CN101404413A - Idle work optimization method suitable for on-line application - Google Patents

Abstract

The invention provides a powerless optimizing method based on a primal-dual interior-point method and a taboo search method. The method combines the primal-dual interior-point method and the taboo search method together and can quickly and conveniently process the powerless optimizing calculation of a large scale electric network with a discrete variable by utilizing the characteristics of good astringency of the primal-dual interior-point method as well as the characteristic of convenient processing on the discrete variable of the taboo search method. The optimizing method is also suitable for an automatic voltage control system.

Description

A kind of idle work optimization method that is suitable for online application

Technical field

The present invention relates to electric power system and calculate the field, be specifically related to a kind of idle work optimization method that is suitable for online application.

Background technology

The quality of power supply is an important indicator of power supply quality.Guarantee that quality of voltage also will put on the agenda, and voltage control is mainly by the idle adjusting of electric company at present, idle deficiency will cause system voltage to reduce, and power consumption equipment can not make full use of, even can cause a series of accidents such as voltage collapse; And idle surplus also can worsen system voltage, the safety of harm system and equipment, and too much idle standbyly can waste unnecessary investment again.In a word,, reduce network loss, guarantee quality of voltage, power supply enterprise is very important by online idle work optimization.

Idle work optimization is one of important research content of electric power system always.For many years, people have carried out a large amount of research to this, and have obtained certain achievement.The idle work optimization computational methods are a lot, wherein mainly comprise two class methods, are the artificial intelligence approach of representative with the former antithesis interior point method traditional mathematics method that is representative with computational methods such as genetic method, TABU search promptly.Former antithesis interior point method has the fast and slack-off advantage with the increase of network size not of computational speed, but that discrete variable is handled is not strong; Though and the TABU search computational methods have the strong advantage of discrete variable disposal ability, compare with other artificial intelligence class methods computational efficiency compare very fast, but to compare computational efficiency still lower with the traditional mathematics method.With regard to solving the online idle work optimization problem of large-scale electrical power system, former antithesis interior point method can be easy to realize the requirement to speed, but is difficult to realize the selection of load tap changer and the best switching scheme of capacitor group; Though the TABU search computational methods can realize the selection of load tap changer and the best switching scheme of capacitor group at an easy rate, its iteration is a veryer long process, can not satisfy the real-time requirement of online idle work optimization.

Summary of the invention

The advantage of the comprehensive two kinds of methods of the present invention has provided two kinds of idle work optimization methods that computational methods combine, and has better solved discrete variable and speed issue, and has obtained satisfied result.

1 idle work optimization method principle based on former antithesis interior point method and TABU search computational methods

At first, determine with system's active loss minimum to be that the idle work optimization Mathematical Modeling of target is:

minf(x ₁，x ₂，x ₃) (1)

s.t.h(x ₁，x ₂，x ₃) (2)

x _1min≤x ₁≤x _1max

x _2min≤x ₂≤x _2max (3)

Formula (1) is system's active loss, and formula (2) is an equality constraint, is the power balance equation of each node.X in the formula ₁And x ₂Be constrained optimization variable, x ₁Be discrete variable, comprise that on-load tap-changing transformer decomposes head and capacitor/reactor group drops into capacity, quantity is p; x ₂Be continuous variable, but comprise reactive apparatus number and node voltage that the generator continuous reactive is regulated, quantity is q.Formula (3) is inequality constraints, is respectively x ₁And x ₂Constraint.x ₃Be unconfined optimization variable, exert oneself and other node voltage phase angle except that balance node constitutes by balancing machine meritorious.

Introduce slack variable (su, sl, sh, sw＞0), inequality constraints is transformed into equality constraint, introduce the nonnegativity restriction of logarithm barrier function cancellation slack variable, and introduce Lagrange multiplier vector y, yu, yl, yh, yw obtains Lagrangian and is:

$L=f(x_1,x_2,x_3)-y^{T}h(x_1,x_2,x_3)-y_u^T(x_1+s_u-x_{1\max })-y_1^T(x_1-s_1-x_{1\min })-y_h^T(x_2+s_h-x_{2\max })$

$-y_w^T(x_2-s_w-x_{2\min })-\mathrm{\μ}(\underset{j=1}{\overset{p}{\mathrm{\Σ}}}{\ln s}_{\mathrm{lj}}+\underset{j=1}{\overset{p}{\mathrm{\Σ}}}{s}_{\mathrm{uj}}+\underset{j=1}{\overset{q}{\mathrm{\Σ}}}{\ln s}_{\mathrm{wj}}+\underset{j=1}{\overset{q}{\mathrm{\Σ}}}{\ln s}_{\mathrm{hj}})---\left(4\right)$

Yu in the formula, yh＜0, yl, yw＞0; μ is the barrier parameter, and μ 〉=0.

