CN103219743A - Pilot node selecting method based on wind electric power fluctuation probability characters - Google Patents

Abstract

The invention provides a pilot node selecting method based on wind electric power fluctuation probability characters. Due to the fact that large-scale renewable energy resources are merged into a control area, Fluctuations and trends of currents are changed greatly, and many challenges are brought to area voltage reactive power control. Pilot nods are as primary goals of secondary voltages, when the pilot nods are selected, access of renewable energy resources needed to be considered. According to the pilot node selecting method based on the wind electric power fluctuation probability characters, access of wind power is as an example, statistics of power injection probability distribution characters of a wind power plant in different periods is carried out, various system random working states with different probabilities are formed through superposition of the power injection probability distribution characters and load working modes of peaks, waists and valleys of a system, selected pilot nodes satisfy that after the system is subject to random disturbance, voltage deviation expectation of the rest load nodes in the whole net is made to be minimum under various random working states through control and elimination of voltage deviations of the selected pilot nodes. Feasibility and validity of the pilot node selecting method based on the wind electric power fluctuation probability characters are verified by IEEE three-machine-nine-node system simulation results and New England 39-node system simulation results.

Description

Leading node selecting method based on wind power fluctuation probability nature

Technical field

The present invention relates to a kind of leading node selecting method based on wind power fluctuation probability nature.

Background technology

Secondary voltage control is as the important step of forming a connecting link in the automatism voltage control system, and keeping leading node voltage is its primary controlled target.The selection direct relation secondary voltage control of therefore leading node is for organizing and coordinating regional idle resource, keeping regional voltage levvl effect.Leading node selecting method from initial be foundation with the electrical distance, select the load bus (document 1 sees reference) of regional electrical distance center or capacity of short circuit maximum, to the node (document 2-4 sees reference) that utilizes the off-line optimization method to determine in the control area to eliminate other node voltage deviation minimums after the disturbance deviation, to the optimized choice method (document 5 sees reference) of considering the controllability factor in the recent period, comparatively complete again based on the leading node selecting method that offline optimization combinatorial problem is found the solution.The tradition electric network swim moves towards relative fixed, load prediction is higher through the years development precision, generation schedule is formulated also ripe pattern, implement in the tapping voltage control procedure, the node relative fixed of regional voltage levvl maybe can be embodied in electrical distance center in the control area, and the associated sensitivity data variation of control measure is less.Therefore according to the sensitivity data of typical operation modes, calculate by offline optimization and to select leading node comparatively ripe in theory, the engineering construction effect show in the past system of selection be feasible effectively.

Because the highlighting day by day of environmental problem, it has been trend of the times that regenerative resource is constantly incorporated in recent years.Regenerative resource such as the injecting power randomnesss such as wind-powered electricity generation, photovoltaic incorporated into are stronger, and it is extensive to insert and distributed infiltration makes electric network source present diversification and decentralized trend, have aggravated trend the control area in and have fluctuateed and move towards variation.Add that the operating point of electric power system is the convergence limit day by day because the electric power system is changed, the heavy load state down control sensitivity meeting with meritorious trend difference have greatly changed (document 6 sees reference).These factors cause electrical distance center or the regional voltage levvl typical case representation node in the control area to change and dynamic migration with trend, as can not embody the regional voltage representativeness under all running statuses by the leading node of original system of selection.

List of references:

[1]Paul?J?P,Leost?J?Y,Tesseron?J?M.Survey?of?the?Secondary?Voltage?Control?in?France:Present?Realization?and?Investigations[J].IEEE?Transactions?on?Power?Systems,1987,2(2):505-512。

[2]Conejo,A.Secondary?Voltage?Control:Nonlinear?Selection?of?Pilot?Buses,Design?of?an?Optimal?Control?Law,and?Simulation?Results[J].IEE?Proceedings-Generation,Transmission?and?Distribution,1998,145(1):77-81。

[3] Sun Yuanzhang, Wang Zhifang, Yao Xiaoyin. the research of electric power system secondary voltage control [J]. Automation of Electric Systems, 1999,23 (9): 9-14.

[4] model is of heap of stone, Chen Hang. secondary voltage Control Study (two) [J], Automation of Electric Systems, 2000,24 (12): 20-24.

[5] Dai Fei, yellow of heap of stone, Xu's arrow, Sun Yuanzhang. Henan Electric Power System subregion and leading node based on secondary voltage control are selected. protecting electrical power system and control [J], 2011,39 (24): 101-105.

[6] Guo Qing comes. electric power system classification reactive voltage Research on Closed Loop Control [D]. and Beijing: Tsing-Hua University, 2005.

[7] woods satellite, Wen Jingyu, Ai Xiaomeng, Cheng Shijie and Li Weiren. the probability distribution research [J] of wind power wave characteristic. Proceedings of the CSEE, 2012,32 (1): 38-46.

[8] Cui Yang, Mu Gang, Liu Yu and Yan Gangui. the spatial and temporal distributions characteristic [J] of wind power fluctuation. electric power network technique, 2011,35 (2): 110-114.

[9] in ocean, Han Xueshan, Liang Jun and Song Shuguang. the regional wind power wave characteristic based on NASA earth observation database is analyzed [J]. Automation of Electric Systems, 2011,35 (5): 77-81.

[10] Xiao Chuanying, Wang Ningbo, ascend crystalline substance and fourth are female. and Jiuquan wind-powered electricity generation power producing characteristics is analyzed [J]. Automation of Electric Systems, 2010,34 (17): 64-67.

