CN106505582A - A kind of dynamic reactive power voltage cooperative control method based on neural network forecast mode - Google Patents

Abstract

The invention discloses a kind of dynamic reactive power voltage cooperative control method based on neural network forecast mode, first, using Wide-area Measurement Information set up power system fall into a trap and mains side generator and grid side dynamic passive compensation equipment dynamic electric voltage response characteristic Mathematical Modeling；Next, with load busbar voltage deviation and controls the quadratic form of cost as object function, sets up the Collaborative Control model of electrical power system dynamic reactive power power voltage；Then, by the Collaborative Control model conversation of electrical power system dynamic reactive power power voltage be quadratic programming problem；Finally, quadratic programming problem is solved, obtains the optimal sequence of electric power system control vector, realize dynamic reactive power voltage Collaborative Control；High-precision Wide-area Measurement Information is taken full advantage of, preferably adapts to the real-time change of operation of power networks state；Using neural network forecast mode, the time delay of Wide-area Measurement Information is preferably compensate for, quickly obtained a kind of simple and effective voltage coordination control strategy, with good application value and prospect.

Description

A kind of dynamic reactive power-voltage cooperative control method based on neural network forecast mode

Technical field

The invention belongs to field of power, more particularly to a kind of dynamic reactive power-electricity based on neural network forecast mode Pressure cooperative control method.

Background technology

Power system voltage stabilization is one of basic demand of electric power netting safe running.In the operation of modern bulk power grid, it is based on Single, distributed voltage control mode, it is difficult to meet increasingly complicated line voltage safe and stable operation and require.Set up power supply Side generator and the reactive power-voltage Collaborative Control mode of grid side dynamic passive compensation equipment, to improve the fast quick-action of system State reactive power and the response speed of voltage-regulation, significant to the safe and stable operation for ensureing bulk power grid.

Both at home and abroad, the control of dynamic reactive voltage is mostly using the feedback controling mode based on single local message.In dynamic Reactive voltage coordinates control aspect, mainly using feedback linearization, control method based on Lyapunov stability etc..These sides Method is more complicated, it is difficult to adapt to the real-time change of operation of power networks state completely, with certain risk.

In art methods, the reactive power-voltage control of mains side generator and grid side dynamic passive compensation equipment Mode processed is mostly using the control mode of local, decen, and voltage control strategy is difficult to the reality of adaptation operation of power networks state completely Shi Bianhua.

Content of the invention

The present invention proposes a kind of dynamic reactive power-voltage cooperative control method based on neural network forecast mode, its purpose The problem for being to overcome voltage control strategy in prior art to be difficult to adapt to the real-time change of operation of power networks state completely.

As electrical power system wide-area measurement system (Wide Area Measurement System, WAMS) is in bulk power grid Extensive utilization, by using high-precision real-time synchronization metric data, by the collaboration for electrical power system dynamic reactive power power vs. voltage Control provides favourable technological means.

A kind of dynamic reactive power-voltage cooperative control method based on neural network forecast mode, including following step：

Step one, sets up power system using Wide-area Measurement Information and falls into a trap and mains side generator and grid side dynamic passive compensation The Mathematical Modeling of equipment dynamic electric voltage response characteristic；

Wherein, x, u and y are respectively state vector, dominant vector and the algebraically vector of power system；

POWER SYSTEM STATE vector x includes the q axle transient potential E ' of mains side synchronous generator_q, grid side synchronization phase modulation The q axle transient potentials of machine are with E '_qcAnd the slippage s of induction conductivity, i.e. x=[E '_q,E′_qc,s]^T；

Electric power system control vector u includes the excitation voltage E of mains side synchronous generator_fWith grid side synchronous capacitor Excitation voltage E_fc, i.e. u=[E_f,E_fc]^T

The algebraically vector y of power system includes electric power networks busbar voltage and power；

Step 2, with load busbar voltage deviation and controls the quadratic form of cost as object function, with described in step one Model and variable set excursion as constraints, set up a kind of collaboration control of electrical power system dynamic reactive power power vs. voltage Simulation：

