CN102841965A - Modeling method of optimal power flow model of receiving end power grid security domain - Google Patents

Abstract

The invention discloses a modeling method of an optimal power flow model of a receiving end power grid security domain. The method includes a first step: constructing a system security domain through analysis of stability of a power flow equation, namely stability of quiescent voltage, phase angles and oscillation frequency of a system, and N-1 operation standards, wherein N loads of the system can be converted into M different sets composed in the load direction according to given power generation dispatching standards to generate a critical load matrix for approximate treatment of the security domain; a second step: constructing a dynamic security constraint optimal power flow model; a third step: constructing a self-adaptive neural fuzzy inference system; and a fourth step: training the self-adaptive fuzzy inference system and constructing the optimal power flow model. The optimal power flow model well expounds the operation standards of current electric power system dispatching. The optimal power flow security domain approximation technology based on the self-adaptive fuzzy inference system can be applied to dispatching optimization between receiving end power grid areas after distribution of hierarchy and areas.

Description

Held the modeling method of power grid security territory optimal load flow model

Technical field

The present invention relates to a kind of modeling method of optimal load flow model, relate in particular to a kind of modeling method of being held power grid security territory optimal load flow model.

Background technology

In electric system long-run development process, it is complementary that operation of power networks planning and system load flow calculate.On the one hand, for guaranteeing the safety and stability of whole electrical network, the formation of Electric Power Network Planning need be based upon system load flow and calculate on the basis; On the other hand, the system optimal trend is calculated the stability and the economy that can improve Electric Power Network Planning.Optimal load flow through control device capable of using in the Adjustment System, can be realized the system stable operation state that intended target is optimum satisfying under specific the system's operation and security constraints.Therefore, the system safety operation problem sharp-pointed day by day with complicated situation under, optimal load flow calculates nature becomes the system stability analysis aid indispensable with optimization.

After the access of extra-high voltage grid, the voltage unstability that power shifted initiation on a large scale when system broke down will become one of subject matter of being held electricity net safety stable.Especially after the extra-high voltage backbone network builds up basically, will increase the system short-circuit levels of current.Because the importance of 500kV subregion electrical network, it will be inevitably exists with the form of looped network, and this has just weakened the benefit that layering and zoning can bring.What the fault probability of happening of electric system was the highest is single-line to ground fault.Receiving-end system is comparatively weak, and load center lacks the forceful electric power source to be supported, especially a little less than the support of 500kV electrical network; Cause the 220kV electrical network too intensive; Short-circuit current exceeds standard, and extra-high voltage insert receive greatly to hold electrical network after, the electrical network layering and zoning operates to electricity net safety stable property and brings stern challenge.

Summary of the invention

The object of the invention is exactly in order to address the above problem, and a kind of computing method of being held power grid security territory optimal load flow are provided, and it has the advantage of the operation criterion of having explained current electric power system dispatching well.

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

A kind of modeling method of being held power grid security territory optimal load flow model, concrete steps are:

The first step: the stability analysis through to power flow equation promptly comes the constructing system security domain to the stability and the N-1 criterion of system's quiescent voltage, phase angle, oscillation frequency.Be the security domain expression formula that obtains to disperse; The N of a system load can be transformed to M the different sets that load direction is formed by given power generation dispatching criterion; N, M more than or equal to 1 be natural number; N is the quantity of loading in the system, and given system has fixing load quantity, generates following critical load matrix and comes security domain is carried out approximate processing:

Second step: it is following to make up dynamic security constrained optimum tide model:

Objective function: $S_b=-(C_d^T{P}_d-C_s^T{P}_s)---\left(2\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d)=0 (3)

$0\≤P_s\≤P_{s_{\max }}---\left(4\right)$

$0\≤P_d\≤P_{d_{\max }}---\left(5\right)$

is a natural number greater than or equal to 1 (6)

$Q_{g_{\min }}\≤Q_g\≤Q_{g_{\max }}---\left(7\right)$

V _min≤V≤V _max (8)

$f_{{\mathrm{NR}}_0}-f_{\mathrm{NR}}(V,\mathrm{\δ})\≤0---\left(9\right)$

Wherein, C _sAnd C _dBe respectively the bid of electric power supply and demand, the unit $/MWh of unit; System's supply and demand power is respectively P _sAnd P _d, unit is MW; F _PF() is the system load flow equation; V and δ are respectively that node voltage is with crossing; I _IjBe electric current through transmission line of electricity ij, this constraint definition the thermally-stabilised limit of system; Q _gBe generator reactive power; f _NR() is used to represent the security of system territory, Be its suitable critical value.