Can get according to the Karush-Kuhn-Tucker optimality condition

$L_{x_1}=\▿f_{x_1}(x_1,x_2,x_3)-{\▿h}_{x_1}^T(x_1,x_2,x_3)y-y_u-y_1=0---\left(5\right)$

$L_{x_2}=\▿f_{x_2}(x_1,x_2,x_3)-{\▿h}_{x_2}^T(x_1,x_2{x}_3)y{-y}_h-y_w=0---\left(6\right)$

$L_{x_3}={\▿f}_{x_3}(x_{1,}{x}_2,x_3)-{\▿h}_{x_3}^T(x_1,x_2,x_3)y=0---\left(7\right)$

L _y＝-h(x ₁，x ₂，x ₃)＝0 (8)

L _yu＝x ₁+s _u-x _1max＝0 (9)

L _y1＝x ₁-s ₁-x _1max＝0 (10)

L _yh＝x ₂+s _h-x _2max＝0 (11)

L _yw＝x ₂-s _w-x _2max＝0 (12)

Ls _u＝S _uY _ue ₁+μe ₁＝0 (13)

Ls ₁＝S ₁Y ₁e ₁-μe ₁＝0 (14)

Ls _h＝S _hY _he ₂+μe ₂＝0 (15)

Ls _w＝S _wY _we ₂-μe ₂＝0 (16)

E in the formula ₁, e ₂Representing dimension respectively is the unit column vector of p and q; Y _u, Y _l, Y _h, Y _w, S _u, S ₁, S _h, S _wBe respectively with y _u, y _l, y _h, y _w, s _u, s _l, s _h, s _wComponent be the diagonal matrix of diagonal element.

Find the solution formula (5)～(16) with Newton method, obtain update equation and be:

$\left[\begin{array}{cccc}{w}_{11}& w_{12}& w_{13}& -\▿h_{x_1}^T\\ w_{21}& w_{22}& w_{23}& -\▿h_{x_2}^T\\ w_{31}& w_{32}& w_{33}& -\▿h_{x_3}^T\\ -\▿h_{x_1}& -\▿h_{x_2}& -\▿h_{x_3}& 0\end{array}\right]\left[\begin{array}{c}{\mathrm{\Δx}}_1\\ \mathrm{\Δ}{x}_2\\ \mathrm{\Δ}{x}_3\\ \mathrm{\Δy}\end{array}\right]=\left[\begin{array}{c}-L_{x_{1}0}-S_{u0}^{-1}(L_{s_{u}0}-Y_{u0}{L}_{y_{u}0})-S_{10}^{-1}(L_{s_{1}0}+Y_{10}{L}_{y_{1}0})\\ -L_{x_{2}0}-S_{h0}^{-1}(L_{s_{h}0}-Y_{h0}{L}_{y_{h}0})-S_{w0}^{-1}(L_{s_{w}0}+Y_{w0}{L}_{y_{w}0})\\ -L_{x_{3}0}\\ -L_{y0}\end{array}\right]---\left(17\right)$

$\mathrm{\Δ}{s}_u=-L_{y_{u}0}-{\mathrm{\Δx}}_1---\left(18\right)$

${\mathrm{\Δs}}_1=L_{y_{1}0}+{\mathrm{\Δx}}_1---\left(19\right)$

$\mathrm{\Δ}{s}_h=-L_{y_{h}0}-\mathrm{\Δ}{x}_2---\left(20\right)$

$\mathrm{\Δ}{s}_w=L_{y_{w}0}+\mathrm{\Δ}{x}_2---\left(21\right)$

${\mathrm{\Δy}}_u=-S_{u0}^{-1}[L_{s_{u}0}+Y_{u0}\mathrm{\Δ}{s}_u]---\left(22\right)$