[11]DupaovǒáJ.,

?N.,

?W.Scenario?Reduction?in?Stochastic?Programming:an?Approach?Using?Probability?Metrics[J].Mathematical?Programming,2003,95(3):493-511。

[12]Heitsch?H.,

?W.Scenario?Reduction?Algorithms?in?Stochastic?Programming[J].Computational?Optimization?and?Applications,2003,24(2-3):187-206。

[13] Wang Hongtao, Liu Yutian. based on the optimum reconstruct [J] of multiple target transmission of electricity rack of NSGA-II. Automation of Electric Systems, 2009,33 (23): 14-18.

[14]Kalyanmoy?Deb,Amrit?Pratap,Sameer?Agarwal.A?fast?and?elitist?multi-objective?genetic?algorithm:NSGA-II[J].IEEE?Transactions?on?Evolutionary?Computation,2002,6(2):182-197。

Summary of the invention

For make under the current form the relative regenerative resource of selected leading node at random injecting power have certain robustness, the present invention is the leading node selecting method that example has proposed to take into account wind power fluctuation probability nature with the wind-powered electricity generation access.

To achieve these goals, the present invention adopts following technical scheme.

At first add up wind energy turbine set injecting power in the control area at the peak, waist and the probability distribution of paddy load period, obtain the probability distributing density function of injecting power by the function match, and the injecting power at random of the wind energy turbine set that on peak, waist and paddy load operation mode basis, superposes, the various random walk states of formation system, its probability of occurrence is by each wind energy turbine set injecting power probability density characteristics decision.Obtain the sensitivity matrix under each running status then, choose and eliminate voltage deviation after the random perturbation and can make all the other load bus variations expect that under various running statuses minimum node is as dominating node.Simulation result shows, this method is feasible effectively, and selected result has under wind-powered electricity generation is incorporated condition into and controls effect preferably.

The core concept that leading node is selected is to be subjected to after the disturbance to make all the other node voltage deviation minimums by eliminating leading node voltage deviation.System of selection in the past only considers that typical operation modes selects at single scene.Nowadays wind-powered electricity generation is incorporated into and is caused the running status change at random frequent, and selected leading node should be able to have under various running statuses totally controls effect preferably.Because different random walk states has different probability, accumulative total is controlled the size that effect not only will be seen voltage deviation under each state, also will consider corresponding probability.Therefore, the target function selected of the leading node mathematic expectaion minimum of getting each node voltage skew under whole random walk states embodies accumulative total and controls the effect optimum.

Concrete steps comprise:

(1) at first according to incorporating the injecting power probability distribution of wind energy turbine set into, the stack peak waist paddy basic operational mode of loading is calculated each random walk state and corresponding probability distribution;

(2) on each running status flow data basis, obtain voltage power-less control associated sensitivity information then;

(3) make up the leading node selection Mathematical Modeling that adds up to control the effect optimum under the various running statuses again;

(4) use ripe optimized Algorithm at last and find the solution definite final selection result.

In the described step (1), electrical network was according to the long-time running experience accumulation in the past, and its running status is divided into peak, waist and paddy load operation mode by load level, and sums up the corresponding period comparatively accurately.Wind-powered electricity generation is incorporated in voltage power-less control field, and to bring main challenge be the change at random that the randomness of its injecting power fluctuation causes system running state.Therefore, the running status of system is divided and can be introduced the stochastic variable of representing the wind-powered electricity generation injecting power on original peak, waist and paddy load operation mode basis, forms the stochastic system running status, and its probability of occurrence is determined by wind-powered electricity generation injecting power probability density characteristics.Specifically comprise the steps:

(1-1) wind-powered electricity generation injecting power probability density characteristics statistics and match.

According to the system running state division rule, the probability nature of system's random walk state is mainly determined by the fluctuation probability density characteristics of wind-powered electricity generation injecting power.Because will superpose with the peak waist paddy load period state of system, at first should solve the probability density function of wind-powered electricity generation at corresponding period injecting power.At present the wave characteristic analysis of exerting oneself mainly concentrates on the summary (document 7-10 sees reference) of whole fluctuation pattern for wind-powered electricity generation, yet there are no the pertinent literature report for the injecting power probability distribution analysis of concrete specific period.As follows for obtaining the concrete grammar that this probability density characteristics adopts:

Add up the frequency that peak waist paddy load period wind-powered electricity generation injecting power occurs according to historical data in the different capacity scope.Be intervals of power for example, add up the meritorious frequency of occurrences of injecting of inherent specific period of each intervals of power scope, obtain the discrete probability distribution feature as its corresponding probability with 10% of rated capacity.

According to statistics, utilize function match peak waist paddy load period injecting power probability density function, through the applicability verification of historical data, formation can reflect the probability density function of injecting power fluctuation essential laws.

The probability density function that each wind-powered electricity generation injecting power has been arranged, can obtain probability distribution by intervals of power integration according to actual needs, and calculate the interior average of each intervals of power as the injecting power in this interval, calculate the trend and the corresponding probability of various random walk states so that superpose with original peak waist paddy load operation mode, selected leading node should be able to have certain robustness to the change at random of running status.

(1-2) scene of system's random walk state is cut down with probability distribution and is upgraded.

After introducing the injecting power stochastic variable, in case the control area contains a plurality of wind fields, the random walk number of states that directly makes up gained will be a magnanimity, and the Combinatorial Optimization algorithm that can cause follow-up leading node to be selected is difficult to find the solution.Can adopt two kinds of approach restriction system random walk amount of state to this.

