CN102355010A - Method for synchronously identifying parameters of multiple generators - Google Patents

Abstract

The invention discloses a method for identifying parameters of multiple generators, belonging to the professional field of electric system and automation thereof under the first-level Disciplines of electrical engineering. According to the method provided by the invention, the systematic variable, particularly a generator relative power angle, a node phase angle variable and a connecting line power and the like obtained by a wide area measurement system (WAMS) based on a PMU (performance monitor unit) technology is fully utilized, and an identifiable analyzing method based on a rail sensitivity is adopted aiming at a Park model with different orders to synchronously identify the parameters of multi-generator. The invention discloses a policy and a method for synchronously identifying the parameters of the multi-generator, comprising the following steps of firstly classifying the generators, secondly identifying the parameters in steps, thirdly identifying the key parameters only, and finally emphasizing the overall characteristics. In the method, an interactive calculation between an ant colony algorithm and a BPA (Bonneville Power Administration) program is adopted to identify the parameters.

Description

Discrimination method in the time of many generator parameters

Technical field

Discrimination method when the present invention relates to a kind of many generator parameters belongs to electrical engineering field Power System and its Automation research direction.

Background technology

Synchronous generator is one of most important element in the electric power system, and operation action that we can say it all produces either large or small influence to the various aspects of electric power system.But, realize element, its structure and the dynamic property ten minutes complicacy of electric energy and mechanical energy conversion because synchronous generator is a kind of integrate rotation and static, electromagnetism transform and mechanical movement.Therefore, the model of synchronous generator and parameter are the important contents in the electric power system Modeling Research always.Owing to lack actual parameter, power system analysis is calculated data or the representative value that the used synchronous generator parameter of emulation adopts producer to provide more at present, or haves no alternative but adopt simplified model.Cause calculating emulation gained result and actual dynamic process to have greater difference, had a strong impact on accuracy and the confidence level calculated, these situation are confirmed by some external documents.In the past, it is generally acknowledged that the complicated more detailed model of use can obtain better result in researching and analysing.Yet NPCC (Northeast Power Coordinating Council) points out that it is more important than using detailed model in stability analysis, in general to use the accurate parameter of electric machine.

Concerned power generation machine Study on Parameter Identification all was to the separate unit generator in the past, measurement be the variable of this machine self, such as port voltage, electric current and merit angle etc.For the separate unit generator, under the measurable situation in merit angle, the dq component of voltage and current all can obtain.If with port voltage as input variable, port current as output variable, then port current can be understood as the function of port voltage and parameter, so parameter that in view of the above can this generator of identification.Adopt this mode to obtain generator parameter, need carry out the parameter measurement identification one by one all generators.In fact, electric power system is an integral body, and whether the system model that each element, each node link together is accurate on the whole, then is to be concerned about the most during electric system simulation calculates.The inventor has proposed the thinking of the whole modeling of electric power system, and is applied to the electric load modeling.According to this thinking, use it for many generator parameters of identification simultaneously on the whole.

The research of relevant electric power system parameter identifiability in the past all is based on analytic method, with the model linearisation, carries out the analysis of parameter identifiability such as earlier then, carries out simulating, verifying in non linear system more at last.Analytic method can obtain result clearly, but for the high-order dynamic equation, is difficult to carry out, the parameter identification of multi-machine power system especially, and it is impossible adopting its identifiability of analytic method analysis.Simultaneously, it is satisfactory to come the precision of identification practical parameter to be difficult to based on utility model, and identification gained practical parameter is steady inadequately, and influencing factor is more.Because the practical parameter of synchronous generator not all is independently also, and the circuit parameter in the Park model magnetic linkage equation is independently each other.Along with the continuous development of China's electric power system, high capacity synchronous generator group, the transmission of electricity of extra-high pressure grade and trans-regional big networking become the principal character of modern power systems.These new developments make the importance of power grid security and stable operation problem become increasingly conspicuous.Synchro generator model and parameter all have decisive meaning to power system analysis, operation, control etc. rationally and accurately.For this reason; The present invention makes full use of the system-wide vector that obtains based on the WAMS of PMU technology (WAMS); Especially the relative merit of generator angle, node phase angle variable and interconnection power etc.; Park model to different orders; Employing is based on the identifiability analytical method of trace sensitivity, identification when carrying out many generator parameters.

