Summary of the invention
The object of the invention is the not good problem of Control platform caused to solve the reasons such as non-linear, uncertain in existing gas turbine temperature controlled processes, disturbing factor, use fuzzy immunization proportional plus integral control, comprise fuzzy control on-line tuning integral action coefficient, also comprise immune fuzzy and control on-line tuning proportional action coefficient, make controller parameter automatically adapt to complexity, time become, nonlinear controlled device, improve control accuracy and dynamic property.
The present invention compared with prior art, have the following advantages and breakthrough technique effect: 1. adopt immune fuzzy to control, according to the feedback regulation principle of Immune System, use fuzzy controller Nonlinear Function Approximation, according to the Drazin inverse proportional action coefficient of controller, the adaptive ability of system is strengthened greatly, ensure that the stability of system, in Dynamic Regulating Process, eliminate the deviation of target with the fastest speed.2. in order to error that compensate for disturbances causes, fuzzy control is adopted according to the deviation of outlet air temperature set value and measured value and deviation variation rate, automatically the optimum apjustment of integral action coefficient is realized, can become during self-adaptation, the gas turbine complication system that non-linear, disturbing factor is many, there is higher control accuracy, less overshoot, stronger antijamming capability, stronger robustness.
In order to realize above-mentioned object, technical scheme of the present invention is as follows:
1) install N number of exhaust gas temperature sensor at equal intervals at gas turbine turbine exhaust mouth, wherein, the kth moment delivery temperature of the 1st ~ N number of exhaust gas temperature sensor measurement is designated as T respectively
survey 1(k), T
survey 2(k) ... T
survey Nk (), N gets the positive integer being more than or equal to 16, and the delivery temperature mean value in kth moment is designated as T
on averagek (), the outlet air temperature set value in kth moment is designated as T
if(k), the outlet air temperature set value T in kth moment
if(k) and delivery temperature mean value T
on averagek the deviation of () is designated as e (k), as the input value of fuzzy immunization pi controller;
2) at the integral action coefficient of line computation fuzzy immunization pi controller: design fuzzy controller, the outlet air temperature set value T in kth moment
if(k) and delivery temperature mean value T
on averagethe deviation e (k) of (k) as the input value of fuzzy controller, the outlet air temperature set value T in kth moment
if(k) and delivery temperature mean value T
on averagek the deviation of () obtains deviation variation rate through differential and is designated as ec (k), also as the input value of fuzzy controller;
Determine kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagethe fuzzy domain of the deviation of (k)-3 ,-2 ,-1,0,1,2,3}, fuzzy language value { NB, NM, NS, Z, PS, PM, PB}, kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagethe fuzzy domain {-3 ,-2 ,-1,0,1 of the deviation variation rate of (k), 2,3}, fuzzy language value { NB, NM, NS, Z, PS, PM, PB}, the membership function of deviation and deviation variation rate is triangle, determine the fuzzy domain {-0.06 ,-0.04 ,-0.02,0,0.02 of integral action coefficient increment, 0.04,0.06}, fuzzy language value { NB, NM, NS, Z, PS, PM, PB}, membership function adopts triangle;
According to engineering experience and gas turbine temperature control requirement, design fuzzy control integral action coefficient increment:
If e (k) is NB, and ec (k) is NB, then integral action coefficient increment △ ki (k) in kth moment is NB,
If e (k) is NB, and ec (k) is NM, then integral action coefficient increment △ ki (k) in kth moment is NB,
If e (k) is NB, and ec (k) is NS, then integral action coefficient increment △ ki (k) in kth moment is NM,
If e (k) is NB, and ec (k) is Z, then integral action coefficient increment △ ki (k) in kth moment is NM,
If e (k) is NB, and ec (k) is PS, then integral action coefficient increment △ ki (k) in kth moment is NS,
If e (k) is NB, and ec (k) is PM, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is NB, and ec (k) is PB, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is NM, and ec (k) is NB, then integral action coefficient increment △ ki (k) in kth moment is NB,
If e (k) is NM, and ec (k) is NM, then integral action coefficient increment △ ki (k) in kth moment is NB,
