\right)=\{\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{a_{i,t}}^{\&prime;}{p}_{i,\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">t</figure-callout>}}^2+\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{b_{i,t}}^{\&prime;}{p}_{i,t}|p_{i,t}\&Element;D_e\}$

Solve the first stage, can obtain

$p_{i,1,s}=-\frac{{b_{i,1}}^{\&prime;}}{2{a_{i,1}}^{\&prime;}}$

Be the extreme value of quadratic function, this value obtains at axis of symmetry place, considers the constraint of unit bound simultaneously, and the optimizing decision in the 1st stage is

p _i,1,o＝min(max(p _i,1,min′,p _i,1,s),p _i,1,max′)

${p_{i,1,\min }}^{\&prime;}=\max (p_{i,1,\min },p_{i,t_0}-\mathrm{\&Delta;}{p}_{i,1,\mathrm{up}})$

${p_{i,1,\max }}^{\&prime;}=\min (p_{i,1,\max },p_{i,t_0}+\mathrm{\&Delta;}{p}_{i,1,\mathrm{dn}})$

P in the known t stage _{i, t, min}', p _{i, t, max}', p _{i, t, o}basis on, the decision-making interval that can obtain the t+1 stage is [p _{i, t+1, min}', p _{i, t+1, max}'];

p _{i，t+1，min}′＝max(p _{i，t+1，min}，p _i，t，min′-Δp _i，t+1，dn)

p _{i，t+1，max}′＝min(p _{i，t+1，max}，p _i，t，max′+Δp _i，t+1，up)

Optimizing decision p _{i, t, o}the interval at place is

calculate all subregions in t+1 stage:

$[\underset{\&OverBar;}{p_{i,t}}-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}},p_{i,t,o}-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}}]$

[p _i,t,o-Δp _{i,t+1,dn,pi,t,o}+Δp _i,t+1,up]

$[p_{i,t,o}+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}},\stackrel{\&OverBar;}{p_{i,t}}+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}}]$

The form that is quadratic function in each interval, its quadratic term coefficient and Monomial coefficient are respectively:

Interval [B wherein ₂, B ₃], [B ₃, B ₄], [B ₄, B ₅] as shown in Figure 3, try to achieve after all optimal strategies of N unit subproblem, direction and compensation are revised in circulation, after final convergence, by the interval comparison of Piecewise Quadratic Functions, can obtain the optimizing decision p in t+1 stage _{i, t+1, o}piecewise function, be the optimal strategy of required unit

$f_i\left(p\right)=\left\{\begin{array}{c}{f_i}^{t+1}\left(p\right)+{f_i}^t(p+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}}),p\&Element;[{\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">B</figure-callout>}}_2,B_3]\\ {f_i}^{t+1}\left(p\right)+{f_i}^t\left(p_{i,t,o}\right),p\&Element;[{\mathrm{<figure-callout\; id="3"\; label="B"\; filenames="HDA0000416642040000022.png"\; state="\{\{state\}\}">B</figure-callout>}}_3,B_4]\\ {f_i}^{t+1}\left(p\right)+{f_i}^t(p-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}}),p\&Element;[B_4,B_5]\end{array}\right.$

。

Solving of described primal problem is specially:

The variable of primal problem is w.When the j time iteration, the value of w is w ^(j).Subproblem, obtained optimization solution

prerequisite under, primal problem need to be determined the correction direction dw of variable w ^(j)with correction step-length λ ^(j); Revise the principle of determining the direction of steepest descent based on negative gradient of direction, can think convergence when its mould value is less than setting threshold, the unit output that now primal problem is taked is the optimum solution of former problem.