The 3rd step: set up the white neural fuzzy inference system that adapts to.

The 4th step: training is white to be adapted to fuzzy inference system and sets up the optimal load flow model, and security domain constrained optimum tide model is:

Objective function: $S_b=C_s^T{P}_s-C_d^T{\mathrm{\Δ\; P}}_d---\left(10\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d, Q _d)=0 (11)

0≤P _s≤P _smax (12)

Q _smin≤Q _s≤Q _smax (13)

V _min≤V≤V _max (14)

${\mathrm{\λ}}_l-\frac{\underset{i}{\mathrm{\Σ}}{w}_{i_m}{f}_{i_m}}{\underset{i}{\mathrm{\Σ}}{w}_{i_m}}\≤0$ $\∀m=1,...,G---\left(15\right)$

Δ P _Dj≤0 j is the natural number (16) more than or equal to 1

${\mathrm{\Δ\; P}}_{d_j}=(\mathrm{\α}{d}_j-{\mathrm{\α}}_0{d}_{j0})P_{\mathrm{Dj}0}$ J is the natural number (17) more than or equal to 1

Wherein: Q _dBe the system requirements reactive powers, Q _sBe that system supplies with reactive power,

Be the changing load amount of j node, P _Dj0Be the initial load of j node, loading coefficient is represented in scalar ce>=0, α ₀Be initial load coefficient, d _J0The expression initial load increases vector, d _jRepresent the load growth vector of all loads under i load growth rate: d _j=[d _J1, d _J2... D _JN] ^T

j is the natural number (18) more than or equal to 1

0≤d _j≤1 j is the natural number (19) more than or equal to 1

$\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_j=1---\left(20\right)$

α≥0(21)

Wherein, Δ P _dBe the changing load amount; M is m security of system territory of all G scheduling scheme.Constraint condition (11) makes Δ P by force _dBe 0 or negative.If Δ P _dBe 0, then the optimal load flow model is separated; Otherwise, if Δ P _dBe negative, represent then that the optimal load flow model does not have to separate.Therefore; This optimal load flow model has been explained the operation criterion of current electric power system dispatching well;

is the merit angle, and active power, reactive power and applied power are formed right-angle triangle, and applied power is a hypotenuse; Angle between the meritorious and applied power is at the merit angle; General G is an imaginary number with its cosine value representation power factor, and m is the number among 1 ~ G.

The concrete steps of the said first step are:

(1) setting up electric system can little algebraic equation be: $\begin{array}{c}\stackrel{\·}{x}=h(x,y,\mathrm{\ρ},\mathrm{\λ})\\ 0=g(x,y,\mathrm{\ρ},\mathrm{\λ})\end{array}---\left(22\right)$ Wherein, x is a system state variables, and common have generator speed and a corner; Y represents the algebraically variable, like load side voltage etc.; ρ representes system's controllable variable, like the generator voltage grade; λ is one group of uncontrollable parameter, and common have load to gain merit and reactive power.

(2) confirm load direction d _i=[d _I1d _I2D _IN] ^TWhen load increases along a certain specific direction, electric system will reach operational limit, come the constructing system security domain through the stability analysis to power flow equation; Promptly come the constructing system security domain, establish λ according to system's quiescent voltage, phase angle, oscillation frequency being carried out stability and N-1 operation criterion _i=[λ _I1λ _I2λ _IN] ^TBe i load increasing rate in N the load, i is the natural number more than or equal to 1, and λ is expressed as

λ _il＝αd _i1

λ _i2＝αd _i2 （23）

...

λ _iN＝αd _iN

Wherein, loading coefficient is represented in scalar ce>=0, d _IjThe load growth direction of expression load j under i load growth rate, i and j are the natural number more than or equal to 1, and load direction satisfies following condition:

$0\≤d_{\mathrm{ij}}\≤1,\∀j$

$\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_{\mathrm{ij}}=1---\left(24\right)$

(3) can make system reach the safety and stability border gradually through increasing loading coefficient α; And then N load of definite stability boundaris ultimate value

system can be transformed to M the different sets that load direction is formed by given power generation dispatching criterion, generates following critical load matrix and come security domain is similar to:

In said the 3rd step, Adaptive Neuro-fuzzy Inference is divided into six layers: X ₁, X ₂Be the input of system, the output of y inference system; Network has similar function with each node of one deck, uses O _1i, O _2iRepresent i node output, i is the natural number more than or equal to 1;

Ground floor: will import data and carry out Fuzzy Processing:

O _1i＝μ _Ai(x ₁),O _2i＝μ _Bi(x ₂),i＝1,2 （25）

Wherein, A _iOr B _iIt is fuzzy set; μ _Ai(x ₁), μ _Bi(x ₂) be the membership function of fuzzy set.