Wherein：J is object function, t₀There is moment, t for disturbance_mFor finish time, t_fFor controlling finish time；ΔV_l T (), Δ u (t) are respectively load busbar voltage bias vector and control variables bias vector；ΔV_l ^T(t) and Δ u^TT () is respectively ΔV_lThe transposed matrix of (t), Δ u (t)；

Matrix R and Q are respectively load busbar voltage deviation weighting matrix and control cost weighting matrix, are to angular moment Battle array；u_max、u_minFor the bound of electric power system control vector u, y_max、y_minBound for power system algebraically vector y；

Step 3, sets up the relation of power system output vector predicted value and dominant vector, and is t by sampling instant_k's The Collaborative Control model conversation of electrical power system dynamic reactive power power vs. voltage is quadratic programming problem, as follows：

Wherein：Q_p=diag (Q ..., Q)_M

R_p=diag (R ..., R)_N,

W is constraint coefficient matrix, and U is constrained vector,For sampling instant t_kWhen, the difference of reference voltage and initial voltage Value vector, D_lvpFor predicting load voltage vector and sampling instant t_kWhen dominant vector Δ u_pkRelational matrix；

J_kRepresent sampling instant t_kWhen voltage cooperative control system control targe, Δ u_pkFor sampling instant t_kWhen to be asked Voltage optimal control sequence matrix.

Step 4：Using the quadratic programming problem described in quadratic programming solution procedure three, obtain electric power system control to The optimal sequence Δ u of amount^* _pk, realize dynamic reactive power-voltage Collaborative Control.

Further, when power system feedback delay value is τ₂When, the optimal sequence Δ u of electric power system control vector^* _pkIn L-th element be u^*(t_k-τ₁+(L-1)T_s|t_k-τ₁), by voltage coordination control strategy u^*(t_k-τ₁+(L-1)T_s|t_k-τ₁) make For power system, when load busbar voltage returns to normal range (NR) and keeps stable, then end voltage closed-loop control；

Wherein, L=((τ₁+τ₂)/T_s)+1, τ₁In order to feedover, time delay value is, T_sFor the Wide-area Measurement Information sampling period.

Further, the prediction load voltage vector and dominant vector Δ u_pkRelational matrix D_lvpExpression formula as follows：

Wherein, N represents prediction step, N=T_p/T_s, the intelligence sample cycle is T_s, the PREDICTIVE CONTROL cycle is T_p, A, B, C are respectively state matrix, control matrix and the output matrix of power system；

Wherein：Δ x, Δ u, Δ y divide for system mode bias vector, control deviation vector sum output bias vector, i.e. Δ x =x-x_k, Δ u=u-u_k, Δ y=y-y_k, x_k、u_k、y_kSampling instant t is represented successively_kWhen POWER SYSTEM STATE vector in real time Value, dominant vector instantaneous value and algebraically vector instantaneous value.

Beneficial effect

The present invention proposes a kind of dynamic reactive power-voltage cooperative control method based on neural network forecast mode, first, Power system is set up using Wide-area Measurement Information to fall into a trap and mains side generator and grid side dynamic passive compensation equipment dynamic electric voltage sound Answer the Mathematical Modeling of characteristic；Next, with load busbar voltage deviation and controls the quadratic form of cost as object function, with aforesaid The value excursion of Mathematical Modeling and variable is constraints, sets up a kind of association of electrical power system dynamic reactive power power vs. voltage Same Controlling model；Then, set up the relation of power system output vector predicted value and dominant vector, and by Electrical Power System Dynamic without The Collaborative Control model conversation of work(power vs. voltage is quadratic programming problem；Finally, solve quadratic programming using quadratic programming to ask Topic, obtains the optimal sequence of electric power system control vector, realizes dynamic reactive power-voltage Collaborative Control；Fully profit of the invention With high-precision Wide-area Measurement Information, preferably adapt to the real-time change of operation of power networks state；Using neural network forecast mode, preferably mend The time delay of Wide-area Measurement Information has been repaid, a kind of simple and effective voltage coordination control strategy can have quickly been obtained, should with good popularization With value and prospect；The dynamic electric voltage that mains side synchronous generator is given full play to grid side dynamic passive compensation equipment supports energy Power.