${\mathrm{\Δy}}_1=-S_{10}^{-1}[L_{s_{1}0}+Y_{10}\mathrm{\Δ}{s}_1]---\left(23\right)$

$\mathrm{\Δ}{y}_h=-S_{h0}^{-1}[L_{s_{h}0}+Y_{h0}\mathrm{\Δ}{s}_h]---\left(24\right)$

$\mathrm{\Δ}{y}_w=-S_{w0}^{-1}[L_{s_{w}0}+Y_{w0}\mathrm{\Δ}{s}_w]---\left(25\right)$

In the following formula:

$w_{11}=\▿f_{x_1{x}_1}^2(x_1,x_2,x_3)-\underset{i=1}{\overset{2n}{\mathrm{\Σ}}}{y}_i\▿h_{{\mathrm{ix}}_1{x}_1}^2(x_1,x_2,x_3)+S_{10}^{-1}{Y}_{10}-S_{u0}^{-1}{Y}_{u0}---\left(26\right)$

$w_{22}=\▿f_{x_2{x}_2}^2(x_1,x_2,x_3)-\underset{i=1}{\overset{2n}{\mathrm{\Σ}}}{y}_i\▿h_{{\mathrm{ix}}_2{x}_2}^2(x_1,x_2,x_3)+S_{w0}^{-1}{Y}_{w0}-S_{h0}^{-1}{Y}_{h0}---\left(27\right)$

Other situations are $w_{\mathrm{kj}}=\▿f_{x_k{x}_j}^2(x_1,x_2,x_3)-\underset{i=1}{\overset{2n}{\mathrm{\Σ}}}{y}_i\▿h_{{\mathrm{ix}}_k{x}_j}^2(x_1,x_2,x_3)---\left(28\right)$

Find the solution formula (17)～(25) in succession, can obtain the correction direction Δ x of former variable and dual variable ₁, Δ x ₂, Δ x ₃, Δ y, Δ s _u, Δ s _l, Δ s _h, Δ s _w, Δ y _u, Δ y _l, Δ y _h, Δ y _w

Next step carries out the processing of discrete variable, and processing method has two kinds, and a kind of is the consolidation processing, and a kind of is to utilize the TABU search computational methods to handle.Characteristics according to them can be in the following way:

After optimize finishing again with x ₁Consolidation is carried out computation optimization one time again to nearest discrete point.Because optimize the result near optimization solution, the optimal speed after the consolidation is quite fast.But this method can only obtain an approximate suboptimal solution, even may at this time can change the TABU search computational methods over to because consolidation makes original suboptimal solution become discrete infeasible solution.

1) target function and constraint

Originally on the basis of target function, add with voltage and cross the border and generator reactive exerts oneself that to cross the border be penalty function,

$\min f^{\′}(x_1,x_2,x_3)=f(x_1,x_2,x_3)+{\mathrm{\λ}}_1\underset{i=1}{\overset{\mathrm{Nd}}{\mathrm{\Σ}}}{\left(\frac{{\mathrm{\ΔV}}_i}{V_{i\max }-V_{i\min }}\right)}^2+{\mathrm{\λ}}_2\underset{i=1}{\overset{\mathrm{Nq}}{\mathrm{\Σ}}}{\left(\frac{{\mathrm{\ΔQ}}_j}{Q_{j\max }-Q_{j\min }}\right)}^2---\left(29\right)$

${\mathrm{\&Delta;V}}_i=\left\{\begin{array}{cc}{V}_i-V_{i\max }& V_i>V_{i\max }\\ 0& V_{i\min }<V_i<V_{i\max }\\ V_{i\min }-V_i& V_i<V_{i\min }\end{array}\right.---\left(30\right)$

${\mathrm{\&Delta;Q}}_j=\left\{\begin{array}{cc}{Q}_j-Q_{j\max }& Q_j>Q_{j\max }\\ 0& Q_{j\min }<Q_j<Q_{j\max }\\ Q_{j\min }-Q_j& Q_j<Q_{j\min }\end{array}\right.---\left(31\right)$

F (x in the formula ₁, x ₂, x ₃) be network loss, Nd, Nq are respectively load bus number and generator node number; λ ₁Be the load bus voltage penalty coefficient that crosses the border; λ ₂Be the generator reactive penalty coefficient that crosses the border of exerting oneself; V _iVoltage magnitude for load bus i; Q _jFor the idle of generator j exerted oneself; Subscript m ax, min represent the bound to dependent variable respectively.