The second layer: each input data subordinate function is multiplied each other, as the relevance grade w of this layer rule _i:

w _i＝μ _Ai(x ₁)μ _Bi(x ₂),i＝1,2 （26）

The 3rd layer: the w that calculates i bar rule _iAnd all relevance grade sum w ₁+ w ₂, and pass through the normalization that both ratios are accomplished each bars rule relevance grade:

$w_i^{\′}=\frac{w_i}{w_1+w_2},$ i＝1,2 （27）

The 4th layer: the output that is used to calculate each bar rule:

O′ _4i＝w′ _if _i＝w′ _i(p _ix ₁+q _ix ₂+r _i)，i＝1,2 （28）

Wherein, f _iBe the consequent conclusion output function of fuzzy system, p _i, q _iBe the system's weighting coefficient under the i bar rule, r _iBe the constant under the i rules and regulations, when this output function is linear function, be called " first-order system "; If constant is called " zeroth order system ".

Layer 5: be used for total output of computing system:

$y=\underset{i}{\mathrm{\Σ}}{w}_i^{\′}{f}_i=\frac{\underset{i}{\mathrm{\Σ}}{w}_i{f}_i}{\underset{i}{\mathrm{\Σ}}{w}_i}$ i=1,2 （29）

Layer 6 should be exported the result through method of weighted mean and carry out the defuzzification processing, and made the error between input and the output minimum through back propagation and least square method.

Concrete steps in said the 4th step are:

(1) training adaptive fuzzy inference system; Make of the input of the border formed of M ultimate value of N-1 load as the adaptive fuzzy inference system; N is the natural number greater than 1; M is the natural number more than or equal to 1, and the safe edge dividing value that defines i load bus is expressed as:

${\mathrm{\λ}}_{\mathrm{Il}}^c\≈f({\mathrm{\λ}}_{i1},{\mathrm{\λ}}_{i2},...,{\mathrm{\λ}}_{\mathrm{Il}-1},{\mathrm{\λ}}_{\mathrm{Il}+1},...,{\mathrm{\λ}}_{\mathrm{IN}})\≈f\left({\widehat{\mathrm{\λ}}}_i\right)$ I is the natural number more than or equal to 1: (30)

The mapping function that is got the load growth rate by formula (9) and formula (25) is:

${\mathrm{\λ}}_l^c=\frac{\underset{i}{\mathrm{\Σ}}{w}_i{f}_i}{\underset{i}{\mathrm{\Σ}}{w}_i}---\left(31\right)$

(2) it is intrafascicular approximately formula (26) to be used for the security domain of optimal load flow equation, forms security domain constrained optimum tide model to be:

Objective function: $S_b=C_s^T{P}_s-C_d^T{\mathrm{\Δ\; P}}_d---\left(10\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d, Q _d)=0 (11)

0≤P _s≤P _smax (12)

Q _smin≤Q _s≤Q _smax (13)

V _min≤V≤V _max (14)

${\mathrm{\λ}}_l-\frac{\underset{i}{\mathrm{\Σ}}{w}_{i_m}{f}_{i_m}}{\underset{i}{\mathrm{\Σ}}{w}_{i_m}}\≤0,\∀m=1,...,G---\left(15\right)$

${\mathrm{\ΔP}}_{\mathrm{dj}}\≤0,\∀j=1,...,N---\left(16\right)$

${\mathrm{\ΔP}}_{d_j}=(\mathrm{\α}{d}_j-{\mathrm{\α}}_0{d}_{j0})P_{\mathrm{dj}0},\∀j=1,...,N---\left(17\right)$

$0\≤d_j\≤1,\∀j=1,...,N---\left(19\right)$

$\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_j=1---\left(20\right)$

α≥0 (21)

Beneficial effect of the present invention: the present invention is based on the adaptive fuzzy inference system and confirm the security of system territory.Adopt IEEE 118 node standard testing systems that the feasibility and the efficient of adaptive fuzzy inference system are carried out simulation calculation and checking, and verify that on layering and zoning theoretical foundation this method is to being held the feasibility of electrical network.The result shows, can apply to held the interregional optimizing scheduling of electrical network behind the layering and zoning based on the optimal load flow security domain approximation technique of adaptive fuzzy inference system.

Description of drawings

Fig. 1 is typical Bai Shiying fuzzy inference system structure;

Fig. 2 is IEEE three district system security domains.

Embodiment

Below in conjunction with accompanying drawing and embodiment the present invention is described further.

