CN103051272B - Magnification factor correction method for power stabilizer - Google Patents

Abstract

The invention discloses a magnification factor correction method for a power stabilizer. The magnification factor correction method comprises the following steps of: predicting the theoretical value of a rotating speed deviation peak value through theoretical derivation; determining an adjustment coefficient according to the ratio of a measured value to the theoretical value of the rotating speed deviation peak value; dynamically adjusting the magnification factor benchmark values of a corresponding motor in different running states according to the adjustment coefficient; and selecting an appropriate magnification factor benchmark value according to the practical running state of the motor to control the running state of the motor. The magnification factor correction method has the beneficial technical effects that the magnification factor of the stabilizer is adjusted dynamically by using a generator system according to a running state requirement under interference of different magnitudes, so that the adaptability is enhanced, oscillation amplitudes among different oscillation periods tend to attenuate rapidly, and system oscillation is eliminated rapidly. Due to the adoption of the magnification factor correction method, the steady-state performance and transient-state performance of the system are enhanced in a larger running range, and the safe running of a power system is ensured.

Description

Magnification factor correction method for power stabilizer

Technical field

The present invention relates to a kind of generator speed that makes and recover stable control technology under interference, particularly relate to a kind of magnification factor correction method for power stabilizer.

Background technology

At present, along with the expansion of electric power system scale, employing that is interconnected and Large-scale machine set quick response excitation system make low-frequency oscillation problem become increasingly conspicuous, serious threat is to the safe operation of electric power system; Theory and practice shows, adopts power system stabilizer, PSS (Power System Stabilizer, PSS) power oscillation damping to be a kind of simple effective method.

Existing PSS is primarily of unit compositions such as signal transacting, filtering, advanced-delayed and amplifications; In its design process, its multiplication factor is adjusted very important, because the damping action of PSS decides primarily of the size of stabilizer multiplication factor.

Electric power system is the non-linear complex large system of a dynamic equilibrium, its service conditions is variable, if stabilizer multiplication factor is the fixed value obtained of adjusting under a certain service conditions, then when power system operation condition changes, may no longer meet the requirement that enough dampings are provided because multiplication factor is too small, or multiplication factor is excessive and worsen and control and the performance of voltage correction.

In order to improve the adaptability to service conditions change, the line reactance identifier that the employing microcomputer had in prior art is formed, it can along with the change of system reactance, by the mode of tabling look-up, find in advance by calculate stored in the stabilizer multiplication factor corresponding to reactance value; In actual motion, reactance identifier is not only needed to measure a large amount of parameters (active power, reactive power, generator voltage and merit angle etc.), but also a series of calculating need be carried out, compare, operation is selected and to be tabled look-up etc. in reactance, in addition, if table will do more detailed by the look-up table wherein related to, then workload is huge and loaded down with trivial details; And power system operation condition is Protean, look-up table also exists bad adaptability, very flexible, the unmanageable shortcoming of error.

Because reactance identifier exists many disadvantages, researchers also been proposed and adopt the method for on-line control to regulate stabilizer multiplication factor, but these schemes have a common point: first, need in stabilizer, arrange a multiplication factor fiducial value, this fiducial value can be worked in larger range of operation.Secondly, then adopt all kinds of control device to carry out on-line amending to this fiducial value, to reach the object of the adaptation to service conditions change.In August, 2004, IEEE formulates definition and the classification of up-to-date stability of power system, and the stable of merit angle be divided into static stability (minor interference power-angle stability) and transient stability (disturbing power-angle stability greatly) two class.Such scheme normally runs in electric power system, under the different situations of minor interference and large interference, multiplication factor is all revised same fiducial value.Particularly in electric power system two kinds of very large situations of running status gap, as normally run and disturbing two states greatly, what multiplication factor adopted is same fiducial value.Through experimental verification, the adjustment of the method to stabilizer multiplication factor seems too coarse, can not obtain best stability contorting effect.

Summary of the invention

For the problem in background technology, the present invention, by a series of means, makes system according to the running status needs of generator, dynamically can adjust stabilizer multiplication factor, makes the trend of the oscillation amplitude between different cycle of oscillation in decay; Concrete scheme is: a kind of magnification factor correction method for power stabilizer, the method comprises the steps:

1) respectively corresponding stabilizer multiplication factor fiducial value is set for the different running statuses of generator: generator operation corresponding stabilizer multiplication factor fiducial value when normal operating conditions is designated as the stabilizer multiplication factor fiducial value that generator operation is corresponding when minor interference state is designated as the stabilizer multiplication factor fiducial value that generator operation is corresponding when large disturbance state is designated as $K_C^{*};$