Summary of the invention

The present invention proposes the parameter while discrimination method of many generators; Employing is based on the identifiability analytical method of trace sensitivity; Strategy and method to generator parameter employing classification, substep identification adopt the interacting operation identification emphasis parameter between ant group algorithm and the BPA program at last.

Identifiability analytical method based on trace sensitivity: if proportional relation between the trace sensitivity of Several Parameters (zero passage simultaneously) or linear correlation, then these parameters are undistinguishable identifications.

The method of many generator parameter identifications simultaneously:

(1) evaluation index of selective systemization.

$J=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{[Y(\mathrm{\θ},k)-Y_m\left(k\right)]}^{T}W\left(k\right)[Y(\mathrm{\θ},k)-Y_m\left(k\right)]$

In the formula, Y is the dynamic response of the main observational variable of system, and subscript m is represented the WAMS measured value, and θ is a parameter of treating identification, and W (k) is a weight coefficient matrix.

(2) select observational variable.Here many generator parameters are carried out whole identification, the systematic perspective of selecting WAMS to collect is measured.The systematic observation variable should be able to reflect the main behavioral characteristics of system, and relatively sensitiveer to parameter.Such as, two or the two relative merit angles of mass-sending between the motors.

(3) select the emphasis parameter.

The strategy of many generator parameter identifications:

1. the identification of classifying.Generator is classified, think that for every type of generator its parameter is identical.Classification main kind and capacity according to generator.At first, can be divided into the hydraulic turbine and steam turbine according to the kind of generator.Classify according to generator capacity then, such as 600MW be one type, 300MW's is one type.

2. substep identification.Confirm the partial parameters of generator earlier according to the steady state measurement data, confirm rest parameter according to the dynamic data that WAMS collects again.

3. emphasis identification.According to trace sensitivity, select the emphasis parameter that sensitivity is big, influence is also big.

4. pay attention to whole.In electric system simulation calculated, what people paid close attention to the most was the overall permanence that can model reflect system.Such as, system vibrates after large disturbance, major concern zone power-angle oscillation, interconnection power oscillation.The fast development of WAMS technology provides condition for this reason.

(4) select the proper optimization method.Because the parametric solution space is quite complicated, needs the strong method of global optimization ability, the simulated evolution method is suitable selection.Through contrast on probation, adopt ant group algorithm.

(5) realization of algorithm.In multimachine system synchronous generator parameter identification, an aspect is to optimize when carrying out many generator parameters, and the another one aspect is to calculate the target function value of reflection entire system dynamic behaviour.Between optimizer and BPA program, set up exchanges data; Promptly optimize the middle generator parameter result who once obtains and substitute the generator parameter among the BPA automatically; BPA Automatic computing system dynamic response and export to optimizer then; Optimizer calculates and obtains target function value; And further optimize the new generator parameter value of acquisition, up to finding optimal solution.As shown in Figure 1.

Description of drawings

Accompanying drawing is the parameter identification flow chart

Embodiment

Discrimination method when the present invention proposes many generator parameters relates generally to the Park model of generator, identification based on the selection of the substep identification of the classification identification of the identifiability analytical method of trace sensitivity, parameter, parameter, observational variable and many generator parameters the time.

1, generator model and parameter

For the steam turbine solid rotor; The transient process of rotor q axle can adopt two equivalent damping winding to describe; Promptly except with the less equivalent damping winding Q of inferior transient process time corresponding constant; Also consider and the bigger equivalent damping winding g of transient process time corresponding constant, obtain 6 rank Park models thus.Its magnetic linkage equation does