Different with former antithesis interior point method, divide control variables constraint and state variable constraint with variable bound.

Being constrained to of control variables:

$\left\{\begin{array}{c}{V}_{\mathrm{Gi}\min }\&le;V_{\mathrm{Gi}}\&le;V_{\mathrm{Gi}\max }\\ T_{i\min }\&le;T_i\&le;T_{i\max }\\ Q_{\mathrm{Ci}\min }\&le;Q_{\mathrm{Ci}}\&le;Q_{\mathrm{Ci}\max }\end{array}\right.---\left(32\right)$

Being constrained to of state variable:

$\left\{\begin{array}{c}{Q}_{\mathrm{Gi}\min }\&le;Q_{\mathrm{Gi}}\&le;Q_{\mathrm{Gi}\max }\\ V_{\mathrm{Di}\min }\&le;V_{\mathrm{Di}}\&le;V_{\mathrm{Di}\max }\end{array}\right.---\left(33\right)$

V in the formula _Gi, T _i, V _DiBe respectively generator node voltage, on-load tap-changing transformer no-load voltage ratio, load bus voltage; Subscript m ax, min are respectively to the upper and lower limit of dependent variable.

2) Bian Ma processing

The present invention adopts the decimal integer coding, and utilizes mapping method, in advance the actual parameter corresponding with each code value is put into a special array; Discrete variable can be listed all according to its actual array may value, and continuous variable is then carried out discretization by certain required precision and handled, and has simplified decode procedure like this, has saved a lot of computing times.

3) interlace operation

Adopt heuristic arithmetic to intersect at hybrid coding.Suppose that vector X represents a certain individuality in the population, each component of X is the encoded radio of control variables.If parent two individualities are X ₁, X ₂, and target function value f (X ₁)＜f (X ₂), filial generation X then ₁=(X ₁+ X ₂)/2, X ₂=X ₁-a (X ₂-X ₁)), a is the random number of interval [0,1].X ' ₂In a certain component get the boundary value (can determine) of this component when crossing the border by the constraints of corresponding control variables.Because the X ' that obtains ₁, X ' ₂In component differ to establish a capital and be integer because when adopting the decimal integer coding, thus just the numberization between [x-0.5, x+0.5] (x is an integer) to put in order be x.

4) mutation operation

Consider after trend after the consolidation is calculated, the situation of crossing the border might occur retraining, the present invention adopts even variation so that the search point can freely move in whole search volume, thereby enters area of feasible solution as early as possible.Evenly the specific operation process of variation is as follows.Suppose to have an individual X=(x of being ₁, x ₂..., x _k..., x _l), if x _kBe change point, its span is [U _{K, min}, U _{K, max}] after this point carries out even mutation operation to individual X, can obtain a new individual X=(x ₁, x ₂..., x ' _k..., x _l), wherein the new genic value of change point is x ' _k=U _{K, min}+ r (U _{K, max}-U _{K, min}), r is a random number in [0,1] scope.

5) TS moves design

The present invention adopts the single combined strategy that moves that moves and exchange to the decimal integer coding, and wherein, it is exactly two single mobile combinations that exchange is moved.The mobile design of the coded system that the present invention is adopted: 1. single moving: a certain position of picked at random sign indicating number string, and increase 1 or subtract 1 operation.2. exchange is moved: two of certain of picked at random sign indicating number string, and increase (or subtracting) 1 and subtract (or increasing) 1 operation.

6) taboo table

The taboo table is the key point of taboo computational methods, the maximum mobile number of permission access is called the scale of taboo table in the taboo table, because the taboo table all needs to upgrade in each iteration, and the general way to manage that adopts " first in first out ", so adopt the data structure of round-robin queue as the taboo table.Coded system to the present invention's employing: 1. move single, the position of record code string in the taboo table, and write down the opposite direction that it moves.2. mobile to exchanging, then write down the position of two sign indicating number strings and mobile opposite direction thereof.