Revise step-length λ ^(j)along with the increase of iterations j, diminish gradually, therefore revise step-length and be taken as:

${\mathrm{\&lambda;}}^{\left(j\right)}=\frac{1}{\mathrm{Aj}+B}$

In formula, A and B are rule of thumb taken as A=1, B=4, thus can revise multiplier w:

w ^(j+1)＝w ^(j)+λ ^(j)dw ^(j)?。

Claims (5)

1. a fixed interval rolling scheduling method for the wind-powered electricity generation of dissolving access, is characterized in that, comprises the following steps:

1) set up the in a few days rolling scheduling model of the wind-powered electricity generation access electrical network of dissolving based on plan a few days ago;

2) set up the bound for objective function of rolling scheduling model;

3) the solving of subproblem in unit output optimization problem;

4) unit output primal problem solves.

2. fixed interval rolling scheduling method according to claim 1, is characterized in that, the in a few days rolling scheduling model of the dissolve wind-powered electricity generation access electrical network of described foundation based on plan is a few days ago specially:

Take 15min as 1 period, within 1 day, have 96 periods, each period is corresponding one by one with each stage, take T period as the 1st stage, and t period is corresponding to T-t+1 stage;

The objective function of the optimizing decision that unit solves is:

$\min {\mathrm{\&theta;}}_i\left(p_{i,t}\right)=\{\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{a_{i,t}}^{\&prime;}{p}_{i,t}^2+\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{b_{i,t}}^{\&prime;}{p}_{i,t}|p_{i,t}\&Element;D_e\}$

P wherein _{i, t}for upgrading in the works the conventional unit of i platform at plan value of exerting oneself of t period, D _econstraint set, a _{i, t}' and b _{i, t}' be respectively the conventional unit of i platform from t ₀period is to quadratic term and the Monomial coefficient of total generating expense of t period;

B _{i, t}' by following formula, determined:

b _i，t′＝b _i-w _t

W _ifor Lagrange multiplier vector corresponding to equality constraint, b _iit is the Monomial coefficient of the conventional unit of i platform.

3. fixed interval rolling scheduling method according to claim 2, is characterized in that, the described bound for objective function of setting up rolling scheduling model is specially:

21) unit output bound constraint

p _i，t，min≤p _i，t≤p _i，t，max

P wherein _{i, t, min}and p _{i, t, max}be respectively the upper and lower bound that i platform unit was exerted oneself in the t period;

22) unit climbing rate constraint

p _i，t-1-Δp _i，t，dn≤p _i，t≤p _i，t-1+Δp _i，t，up

Δ p wherein _{i, t, dn}with Δ p _{i, t, up}be respectively the maximal value of exerting oneself and rising from power of falling that i platform unit allows from the t-1 period to the t period;

23) balancing the load constraint

$\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{p}_{i,t}={\stackrel{\&OverBar;}{D}}_t-{\stackrel{\&OverBar;}{W}}_t$

Wherein

with be respectively the residue system loading predicted value of period and the wind-powered electricity generation predicted value of exerting oneself.

4. fixed interval rolling scheduling method according to claim 3, is characterized in that, in described unit output optimization problem, solving of subproblem is specially:

First calculate the optimizing decision of the first period, now objective function is

$\min {\mathrm{\&theta;}}_i\left(p_{i,t}\right)=\{\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{a_{i,t}}^{\&prime;}{p}_{i,t}^2+\underset{t=t_0+1}{\overset{T}{\mathrm{\&Sigma;}}}{b_{i,t}}^{\&prime;}{p}_{i,t}|p_{i,t}\&Element;D_e\}$

Solve the first stage, can obtain

$p_{i,1,s}=-\frac{{b_{i,1}}^{\&prime;}}{2{a_{i,1}}^{\&prime;}}$

P _{i, 1, s}for the extreme value of quadratic function, this value obtains at axis of symmetry place, considers the constraint of unit bound simultaneously, and the optimizing decision in the 1st stage is

p _i,1,o＝min(max(p _i,1,min′,p _i,1,s),p _i,1,max′)