$\left[\begin{array}{c}{\mathrm{\ψ}}_d\\ {\mathrm{\ψ}}_q\\ {\mathrm{\ψ}}_0\\ {\mathrm{\ψ}}_f\\ {\mathrm{\ψ}}_D\\ {\mathrm{\ψ}}_g\\ {\mathrm{\ψ}}_Q\end{array}\right]=\left[\begin{array}{ccccccc}{X}_d& 0& 0& X_{\mathrm{ad}}& X_{\mathrm{ad}}& 0& 0\\ 0& X_q& 0& 0& 0& X_{\mathrm{aq}}& X_{\mathrm{aq}}\\ 0& 0& X_0& 0& 0& 0& 0\\ X_{\mathrm{ad}}& 0& 0& X_f& X_{\mathrm{ad}}& 0& 0\\ X_{\mathrm{ad}}& 0& 0& X_{\mathrm{ad}}& X_D& 0& 0\\ 0& X_{\mathrm{aq}}& 0& 0& 0& X_g& X_{\mathrm{aq}}\\ 0& X_{\mathrm{aq}}& 0& 0& 0& X_{\mathrm{aq}}& X_Q\end{array}\right]\left[\begin{array}{c}-i_d\\ {-i}_q\\ -i_0\\ i_f\\ i_D\\ i_g\\ i_Q\end{array}\right]---\left(1\right)$

Voltage equation does

$\left[\begin{array}{c}{u}_d\\ u_q\\ u_0\\ u_f\\ u_D\\ u_g\\ u_Q\end{array}\right]=\frac{d}{\mathrm{dt}}\left[\begin{array}{c}{\mathrm{\ψ}}_d\\ {\mathrm{\ψ}}_q\\ {\mathrm{\ψ}}_0\\ {\mathrm{\ψ}}_f\\ {\mathrm{\ψ}}_D\\ {\mathrm{\ψ}}_g\\ {\mathrm{\ψ}}_Q\end{array}\right]+\left[\begin{array}{c}-{\mathrm{\ω\ψ}}_q\\ {\mathrm{\ω\ψ}}_d\\ 0\\ 0\\ 0\\ 0\\ 0\end{array}\right]+\left[\begin{array}{c}-r_a{i}_d\\ -r_a{i}_q\\ -r_a{i}_0\\ r_f{i}_f\\ r_D{i}_D\\ r_g{i}_g\\ r_Q{i}_Q\end{array}\right]---\left(2\right)$

U wherein _D=0, u _Q=0, u _g=0.In the formula, ψ representes the magnetic linkage of each winding; U representes the voltage of each winding; I representes the electric current of each winding; X _d, X _q, X _g, X _Q, X _D, X _fRepresent respectively d axle winding, q axle winding, damping winding and excitation winding from induction reactance, X wherein _d, X _qAlso claim synchronous reactance; X _Ad(D) mutual inductance between is anti-for d, f, claims d armature axis reaction synchronous reactance again for last three windings of d axle; X _Aq(Q) mutual inductance between is anti-for q, g, claims q armature axis reaction synchronous reactance again for last three windings of q axle; R is the resistance of each winding.

For the hydraulic turbine, because rotor is salient pole, can not consider the g winding, obtain 5 rank Park models thus, only need be in the equation of 6 rank Park models the equation of subscripting g and variable be removed and get final product.

Circuit parameter in the Park model magnetic linkage equation is independently each other.For this reason, adopt the Park model to come the method for identification synchronous generator parameter.

2, based on the identifiability analytical method of trace sensitivity

If proportional relation between the trace sensitivity of Several Parameters (zero passage simultaneously) or linear correlation, then these parameters are undistinguishable identifications.

There is an implicit function relation between (1) two parameter

If two parameter θ are arranged in the model ₁, θ ₂If the trajectory indistinguishable according to two parameters identified, indicating that the two parameters showed an implicit function common to track work, ie

Assuming

on the two parameters can lead, according to the chain rule of calculus can be obtained

Then have

It should be noted that;

becomes when being, and become during

right and wrong.Therefore, the sensitivity on the timeline track

proportional to each other the same time a zero-crossing, if the sensitivity curve of the oscillation track, the performance of the in-phase or anti-phase.

(2) parameter divides into groups to exist the implicit function relation

If parameter is total M in the model, wherein I relevant parameter divides K group

${\mathrm{\θ}}_k=[{\mathrm{\θ}}_{k1},{\mathrm{\θ}}_{k2},...,{\mathrm{\θ}}_{{\mathrm{kM}}_k}],$ k＝1，2，…，K (6)

If incoherent independent parameter does

θ _K+1＝[θ _I+1，θ _I+2，…，θ _M] (7)

In the output variable track, each group parameter influences track, promptly jointly through certain implicit function relation

For relevant between the same group of parameter and do not have the situation of repetition between on the same group the parameter, promptly $\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{M}_k=I.$

Let

right parameters can lead, there

For the same set of parameters

And when the same variable, and

different but constant, so the same set of parameters of the trajectory sensitivity proportional to each other that both crossings.However, due to the different set of parameters

is not the same, different set of parameters are not simultaneously zero crossing trajectory sensitivities.