The firstth, according to the actual conditions of system, the variation of considering injecting power has only above certain limit and just can select exert an influence to leading node, can carry out rational discretization to random variable of continuous type and handle.Be benchmark promptly, be that intervals of power is got corresponding probability with the reasonable percentage of rated capacity, and in intervals of power, get average as the injecting power under the corresponding probability with the wind field rated capacity.If the wind field that inserts in the control area is less, intervals of power can be less so that make selected leading node adapt to various random walk states as far as possible.When inserting more for a long time, the corresponding increase of intervals of power is to reduce number of combinations.For example system node i and j insert wind field, establish separately that rated capacity is respectively Sr1 and Sr2, as get 10% and be intervals of power, if the peak probability of load period node i injecting power in [0,10%Sr1] scope is ρ _Xi, average is P _I1, the probability of node j injecting power in [0,10%Sr2] scope is ρ _Xj, average is Pj1, supposes that a node i and j injecting power of giving a dinner for a visitor from afar is separate, then to be respectively the running status probability of Pi1 and Pj1 be ρ for peak load period node i and node j injecting power _x=ρ _Xiρ _XjBy this rule, consider the basic operational mode of peak waist paddy load, the random walk number of states of final combination is 3 * 10 * 10.If intervals of power gets 25%, the running status number then becomes 3 * 4 * 4.

The secondth, can cut down the balance that technology (scenario reduction) realizes precision and amount of calculation by application scenarios.Because the random walk state depends on this stochastic variable of wind-powered electricity generation injecting power, be a kind of scene therefore with every kind of wind-powered electricity generation injecting power combination and corresponding definition of probability.Each scene determines a kind of random walk state, and both are one to one.It is exactly the reduction of random walk state that scene is cut down, and purpose is to be control random walk amount of state, reduces the amount of calculation of optimized choice.At first define the probability metrics between each scene, determine the good scene quantity that needs reservation, choose again with the nearest scene of other scene probability metricses and kept, the probability that will reject scene at last integrate with the nearest reservation scene probability of its probability metrics in, form new probability distribution.The number of combinations of wind-powered electricity generation injecting power can be reduced according to this method, also subduing of random walk number of states can be realized, between amount of calculation and precision, compromise comparatively flexibly (document 11-12 sees reference).

Elect before the present invention is preferably quick and select (fast forward selection) algorithm.Flow process is as follows:

At first establish the set of whole scene sequence numbers for Ω=1,2,3 ..., S}, scene ξ ^kWith scene ξ ^uAll represent k wherein, u ∈ Ω with the vector that comprises W wind-powered electricity generation injecting power stochastic variable.The definition scene is to (ξ ^k, ξ ^u) between distance

Be respectively ξ ^kAnd ξ ^uIn element.Keep scene sequence number set initial value

Be sky, deletion scene sequence number set initial value

Step1. calculate all scenes between distance

Scene ξ ^kProbability be p _k, and calculate:

$z_u^{\left[1\right]}:=\underset{\stackrel{k=1}{k\≠u}}{\overset{S}{\mathrm{\Σ}}}{p}_k{c}_{\mathrm{ku}}^{\left[1\right]},u=\mathrm{1,2,3},...,S$

Choose $u_1=\arg \underset{u\∈\{\mathrm{1,2,3}...,S\}}{\min }{z}_u^{\left[1\right]},$

Order ${\mathrm{\Ω}}_J^{\left[1\right]}={\mathrm{\Ω}}_J^{\left[0\right]}\backslash \left\{u_1\right\},$ ${\mathrm{\Ω}}_s^{\left[1\right]}={\mathrm{\Ω}}_s^{\left[0\right]}\∪\left\{u_1\right\},$

Stepi. iterate calculating:

$c_{\mathrm{ku}}^{\left[i\right]}:=\min \{c_{\mathrm{ku}}^{[i-1]},c_{{\mathrm{ku}}_{i-1}}^{[i-1]}\},k,u\∈{\mathrm{\Ω}}_J^{[i-1]},u_{i-1}\∈{\mathrm{\Ω}}_S^{[i-1]},$

With Be respectively the set of deletion scene and the set of reservation scene, i 〉=2 in i-1 step.

$z_u^{\left[i\right]}:=\underset{k\∈{\mathrm{\Ω}}_J^{[i-1]}\backslash \left\{u\right\}}{\mathrm{\Σ}}{p}_k{c}_{\mathrm{ku}}^{\left[i\right]},u\∈{\mathrm{\Ω}}_J^{[i-1]}$

Select $u_i=\arg \underset{u\∈{\mathrm{\Ω}}_J^{[i-1]}}{\min }{z}_u^{\left[i\right]}$ And order

${\mathrm{\Ω}}_J^{\left[i\right]}={\mathrm{\Ω}}_J^{[i-1]}\backslash \left\{u_i\right\},{\mathrm{\Ω}}_s^{\left[i\right]}={\mathrm{\Ω}}_s^{[i-1]}\∪\left\{u_i\right\},$ It is the reservation scene set in i-1 step.

Stepi+1. judge

Whether interior element quantity reaches set point, if reach then stop.Otherwise turn to Stepi.

After finishing the selection that keeps scene, deletion scene probability is added on the scene probability nearest with it in the reservation scene, forms new probability distribution.