If e (k) is NM, and ec (k) is NS, then integral action coefficient increment △ ki (k) in kth moment is NM,
If e (k) is NM, and ec (k) is Z, then integral action coefficient increment △ ki (k) in kth moment is NS,
If e (k) is NM, and ec (k) is PS, then integral action coefficient increment △ ki (k) in kth moment is NS,
If e (k) is NM, and ec (k) is PM, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is NM, and ec (k) is PB, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is NS, and ec (k) is NB, then integral action coefficient increment △ ki (k) in kth moment is NB,
If e (k) is NS, and ec (k) is NM, then integral action coefficient increment △ ki (k) in kth moment is NM,
If e (k) is NS, and ec (k) is NS, then integral action coefficient increment △ ki (k) in kth moment is NS,
If e (k) is NS, and ec (k) is Z, then integral action coefficient increment △ ki (k) in kth moment is NS,
If e (k) is NS, and ec (k) is PS, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is NS, and ec (k) is PM, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is NS, and ec (k) is PB, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is Z, and ec (k) is NB, then integral action coefficient increment △ ki (k) in kth moment is NM,
If e (k) is Z, and ec (k) is NM, then integral action coefficient increment △ ki (k) in kth moment is NM,
If e (k) is Z, and ec (k) is NS, then integral action coefficient increment △ ki (k) in kth moment is NS,
If e (k) is Z, and ec (k) is Z, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is Z, and ec (k) is PS, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is Z, and ec (k) is PM, then integral action coefficient increment △ ki (k) in kth moment is PM,
If e (k) is Z, and ec (k) is PB, then integral action coefficient increment △ ki (k) in kth moment is PM,
If e (k) is PS, and ec (k) is NB, then integral action coefficient increment △ ki (k) in kth moment is NM,
If e (k) is PS, and ec (k) is NM, then integral action coefficient increment △ ki (k) in kth moment is NS,
If e (k) is PS, and ec (k) is NS, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is PS, and ec (k) is Z, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is PS, and ec (k) is PS, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is PS, and ec (k) is PM, then integral action coefficient increment △ ki (k) in kth moment is PM,
If e (k) is PS, and ec (k) is PB, then integral action coefficient increment △ ki (k) in kth moment is PB,
If e (k) is PM, and ec (k) is NB, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is PM, and ec (k) is NM, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is PM, and ec (k) is NS, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is PM, and ec (k) is Z, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is PM, and ec (k) is PS, then integral action coefficient increment △ ki (k) in kth moment is PM,
If e (k) is PM, and ec (k) is PM, then integral action coefficient increment △ ki (k) in kth moment is PB,
If e (k) is PM, and ec (k) is PB, then integral action coefficient increment △ ki (k) in kth moment is PB,
If e (k) is PB, and ec (k) is NB, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is PB, and ec (k) is NM, then integral action coefficient increment △ ki (k) in kth moment is Z,
If e (k) is PB, and ec (k) is NS, then integral action coefficient increment △ ki (k) in kth moment is PS,
If e (k) is PB, and ec (k) is Z, then integral action coefficient increment △ ki (k) in kth moment is PM,
If e (k) is PB, and ec (k) is PS, then integral action coefficient increment △ ki (k) in kth moment is PM,
If e (k) is PB, and ec (k) is PM, then integral action coefficient increment △ ki (k) in kth moment is PB,
If e (k) is PB, and ec (k) is PB, then integral action coefficient increment △ ki (k) in kth moment is PB;
Kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagek the deviation e (k) of (), deviation variation rate ec (k) are multiplied by quantizing factor respectively and are quantified as fuzzy quantity on fuzzy domain, according to fuzzy control rule, obtain the fuzzy control quantity of integral action coefficient increment △ ki (k) in kth moment, be multiplied by quantizing factor by its precision, add the conventional proportional integral controller integral action coefficient initial value that conventional proportional integral attitude conirol method is adjusted out, obtain the integral action coefficient of kth moment fuzzy immunization pi controller
ki(k)=ki0+△ki(k)×α