Description of the drawings

Fig. 1 is the workflow diagram of the inventive method；

Fig. 2 is the dynamic reactive power-voltage Collaborative Control based on neural network forecast mode；

Fig. 3 is 10 machine of New England, 39 node system structural representation；

Fig. 4 is the voltage control mode load bus voltage change curve only with local information feedback；

Fig. 5 is the voltage control mode synchronous capacitor reactive power change curve only with local information feedback；

Fig. 6 is the dynamic electric voltage control mode load bus voltage change curve based on neural network forecast mode；

Fig. 7 is the dynamic electric voltage control mode synchronous capacitor reactive power change curve based on neural network forecast mode.

Specific embodiment

Below in conjunction with accompanying drawing and it is embodied as that the invention will be further described.

Step one, the present invention utilize Wide-area Measurement Information, set up meter and mains side generator is set with grid side dynamic passive compensation For (such as：Synchronous capacitor) dynamic electric voltage response characteristic Mathematical Modeling, as follows：

In formula：Y represents the vector of system algebraically vector, i.e. electric power networks busbar voltage and power composition；X and u are respectively The state vector of system and dominant vector, which is expressed as

X=[E '_q,E′_qc,s]^T, u=[E_f,E_fc]^T

Wherein：E′_q、E_fRespectively mains side synchronous generator q axles transient potential and excitation voltage are vectorial；E′_qc、E_fcIt is divided into Grid side synchronous capacitor q axles transient potential and excitation voltage vector；S is induction conductivity slippage vector.

System model derivation based on Wide-area Measurement Information is as follows.

PMUs can provide high-precision real-time synchronization data, including：Active and reactive power, node voltage, generator work( The information such as angle.On the one hand, generator's power and angle δ is obtained in real time from synchronous generator pusher side PMU_i, angular frequency_i, active-power P_gi, machine Terminal voltage V_giAnd phase angle theta_giAfter information, q axle transient potential E ' can be tried to achieve by formula (2) and formula (3)_qi.

Meanwhile, generator d, q shaft current i_diAnd i_qiCan be obtained by formula (4).

Generator electromagnetic torque T_eiCan be obtained by formula (5), i.e.,

T_ei=E '_qii_qi-(X′_di-X_qi)i_dii_qi(5)

And the 3 order mode type of synchronous generator of the dynamic excitation voltage characteristic of tradition consideration is represented by

Wherein：Formula (6) represents that generator excitation voltage dynamic characteristic equation, formula (7)-formula (8) represent that generator mechanical is moved Step response equation.i_di、T_ji、ω₀、D_iAnd T_miBiao Shi not the d shaft currents of generator, generator inertia time constant, specified angular frequency Rate, generator rotor angle, damped coefficient and machine torque.I=1 ..., m, m are generator number of units.

After using Wide-area Measurement Information, quantity of state and electric parameters in equation (7) and (8) are known quantity.So, wide area can be used The variable instantaneous value that information is obtained replaces dynamical equation (7) and (8).That is, a retention equation (6) in 3 rank dynamical equation of generator.

On the other hand, network load side adopts induction motor model.Load side wattful power is obtained in real time from WAMS systems Rate P_liAnd reactive power Q_li, node voltage V_liAnd phase angle theta_liAfterwards, induction conductivity slippage s is obtained_j, as follows：

Wherein：

R₁=r_m+r_s-R_j/K_Hj, X₁=X_m+X_s-X_j/K_Hj

r_s、X_s、r_r、X_r、r_m、X_mAnd K_HjRespectively induction conductivity stator resistance and reactance, rotor resistance and reactance, excitation Resistance and reactance, capacity conversion factor.J=1 ..., n, n are induction motor load number.

Additionally, the dynamic model of synchronous capacitor is represented by

Wherein：T′_d0ci、E′_qci、E_fci、X_dci,X′_dci,X_qciAnd i_dciRespectively the d axle time constants of synchronous capacitor, q Axle transient potential, excitation voltage, d axle reactance, d axle transient state reactance, the reactance of q axles and d shaft currents.