(1-3) injecting power at random of wind energy turbine set and original peak waist paddy load operation mode are superposeed calculate system's random walk state and the corresponding probability distribution of obtaining reflection wind-powered electricity generation injecting power fluctuation probability nature.The change at random of selected leading node answering system state has certain robustness, to guarantee that accumulative total is controlled the effect optimum under the various random walk states.

In the described step (2), the sensitivity relation that is located under the system running state i is:

$\left[\begin{array}{cc}{S}_{\mathrm{GG}}^i& S_{\mathrm{GL}}^i\\ S_{\mathrm{LG}}^i& S_{\mathrm{LL}}^i\end{array}\right]\left[\begin{array}{c}\mathrm{\Δ}{V}_G^i\\ \mathrm{\Δ}{V}_L^i\end{array}\right]=\left[\begin{array}{c}\mathrm{\Δ}{Q}_G^i\\ \mathrm{\Δ}{Q}_L^i\end{array}\right]---\left(1\right)$

With

Be respectively the voltage and the idle variation of generator node under the system running state i;

With

Voltage and idle variation for load bus;

With

Be sensitivity matrix, promptly in the power flow equation Jacobian matrix with voltage, idle relevant part.

In the described step (3), establish and comprise N PQ node in the control area, choose P leading node.On the basis of formula (1), be subjected to load or burden without work disturbance at random under the various random walk states after, the target of selecting leading node is how to choose P * leading node selection matrix C=[c of N dimension _Ij], make the variation Δ V of all the other load buses of the whole network _LExpectation is minimum.Promptly make target function I (C):

$\begin{array}{c}I\left(C\right)=E\left\{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\right[{\left(\mathrm{\Δ}{V}_L^i\right)}^T{Q}_x^i\mathrm{\Δ}{V}_L^i\left]\right\}\\ =\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_{i}E\left\{\right[{\left(\mathrm{\Δ}{V}_L^i\right)}^T{Q}_x^i\mathrm{\Δ}{V}_L^i\left]\right\}\end{array}---\left(2\right)$

Reach minimum.Wherein E{} is for asking for mathematic expectaion, ρ _iBe the probability that running status i occurs, N is the load bus number, and M is the total quantity of running status, and T is a transposition.The value rule of cij be with all load buses from 1 to N numbering, if j load bus is chosen as the individual leading node of i in the system, cij=1 then, otherwise cij=0.

Be the diagonal angle weight matrices, can determine its concrete numerical value according to the relative importance of loading under the different system running status.According to the derivation of (list of references 3), as the formula (3) for the relation between voltage deviation under the running status i and leading node selection matrix and the disturbance quantity:

${\mathrm{\ΔV}}_L^i=(I-B_i{F}_{i}C)M_i{\mathrm{\ΔQ}}_L^i---\left(3\right)$

Wherein, I is a unit matrix, $M_i=S_{\mathrm{LL}}^{i-1},$ $B_i=-S_{\mathrm{LL}}^{i-1}{S}_{\mathrm{LG}}^i,F_i={\left({\mathrm{CB}}_i\right)}^T{\left({\mathrm{CB}}_i{B}_i^T{C}^T\right)}^{-1},$ The suffered disturbance of system is the gaussian random load disturbance under this state, and its desired value is zero, and standard deviation is proportional to disturbance front nodal point load or burden without work.Covariance matrix under the definition status i:

$C_{\mathrm{LL}}^i=E\left\{{\mathrm{\ΔQ}}_L^{\mathrm{iT}}{\mathrm{\ΔQ}}_L^i\right\}---\left(4\right)$

In formula (3), (4) substitution formula (2), can carry out following derivation by the character of mathematic expectaion again:

$\begin{array}{c}I\left(C\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_{i}E\left\{\right[(I-B_i{F}_{i}C)M_i\mathrm{\Δ}{Q}_L^i{]}^T{Q}_x^i\\ \left[(I-B_i{F}_{i}C)M_i\mathrm{\Δ}{Q}_L^i\right]\}\end{array}=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\mathrm{trace}\left\{P_L^i{Q}_x^i\right\}-\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\mathrm{trace}\left\{(2H_1^i-H_2^i{H}_3^{i^{-1}}{H}_4^i)H_3^{i^{-1}}\right\}$ Wherein, $P_L^i=M_i{C}_{\mathrm{LL}}^i{M}_i^T,$ $H_1^i={\mathrm{CP}}_L^i{Q}_x^i{B}_i{B}_i^T{C}^T,$ $H_2^i={\mathrm{CB}}_i{B}_i^T{Q}_x^i{B}_i{B}_i^T{C}^T,$ $H_3^i={\mathrm{CB}}_i{B}_i^T{C}^T,$ $H_4^i={\mathrm{CP}}_L^i{C}^T.$ Order $f\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\mathrm{trace}\left\{(2H_1^i-H_2^i{H}_3^{i^{-1}}{H}_4^i)H_3^{i^{-1}}\right\},$ Because Irrelevant with leading node selection matrix, so former target function is equivalent to maxf (C), promptly how chooses leading node selection matrix C and make target function f (C) value reach maximum.