Wherein, the integral action coefficient of the fuzzy immunization pi controller that ki (k) is the kth moment, ki0 is the conventional proportional integral controller integral action coefficient initial value that conventional proportional integral attitude conirol method is adjusted out, and α is the quantizing factor of integral action coefficient increment;
3) at the proportional action coefficient of line computation fuzzy immunization pi controller: design immune fuzzy controller, if the antigen levels in the i-th generation is ε (i), the output of antigen stimulated cells is T
h(i), T
h(i)=k
1ε (i), k
1for exitation factor, the output of antigen T suppression cell is T
s(i), T
s(i)=k
2f () ε (i), k
2for inhibiting factor, f () is nonlinear function, and total stimulation of cell antigen is S (i)=T
h(i)-T
s(i)=(k
1-k
2f ()) ε (i)=K (1-η f ()) ε (i), with kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagek the deviation e (k) of () is as antigen levels ε (i), then immune fuzzy controller exports as K (1-η f ()) e (k), the proportional action coefficient of fuzzy immunization pi controller is K (1-η f ()), wherein, K=k
1control reaction velocity,
control stablizing effect, f () is selected nonlinear function;
Adopt method Nonlinear Function Approximation f () of fuzzy reasoning, the fuzzy immunization pi controller in kth moment exports and is designated as G (k), fuzzy immunization pi controller exports and obtains exporting change rate through differential, the value in kth moment is designated as △ G (k), G (k), △ G (k) is all as the input value of immune fuzzy controller, determine the fuzzy domain {-1 that fuzzy immunization pi controller exports, 1}, fuzzy language value { N, P}, the fuzzy domain {-1 of fuzzy immunization pi controller exporting change rate, 1}, fuzzy language value is { N, P}, the membership function of output and exporting change rate all adopts triangle, determine the fuzzy domain {-1 of nonlinear function f () value, 1}, fuzzy language value { N, Z, P}, membership function adopts triangle, design nonlinear function f ():
If G (k) is N, and △ G (k) is N, then the kth moment nonlinear function f (, k) be P;
If G (k) is N, and △ G (k) is P, then the kth moment nonlinear function f (, k) be Z;
If G (k) is P, and △ G (k) is N, then the kth moment nonlinear function f (, k) be Z;
If G (k) is P, and △ G (k) is P, then the kth moment nonlinear function f (, k) be N;
The fuzzy immunization pi controller in kth moment exports G (k), exporting change rate △ G (k) is multiplied by quantizing factor obfuscation respectively, according to above-mentioned fuzzy rule, obtain kth moment nonlinear function f (, k) fuzzy control quantity, is multiplied by quantizing factor by its precision, obtain kth moment nonlinear function f (, k) exact value, thus the proportional action coefficient obtaining kth moment fuzzy immunization pi controller
kp(k)=K(1-ηf(·,k))
Wherein, the proportional action coefficient of the fuzzy immunization pi controller that kp (k) is the kth moment, K, η are parameter;
4) the fuzzy immunization pi controller calculating the kth moment exports,
Wherein, T represents the sampling time;
5) output of the fuzzy immunization pi controller in kth moment is temperature control fuel stroke benchmark G (k) and sends into fuel stroke selection of reference frame module, participate in the control of fuel flow rate, obtain the fuel quantity w in kth moment
fk (), sends into firing chamber by fuel control valve door.
For step 2), kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagek the quantizing factor of the deviation of () adopts the method for gathering through test to determine, its scope is 0.01 ~ 0.1, kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagek the quantizing factor of the deviation variation rate of () adopts the method for gathering through test to determine, its scope is 0.01 ~ 0.05, the quantizing factor of kth moment fuzzy immunization pi controller integral action coefficient increment adopts the method for gathering through test to obtain, and its scope is 0.01 ~ 0.03;
For step 3), the quantizing factor that kth moment fuzzy immunization pi controller exports adopts the method for gathering through test to obtain, its scope is 0.01 ~ 1, the quantizing factor of kth moment fuzzy immunization pi controller exporting change rate adopts the method for gathering through test to obtain, its scope is 0.01 ~ 1, kth moment nonlinear function f (, k) quantizing factor adopts the method for gathering through test to obtain, its scope is 0.01 ~ 1, parameter K in the proportional action coefficient K (1-η f ()) of fuzzy immunization pi controller, η adopts the method for gathering through test to obtain, the scope of K is 0.0005 ~ 0.0007, the scope of η is 0.1 ~ 0.3,
For step 4), sampling time T value is 0.001 second.