Association type (6), formula (9), formula (10) and electric power networks equation obtain system equivalent dynamic model, i.e. formula (1).

Step 2, with load busbar voltage deviation and controls the quadratic form of cost as object function, with system dynamical equation (1) and variable excursion be constraints, set up a kind of Collaborative Control mould of electrical power system dynamic reactive power power vs. voltage Type is as follows：

Wherein：J is object function, t₀There is moment, t for disturbance_mFor finish time, t_fFor controlling finish time；ΔV_l T (), Δ u (t) are respectively load busbar voltage bias vector and control variables bias vector；Matrix R and Q are respectively load bus Voltage deviation weighting matrix and control cost weighting matrix, they are diagonal matrix；u_max、u_minFor the bound of controlled quentity controlled variable u, y_max、y_minBound for algebraic quantity y.

Step 3, sets up the relation of power system output vector predicted value and dominant vector, and by Electrical Power System Dynamic without The Collaborative Control model conversation of work(power vs. voltage is quadratic programming problem, as follows：

Define in PREDICTIVE CONTROL, the intelligence sample cycle is T_s, the PREDICTIVE CONTROL cycle is T_p, step-length is controlled for M, prediction step For N.Wherein：M, N are integer, M≤N, and N=T_p/T_s.

First, in sampling instant t_k, using state vector x of system_k, dominant vector u_kWith output vector y_kReal-time wide Domain information, will be quantitation for equation (1) after obtain system model equation and be

Wherein：Δ x, Δ u, Δ y divide for system mode bias vector, control deviation vector sum output bias vector, i.e. Δ x =x-x_k, Δ u=u-u_k, Δ y=y-y_k；A, B, C are respectively state matrix, control matrix and the output matrix of system.

Further by formula (12) differencing after, can obtain

Wherein：Δx_k+1Predicted value for+1 moment state deviation vector of kth；The size of matrix G and H is

Using formula (13), derivation system future N walks output vector Δ y_pkWith dominant vector Δ u_pkRelation, that is, have

Δy_pk=D_lvpΔu_pk(14)

Wherein：D_lvpFor Δ y_pkWith Δ u_pkRelational matrix, its expression formula is

Then, using formula (14), it is secondary rule by the Collaborative Control model conversation of electrical power system dynamic reactive power power vs. voltage The problem of drawing.In predetermined period T_pInterior, in formula (2), object function is converted into

Wherein：

Q_p=diag (Q ..., Q)_M

R_p=diag (R ..., R)_N,

For reference voltage and the difference value vector of initial voltage, matrix D_lvpFor predicting load voltage vector and dominant vector Δu_pkRelational matrix.

Formula (15) and formula (14) are substituted into formula (11), can be obtained

Finally, formula (16) is solved using quadratic programming, obtain the optimal sequence Δ u of dominant vector^* _pk.

Step 4, the time delay for compensating Wide-area Measurement Information using neural network forecast mode obtain a kind of voltage of consideration delay compensation Coordination control strategy, its detailed construction are as shown in Figure 2.

It is τ to define Predictive Control System overall delay value, and feedforward time delay value is τ₁, feedback delay value is τ₂, i.e. τ=τ₁+τ₂. Delay during the information of consideration control system, in sampling instant t_kVoltage optimal control sequence Δ u^* _pkIt is expressed as

Further consideration feedback delay value is τ₂, take L=(τ₁+τ₂)/T_s+ 1, then optimal control sequence Δ u^* _pkMiddle l-th Element is

Voltage coordination control strategy after consideration delay compensationAs

By voltage coordination control strategyAfter acting on system, load busbar voltage returns to normal range (NR) and keeps steady Fixed, then end voltage closed-loop control.

Hereinafter the advantages of the present invention will be further illustrated as example with certain application.