${p_{i,1,\min }}^{\&prime;}=\max (p_{i,1,\min },p_{i,t_0}-\mathrm{\&Delta;}{p}_{i,1,\mathrm{up}})$

${p_{i,1,\max }}^{\&prime;}=\min (p_{i,1,\max },p_{i,t_0}+\mathrm{\&Delta;}{p}_{i,1,\mathrm{dn}})$

P wherein _{i, 1, o}be the optimizing decision value in the 1st stage, p _{i, 1, min}' and p _{i, 1, max}' consider the unit decision-making interval of the 1st period after the constraint of lower limit in fact for unit;

Decision-making interval and optimizing decision value p in the known t stage _{i, t, min}', p _{i, t, max}', p _{i, t, o}basis on, the decision-making interval that can obtain the t+1 stage is [p _{i, t+1, min}', p _{i, t+1, max}'];

p _{i，t+1，min}′＝max(p _{i，t+1，min}，p _i，t，min′-Δp _i，t+1，dn)

p _{i，t+1，max}′＝min(p _{i，t+1，max}，p _i，t，max′+Δp _i，t+1，up)

Δ p _{i, t+1, dn}with Δ p _{i, t+1, up}being respectively unit i falls and exerts oneself and maximum emersion power in the maximum of t+1 period, p _{i, t+1, max}and p _{i, t+1, min}be respectively unit i in the bound of exerting oneself of t+1 period, unit i is at the optimizing decision p of t period _{i, t, o}the interval at place is

, calculate all subregions in t+1 stage:

$[\underset{\&OverBar;}{p_{i,t}}-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}},p_{i,t,o}-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}}]$

[p _i,t,o-Δp _{i,t+1,dn,pi,t,o}+Δp _i,t+1,up]

$[p_{i,t,o}+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}},\stackrel{\&OverBar;}{p_{i,t}}+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}}]$

The form that is quadratic function in each interval, its quadratic term coefficient and Monomial coefficient are respectively:

[B wherein ₂, B ₃], [B ₃, B ₄], [B ₄, B ₅] represent that correction is interval, a _{i, t}and b _{i, t}for revising front unit i at quadratic term, the Monomial coefficient of the generating expense of t period, a _{i, t}' and b _{i, t}' for to revise rear unit i at quadratic term, the Monomial coefficient of the generating expense of t period;

Try to achieve after all optimal strategies of N unit subproblem, direction and compensation are revised in circulation, after final convergence, by the interval comparison of Piecewise Quadratic Functions, can obtain the optimizing decision p in t+1 stage _{i, t+1, o}piecewise function, be the optimal strategy of required unit

$f_i\left(p\right)=\left\{\begin{array}{c}{f_i}^{t+1}\left(p\right)+{f_i}^t(p+\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{dn}}),p\&Element;[B_2,B_3]\\ {f_i}^{t+1}\left(p\right)+{f_i}^t\left(p_{i,t,o}\right),p\&Element;[B_3,B_4]\\ {f_i}^{t+1}\left(p\right)+{f_i}^t(p-\mathrm{\&Delta;}{p}_{i,t+1,\mathrm{up}}),p\&Element;[B_4,B_5]\end{array}\right.$

Wherein 3 interval generating expenses of correspondence of i unit are the piecewise functions about unit output p.

5. fixed interval rolling scheduling method according to claim 4, is characterized in that, described solving of unit output primal problem is specially:

Subproblem, obtained optimization solution

prerequisite under, primal problem need to be determined the correction direction dw of variable w ^(j)with correction step-length λ ^(j);

Revise step-length λ ^(j)along with the increase of iterations j, diminish gradually, therefore revise step-length and be taken as:

${\mathrm{\&lambda;}}^{\left(j\right)}=\frac{1}{\mathrm{Aj}+B}$

In formula, A and B are rule of thumb taken as A=1, B=4, thus can revise multiplier w:

w ^(j+1)＝w ^(j)+λ ^(j)dw ^(j)。

Images