A kind of modeling method of being held power grid security territory optimal load flow model, concrete steps are:

The first step: the stability analysis through to power flow equation promptly comes the constructing system security domain to the stability and the N-1 criterion of system's quiescent voltage, phase angle, oscillation frequency; Be the security domain expression formula that obtains to disperse; The N of a system load can be transformed to M the different sets that load direction is formed by given power generation dispatching criterion; N is the quantity of loading in the system, and given system has fixing load quantity, generates following critical load matrix and comes security domain is carried out approximate processing:

Second step: it is following to make up dynamic security constrained optimum tide model:

Objective function: $S_b=-(C_d^T{P}_d-C_s^T{P}_s)---\left(2\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d)=0 (3)

$0\≤P_s\≤P_{s_{\max }}---\left(4\right)$

$0\≤P_d\≤P_{d_{\max }}---\left(5\right)$

$I_{\mathrm{Ij}}(\mathrm{\δ},V)\≤I_{i{j}_{\mathrm{Max}}}$ $\∀i,j,i\≠\mathrm{J\; i},j$ For more than or equal to 1 natural number (6)

$Q_{g_{\min }}\≤Q_g\≤Q_{g_{\max }}---\left(7\right)$

V _min≤V≤V _max (8)

$f_{{\mathrm{NR}}_0}-f_{\mathrm{NR}}(V,\mathrm{\δ})\≤0---\left(9\right)$

Wherein, C _sAnd C _dBe respectively the bid of electric power supply and demand, the unit $/MWh of unit; System's supply and demand power is respectively P _sAnd P _d, unit is MW; F _PF() is the system load flow equation; V and δ are respectively that node voltage is with crossing; I _IjBe electric current through transmission line of electricity ij, this constraint definition the thermally-stabilised limit of system; Q _gBe generator reactive power; f _NR() is used to represent the security of system territory,

Be its suitable critical value.

The 3rd step: set up Adaptive Neuro-fuzzy Inference.

The 4th step: train the adaptive fuzzy inference system and set up the optimal load flow model, security domain constrained optimum tide model is:

Objective function: $S_b=C_s^T{P}_s-C_d^T{\mathrm{\Δ\; P}}_d---\left(10\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d, Q _d)=0 (11)

0≤P _s≤P _smax (12)

Q _smin≤Q _s≤Q _smax (13)

V _min≤V≤V _max (14)

${\mathrm{\λ}}_l-\frac{\underset{i}{\mathrm{\Σ}}{w}_{i_m}{f}_{i_m}}{\underset{i}{\mathrm{\Σ}}{w}_{i_m}}\≤0$ $\∀m=1,...,G---\left(15\right)$

Δ P _Dj≤0j is the natural number (16) more than or equal to 1

${\mathrm{\Δ\; P}}_{d_j}=(\mathrm{\α}{d}_j-{\mathrm{\α}}_0{d}_{j0})P_{\mathrm{Dj}0}$ J is the natural number (17) more than or equal to 1

j is the natural number (18) more than or equal to 1

0≤d _j≤1 j is the natural number (19) more than or equal to 1

$\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_j=1---\left(20\right)$

α≥0 (21)

Wherein, Δ P _dBe the changing load amount; M is m security of system territory of all G scheduling scheme.Constraint condition (11) makes Δ P by force _dBe 0 or negative.If Δ P _dBe 0, then the optimal load flow model is separated; Otherwise, if Δ P _dBe negative, represent then that the optimal load flow model does not have to separate.Therefore; This optimal load flow model has been explained the operation criterion of current electric power system dispatching well;

is the merit angle, and active power, reactive power and applied power are formed right-angle triangle, and applied power is a hypotenuse; Angle between the meritorious and applied power is at the merit angle; General G is an imaginary number with its cosine value representation power factor, and m is the number among 1～G.

The concrete steps of the said first step are:

(1) setting up electric system can little algebraic equation be: $\begin{array}{c}\stackrel{\·}{x}=h(x,y,\mathrm{\ρ},\mathrm{\λ})\\ 0=g(x,y,\mathrm{\ρ},\mathrm{\λ})\end{array}---\left(22\right)$ Wherein, x is a system state variables, and common have generator speed and a corner; Y represents the algebraically variable, like load side voltage etc.; ρ representes system's controllable variable, like the generator voltage grade; λ is one group of uncontrollable parameter, and common have load to gain merit and reactive power.