NETWORK STRUCTURE PRESERVING POWER SYSTEM figures of the Fig. 3 for 10 machine of New England, 39 node.In the system, except load bus 31 and node 39 Outside using constant-impedance model, other 17 load buses adopt induction motor model；And in node 3, node 4, node 7th, synchronous capacitor of the configuration rated capacity for 300Mvar at node 8, the phase modifier of each node be correspondingly abbreviated as SC3, SC4, SC7 and SC8.The range of operation that node voltage is allowed is 0.95p.u.～1.10p.u., the scope of excitation voltage be -6p.u.～ 6p.u., the reference capacity of system is 100MVA.It is assumed that the sufficient PMU of the system configuration, it is ensured that the observability of system, and Overall delay value τ of control system is 0.6s.In PREDICTIVE CONTROL, signal sampling period T_s=0.2s, controls step-length M=5, prediction step Long N=6.

Failure is assumed：In Fig. 3, as t=2s, the power of load bus 8 increases by 1.5 times, while 8 institute of node 7 and node Connect line tripping.

The dynamic electric voltage fed back only with local information is controlled, strategy 1 is designated as；Carried pre- based on network using the present invention The dynamic electric voltage Collaborative Control of survey mode, is designated as strategy 2.

Using the lotus node voltage of strategy 1, the change of synchronous generator reactive power respectively as shown in Figure 4 and Figure 5.By Fig. 4 Can obtain, the voltage of failure front nodal point 7 and node 8 is in normal range (NR).After failure, using the voltage of strategy 1, node 7 and node 8 0.908p.u. and 0.916p.u. is remained respectively, and magnitude of voltage is below 0.95p.u..

Using the lotus node voltage of strategy 2, the change of synchronous generator reactive power respectively as shown in Figure 6 and Figure 7.By Fig. 6 Can obtain, using strategy 2 after, the fast quick-recovery of the voltage of node 7 and node 8, finally remain respectively 0.980p.u. and 0.991p.u..Magnitude of voltage is in normal range (NR).

Can be obtained by Fig. 5 and Fig. 6, under above two voltage control strategy, the reactive power of synchronous capacitor after system stability As shown in table 1.Can be obtained by table 1, in the presence of strategy 2, the excitation voltage of each generator is above the result of strategy 1；Therefore, Strategy 2 can more preferably cooperate with the dynamic reactive power-voltage control of generator and synchronous capacitor than strategy 1, improve system dynamic Voltage responsive characteristic, holding system voltage levvl is in normal range (NR).

Synchronous capacitor reactive power after table 1 is stable

In the present embodiment, can be with using a kind of dynamic reactive power-voltage collaboration that implements based on neural network forecast mode Realizing method of the present invention step, it is single which includes that the power system dynamic model being sequentially connected is set up to the device of control method Unit, dynamic electric voltage Collaborative Control model form unit, and PREDICTIVE CONTROL unit and delay compensation unit and control strategy ask for list Unit.

Specific embodiment described herein is only to the spiritual explanation for example of the present invention.Technology neck belonging to of the invention The technical staff in domain can be made various modifications or supplement or replaced using similar mode to described specific embodiment Generation, but without departing from the spiritual of the present invention or surmount scope defined in appended claims.

Claims (3)

1. a kind of dynamic reactive power-voltage cooperative control method based on neural network forecast mode, it is characterised in that including following Several steps：

Step one, sets up power system using Wide-area Measurement Information and falls into a trap and mains side generator and grid side dynamic passive compensation equipment The Mathematical Modeling of dynamic electric voltage response characteristic；

$\left\{\begin{array}{c}\stackrel{\·}{x}=f(x,y,u)\\ 0=g(x,y,u)\end{array}\right.$

Wherein, x, u and y are respectively state vector, dominant vector and the algebraically vector of power system；

POWER SYSTEM STATE vector x includes the q axle transient potential Ε ' of mains side synchronous generator_q, grid side synchronous capacitor q Axle transient potential is with Ε '_qcAnd the slippage s of induction conductivity, i.e. x=[E '_q,E′_qc,s]^T；

Electric power system control vector u includes the excitation voltage E of mains side synchronous generator_fExcitation with grid side synchronous capacitor Voltage E_fc, i.e. u=[E_f,E_fc]^T

The algebraically vector y of power system includes electric power networks busbar voltage and power；