2) when generator is interfered in running, speed difference between generator actual speed and expectation rotating speed is periodic swinging, speed difference between actual speed and expectation rotating speed is also offspeed value, obtains crest value and the trough value of continuous print offspeed value;

3) single crest value or trough value are defined as offspeed peak value, two offspeed peak values adjacent in sequential, the theoretical value that after calculating according to last offspeed peak value, an offspeed peak value is corresponding: establish two the offspeed peak values occurred in turn in section to be sometime followed successively by Δ ω _maxand Δ ω (I) _max(I+1); According to Δ ω _max(I), calculate under standard damping condition with Δ ω _max(I+1) corresponding in sequential theoretical value:

Δω _max(I+1) ^*＝Δω _max(I)·exp{-d ^*·2πf[t(I+1)-t(I)]}

Wherein, Δ ω _max(I+1) ^*the theoretical value that after being, an offspeed peak value is corresponding; Exp is exponential function; d ^*for standard damping; F is frequency of oscillation; T (I+1) is for getting Δ ω _max(I+1) time point time; T (I) is for getting Δ ω _max(I) time point time;

4) regulation coefficient λ is calculated according to following formula:

$\mathrm{\λ}=\frac{\left|{\mathrm{\Δ\ω}}_{\max }(I+1)\right|}{\left|{\mathrm{\Δ\ω}}_{\max }{(I+1)}^{*}\right|};$

5) if λ > 1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}(1+\mathrm{\λ})$

$K_B=K_B^{*}(1+\mathrm{\λ})$

$K_C=K_C^{*}(1+\mathrm{\λ})$

If λ < 1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}(1-\mathrm{\&lambda;})$

$K_B=K_B^{*}(1-\mathrm{\&lambda;})$

$K_C=K_C^{*}(1-\mathrm{\&lambda;})$

If λ=1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}$

$K_B=K_B^{*}$

$K_C=K_C^{*}$

Wherein, K _abe corresponding adjusted value; K _bbe corresponding adjusted value; K _cbe corresponding adjusted value;

6) according to the actual motion state of generator, from K _a, K _band K _cin choose corresponding parameter as stabilizer multiplication factor, the running status of generator is controlled;

Repeat aforementioned operation, continuous control is carried out to the running status of generator.

Aforementioned control method can make between adjacent offspeed peak value (being also oscillation amplitude) in attenuation trend; On the basis of aforementioned schemes, in order to make the solution of the present invention be optimized further, the invention allows for following improvement project:

Step 6) in, from K _a, K _band K _cin choose corresponding parameter before, first as follows to K _a, K _band K _cprocess:

(1) within single cycle of oscillation, the offspeed value Δ ω of Real-time Obtaining generator, calculates the rate of change of offspeed value relative to time variations according to Δ ω

(2) preset a two dimension fuzzy checking list in system, Δ ω and as two input variables of two dimension fuzzy checking list, according to the output variable N of two dimension fuzzy checking list to K _a, K _band K _ccarry out dynamic conditioning, finally make the stabilizer multiplication factor fiducial value acted in single cycle of oscillation be dynamic change, to adapt in single cycle of oscillation, the real-time demand for control of generator;

To K _a, K _band K _cthe method of carrying out dynamic conditioning is undertaken by following formula:

K _A(N)＝K _A·N

K _B(N)＝K _B·N

K _C(N)＝K _C·N

Wherein, N is the output variable of two dimension fuzzy checking list; K _a(N) be K _acorresponding numerical value after the adjustment of two dimension fuzzy checking list; K _b(N) be K _bcorresponding numerical value after the adjustment of two dimension fuzzy checking list; K _c(N) be K _ccorresponding numerical value after the adjustment of two dimension fuzzy checking list;

After completing the operation of step (1) and (2), from K _a(N), K _band K (N) _c(N) choose in corresponding parameter as stabilizer multiplication factor (also by aforesaid " from K _a, K _band K _cin choose corresponding parameter " K in operation _a, K _band K _creplace with K _a(N), K _band K (N) _c(N)), the running status of generator is controlled.

This improvement project can solve the adaptability problem of inner balancer multiplication factor single cycle of oscillation, make the stabilizer multiplication factor fiducial value acted in single cycle of oscillation be dynamic change, to adapt in single cycle of oscillation, the Δ ω of generator is in demand for control during different numerical value; After this improvement project is combined with aforementioned schemes, both the problem that oscillation amplitude between cycle of oscillation is excessive can have been solved, the adaptability problem of inner balancer multiplication factor single cycle of oscillation can also be solved, thus the adaptive ability of control system is got a promotion, to calm down the vibration of system as early as possible, improve bad adaptability appears in PSS deficiency when system running state changes, ensure that the safe operation of electric power system to a greater extent.