In the described step (4), it is a typical combinatorial optimization problem that the leading node of considering various random walk states is chosen Mathematical Modeling, it is found the solution and can adopt in the document in the past comparatively ripely such as optimized Algorithm such as intelligent search, obtains leading node selection result.The present invention has selected the genetic algorithm NSGA-II that comprises elitism strategy for use.This algorithm is the SGA(standard genetic algorithm) improvement, the defect individual that its elitism strategy can keep in the parent directly enters filial generation, loses with the optimal solution that prevents to obtain." precocity " problem that this algorithm not only can overcome in the genetic algorithm in handling the single goal problem is avoided locally optimal solution, and possesses the multiple-objection optimization ability, quick non-domination ordering in the algorithm can be according to the noninferior solution level of individuality to the population layering, undertaken guiding search to carry out by calculating individual crowding distance again to the optimal solution set direction with layer ordering.This algorithm is comparatively practical for actual engineering construction, is the optimal solution set ordering owing to what finally provide, rather than an independent optimal solution, therefore can carry out the secondary decision-making according to system's actual conditions in the engineering construction and select to need not to recomputate.The algorithm detail is referring to list of references 13-14.

Beneficial effect of the present invention: wind-powered electricity generation is incorporated the randomness that operation brings running status to change to system into, and for guaranteeing regional voltage control effect, leading node is as the secondary voltage control target, and it selects the change at random of relative wind-powered electricity generation injecting power should have robustness.For this reason, the present invention is based on wind-powered electricity generation injecting power in the statistics control area at the peak, the load probability distribution of period of waist and paddy, stack peak, waist and paddy load operation mode form the various random walk states of system, have proposed to consider the leading node selecting method of wind-powered electricity generation injecting power fluctuation probability nature.This method has overcome the deficiency that traditional system of selection can only be chosen at single running status, take into full account wind-powered electricity generation and inject the system running state change at random that causes, selected result can better adapt to because regenerative resource is incorporated the system running state random fluctuation that causes into.Simulation result shows, this method is feasible effectively, and wind-powered electricity generation is incorporated into and compared conventional method under the condition and have better control effect.

Description of drawings

Fig. 1 is a schematic diagram of the present invention.

Fig. 2 is the IEEE3 machine 9 node test system diagrams that insert wind energy turbine set.

Fig. 3 is a New England39 node system structure chart.

Fig. 4 is a variation expectation curve comparison diagram.

Fig. 5 adds the variation expectation curve comparison diagram that scene is cut down.

Embodiment

The invention will be further described below in conjunction with accompanying drawing and embodiment.

For checking institute of the present invention extracting method, IEEE3 machine 9 nodes and New England39 node system have been carried out simulation calculation.IEEE3 machine 9 node system scales are less, add the comparatively careful probability distribution of wind-powered electricity generation, by the influence of traversal search analytical system random walk state variation to leading node selection.Add rough relatively wind-powered electricity generation probability distribution by New England39 node system again, form multiple random walk state, and application scenarios reduction technology control random walk amount of state, adopt the optimizing of NSGA-II algorithm to choose leading node at last and list of references 2 compares, show the feasibility and the validity of the inventive method.

Embodiment 1:IEEE3 machine 9 node system examples.

Node 3 in IEEE3 machine 9 node systems is changed to wind energy turbine set, and system configuration as shown in Figure 2.

Table 1 peak, waist and paddy load be each node burden with power table down.The peak of node 5, node 6 and node 8, waist and paddy typical load perunit value are as shown in table 1.Power factor gets 0.98.The rated capacity of wind-powered electricity generation is got 30% of total load under the peak load pattern.

Table 1

According to the measured data of Shandong Power wind field, injecting power is that intervals of power statistics peak, waist paddy load period probability distribution are as shown in table 2 by 10% wind field rated capacity, and table 2 is the probability distribution table of peak, waist and paddy load period injecting power.Because of having only a stochastic variable, the probability of system's random walk state is identical with table 2.The peak waist paddy load period is respectively: the peak load period be 8 of mornings of every day to 12 noon, at 14 in afternoon is to point in evenings 21.The paddy load is that 0 of night is to point in mornings 6 period.All the other periods are the waist load period.For simplicity, do not carry out the function match of probability distribution, directly the applied statistics result is as basis.

Table 2

Because injecting power is 0 greater than the probability of 80% rated capacity, so the interior at interval injecting power of corresponding power is no longer considered.It is 0 and 9 kinds of running statuses of each intervals of power average that system running state can be divided into injecting power the period respectively at the Feng Yaogu load.Injecting power average in peak, waist and each intervals of power of paddy load day part is as shown in table 3, establishes the complete local compensation of reactive power that wind field consumes.Table 3 is the injecting power average table (unit: rated capacity percentage) in peak, waist and each intervals of power of paddy load day part.

Table 3

If the control unit is G1, leading node selects one in node 3,5,6,8.Control unit G1 meritorious exert oneself be made as total burden with power deduct wind-powered electricity generation exert oneself the back surplus burden with power 55%, all the other burdens with power come balance by node 1.After trend calculating, available " perturbation method " calculates sensitivity matrix B and the M under each running status.

The load or burden without work disturbance that each running status lower node 3,5,6,8 is applied is that one group of expectation is 0, and standard deviation is proportional to the gaussian random disturbance of load or burden without work, and calculates the covariance matrix CLL under each state.Be the access control effect, after applying random perturbation under each running status, eliminate the voltage deviation of node 3,5,6,8 successively, calculate the preceding voltage deviation of the relative disturbance of all the other node voltages.