Step 2, with load busbar voltage deviation and controls the quadratic form of cost as object function, with the model described in step one Excursion is set as constraints with variable, a kind of Collaborative Control mould of electrical power system dynamic reactive power power vs. voltage is set up Type：

$\min J=\frac{1}{2}{\∫}_{t_0}^{t_m}\left({{\mathrm{\ΔV}}_l}^T\right(t\left){\mathrm{R\ΔV}}_l\right(t)+{\mathrm{\Δu}}^T(t\left)Q\mathrm{\Δ}u\right(t\left)\right)dt$

$\begin{array}{cc}s.t.& \left\{\begin{array}{c}\stackrel{\·}{x}=f(x,y,u)\\ 0=g(x,y,u)\\ u_{\min }\≤u\left(t\right)\≤u_{max},t_0\≤t\≤t_m\\ y_{\min }\≤y\left(t\right)\≤y_{max},t_f\≤t\≤t_m\end{array}\right.\end{array}$

Wherein：J is object function, t₀There is moment, t for disturbance_mFor finish time, t_fFor controlling finish time；ΔV_l(t)、Δu T () is respectively load busbar voltage bias vector and control variables bias vector；ΔV_l ^T(t) and Δ u^TT () is respectively Δ V_l The transposed matrix of (t), Δ u (t)；

Matrix R and Q are respectively load busbar voltage deviation weighting matrix and control cost weighting matrix, are diagonal matrix； u_max、u_minFor the bound of electric power system control vector u, y_max、y_minBound for power system algebraically vector y；

Step 3, sets up the relation of power system output vector predicted value and dominant vector, and is t by sampling instant_kPower train The Collaborative Control model conversation of system dynamic reactive power-voltage is quadratic programming problem, as follows：

Wherein：Q_p=diag (Q ..., Q)_M

R_p=diag (R ..., R)_N,

W is constraint coefficient matrix, and U is constrained vector,For sampling instant t_kWhen, the difference of reference voltage and initial voltage to Amount, D_lvpFor predicting load voltage vector and sampling instant t_kWhen dominant vector Δ u_pkRelational matrix；

J_kRepresent sampling instant t_kWhen voltage cooperative control system control targe, Δ u_pkFor sampling instant t_kWhen voltage to be asked Optimal control sequence matrix.

Step 4：Using the quadratic programming problem described in quadratic programming solution procedure three, electric power system control vector is obtained Optimal sequence Δ u^* _pk, realize dynamic reactive power-voltage Collaborative Control.

2. method according to claim 1, it is characterised in that when power system feedback delay value is τ₂When, power system control The optimal sequence Δ u of system vector^* _pkIn l-th element be u^*(t_k-τ₁+(L-1)T_s|t_k-τ₁), by voltage coordination control strategy u^*(t_k-τ₁+(L-1)T_s|t_k-τ₁) power system is acted on, when load busbar voltage returns to normal range (NR) and keeps stable, then End voltage closed-loop control；

Wherein, L=((τ₁+τ₂)/T_s)+1, τ₁In order to feedover, time delay value is, T_sFor the Wide-area Measurement Information sampling period.

3. method according to claim 1 and 2, it is characterised in that the prediction load voltage vector and dominant vector Δ u_pkRelational matrix D_lvpExpression formula as follows：

Wherein, N represents prediction step, N=T_p/T_s, the intelligence sample cycle is T_s, the PREDICTIVE CONTROL cycle is T_p, A, B, C are respectively state matrix, control matrix and the output matrix of power system；

$\left\{\begin{array}{c}\mathrm{\Δ}\stackrel{\·}{x}=A\mathrm{\Δ}x+B\mathrm{\Δ}u\\ \mathrm{\Δ}y=C\mathrm{\Δ}x\end{array}\right.$

Wherein：Δ x, Δ u, Δ y divide for system mode bias vector, control deviation vector sum output bias vector, i.e. Δ x=x- x_k, Δ u=u-u_k, Δ y=y-y_k, x_k、u_k、y_kSampling instant t is represented successively_kWhen POWER SYSTEM STATE vector instantaneous value, control The vectorial instantaneous value of system and algebraically vector instantaneous value.