(2) confirm load direction d _i=[d _I1d _I2D _IN] ^TWhen load increases along a certain specific direction, electric system will reach operational limit, come the constructing system security domain through the stability analysis to power flow equation; Promptly come the constructing system security domain, establish λ according to system's quiescent voltage, phase angle, oscillation frequency being carried out stability and N-1 operation criterion _i=[λ _I1λ _I2λ _IN] ^TBe i load increasing rate in N the load, i is the natural number more than or equal to 1, and λ is expressed as

λ _il＝αd _i1

λ _i2＝αd _i2 （23）

...

λ _iN＝αd _iN

Wherein, loading coefficient is represented in scalar ce>=0, d _IjThe load growth direction of expression load j under i load growth rate, i and j are the natural number more than or equal to 1, and load direction satisfies following condition:

0≤d _ij≤1 $\∀j$

$\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_{\mathrm{ij}}=1---\left(24\right)$

(3) can make system reach the safety and stability border gradually through increasing loading coefficient α; And then N load of definite stability boundaris ultimate value system can be transformed to M the different sets that load direction is formed by given power generation dispatching criterion, generates following critical load matrix and come security domain is similar to:

In said the 3rd step, Adaptive Neuro-fuzzy Inference is divided into six layers: X ₁, X ₂Be the input of system, the output of y inference system; Network has similar function with each node of one deck, uses O _1iRepresent i node output, i is the natural number more than or equal to 1;

Ground floor: will import data and carry out Fuzzy Processing:

O _1i＝μ _Ai(x ₁),O _2i＝μ _Bi(x ₂),i＝1,2 （25）

Wherein, A _iOr B _iIt is fuzzy set; μ _Ai(x ₁) be the membership function of fuzzy set.

The second layer: each input data subordinate function is multiplied each other, as the relevance grade w of this layer rule _i:

w _i＝μ _Ai(x ₁)μ _Bi(x ₂),i＝1,2 （26）

The 3rd layer: the w that calculates i bar rule _iAnd all relevance grade sum w ₁+ w ₂, and pass through the normalization that both ratios are accomplished each bars rule relevance grade:

$w_i^{\′}=\frac{w_i}{w_1+w_2},i=\mathrm{1,2}---\left(27\right)$

The 4th layer: the output that is used to calculate each bar rule:

O′ _4i＝w′ _if _i＝w′ _i(p _ix ₁+q _ix ₂+r _i),i＝1,2（28）

Wherein, f _iBe the consequent conclusion output function of fuzzy system, p _i, q _iBe the system's weighting coefficient under the i bar rule, r _iBe the constant under the i rules and regulations, when this output function is linear function, be called " first-order system "; If constant is called " zeroth order system ".

Layer 5: be used for total output of computing system:

$y=\underset{i}{\mathrm{\Σ}}{w}_i^{\′}{f}_i=\frac{\underset{i}{\mathrm{\Σ}}{w}_i{f}_i}{\underset{i}{\mathrm{\Σ}}{w}_i}---\left(29\right)$

Layer 6 should be exported the result through method of weighted mean and carry out the defuzzification processing, and made the error between input and the output minimum through back propagation and least square method.

Concrete steps in said the 4th step are:

(1) training adaptive fuzzy inference system; Make the input of the border formed of M ultimate value

of N-1 load as the adaptive fuzzy inference system, the safe edge dividing value

that defines i load bus is expressed as:

${\mathrm{\λ}}_{\mathrm{Il}}^c\≈f({\mathrm{\λ}}_{i1},{\mathrm{\λ}}_{i2},...,{\mathrm{\λ}}_{\mathrm{Il}-1},{\mathrm{\λ}}_{\mathrm{Il}+1},...,{\mathrm{\λ}}_{\mathrm{IN}})\≈f\left({\widehat{\mathrm{\λ}}}_i\right)$ I is the natural number more than or equal to 1: (30)

The mapping function that is got the load growth rate by formula (9) and formula (25) is:

${\mathrm{\λ}}_l^c=\frac{\underset{i}{\mathrm{\Σ}}{w}_i{f}_i}{\underset{i}{\mathrm{\Σ}}{w}_i}---\left(31\right)$

(2) it is intrafascicular approximately formula (26) to be used for the security domain of optimal load flow equation, forms security domain constrained optimum tide model to be:

Objective function: $S_b=C_s^T{P}_s-C_d^T{\mathrm{\Δ\; P}}_d---\left(10\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d, Q _d)=0 (11)

0≤P _s≤P _smax (12)

Q _smin≤Q _s≤Q _smax (13)

V _min≤V≤V _max (14)

${\mathrm{\λ}}_l-\frac{\underset{i}{\mathrm{\Σ}}{w}_{i_m}{f}_{i_m}}{\underset{i}{\mathrm{\Σ}}{w}_{i_m}}\≤0,\∀m=1,...,G---\left(15\right)$