For the two dimension fuzzy checking list described in above, the invention allows for following preferred version:

Wherein, NB, NS, ZE, PS, PB are 5 fuzzy subsets of input variable; CH, CL, OK, AL, AH are 5 fuzzy subsets of output variable N; NB, NS, ZE, PS, PB represent respectively: negative large, negative little, zero, just little, honest; CH, CL, OK, AL, AH represent respectively: high-shrinkage, low-shrinkage, constant, lowly to put, Gao Fang.

Its fuzzy control rule is:

Can be divided into 1 ~ 4 interval in figure see a cycle of oscillation shown in Fig. 3,4,5, Fig. 3, Fig. 4 is the degree of membership distribution map of input variable, and Fig. 5 is the degree of membership distribution map of output variable;

When 1, being in the 1st section in Fig. 3 when system responses, the absolute value of offspeed value Δ ω rises, vibration aggravation, for eliminating deviation value as early as possible, stabilizer multiplication factor should strengthen, to provide more damping, to make Δ ω return to rated value as early as possible, therefore, there is following control law:

Rule 1: if generator speed disagreement value A ω is PB and its rate of change be PB, then output variable N is AH;

Rule 2: if generator speed disagreement value A ω is PB and its rate of change be PS, then output variable N is AH;

Rule 3: if generator speed disagreement value A ω is PS and its rate of change be PB, then output variable N is AH;

Rule 4: if generator speed disagreement value A ω is PS and its rate of change be PS, then output variable N is AL;

When 2, being in the 2nd section in Fig. 3 when system responses, the absolute value of speed deviations rated value Δ ω declines, oscillations; When Δ ω is larger, stabilizer multiplication factor should remain unchanged; When Δ ω is less, principal contradiction is now converted into the stability of a system, for preventing system overshoot excessive and in order to make system stablize as early as possible, the effect of damping should be reduced, therefore, now should suitably reduce stabilizer multiplication factor, in order to avoid excessively strong control action causes vibration aggravation in next time-domain, therefore, following control law is had:

Rule 5: if generator speed disagreement value A ω is PB and its rate of change be NB, then output variable N is OK;

Rule 6: if generator speed disagreement value A ω is PB and its rate of change be NS, then output variable N is OK;

Rule 7: if generator speed disagreement value A ω is PS and its rate of change be NB, then output variable N is CH;

Rule 8: if generator speed disagreement value A ω is PS and its rate of change be NS, then output variable N is CL;

When 3, being in the 3rd section in Fig. 3 when system responses, the absolute value of Δ ω has the trend of increase, vibration aggravation.For eliminating deviation value as early as possible, stabilizer multiplication factor should strengthen, and to provide more damping, to prevent Δ ω from increasing further, makes it get back to speed rated value as early as possible.Therefore, following control law is had:

Rule 9: if generator speed disagreement value A ω is NB and its rate of change be NB, then output variable N is AH;

Rule 10: if generator speed disagreement value A ω is NB and its rate of change be NS, then output variable N is AH;

Rule 11: if generator speed disagreement value A ω is NS and its rate of change be NB, then output variable N is AH;

Rule 12: if generator speed disagreement value A ω is NS and its rate of change be NS, then output variable N is AL;

When 4, being in the 4th section in Fig. 3 when system responses, the absolute value of Δ ω has a declining tendency, oscillations, and now, when the absolute value of Δ ω is larger, stabilizer multiplication factor should remain unchanged; When Δ ω is less, principal contradiction is now converted into the stability of a system, for preventing system overshoot excessive and making system stablize as early as possible, should suitably reduce stabilizer multiplication factor, in order to avoid excessively strong control action causes vibration aggravation in next time-domain, therefore, there is following control law:

Rule 13: if generator speed disagreement value A ω is NB and its rate of change be PB, then output variable N is OK;

Rule 14: if generator speed disagreement value A ω is NB and its rate of change be PS, then output variable N is OK;

Rule 15: if generator speed disagreement value A ω is NS and its rate of change be PB, then output variable N is CH;

Rule 16: if generator speed disagreement value A ω is NS and its rate of change be PS, then output variable N is CL;

5, when absolute value very little time, now no matter Δ ω is large or little, all should strengthen stabilizer multiplication factor, to provide more damping, makes Δ ω get back to rated value as early as possible, therefore, has following control law:

Rule 17: if generator speed disagreement value A ω is PB and its rate of change be ZE, then output variable N is AH;

Rule 18: if generator speed disagreement value A ω is PS and its rate of change be ZE, then output variable N is AL;

Rule 19: if generator speed disagreement value A ω is NB and its rate of change be ZE, then output variable N is AH;

Rule 20: if generator speed disagreement value A ω is NS and its rate of change be ZE, then output variable N is AL;

6, when the absolute value of Δ ω is very little, represent that vibration has been calmed down or vibrated, now should stop regulating, namely stabilizer multiplication factor should remain unchanged, and can obtain remaining 5 control laws thus.