(3) there are a plurality of implicit functions relations in many group parameters intersections

In general, not only between the same set of parameters related to, but different sets of parameters may be repeated, or that the same argument may be repeated in different groups, namely

Let

right parameters can lead, there

So

In general, K＜I, so

So;

L,

linear correlation.

To sum up visible, the trace sensitivity of parameter that does not repeat to appear at other group in same group of parameter is proportional mutually, that is to say zero crossing simultaneously; But for the parameter group that repetition parameter is not arranged between the parameter on the same group, the trace sensitivity of these group parameters is zero crossing simultaneously not, but has linear dependence.In other words, if the trace sensitivity of several parameters while zero crossing then can be judged these parameter correlations, but promptly not be unique identification; If the sensitivity of several parameters is zero passage simultaneously not, also linear correlation not can judge that then these parameters are uncorrelated, but promptly unique identification.So, as long as the trace sensitivity of all parameters of check whether simultaneously zero crossing or linear correlation, just can judge the identifiability of parameter.If trace sensitivity is an oscillating curve, zero crossing means that it is homophase or anti-phase that oscillatory process looks simultaneously.

3, the classification identification of generator parameter

Generator is classified, think that for every type of generator its parameter is identical.Classification main kind and capacity according to generator.At first, can be divided into the hydraulic turbine and steam turbine according to the kind of generator.Classify according to generator capacity then, such as 600MW be one type, 300MW's is one type.

Identifiability analytical method according to the front based on trace sensitivity; The identifiability of parameter in every generator Park model in the analytical system; Show after deliberation; The same parameter of same type generator cannot be separated identification, and the same parameter of dissimilar generators can be separated identification.

4, the substep identification of generator parameter

Confirm the partial parameters of generator earlier according to the steady state measurement data, confirm rest parameter according to the dynamic data that WAMS collects again.Be the method for substep identification below:

Steady state voltage equation by excitation winding:

u _f0＝r _fi _f0

(12)

Can get excitation winding resistance:

r _f＝u _f0/i _f0

In the formula, subscript f represents the excitation winding variable, and subscript 0 is represented steady-state value.

The steady state voltage equation of stator winding is:

$\left\{\begin{array}{c}{u}_d=-r_a{i}_d+X_q{i}_q\\ u_q=-X_d{i}_d-r_a{i}_q+X_{\mathrm{ad}}{i}_f\end{array}\right.$

(13)

Wherein have 4 unknown parameters, i.e. r _a, X _d, X _q, X _AdAccording to former and later two different steady state datas of dynamic process, can get 4 equations:

$\left\{\begin{array}{c}{u}_{d0}=-r_a{i}_{d0}+X_q{i}_{q0}\\ u_{q0}=-X_d{i}_{d0}-r_a{i}_{q0}+X_{\mathrm{ad}}{i}_{f0}\\ u_{d\∞}=-r_a{i}_{d\∞}+X_q{i}_{q\∞}\\ u_{q\∞}=-X_d{i}_{d\∞}-r_a{i}_{q\∞}+X_{\mathrm{ad}}{i}_{f\∞}\end{array}\right.$

(14)

In the formula, subscript ∞ represents the back steady-state value.Can uniquely obtain r in view of the above _a, X _d, X _q, X _Ad, also can obtain relevant parameter:

$\left\{\begin{array}{c}{X}_l=X_d-X_{\mathrm{ad}}\\ X_{\mathrm{aq}}=X_q-X_l\end{array}\right.$

(15)

We can obtain r by top formula _a, r _f, X _d, X _q, X _Ad, X _Aq, X _lThis Several Parameters.

Can know by last surface analysis,, still need identification X for 6 rank models _f, X _D, X _Q, X _g, r _D, r _Q, r _gAnd T _jTotally 8 parameters.For 5 rank models, still need identification X _f, X _D, X _Q, r _D, r _QAnd T _jTotally 6 parameters.