7) discharge criterion

" release criterion " that the present invention adopts is: if a migration after current separating, can obtain one will good separating than any separate all that searched in the past, then claim to move and satisfy the release criterion.

8) computational methods stop criterion

The termination criterion of computational methods comprises the comparison of two aspect maximum search number of times target function values, reach maximum search number of times or target function value or current search to optimal value when not improving, just stop search.

Because must carrying out electric power system tide in finding the solution the idle work optimization process, the TABU search computational methods calculate, utilize calculation of tidal current to carry out former antithesis interior point method simultaneously and calculate the span that also can improve initial value, thereby improve the adaptability of these computational methods, can also make former antithesis interior point method calculate the iteration convergence number of times reduces, calculate so the idle work optimization computational methods are at first carried out trend, change former antithesis interior point method again over to and calculate.The present invention adopts the state version BPA power system analysis program means PSD-BPA of Chinese DianKeYuan exploitation that system introduces to carry out trend and calculates.

System is carried out idle work optimization with former antithesis interior point method to be calculated, if calculating, interior point method do not restrain, separating with trend is that initial value carries out the calculating of TABU search computational methods, and the iterations of this moment is made as 100 times, can guarantee to try to achieve an original trend of ratio like this and separate more excellent preferably separating; If interior point method calculates convergence, then provide one group of suboptimal solution, carry out the consolidation processing with this suboptimal solution again, carry out computation optimization again one time, if convergence then finishes, if do not restrain then and change the calculating of TABU search computational methods over to as initial value, owing to be in a kind of suboptimum state this moment, so calculating convergence rate, the trend of TABU search computational methods obviously accelerates, electrical network with the IEEE30 node is an example, seeking the optimum calculating that group time needs iteration more than 100 time of planting, after adopting COMPREHENSIVE CALCULATING method (TABU search computational methods behind the former antithesis interior point method earlier), only need less than can restrain for 10 times as direct employing TABU search computational methods, saved a large amount of computing times.

After carrying out former antithesis interior point method calculating, load tap changer and capacitor group all have a definite optimal value, just this optimal value is that continuous variable causes rather than caused by discrete variable, so with this optimal value is that a small range is determined at the center, dwindled the random search scope of TABU search computational methods, made the easier optimum point that searches of TABU search computational methods.

In sum, the advantage that the computational speed that the idle work optimization method that former antithesis interior point method combines with the TABU search computational methods had both had an interior point method soon and does not reduce with the increase of network size also has the strong advantage of processing discrete variable of TABU search computational methods; Under the situation that former antithesis interior point method is not restrained, can try to achieve than original trend calculating by the TABU search computational methods and separate more excellent separating preferably simultaneously.Therefore, be a kind of more satisfactory, practical calculation method for handling online idle work optimization calculating.

2 idle work optimization method technical steps based on former antithesis interior point method and TABU search computational methods

1) input electrical network parameter carries out the PSD-BPA trend and calculates;

2) set up the idle work optimization model;

3) adopt former antithesis interior point method to carry out idle work optimization;

4) if former antithesis interior point method computational methods do not restrain, establish TABU search computational methods iterations K=100, change step 7; If former antithesis interior point method computational methods convergence is then carried out the consolidation discretization and is handled;

5) carrying out idle work optimization with former antithesis interior point method once more calculates;

6) if former antithesis interior point method computational methods do not restrain, establish TABU search computational methods iterations K=10; If former antithesis interior point method computational methods convergence, then integrated approach finishes;

7) adopt the TABU search computational methods to carry out idle work optimization and calculate, integrated approach finishes.