Address above and needed the running status residing for generator reality (i.e. normal operating conditions, minor interference state and large disturbance state) to select corresponding stabilizer multiplication factor fiducial value, and it is adjusted, then stabilizer is acted on, for the problem how choosing corresponding stabilizer multiplication factor fiducial value, the invention allows for following preferred version:

Step 6) in, choose corresponding parameter as follows:

[1] means by experiment, the different running statuses of simulation generator, are obtained by debugging and correspond respectively to the stabilizer multiplication factor fiducial value of generator operation under normal operating conditions, minor interference state and large disturbance state with

[2] in experimentation, gather respectively to the data of generator under three kinds of running statuses, the data collected are the output signal u of stabilizer and the tachometer value ω of generator; For often kind of running status, all obtain one group of burst and one group of rotating speed sequence, burst is the ordered series of numbers u comprising multiple u ₁, u ₂u _i, each u arranges chronologically; Rotating speed sequence is the ordered series of numbers ω comprising multiple ω ₁, ω ₂ω _i, and ω and u is according to sequential one_to_one corresponding;

[3] according to the burst under three kinds of running statuses and rotating speed sequence, set up the neural network prediction model under the three kinds of running statuses corresponding to generator respectively, and set up forecasting model database according to three neural network prediction models;

[4] forecasting model database is put into operation online, real output signal u (k) of Real-time Collection stabilizer and actual speed value ω (k) of generator, k is the sequence number of sampling number; By the I and II time delay value input prediction model library of u (k) and ω (k), the rotor speed forecast value under the three kinds of running statuses corresponding respectively to generator can be obtained with according to following formula, calculate respectively with with between error e _a(k), e _b(k) and e _c(k):

$e_A\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_A\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

$e_B\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_B\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

$e_C\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_C\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

Wherein, ω (k) is the actual speed value of the generator of acquisition when kth time is sampled; with what be respectively that forecasting model database exports corresponds under three kinds of running statuses, possesses the rotor speed forecast value of corresponding relation with ω (k) in sequential; the normal operating conditions of corresponding generator, the minor interference state of corresponding generator, the large disturbance state of corresponding generator;

[5] e is compared _a(k), e _b(k) and e _cthe size of (k), the running status that wherein numerical value reckling is corresponding is the current operating conditions of generator, using the parameter of correspondence as stabilizer multiplication factor (if i.e., e _ak (), then judge that generator current operating conditions is normal operating conditions, if e _bk () is minimum, then judge that generator current operating conditions is minor interference state, if e _ck () is minimum, then judge that generator current operating conditions is large disturbance state), the running status of generator is controlled.

Introduce neural network prediction model in the program, effectively can solve the current operating conditions problem how identifying generator, coordinate aforesaid stabilizer multiplication factor fiducial value method of adjustment, the performance of control system can be made to be optimized further.In the method for the current operating conditions of aforesaid employing neural network prediction model identification generator, only considering error current, in order to make error more accurate, also can do following further improvement to it:

Step compares e in [5] _a(k), e _b(k) and e _cbefore the size of (k), according to following formula to e _a(k), e _b(k) and e _ck () is for further processing:

$e_A^{\&prime;}\left(k\right)=e_A\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_A\left(j\right)$

$e_B^{\&prime;}\left(k\right)=e_B\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_B\left(j\right)$

$e_C^{\&prime;}\left(k\right)=e_C\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_C\left(j\right)$

Wherein, e' _ak () is for corresponding to e _athe overall error of (k); E' _bk () is for corresponding to e _bthe overall error of (k); E' _ck () is for corresponding to e _cthe overall error of (k); e _aj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; e _bj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; e _cj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; Q is matching length, and q is positive integer, q>=2; K, j are positive integer, and k-q≤j≤k-1; β is matching error forgetting factor, 0 < β < 1;

Then, in step [5], e is compared _a(k), e _b(k) and e _cduring the size of (k), by e _a(k), e _b(k) and e _ck () replaces with e' respectively _a(k), e' _b(k) and e' _c(k), then compare e' _a(k), e' _b(k) and e' _ck the size of () also continues subsequent treatment.