${\mathrm{\ΔP}}_{\mathrm{dj}}\≤0,\∀j=1,...,N---\left(16\right)$

${\mathrm{\ΔP}}_{d_j}=(\mathrm{\α}{d}_j-{\mathrm{\α}}_0{d}_{j0})P_{\mathrm{dj}0},\∀j=1,...,N---\left(17\right)$

$0\≤d_j\≤1,\∀j=1,...,N---\left(19\right)$

$\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_j=1---\left(20\right)$

α≥0 (21)

Utilize PST, PSAT and UWPFLOW software carry out emulation to system.PSAT is used to calculate the conventional trend parameter of electric system, and with the input of this flow data as UWPFLOW and PST, obtains the voltage stability boundaris and concussion stability boundaris of system respectively; Again by the output result of PST and UWPFLOW gained as the input among the PSAT, calculate remaining voltage-regulation coefficient with this, thereby confirm total system safety and stability territory.Generate the sealing differentiable function through adaptive fuzzy inference system software by gained safety and stability territory, and with this function embedded system optimal load flow model, respectively through Newton method and interior point method compute optimal tide model.

Select for use IEEE 118 node standard testing system verifications to put forward the practicality of optimal load flow model; Analyze and do not lose ubiquity for simplifying; Optimal load flow model security domain obtains through typical scheduling method; G=1 in the formula 31 for example, IEEE 118 node standard testing systems are made up of 53 generators and 91 loads, and table 1 is genset bid data.

Table 1 genset bid data

Press layering and zoning principle and electricity market related notion; Respectively this test macro is divided into three and four operation areas; Corresponding security domain is promptly represented each interregional power transmission limitations, with 631 different load flows that obtain to as adaptive fuzzy inference system training data.

Embodiment one:

As shown in Figure 2, former 118 node standard testing systems are divided into three zones, and wherein zone 1 respectively comprises 31 loads with zone 2, and zone 3 comprises 29 loads.To a station symbol semicomputer (Duo 2 double-core 2.2Ghz processors, 2G internal memory), remain in the 10-5 for making the error between input and the output, obtained the security of system territory altogether in 156 seconds consuming time.

Be the validity of the test optimal load flow model of putting forward, press the selected P of table 2 _DA1, P _DA2And P _DA3Force the system running state security domain that jumps out, mark three among Fig. 2 respectively and jump out a little, the bid of wherein supposing the changing load amount is C _DA1=200$/MWh, C _DA2=400$/MWh and C _DA3=600$/MWh, separating of corresponding optimal load flow model is as shown in table 3; Analysis can know that in the ordinary course of things, the highest load P bids _DA1Reduction Δ Pd _A1Minimum is so that system running state is got back in the security domain.

Table 2 system testing scheme

Table 3 system loading change amount

Embodiment two:

Former 118 node standard testing systems are divided into four zones, and wherein zone 1, zone 2 respectively comprise 22 loads with zone 3, and zone 4 comprises 25 loads.For making the error between input and the output remain on 10 ^-5In, obtained the security of system territory altogether in 225 seconds consuming time.Know selected P by three regional sample calculation analysis _DA1, P _DA2, P _DA3And P _DA4Force the system running state security domain that jumps out, like table 4; The bid of supposing the changing load amount is C _DA1=800$/MWh, C _DA2=100$/MWh, C _DA3=300$/MWh and C _DA4=600$/MWh;

Table 5-4 system testing scheme

Table 5-5 system loading change amount

Separating of corresponding optimal load flow model is as shown in table 5; In the four regional examples, the highest load P bids _DA1The load P relatively low with bid _DA2And P _DA3Reduction all approaches 0, and the zone four load reduction Δ Ps bigger to the security of system influence _DA4Maximum, analysis can be known, in the ordinary course of things, measure valency than changing load, and security of system is bigger to the influence of load reduction.

Though the above-mentioned accompanying drawing specific embodiments of the invention that combines is described; But be not restriction to protection domain of the present 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 need not pay various modifications that creative work can make or distortion still in protection scope of the present invention.

Claims (5)

1. a modeling method of being held power grid security territory optimal load flow model is characterized in that, concrete steps are:

The first step: the stability analysis through to power flow equation promptly comes the constructing system security domain to the stability and the N-1 criterion of system's quiescent voltage, phase angle, oscillation frequency; Be the security domain expression formula that obtains to disperse; The N of system load is transformed to M the different sets that load direction is formed by given power generation dispatching criterion, and N, M are the natural number more than or equal to 1, and N is the quantity of loading in the system; Given system has fixing load quantity, generates the critical load matrix and comes security domain is carried out approximate processing;

Second step: make up dynamic security constrained optimum tide model;

The 3rd step: set up Adaptive Neuro-fuzzy Inference;

The 4th step: train the adaptive fuzzy inference system and set up the optimal load flow model.