If according to system of selection in the past, under the different running statuses, the leading node of selection is dynamic change.For example peak load each running status of period injecting power by 0 to 10% rated capacity scope in the time, node 3 is best as leading node control effect, i.e. the voltage deviation minimum of all the other nodes.But when injecting power increases to 20%, select node 8 control effects best.Become node 3 when 30%-70% again, 70%-80% then transfers node 8 once more to.The reason that this change at random produces is: in four nodes to be selected, relative other node electrical distances constantly reduce recently and with node 3 meritorious increases of injecting between node 3 and the node 8, because disturbance is one group of random number, make that the deviation size has randomness after each node voltage disturbance, the control unit eliminates that the accumulative total of all the other node voltage deviations has just had this randomness after the relevant voltage deviation.As not considering the Comprehensive Control effect of each running status, be difficult to determine final result.

According to the extracting method calculating target function f(C of institute of the present invention), Theoretical Calculation result should select node 3 as leading node f(C) maximum.Be simulating, verifying Theoretical Calculation result's correctness, consider each running status probability of occurrence, the mathematic expectaion of voltage deviation absolute value before the relative disturbance of all the other node voltages after the leading node voltage disturbance of calculating elimination.Simulation result shows, selects node 3 as all the other node accumulation voltage deviation expectation ES minimums of leading node, with accord with theoretical analysis, shows the feasibility and the correctness of institute of the present invention extracting method.Simulation result and Theoretical Calculation f(C) result contrasts as shown in table 4.Table 4 is taken the node Choice Theory as the leading factor and is calculated and the simulation result contrast table.

Table 4

Embodiment 2:New England39 node test system example.

New England39 node system structure as shown in Figure 3.

The initial load of each node all multiply by proportionality coefficient 1.3 and 1.7 and constitutes waist load and peak load operation state as paddy load operation state.Picked at random node 2,7 and 16 inserts wind field.Each wind field injecting power adopts identical historical data with 3.1 examples, and each wind field rated capacity all is made as 12% of the meritorious total amount of peak load.

If press the intervals of power of embodiment 1 example, then whole stochastic regime quantity of system are 3 * 9 * 9 * 9=2187, calculate for follow-up optimizing and bring bigger amount of calculation.Be control random walk number of states, intervals of power adopts 50%, and promptly the wind field injecting power is only considered three kinds of situations, promptly injecting power be 0, less than 50% with greater than 50% rated capacity.The random walk number of states of system becomes 3 * 3 * 3 * 3=81.Peak waist paddy load period statistical probability distributes as shown in table 5.Table 5 is the average and the probability distribution statistical table of peak waist paddy load each wind field injecting power of period.

Table 5

Can get the flow data and the probability distribution of 81 kinds of system's random walk states thus.Utilize perturbation method to ask for sensitivity data under each state, using the NSGA-II algorithm be target with maxf (c), selects 4 conducts to dominate node in 29 PQ nodes.The optimizing result takes node as the leading factor, and to be that node 9,11,17 and accumulative total is controlled effect at 28 o'clock best.List of references 2 has provided the selection result of single scene, and promptly node 18,20,22 and 26 is taken node as the leading factor.This selection be in system near making under the scene of operational limit, be equivalent to test in the example whole wind field injecting powers and be zero peak load operation state.Control effect for contrast the inventive method and list of references 2, calculate and apply the variation expectation of eliminating leading all the other load buses of node voltage deviation gained after the disturbance under all random walk states, the variation expectation that connects all PQ nodes forms curve as shown in Figure 4.

Variation expectation curve 1 is the control effect of the selected leading node of the present invention among the figure, and variation expectation curve 2 is the control effect of list of references 2 selected leading nodes.As seen from the figure, the selected leading node control effect integral body of the present invention is better than the result of list of references 2.The expectation of the variation of 1-17 node is all less than curve 2, in the middle of the node 18-29, except that 18,20,22 and 26 (these four 2 leading node voltage Deviation Control be 0) as a reference, has only 19,21,23,27 4 nodes a little more than curve 2.And curve 2 is expected then to have reached 17 greater than the node number of curve 1 except that leading node 9,11,17 and 28.Variation expectation average by each node also can be found out whole structure.Taking node voltage skew expectation average as the leading factor with 18,20,22 and 26 is 0.0413, taking node voltage skew expectation average as the leading factor with 9,11,17 and 28 is 0.0296, so the leading node selected of institute of the present invention extracting method is totally controlled effect and is better than the leading node of selecting under the single scene.

Be to reduce the amount of calculation of optimized choice, use described in 2.3 fast before elect and select algorithm, 81 scenes are reduced to 27.Be that peak, waist paddy load respectively keep 9 scenes.With the peak load is that each scene of example and probability distribution thereof are as shown in table 6 the period.Table 6 keeps scene and probability distribution table for the peak load period of scene after cutting down.

Table 6

Promptly only kept random walk state in 27 after cutting down scene.Utilize perturbation method to ask for sensitivity equally, using the NSGA-II algorithm is target with maxf (c), and the leading node of selection is 9,14,16 and 28.The contrast of control effect and whole scenes of consideration and single scene as shown in Figure 5.

Variation is expected 1 curve for considering the control effect of whole 81 scenes among the figure, and skew expectation 2 is the control effect of single scene in the list of references 2, the control effect that skew expectation 3 is cut down for scene.Selection result after scene is cut down is with to consider that whole scenes are compared the control effect slightly poor, but computation amount, has realized the balance of amount of calculation and computational accuracy preferably.To cut down each node voltage skew expectation average of back be 0.0304 to scene in addition, compare whole scenes 0.0296 and single scene 0.0413, also verified above-mentioned conclusion.