This has considered error current and history error, has made the basis for estimation of control system more accurate, further increase systematic function in improving.

Advantageous Effects of the present invention is: make generator system according to running status needs, under the interference of different size, dynamic conditioning stabilizer multiplication factor, strengthens its adaptivity, make the oscillation amplitude between different cycle of oscillation be rapid decay trend, promptly calm down system oscillation.In larger range of operation, improve steady-state behaviour and the transient performance of system, guarantee the safe operation of electric power system.

Accompanying drawing explanation

Fig. 1, the neural network prediction model principle schematic (ω in figure _i(k) rotating speed for collecting, u _i(k) stabilizer output signal for collecting, for the rotor speed forecast value exported; I=A, B, C one of them);

Fig. 2, control principle schematic diagram of the present invention;

Fig. 3, be disturbed rear spinner velocity and depart from response typical plot;

The degree of membership distribution situation figure of two input variables of Fig. 4, two dimension fuzzy checking list;

The degree of membership distribution situation figure of the output variable of Fig. 5, two dimension fuzzy checking list.

Embodiment

1, a magnification factor correction method for power stabilizer, is characterized in that: the method comprises the steps:

1) respectively corresponding stabilizer multiplication factor fiducial value is set for the different running statuses of generator: generator operation corresponding stabilizer multiplication factor fiducial value when normal operating conditions is designated as the stabilizer multiplication factor fiducial value that generator operation is corresponding when minor interference state is designated as the stabilizer multiplication factor fiducial value that generator operation is corresponding when large disturbance state is designated as $K_C^{*};$

2) when generator is interfered in running, speed difference between generator actual speed and expectation rotating speed is periodic swinging, speed difference between actual speed and expectation rotating speed is also offspeed value, obtains crest value and the trough value of continuous print offspeed value;

3) single crest value or trough value are defined as offspeed peak value, two offspeed peak values adjacent in sequential, the theoretical value that after calculating according to last offspeed peak value, an offspeed peak value is corresponding: establish two the offspeed peak values occurred in turn in section to be sometime followed successively by Δ ω _maxand Δ ω (I) _max(I+1); According to Δ ω _max(I), calculate under standard damping condition with Δ ω _max(I+1) corresponding in sequential theoretical value:

Δω _max(I+1) ^*＝Δω _max(I)·exp{-d ^*·2πf[t(I+1)-t(I)]}

Wherein, Δ ω _max(I+1) ^*the theoretical value that after being, an offspeed peak value is corresponding; Exp is exponential function; d ^*for standard damping; F is frequency of oscillation; T (I+1) is for getting Δ ω _max(I+1) time point time; T (I) is for getting Δ ω _max(I) time point time;

4) regulation coefficient λ is calculated according to following formula:

$\mathrm{\&lambda;}=\frac{\left|{\mathrm{\&Delta;\&omega;}}_{\max }(I+1)\right|}{\left|{\mathrm{\&Delta;\&omega;}}_{\max }{(I+1)}^{*}\right|};$

5) if λ > 1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}(1+\mathrm{\&lambda;})$

$K_B=K_B^{*}(1+\mathrm{\&lambda;})$

$K_C=K_C^{*}(1+\mathrm{\&lambda;})$

If λ < 1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}(1-\mathrm{\&lambda;})$

$K_B=K_B^{*}(1-\mathrm{\&lambda;})$

$K_C=K_C^{*}(1-\mathrm{\&lambda;})$

If λ=1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}$

$K_B=K_B^{*}$

$K_C=K_C^{*}$

Wherein, K _abe corresponding adjusted value; K _bbe corresponding adjusted value; K _cbe corresponding adjusted value;

6) according to the actual motion state of generator, from K _a, K _band K _cin choose corresponding parameter as stabilizer multiplication factor, the running status of generator is controlled;

Repeat aforementioned operation, continuous control is carried out to the running status of generator.For the ease of describing, preceding method may be defined as one-level precorrection (its link residing is in systems in which as Suo Shi " one-level precorrection " module in Fig. 2).