Therefore, the present invention proposes a kind of idle work optimization method that is suitable for online application, it is characterized in that may further comprise the steps:

1) input comprises relevant load, the associated electrical network parameters of generator, circuit, transformer, and utilization trend computational tool carries out trend and calculates;

2) set up two kinds of idle work optimization Mathematical Modelings:

The one, former antithesis interior point method idle work optimization Mathematical Modeling, promptly

minf(x ₁，x ₂，x ₃) (1)

s.t.h(x ₁，x ₂，x ₃) (2)

x _1min≤x ₁≤x _1max

x _2min≤x ₂≤x _2max (3)

Formula (1) is system's active loss, and formula (2) is an equality constraint, is the power balance equation of each node, the x in the formula ₁And x ₂Be constrained optimization variable, x ₁Be discrete variable, comprise that on-load tap-changing transformer decomposes head and capacitor or reactor group and drops into capacity, quantity is p; x ₂Be continuous variable, but comprise reactive apparatus number and node voltage that the generator continuous reactive is regulated, quantity is q, and formula (3) is inequality constraints, is respectively x ₁And x ₂Constraint, x ₃Be unconfined optimization variable, exert oneself and other node voltage phase angle except that balance node constitutes by balancing machine meritorious;

The 2nd, TABU search computational methods idle work optimization Mathematical Modeling on the basis of original target function, adds with voltage and crosses the border and generator reactive exerts oneself that to cross the border be penalty function,

$\min f^{\&prime;}(x_1,x_2,x_3)=f(x_1,x_2,x_3)+{\mathrm{\&lambda;}}_1\underset{i=1}{\overset{\mathrm{Nd}}{\mathrm{\&Sigma;}}}{\left(\frac{{\mathrm{\&Delta;V}}_i}{V_{i\max }-V_{i\min }}\right)}^2+{\mathrm{\&lambda;}}_2\underset{i=1}{\overset{\mathrm{Nq}}{\mathrm{\&Sigma;}}}{\left(\frac{{\mathrm{\&Delta;Q}}_j}{Q_{j\max }-Q_{j\min }}\right)}^2$

${\mathrm{\&Delta;V}}_i=\left\{\begin{array}{cc}{V}_i-V_{i\max }& V_i>V_{i\max }\\ 0& V_{i\min }<V_i<V_{i\max }\\ V_{i\min }-V_i& V_i<V_{i\min }\end{array}\right.$

$\mathrm{\&Delta;}{Q}_j=\left\{\begin{array}{cc}{Q}_j-Q_{j\max }& Q_j>Q_{j\max }\\ 0& Q_{j\min }<Q_j<Q_{j\max }\\ Q_{j\min }-Q_j& Q_j<Q_{j\min }\end{array}\right.$

F (x in the formula ₁, x ₂, x ₃) being active loss, Nd, Nq are respectively load bus number and generator node number; λ ₁Be the load bus voltage penalty coefficient that crosses the border; λ ₂Be the generator reactive penalty coefficient that crosses the border of exerting oneself; V _iVoltage magnitude for load bus i; Q _jFor the idle of generator j exerted oneself; Subscript m ax, min represent the bound to dependent variable respectively,

Variable bound is divided into control variables constraint and state variable constraint:

Being constrained to of control variables:

$\left\{\begin{array}{c}{V}_{\mathrm{Gi}\min }\&le;V_{\mathrm{Gi}}\&le;V_{\mathrm{Gi}\max }\\ T_{i\min }\&le;T_i\&le;T_{i\max }\\ Q_{\mathrm{Ci}\min }\&le;Q_{\mathrm{Ci}}\&le;Q_{\mathrm{Ci}\max }\end{array}\right.$

Being constrained to of state variable:

$\left\{\begin{array}{c}{Q}_{\mathrm{Gi}\min }\&le;Q_{\mathrm{Gi}}\&le;Q_{\mathrm{Gi}\max }\\ V_{\mathrm{Di}\min }\&le;V_{\mathrm{Di}}\&le;V_{\mathrm{Di}\max }\end{array}\right.$

V in the formula _Gi, T _i, V _DiBe respectively generator node voltage, on-load tap-changing transformer no-load voltage ratio, load bus voltage; Subscript m ax, min are respectively to the upper and lower limit of dependent variable.

3) described former antithesis interior point method being carried out idle work optimization calculates;

4) if described former antithesis interior point method computational methods do not restrain, the TABU search computational methods iterations of establishing in the idle work optimization method is 100, changes step 7; If the former antithesis interior point method computational methods convergence in the described idle work optimization method is then carried out the consolidation discretization and is handled

5) carrying out idle work optimization with the former antithesis interior point method in the described idle work optimization method of claim 1 once more calculates;

6) if the former antithesis interior point method computational methods in the described idle work optimization method do not restrain, the TABU search computational methods iterations of establishing in the described idle work optimization method is 10; If described former antithesis interior point method computational methods convergence, then integrated approach finishes;

7) adopt described TABU search computational methods to carry out idle work optimization and calculate, comprehensive optimization method finishes.