Though above-mentionedly in conjunction with the accompanying drawings the specific embodiment of the present invention is described; but be not limiting the scope of the invention; one of ordinary skill in the art should be understood that; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various modifications that creative work can make or distortion still in protection scope of the present invention.

Claims (9)

1. leading node selecting method based on wind power fluctuation probability nature, it is characterized in that, at first add up wind energy turbine set injecting power in the control area at the peak, waist and the probability distribution of paddy load period, obtain the probability distributing density function of injecting power by the function match, and the injecting power at random of the wind energy turbine set that on peak, waist and paddy load operation mode basis, superposes, the various random walk states of formation system; Obtain the sensitivity matrix under each running status then, choose and eliminate voltage deviation after the random perturbation and can make all the other load bus variations expect that under various running statuses minimum node is as dominating node.

2. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 1 is characterized in that concrete steps comprise:

(1) at first according to incorporating the injecting power probability distribution of wind energy turbine set into, the stack peak waist paddy basic operational mode of loading is calculated each random walk state and corresponding probability distribution;

(2) on each running status flow data basis, obtain voltage power-less control associated sensitivity information then;

(3) make up the leading node selection Mathematical Modeling that adds up to control the effect optimum under the various running statuses again;

(4) last optimizing application algorithm is found the solution and is determined final selection result.

3. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 1 is characterized in that, in the described step (1), specifically comprises the steps:

(1-1) wind-powered electricity generation injecting power probability density characteristics statistics and match;

Add up the frequency that peak waist paddy load period wind-powered electricity generation injecting power occurs according to historical data in the different capacity scope;

According to statistics, utilize function match peak waist paddy load period injecting power probability density function, through the applicability verification of historical data, form the probability density function of reflection injecting power fluctuation essential laws;

Intervals of power integration according to actual needs obtains probability distribution, and calculates the interior average of each intervals of power as the injecting power in this interval;

(1-2) scene of system's random walk state is cut down with probability distribution and is upgraded;

(1-3) injecting power at random of wind energy turbine set and original peak waist paddy load operation mode are superposeed calculate system's random walk state and the corresponding probability distribution of obtaining reflection wind-powered electricity generation injecting power fluctuation probability nature.

4. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 3 is characterized in that, in the described step (1-2), random variable of continuous type is carried out rational discretization handle; Be benchmark promptly, be that intervals of power is got corresponding probability with the reasonable percentage of rated capacity, and in intervals of power, get average as the injecting power under the corresponding probability with the wind field rated capacity.

5. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 3 is characterized in that in the described step (1-2), application scenarios reduction technology realizes the balance of precision and amount of calculation; At first define the probability metrics between each scene, determine the good scene quantity that needs reservation, choose again with the nearest scene of other scene probability metricses and kept, the probability that will reject scene at last integrate with the nearest reservation scene probability of its probability metrics in, form new probability distribution.

6. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 5 is characterized in that, elects before selecting for use fast and selects algorithm, and flow process is as follows:

At first establish the set of whole scene sequence numbers for Ω=1,2,3 ..., S}, scene ξ ^kWith scene ξ ^uAll represent k wherein, u ∈ Ω with the vector that comprises W wind-powered electricity generation injecting power stochastic variable; The definition scene is to (ξ ^k, ξ ^u) between distance

Be respectively ξ ^kAnd ξ ^uIn element; Keep scene sequence number set initial value

Be sky, deletion scene sequence number set initial value

Step1. calculate all scenes between distance

Scene ξ ^kProbability be p _k, and calculate:

$z_u^{\left[1\right]}:=\underset{\stackrel{k=1}{k\≠u}}{\overset{S}{\mathrm{\Σ}}}{p}_k{c}_{\mathrm{ku}}^{\left[1\right]},u=\mathrm{1,2,3}...,S$

Choose $u_1=\arg \underset{u\∈\{\mathrm{1,2,3}...,S\}}{\min }{z}_u^{\left[1\right]},$

Order ${\mathrm{\Ω}}_J^{\left[1\right]}={\mathrm{\Ω}}_J^{\left[0\right]}\backslash \left\{u_1\right\},$ ${\mathrm{\Ω}}_s^{\left[1\right]}={\mathrm{\Ω}}_s^{\left[0\right]}\∪\left\{u_1\right\}$

Stepi. iterate calculating:

$c_{\mathrm{ku}}^{\left[i\right]}:\min \{c_{\mathrm{ku}}^{[i-1]},c_{{\mathrm{ku}}_{i-1}}^{[i-1]}\},k,u\∈{\mathrm{\Ω}}_J^{[i-1]},u_{i-1}\∈{\mathrm{\Ω}}_S^{[i-1]},$

With

Be respectively the set of deletion scene and the set of reservation scene, i 〉=2 in i-1 step;

$z_u^{\left[i\right]}:=\underset{k\∈{\mathrm{\Ω}}_J^{[i-1]}\backslash \left\{u\right\}}{\mathrm{\Σ}}{p}_k{c}_{\mathrm{ku}}^{\left[i\right]},u\∈{\mathrm{\Ω}}_J^{[i-1]}$

Select $u_i=\arg \underset{u\∈{\mathrm{\Ω}}_J^{[i-1]}}{\min }{z}_u^{\left[i\right]}$ And order

$Z_u^{\left[i\right]}:=\underset{k\∈{\mathrm{\Ω}}_J^{[i-1]}\backslash \left\{u\right\}}{\mathrm{\Σ}}{p}_k{C}_{\mathrm{ku}}^{\left[i\right]},u\∈{\mathrm{\Ω}}_J^{[i-1]}$

It is the reservation scene set in i-1 step;

Stepi+1. judge

Whether interior element quantity reaches set point, if reach then stop; Otherwise turn to Stepi;

After finishing the selection that keeps scene, deletion scene probability is added on the scene probability nearest with it in the reservation scene, forms new probability distribution.

7. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 2 is characterized in that,

In the described step (2), the sensitivity relation that is located under the system running state i is:

$\left[\begin{array}{cc}{S}_{\mathrm{GG}}^i& S_{\mathrm{GL}}^i\\ S_{\mathrm{LG}}^i& S_{\mathrm{LL}}^i\end{array}\right]\left[\begin{array}{c}\mathrm{\Δ}{V}_G^i\\ \mathrm{\Δ}{V}_L^i\end{array}\right]=\left[\begin{array}{c}\mathrm{\Δ}{Q}_G^i\\ \mathrm{\Δ}{Q}_L^i\end{array}\right]---\left(1\right)$

With

Be respectively the voltage and the idle variation of generator node under the system running state i;

With

Voltage and idle variation for load bus;

With

Be sensitivity matrix, promptly in the power flow equation Jacobian matrix with voltage, idle relevant part.

8. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 7 is characterized in that,

In the described step (3), establish and comprise N PQ node in the control area, choose P leading node; On the basis of formula (1), be subjected to load or burden without work disturbance at random under the various random walk states after, the target of selecting leading node is how to choose P * leading node selection matrix C=[c of N dimension _Ij], make the variation Δ V of all the other load buses of the whole network _LExpectation is minimum; Promptly make target function I (C):

$\begin{array}{c}I\left(C\right)=E\left\{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\right[{\left(\mathrm{\Δ}{V}_L^i\right)}^T{Q}_x^i\mathrm{\Δ}{V}_L^i\left]\right\}\\ =\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{\ρ}}_{i}E\left\{\right[{\left(\mathrm{\Δ}{V}_L^i\right)}^T{Q}_x^i\mathrm{\Δ}{V}_L^i\left]\right\}\end{array}---\left(2\right)$

Reach minimum; Wherein E{} is for asking for mathematic expectaion, ρ _iBe the probability that running status i occurs, N is the load bus number, and M is the total quantity of running status, and T is a transposition; The value rule of cij be with all load buses from 1 to N numbering, if j load bus is chosen as the individual leading node of i in the system, cij=1 then, otherwise cij=0;

Be the diagonal angle weight matrices, determine its concrete numerical value according to the relative importance of loading under the different system running status; According to derivation, as the formula (3) for the relation between voltage deviation under the running status i and leading node selection matrix and the disturbance quantity:

$\mathrm{\Δ}{V}_L^i=(I-B_i{F}_{i}C)M_i{\mathrm{\ΔQ}}_L^i---\left(3\right)$

Wherein, I is a unit matrix, $M_i=S_{\mathrm{LL}}^{i-1},$ $B_i={-S}_{\mathrm{LL}}^{i-1}{S}_{\mathrm{LG}}^i,$ $F_i={\left(C{B}_i\right)}^T{\left({\mathrm{CB}}_i{B}_i^T{C}^T\right)}^{-1},$ The suffered disturbance of system is the gaussian random load disturbance under this state, and its desired value is zero, and standard deviation is proportional to disturbance front nodal point load or burden without work; Covariance matrix under the definition status i:

$C_{\mathrm{LL}}^i=E\left\{\mathrm{\Δ}{Q}_L^{\mathrm{iT}}\mathrm{\Δ}{Q}_L^i\right\}---\left(4\right)$

In formula (3), (4) substitution formula (2), carry out following derivation by the character of mathematic expectaion again:

$\begin{array}{c}I\left(C\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_{i}E\left\{\right[(I-B_i{F}_{i}C)M_i\mathrm{\Δ}{Q}_L^i{]}^T{Q}_x^i\\ \left[(I-B_i{F}_{i}C)M_i\mathrm{\Δ}{Q}_L^i\right]\}\end{array}=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\mathrm{trace}\left\{P_L^i{Q}_x^i\right\}-\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\mathrm{trace}\left\{(2H_1^i-H_2^i{H}_3^{i^{-1}}{H}_4^i)H_3^{i^{-1}}\right\}$ Wherein, $P_L^i=M_i{C}_{\mathrm{LL}}^i{M}_i^T,$ $H_1^i={\mathrm{CP}}_L^i{Q}_x^i{B}_i{B}_i^T{C}^T,$ $H_2^i={\mathrm{CB}}_i{B}_i^T{Q}_x^i{B}_i{B}_i^T{C}^T,$ $H_3^i={\mathrm{CB}}_i{B}_i^T{C}^T,$ $H_4^i={\mathrm{CP}}_L^i{C}^T;$ Order $f\left(C\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i\mathrm{trace}\left\{(2H_1^i-H_2^i{H}_3^{i^{-1}}{H}_4^i)H_3^{i^{-1}}\right\},$ Because

Irrelevant with leading node selection matrix, so former target function is equivalent to maxf (C), promptly how chooses leading node selection matrix C and make target function f (C) value reach maximum.

9. the leading node selecting method based on wind power fluctuation probability nature as claimed in claim 2 is characterized in that, in the described step (4), selects the genetic algorithm NSGA-II that comprises elitism strategy for use.

Images