The present invention is to K _a, K _band K _cthe method that the further adjustment carried out adopts may be defined as secondary ambiguity correction (its link residing is in systems in which as Suo Shi " secondary ambiguity correction " module in Fig. 2), and the concrete scheme of this secondary ambiguity correction is: step 6) in, from K _a, K _band K _cin choose corresponding parameter before, first as follows to K _a, K _band K _cprocess:

(1) within single cycle of oscillation, the offspeed value Δ ω of Real-time Obtaining generator, calculates the rate of change of offspeed value relative to time variations according to Δ ω

(2) preset a two dimension fuzzy checking list in system, Δ ω and as two input variables of two dimension fuzzy checking list, according to the output variable N of two dimension fuzzy checking list to K _a, K _band K _ccarry out dynamic conditioning, finally make the stabilizer multiplication factor fiducial value acted in single cycle of oscillation be dynamic change, to adapt in single cycle of oscillation, the real-time demand for control of generator;

To K _a, K _band K _cthe method of carrying out dynamic conditioning is undertaken by following formula:

K _A(N)＝K _A·N

K _B(N)＝K _B·N

K _C(N)＝K _C·N

Wherein, N is the output variable of two dimension fuzzy checking list; K _a(N) be K _acorresponding numerical value after the adjustment of two dimension fuzzy checking list; K _b(N) be K _bcorresponding numerical value after the adjustment of two dimension fuzzy checking list; K _c(N) be K _ccorresponding numerical value after the adjustment of two dimension fuzzy checking list;

After completing the operation of step (1) and (2), from K _a(N), K _band K (N) _c(N) choose corresponding parameter in as stabilizer multiplication factor, the running status of generator is controlled.

Two dimension fuzzy checking list mentioned above is as shown in the table:

Wherein, NB, NS, ZE, PS, PB are 5 fuzzy subsets of input variable; CH, CL, OK, AL, AH are 5 fuzzy subsets of output variable N; NB, NS, ZE, PS, PB represent respectively: negative large, negative little, zero, just little, honest; CH, CL, OK, AL, AH represent respectively: high-shrinkage, low-shrinkage, constant, lowly to put, Gao Fang.

Aforesaid step 6) in, corresponding parameter (it is made up of " selection of the stabilizer multiplication factor fiducial value " module in Fig. 2 and " forecasting model database " module in systems in which) can be chosen as follows:

[1] means by experiment, the different running statuses of simulation generator, are obtained by debugging and correspond respectively to the stabilizer multiplication factor fiducial value of generator operation under normal operating conditions, minor interference state and large disturbance state with

[2] in experimentation, gather respectively to the data of generator under three kinds of running statuses, the data collected are the output signal u of stabilizer and the tachometer value ω of generator; For often kind of running status, all obtain one group of burst and one group of rotating speed sequence, burst is the ordered series of numbers u comprising multiple u ₁, u ₂u _i, each u arranges chronologically; Rotating speed sequence is the ordered series of numbers ω comprising multiple ω ₁, ω ₂ω _i, and ω and u is according to sequential one_to_one corresponding;

[3] according to the burst under three kinds of running statuses and rotating speed sequence, set up corresponding to the neural network prediction model under three kinds of running statuses of generator (as " the neural network prediction model M in Fig. 2 respectively _a" module, " neural network prediction model M _b" module and " neural network prediction model M _c" module), and set up forecasting model database according to three neural network prediction models; The principle of neural network prediction model as shown in Figure 1;

[4] forecasting model database is put into operation online, real output signal u (k) of Real-time Collection stabilizer and actual speed value ω (k) of generator, k is the sequence number of sampling number; By the I and II time delay value input prediction model library of u (k) and ω (k), the rotor speed forecast value under the three kinds of running statuses corresponding respectively to generator can be obtained with according to following formula, calculate respectively with with between error e _a(k), e _b(k) and e _c(k):

$e_A\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_A\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

$e_B\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_B\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

$e_C\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_C\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

Wherein, ω (k) is the actual speed value of the generator of acquisition when kth time is sampled; with what be respectively that forecasting model database exports corresponds under three kinds of running statuses, possesses the rotor speed forecast value of corresponding relation with ω (k) in sequential; the normal operating conditions of corresponding generator, the minor interference state of corresponding generator, the large disturbance state of corresponding generator;

[5] e is compared _a(k), e _b(k) and e _ck the size of (), the running status that wherein numerical value reckling is corresponding is the current operating conditions of generator, using the parameter of correspondence as stabilizer multiplication factor, controls the running status of generator.

On the basis of introducing neural network prediction model, the present invention has also done following improvement: step compares e in [5] _a(k), e _b(k) and e _cbefore the size of (k), according to following formula to e _a(k), e _b(k) and e _ck () is for further processing:

$e_A^{\&prime;}\left(k\right)=e_A\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_A\left(j\right)$

$e_B^{\&prime;}\left(k\right)=e_B\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_B\left(j\right)$

$e_C^{\&prime;}\left(k\right)=e_C\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_C\left(j\right)$

Wherein, e' _ak () is for corresponding to e _athe overall error of (k); E' _bk () is for corresponding to e _bthe overall error of (k); E' _ck () is for corresponding to e _cthe overall error of (k); e _aj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; e _bj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; e _cj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; Q is matching length, and q is positive integer, q>=2; K, j are positive integer, and k-q≤j≤k-1; β is matching error forgetting factor, 0 < β < 1;

Then, e is compared _a(k), e _b(k) and e _cduring the size of (k), by e _a(k), e _b(k) and e _ck () replaces with e' respectively _a(k), e' _b(k) and e' _c(k), then compare e' _a(k), e' _b(k) and e' _ck the size of () also continues subsequent treatment.