2. a kind of according to claim 1 modeling method of being held power grid security territory optimal load flow model is characterized in that the concrete steps of the said first step are:

(1) setting up electric system can little algebraic equation be: $\begin{array}{c}\stackrel{\·}{x}=h(x,y,\mathrm{\ρ},\mathrm{\λ})\\ 0=g(x,y,\mathrm{\ρ},\mathrm{\λ})\end{array}---\left(22\right)$ Wherein, x is a system state variables, and common have generator speed and a corner; Y represents the algebraically variable, like load side voltage etc.; ρ representes system's controllable variable, like the generator voltage grade; λ is one group of uncontrollable parameter, and common have load to gain merit and reactive power;

(2) confirm load direction d _i=[d _I1d _I2D _IN] ^TWhen load increases along a certain specific direction, electric system will reach operational limit, come the constructing system security domain through the stability analysis to power flow equation; Promptly come the constructing system security domain according to system's quiescent voltage, phase angle, oscillation frequency being carried out stability and N-1 operation criterion

Be the merit angle, active power, reactive power and applied power are formed right-angle triangle, and applied power is a hypotenuse, and the angle between the meritorious and applied power is generally used its cosine value representation power factor at the merit angle, and G is an imaginary number, and m is the number among 1 ~ G;

If λ _i=[λ _I1λ _I2λ _IN] ^TBe i load increasing rate in N the load, i is the natural number more than or equal to 1, and λ is expressed as:

λ _il＝αd _i1

λ _i2＝αd _i2 （23）

...

λ _iN＝αd _iN

Wherein, loading coefficient is represented in scalar ce>=0, d _IjThe load growth direction of expression load j under i load growth rate, i and j are the natural number more than or equal to 1, and load direction satisfies following condition:

$\begin{array}{c}0\≤d_{\mathrm{ij}}\≤1,\∀j\\ \underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_{\mathrm{ij}}=1\end{array}---\left(24\right)$

(3) can make system reach the safety and stability border gradually through increasing loading coefficient α; And then N load of definite stability boundaris ultimate value system can be transformed to M the different sets that load direction is formed by given power generation dispatching criterion, generates following critical load matrix and come security domain is similar to:

3. a kind of according to claim 1 modeling method of being held power grid security territory optimal load flow model is characterized in that, the optimal load flow model in said second step is following:

Objective function: $S_b=-(C_d^T{P}_d-C_s^T{P}_s)---\left(2\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d)=0 (3)

$0\≤P_s\≤P_{s_{\max }}---\left(4\right)$

$0\≤P_d\≤P_{d_{\max }}---\left(5\right)$

$I_{\mathrm{Ij}}(\mathrm{\δ},V)\≤I_{i{j}_{\mathrm{Max}}}$ $\∀i,j,i\≠\mathrm{J\; i},j$ Be natural number (6) more than or equal to l

$Q_{g_{\min }}\≤Q_g\≤Q_{g_{\max }}---\left(7\right)$

V _min≤V≤V _max (8)

$f_{{\mathrm{NR}}_0}-f_{\mathrm{NR}}(V,\mathrm{\δ})\≤0---\left(9\right)$

Wherein, C _sAnd C _dBe respectively the bid of electric power supply and demand, the unit $/MWh of unit; System's supply and demand power is respectively P _sAnd P _d, unit is MW; F _PF() is the system load flow equation; V and δ are respectively that node voltage is with crossing; I _IjBe electric current through transmission line of electricity ij, this constraint definition the thermally-stabilised limit of system; Q _gBe generator reactive power; f _NR() is used to represent the security of system territory,

Be its suitable critical value.