5, the evaluation index of selective systemization

$J=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{[Y(\mathrm{\θ},k)-Y_m\left(k\right)]}^{T}W\left(k\right)[Y(\mathrm{\θ},k)-Y_m\left(k\right)]$

In the formula, Y is the dynamic response of the main observational variable of system, and subscript m is represented the WAMS measured value, and θ is a parameter of treating identification, and W (k) is a weight coefficient matrix.Observational variable should be able to reflect the main behavioral characteristics of system, and sensitive to parameter.For the generator parameter identification, it is proper selecting the relative swing curve between the generator.

6, the selection of observational variable

In multi-machine power system, different observational variables is different for the sensitivity of parameter, so the effect of selecting different observational variables to carry out parameter identification is also just different.For multi-machine power system, can the quality of model mainly see the overall dynamics behavior of the system that simulate, and generally compares the relative merit of generator angle, generator port voltage relative phase angle, interconnection power and hinge busbar voltage between the region-of-interest.Adopt different observational variables, the sensitivity phase place of same parameter might be different, so the parameter identifiability that adopts single observational variable to obtain is incomplete.So we need analyze different observational variables, have only the sensitivity that obtains when all observational variables zero passage simultaneously, could judge that these parameters can not separate identification.The simultaneity factor observational variable should be able to reflect the main behavioral characteristics of system; And it is relatively sensitiveer to parameter; Be used for carrying out parameter identification so choose the observational variable that trace sensitivity is big, identification precision is high; Such as, two or two mass-sendings relative merit angle or the relative phase angle of port voltage between the motors.

7, select the emphasis parameter

In rest parameter, some parameter is very little to measured dynamic trajectory and error effect thereof, reduces the parameter identification number thereby it can be taken as representative value.Select promptly that sensitivity is big, the also big parameter of influence is carried out the emphasis identification.

8, identification many generator parameters the time

The generator of same type adopts same set of parameter in the system, and when system's generation disturbance, as observational variable, the emphasis parameter of all generators in the identification system adopts ant group algorithm simultaneously, calls BPA simultaneously and realizes, sees Fig. 1 with the relative merit of generator angle.

Claims (3)

1. the method for many generator parameters identification simultaneously is characterized in that:

(1) evaluation index of selective systemization:

$J=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{[Y(\mathrm{\θ},k)-Y_m\left(k\right)]}^{T}W\left(k\right)[Y(\mathrm{\θ},k)-Y_m\left(k\right)]$

In the formula, Y is the dynamic response of the main observational variable of system, and subscript m is represented the WAMS measured value, and θ is a parameter of treating identification, and W (k) is a weight coefficient matrix;

(2) select observational variable, many generator parameters are carried out whole identification here, the systematic perspective of selecting WAMS to collect is measured, and the systematic observation variable should be able to reflect the main behavioral characteristics of system, and relatively sensitiveer to parameter;

(3) select the emphasis parameter;

(4) select the proper optimization method, because the parametric solution space is quite complicated, need the strong method of global optimization ability, the simulated evolution method is suitable selection, through contrast on probation, adopts ant group algorithm;

2. the method for a kind of many generator parameters according to claim 1 identification simultaneously, it is characterized in that: said selection emphasis parameter adopts:

1. the identification of classifying is classified to generator, and classification is mainly according to the kind and the capacity of generator;

2. the partial parameters of generator is confirmed earlier in substep identification according to the steady state measurement data, confirm rest parameter according to the dynamic data that WAMS collects again;

3. emphasis identification according to trace sensitivity, is selected the emphasis parameter that sensitivity is big, influence is also big;

4. pay attention to integral body, in electric system simulation calculated, what people paid close attention to the most was the overall permanence that can model reflect system.

3. the method for a kind of many generator parameters according to claim 1 identification simultaneously; It is characterized in that: said ant group algorithm is realized in the following way: in multi-computer system synchronous generator parameter identification; An aspect is to optimize when carrying out many generator parameters; The another one aspect is to calculate the target function value of reflection entire system dynamic behaviour; Between optimizer and BPA program, set up exchanges data; Promptly optimize the middle generator parameter result who once obtains and substitute the generator parameter among the BPA automatically; BPA Automatic computing system dynamic response and export to optimizer then; Optimizer calculates and obtains target function value; And further optimize the new generator parameter value of acquisition, up to finding optimal solution.

Images