Claims (5)

1. a magnification factor correction method for power stabilizer, is characterized in that: the method comprises the steps:

1) respectively corresponding stabilizer multiplication factor fiducial value is set for the different running statuses of generator: generator operation corresponding stabilizer multiplication factor fiducial value when normal operating conditions is designated as the stabilizer multiplication factor fiducial value that generator operation is corresponding when minor interference state is designated as the stabilizer multiplication factor fiducial value that generator operation is corresponding when large disturbance state is designated as

2) when generator is interfered in running, speed difference between generator actual speed and expectation rotating speed is periodic swinging, speed difference between actual speed and expectation rotating speed is also offspeed value, obtains crest value and the trough value of continuous print offspeed value;

3) single crest value or trough value are defined as offspeed peak value, two offspeed peak values adjacent in sequential, the theoretical value that after calculating according to last offspeed peak value, an offspeed peak value is corresponding: establish two the offspeed peak values occurred in turn in section to be sometime followed successively by Δ ω _maxand Δ ω (I) _max(I+1); According to Δ ω _max(I), calculate under standard damping condition with Δ ω _max(I+1) corresponding in sequential theoretical value:

Δω _max(I+1) ^*＝Δω _max(I)·exp{-d ^*·2πf[t(I+1)-t(I)]}

Wherein, Δ ω _max(I+1) ^*the theoretical value that after being, an offspeed peak value is corresponding; Exp is exponential function; d ^*for standard damping; F is frequency of oscillation; T (I+1) is for getting Δ ω _max(I+1) time point time; T (I) is for getting Δ ω _max(I) time point time;

4) regulation coefficient λ is calculated according to following formula:

$\mathrm{\&lambda;}=\frac{\left|\mathrm{\&Delta;}{\mathrm{\&omega;}}_{\max }(I+1)\right|}{\left|{\mathrm{\&Delta;\&omega;}}_{\max }{(I+1)}^{*}\right|};$

5) if λ > 1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}(1+\mathrm{\&lambda;})$

$K_B=K_B^{*}(1+\mathrm{\&lambda;})$

$K_C=K_C^{*}(1+\mathrm{\&lambda;})$

If λ < 1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}(1-\mathrm{\&lambda;})$

$K_B=K_B^{*}(1-\mathrm{\&lambda;})$

$K_C=K_C^{*}(1-\mathrm{\&lambda;})$

If λ=1, then by following formula, multiplication factor fiducial value is adjusted:

$K_A=K_A^{*}$

$K_B=K_B^{*}$

$K_C=K_C^{*}$

Wherein, K _abe corresponding adjusted value; K _bbe corresponding adjusted value; K _cbe corresponding adjusted value;

6) according to the actual motion state of generator, from K _a, K _band K _cin choose corresponding parameter as stabilizer multiplication factor, the running status of generator is controlled;

Repeat step 2) to 6) and operation, continuous control is carried out to the running status of generator.

2. magnification factor correction method for power stabilizer according to claim 1, is characterized in that: step 6) in, from K _a, K _band K _cin choose corresponding parameter before, first as follows to K _a, K _band K _cprocess:

(1) within single cycle of oscillation, the offspeed value Δ ω of Real-time Obtaining generator, calculates the rate of change of offspeed value relative to time variations according to Δ ω

(2) preset a two dimension fuzzy checking list in system, Δ ω and as two input variables of two dimension fuzzy checking list, according to the output variable N of two dimension fuzzy checking list to K _a, K _band K _ccarry out dynamic conditioning, finally make the stabilizer multiplication factor fiducial value acted in single cycle of oscillation be dynamic change, to adapt in single cycle of oscillation, the real-time demand for control of generator;

To K _a, K _band K _cthe method of carrying out dynamic conditioning is undertaken by following formula:

K _A(N)＝K _A·N

K _B(N)＝K _B·N

K _C(N)＝K _C·N

Wherein, N is the output variable of two dimension fuzzy checking list; K _a(N) be K _acorresponding numerical value after the adjustment of two dimension fuzzy checking list; K _b(N) be K _bcorresponding numerical value after the adjustment of two dimension fuzzy checking list; K _c(N) be K _ccorresponding numerical value after the adjustment of two dimension fuzzy checking list;

After completing the operation of step (1) and (2), from K _a(N), K _band K (N) _c(N) choose corresponding parameter in as stabilizer multiplication factor, the running status of generator is controlled.