Description of drawings

Fig. 1 is according to the flow chart that is suitable for the idle work optimization method of online application of the present invention.

Embodiment

Below be an embodiment of the inventive method, carry out l-G simulation test with Chongqing electricity grid real time data section one day in 2008 and make embodiment, further specify as follows:

This real time data section has 305 buses, 132 circuits, and 91 transformers, active loss is 74.5MW.In order just with relatively to describe, integrated approach during the present invention will invent is divided into three flow processs, as show shown in the 1-3, step 1-2-3-4-5-6 in the flow process 1 expression calculation process, step 1-2-3-4-5-6-7 in the flow process 2 expression calculation process, step 1-2-3-4-7 in the flow process 3 expression calculation process, if having one and above flow process to calculate successfully to represent this COMPREHENSIVE CALCULATING method calculates successfully, if have two and above flow process to calculate successfully, this COMPREHENSIVE CALCULATING method is got computing time than the flow process of lacking by flow chart.

1) if the system agreement upper voltage limit of setting up departments is 1.1, this moment, former antithesis interior point method convergence was recomputated success after the discrete variable consolidation, and active loss is 72.8MW, and be 4.12 seconds computing time, and result of calculation sees Table 2.Flow process 1, flow process 2 and flow process 3 all restrain.This integrated approach is got flow process 1.

2) if the system agreement upper voltage limit of setting up departments is 1.09, this moment, the unsuccessful taboo search method that changes over to was recomputated in former antithesis interior point method convergence after the discrete variable consolidation, each node voltage is in acceptability limit, active loss is 73.5MW, and be 10.23 seconds computing time, and result of calculation sees Table 3.Flow process 1 does not restrain, flow process 2 and flow process 3 convergences.This integrated approach is got flow process 2.

3) if the system agreement upper voltage limit of setting up departments is 1.08, this moment, former antithesis interior point method failure changed taboo search method over to, calculated successfully, and all in acceptability limit, active loss is 74.1MW to each node voltage, and be 30.57 seconds computing time, and result of calculation sees Table 4.Flow process 1 and flow process 2 do not restrain, flow process 3 convergences.This integrated approach is got flow process 3.

Table 1 upper voltage limit is 1.1 o'clock idle work optimization result of calculation

Table 2 upper voltage limit is 1.09 o'clock idle work optimization result of calculation

Table 3 upper voltage limit is 1.08 o'clock idle work optimization result of calculation

As seen through the above analysis, utilize this integrated approach to have better convergence performance, be tending towards under the situation of strictness in the voltage constraint, can both draw comparatively ideal result of calculation, though prolong computing time to some extent, but also in the acceptable scope of automatic voltage control system, can be used for online application.

The present invention has been described according to preferred embodiment.Obviously, reading and understanding above-mentioned detailed description postscript and can make multiple correction and replacement.What this invention is intended to is that the application is built into all these corrections and the replacement that has comprised within the scope that falls into the appended claims or its equivalent.

Claims (1)

1, a kind of idle work optimization method that is suitable for online application is characterized in that may further comprise the steps:

1) input comprises relevant load, the associated electrical network parameters of generator, circuit, transformer, and utilization trend computational tool carries out trend and calculates;

2) set up two kinds of idle work optimization Mathematical Modelings:

The one, former antithesis interior point method idle work optimization Mathematical Modeling, promptly

min?f(x ₁，x ₂，x ₃) (1)

s.t.?h(x ₁，x ₂，x ₃) (2)

x _1min≤x ₁≤x _1max

x _2min≤x ₂≤x _2max (3)