4. a kind of according to claim 1 modeling method of being held power grid security territory optimal load flow model is characterized in that, in said the 3rd step, Adaptive Neuro-fuzzy Inference is divided into six layers: X ₁, X ₂Be the input of system, the output of y inference system; Network has similar function with each node of one deck, uses O _1iRepresent i node output, i is the natural number more than or equal to 1;

Ground floor: will import data and carry out Fuzzy Processing:

O _1i＝μ _Ai(x ₁),O _2i＝μ _Bi(x ₂),i＝1,2（25）

Wherein, A _iOr B _iIt is fuzzy set; μ _Ai(x ₁) be the membership function of fuzzy set;

The second layer: each input data subordinate function is multiplied each other, as the relevance grade w of this layer rule _i:

w _i＝μ _Ai(x ₁)μ _Bi(x ₂),i＝1,2（26）

The 3rd layer: the w that calculates i bar rule _iAnd all relevance grade sum w ₁+ w ₂, and pass through the normalization that both ratios are accomplished each bars rule relevance grade:

$w_i^{\′}=\frac{w_i}{w_1+w_2},i=\mathrm{1,2}---\left(27\right)$

The 4th layer: the output that is used to calculate each bar rule:

O′ _4i＝w′ _if _i＝w′ _i(p _ix ₁+q _ix ₂+r _i),i＝1,2（28）

Wherein, f _iBe the consequent conclusion output function of fuzzy system, p _i, q _iBe the system's weighting coefficient under the i bar rule, r _iBe the constant under the i rules and regulations, when this output function is linear function, be called " first-order system "; If constant is called " zeroth order system ";

Layer 5: be used for total output of computing system:

$y=\underset{i}{\mathrm{\Σ}}{w}_i^{\′}{f}_i=\frac{\underset{i}{\mathrm{\Σ}}{w}_i{f}_i}{\underset{i}{\mathrm{\Σ}}{w}_i}---\left(29\right)$

Layer 6 should be exported the result through method of weighted mean and carry out the defuzzification processing, and made the error between input and the output minimum through back propagation and least square method.

5. a kind of according to claim 1 modeling method of being held power grid security territory optimal load flow model is characterized in that, the concrete steps in said the 4th step are:

(1) training adaptive fuzzy inference system; Make of the input of the border formed of M ultimate value λ ic of N-1 load as the adaptive fuzzy inference system; N is the quantity of loading in the system; Given system has fixing load quantity, and M is a natural number, and the safe edge dividing value λ lic that defines i load bus is expressed as:

${\mathrm{\λ}}_{\mathrm{Il}}^c\≈f({\mathrm{\λ}}_{i1},{\mathrm{\λ}}_{i2},...,{\mathrm{\λ}}_{\mathrm{Il}-1},{\mathrm{\λ}}_{\mathrm{Il}+1},...,{\mathrm{\λ}}_{\mathrm{IN}})\≈f\left({\widehat{\mathrm{\λ}}}_i\right)$ I is the natural number more than or equal to 1: (30)

The mapping function that is got the load growth rate by formula (9) and formula (25) is:

${\mathrm{\λ}}_l^c=\frac{\underset{i}{\mathrm{\Σ}}{w}_i{f}_i}{\underset{i}{\mathrm{\Σ}}{w}_i}---\left(31\right)$

(2) it is intrafascicular approximately formula (26) to be used for the security domain of optimal load flow equation, forms security domain constrained optimum tide model to be:

Objective function: $S_b=C_s^T{P}_s-C_d^T{\mathrm{\Δ\; P}}_d---\left(10\right)$

Constraint condition: F _PF(δ, V, Q _g, P _s, P _d, Q _d)=0 (11)

0≤P _s≤P _smax (12)

Q _smin≤Q _s≤Q _smax (13)

V _min≤V≤V _max (14)

${\mathrm{\λ}}_l-\frac{\underset{i}{\mathrm{\Σ}}{w}_{i_m}{f}_{i_m}}{\underset{i}{\mathrm{\Σ}}{w}_{i_m}}\≤0,\∀m=1,...,G---\left(15\right)$

${\mathrm{\ΔP}}_{\mathrm{dj}}\≤0,\∀j=1,...,N---\left(16\right)$

${\mathrm{\ΔP}}_{d_j}=(\mathrm{\α}{d}_j-{\mathrm{\α}}_0{d}_{j0})P_{\mathrm{dj}0},\∀j=1,...,N---\left(17\right)$

$0\≤d_j\≤1,\∀j=1,...,N---\left(19\right)$

$\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{d}_j=1---\left(20\right)$

α≥0(21)；

Wherein, Δ P _dBe the changing load amount; M is m security of system territory of all G scheduling scheme, and constraint condition (11) makes Δ P by force _dBe 0 or negative; If Δ P _dBe 0, then the optimal load flow model is separated; Otherwise, if Δ P _dBe negative, represent then that the optimal load flow model does not have to separate that therefore, this optimal load flow model has been explained the operation criterion of current electric power system dispatching well,

Be the merit angle, active power, reactive power and applied power are formed right-angle triangle, and applied power is a hypotenuse, and the angle between the meritorious and applied power is generally used its cosine value representation power factor at the merit angle, and G is an imaginary number, and m is the number among 1 ~ G.

Images