3. magnification factor correction method for power stabilizer according to claim 2, is characterized in that: described two dimension fuzzy checking list is as shown in the table:

Wherein, NB, NS, ZE, PS, PB are 5 fuzzy subsets of input variable; CH, CL, OK, AL, AH are 5 fuzzy subsets of output variable N; NB, NS, ZE, PS, PB represent respectively: negative large, negative little, zero, just little, honest; CH, CL, OK, AL, AH represent respectively: high-shrinkage, low-shrinkage, constant, lowly to put, Gao Fang.

4. magnification factor correction method for power stabilizer according to claim 1 and 2, is characterized in that: step 6) in, choose corresponding parameter as follows:

[1] means by experiment, the different running statuses of simulation generator, are obtained by debugging and correspond respectively to the stabilizer multiplication factor fiducial value of generator operation under normal operating conditions, minor interference state and large disturbance state with

[2] in experimentation, gather respectively to the data of generator under three kinds of running statuses, the data collected are the output signal u of stabilizer and the tachometer value ω of generator; For often kind of running status, all obtain one group of burst and one group of rotating speed sequence, burst is the ordered series of numbers u comprising multiple u ₁, u ₂u _i, each u arranges chronologically; Rotating speed sequence is the ordered series of numbers ω comprising multiple ω ₁, ω ₂ω _i, and ω and u is according to sequential one_to_one corresponding;

[3] according to the burst under three kinds of running statuses and rotating speed sequence, set up the neural network prediction model under the three kinds of running statuses corresponding to generator respectively, and set up forecasting model database according to three neural network prediction models;

[4] forecasting model database is put into operation online, real output signal u (k) of Real-time Collection stabilizer and actual speed value ω (k) of generator, k is the sequence number of sampling number; By the I and II time delay value input prediction model library of u (k) and ω (k), the rotor speed forecast value under the three kinds of running statuses corresponding respectively to generator can be obtained with according to following formula, calculate respectively ω (k) with with between error e _a(k), e _b(k) and e _c(k):

$e_A\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_A\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

$e_B\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_B\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

$e_C\left(k\right)=\left|\frac{\mathrm{\&omega;}\left(k\right)-{\widehat{\mathrm{\&omega;}}}_C\left(k\right)}{\mathrm{\&omega;}\left(k\right)}\right|$

Wherein, ω (k) is the actual speed value of the generator of acquisition when kth time is sampled; with what be respectively that forecasting model database exports corresponds under three kinds of running statuses, possesses the rotor speed forecast value of corresponding relation with ω (k) in sequential; the normal operating conditions of corresponding generator, the minor interference state of corresponding generator, the large disturbance state of corresponding generator;

[5] e is compared _a(k), e _b(k) and e _ck the size of (), the running status that wherein numerical value reckling is corresponding is the current operating conditions of generator, using the parameter of correspondence as stabilizer multiplication factor, controls the running status of generator.

5. magnification factor correction method for power stabilizer according to claim 4, is characterized in that: step compares e in [5] _a(k), e _b(k) and e _cbefore the size of (k), according to following formula to e _a(k), e _b(k) and e _ck () is for further processing:

$e_A^{\&prime;}\left(k\right)=e_A\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_A\left(j\right)$

$e_B^{\&prime;}\left(k\right)=e_B\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_B\left(j\right)$

$e_C^{\&prime;}\left(k\right)=e_C\left(k\right)+\underset{k-q}{\overset{k-1}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}^{k-j}{e}_C\left(j\right)$

Wherein, e' _ak () is for corresponding to e _athe overall error of (k); E' _bk () is for corresponding to e _bthe overall error of (k); E' _ck () is for corresponding to e _cthe overall error of (k); e _aj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; e _bj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; e _cj () is for the ω (j) obtained when jth time is sampled is with corresponding between error; Q is matching length, and q is positive integer, q>=2; K, j are positive integer, and k-q≤j≤k-1; β is matching error forgetting factor, 0 < β < 1;

Then, e is compared _a(k), e _b(k) and e _cduring the size of (k), by e _a(k), e _b(k) and e _ck () replaces with e' respectively _a(k), e' _b(k) and e' _c(k), then compare e' _a(k), e' _b(k) and e' _ck the size of () also continues subsequent treatment.