Formula (1) is system's active loss, and formula (2) is an equality constraint, is the power balance equation of each node, the x in the formula ₁And x ₂Be constrained optimization variable, x ₁Be discrete variable, comprise that on-load tap-changing transformer decomposes head and capacitor or reactor group and drops into capacity, quantity is p; x ₂Be continuous variable, but comprise reactive apparatus number and node voltage that the generator continuous reactive is regulated, quantity is q, and formula (3) is inequality constraints, is respectively x ₁And x ₂Constraint, x ₃Be unconfined optimization variable, exert oneself and other node voltage phase angle except that balance node constitutes by balancing machine meritorious;

The 2nd, TABU search computational methods idle work optimization Mathematical Modeling on the basis of original target function, adds with voltage and crosses the border and generator reactive exerts oneself that to cross the border be penalty function,

$\min f^{\&prime;}(x_1,x_2,x_3)=f(x_1,x_2,x_3)+{\mathrm{\&lambda;}}_1\underset{i=1}{\overset{\mathrm{Nd}}{\mathrm{\&Sigma;}}}{\left(\frac{\mathrm{\&Delta;}{V}_i}{V_{i\max }-V_{i\min }}\right)}^2+{\mathrm{\&lambda;}}_2\underset{i=1}{\overset{\mathrm{Nq}}{\mathrm{\&Sigma;}}}{\left(\frac{\mathrm{\&Delta;}{Q}_j}{Q_{j\max }-Q_{j\min }}\right)}^2$

$\mathrm{\&Delta;}{V}_i=\left\{\begin{array}{cc}{V}_i-V_{i\max }& V_i>V_{i\max }\\ 0& V_{i\min }<V_i<V_{i\max }\\ V_{i\min }-V_i& V_i<V_{i\min }\end{array}\right.$

$\mathrm{\&Delta;}{Q}_j=\left\{\begin{array}{cc}{Q}_j-Q_{j\max }& Q_j>Q_{j\max }\\ 0& Q_{j\min }<Q_j<Q_{j\max }\\ Q_{j\min }-Q_j& Q_j<Q_{j\min }\end{array}\right.$

F (x in the formula ₁, x ₂, x ₃) being active loss, Nd, Nq are respectively load bus number and generator node number; λ ₁Be the load bus voltage penalty coefficient that crosses the border; λ ₂Be the generator reactive penalty coefficient that crosses the border of exerting oneself; V _iVoltage magnitude for load bus i; Q _jFor the idle of generator j exerted oneself; Subscript m ax, min represent the bound to dependent variable respectively,

Variable bound is divided into control variables constraint and state variable constraint:

Being constrained to of control variables:

$\left\{\begin{array}{c}{V}_{\mathrm{Gi}\min }\&le;V_{\mathrm{Gi}}\&le;V_{\mathrm{Gi}\max }\\ T_{i\min }\&le;T_i\&le;T_{i\max }\\ Q_{\mathrm{Ci}\min }\&le;Q_{\mathrm{Ci}}\&le;Q_{\mathrm{Ci}\max }\end{array}\right.$

Being constrained to of state variable:

$\left\{\begin{array}{c}{Q}_{\mathrm{Gi}\min }\&le;Q_{\mathrm{Gi}}\&le;Q_{\mathrm{Gi}\max }\\ V_{\mathrm{Di}\min }\&le;V_{\mathrm{Di}}\&le;V_{\mathrm{Di}\max }\end{array}\right.$

V in the formula _Gi, T _i, V _DiBe respectively generator node voltage, on-load tap-changing transformer no-load voltage ratio, load bus voltage; Subscript m ax, min are respectively to the upper and lower limit of dependent variable.

3) described former antithesis interior point method being carried out idle work optimization calculates;

4) if described former antithesis interior point method computational methods do not restrain, the TABU search computational methods iterations of establishing in the idle work optimization method is 100, changes step 7; If the former antithesis interior point method computational methods convergence in the described idle work optimization method is then carried out the consolidation discretization and is handled

5) carrying out idle work optimization with the former antithesis interior point method in the described idle work optimization method of claim 1 once more calculates;

6) if the former antithesis interior point method computational methods in the described idle work optimization method do not restrain, the TABU search computational methods iterations of establishing in the described idle work optimization method is 10; If described former antithesis interior point method computational methods convergence, then integrated approach finishes;

7) adopt described TABU search computational methods to carry out idle work optimization and calculate, comprehensive optimization method finishes.

Images