Embodiment
Below in conjunction with the drawings and specific embodiments, the present invention is described further
Fig. 1 is heavy duty gas turbine temperature controlled fuzzy immunization proportional plus integral control process flow diagram, and its specific implementation method comprises the following steps:
1) install N number of exhaust gas temperature sensor at equal intervals at gas turbine turbine exhaust mouth, wherein, the delivery temperature of the 1st ~ N number of exhaust gas temperature sensor measurement is designated as T respectively
survey 1, T
survey 2t
survey N, N gets the positive integer being more than or equal to 16, and delivery temperature mean value is designated as T
on average, the delivery temperature mean value T in kth moment
on averagek () value is N number of measurement value sensor T in kth moment
survey 1(k), T
survey 2(k) ... T
survey Nk () removes maximal value and the later mean value of minimum value, outlet air temperature set value is designated as T
if, the outlet air temperature set value T in kth moment
if(k) and delivery temperature mean value T
on averagek the deviation of () is designated as e (k), as the input value of fuzzy immunization pi controller;
2) outlet air temperature set value is designated as T
if, the outlet air temperature set value T in kth moment
if(k) and delivery temperature mean value T
on averagek the deviation of () is designated as e (k), as the input value of fuzzy immunization pi controller, the k moment is the timer time that gas turbine brings into operation, k=0,1,2,, suppose that the sampling time is 0.01 second, then the 1st moment was the 0.01st second, 2nd moment was the 0.02nd second, by that analogy;
3) at the integral action coefficient of line computation fuzzy immunization pi controller: design fuzzy controller, the outlet air temperature set value T in kth moment
if(k) and delivery temperature mean value T
on averagethe deviation e (k) of (k) as the input value of fuzzy controller, the outlet air temperature set value T in kth moment
ifwith delivery temperature mean value T
on averagedeviation obtain deviation variation rate through differential and be designated as ec (k), also as the input value of fuzzy controller;
Determine outlet air temperature set value T
ifwith delivery temperature mean value T
on averagethe fuzzy domain of deviation {-3 ,-2 ,-1,0,1,2,3}, { NB, NM, NS, Z, PS, PM, PB} also can be expressed as { negative large, in negative, negative little, zero, just little, center, honest }, outlet air temperature set value T fuzzy language value
ifwith delivery temperature mean value T
on averagethe fuzzy domain {-3 of deviation variation rate,-2,-1, 0, 1, 2, 3}, fuzzy language value { NB, NM, NS, Z, PS, PM, PB}, also can be expressed as { negative large, in negative, negative little, zero, just little, center, honest }, the membership function of deviation and deviation variation rate is triangle, respectively as shown in Figures 2 and 3, determine the fuzzy domain {-0.06 of integral action coefficient increment,-0.04,-0.02, 0, 0.02, 0.04, 0.06}, fuzzy language value { NB, NM, NS, Z, PS, PM, PB}, also can be expressed as { negative large, in negative, negative little, zero, just little, center, honest }, membership function adopts triangle, as shown in Figure 4,
According to engineering experience and gas turbine temperature control requirement, the rule of design fuzzy control integral action coefficient increment, principle of design is as follows: when 1. e (k) is larger, integral action can make too by force system overshoot strengthen, to be limited integral action, be got less △ ki (k); 2., during e (k) median size, △ ki (k) value is moderate; 3., when e (k) is less, for reducing steady-state error, △ ki (k) value should be more greatly.Design fuzzy control integral action coefficient increment:
Kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagek the deviation e (k) of (), deviation variation rate ec (k) are multiplied by quantizing factor respectively and are quantified as fuzzy quantity on fuzzy domain, quantizing factor computing method: the 1. quantizing factor Ke=i/e (max) of deviation, wherein i is the quantification gradation of deviation, the domain that e (max) is deviation.2. the quantizing factor Kec=m/ec (max) of deviation variation rate, wherein m is the quantification gradation of deviation variation rate, the domain that ec (max) is deviation variation rate, in implementation process, with reference to above-mentioned computing method, adopt the method determination outlet air temperature set value T gathered through test according to engineering experience
ifwith delivery temperature mean value T
on averagek the quantizing factor of the deviation of (), its scope is 0.01 ~ 0.1, outlet air temperature set value T
ifwith delivery temperature mean value T
on averagethe quantizing factor of the deviation variation rate of (k), its scope is 0.01 ~ 0.05, e (k), after ec (k) obfuscation, search e (k), the position of ec (k) in fuzzy control rule table, obtain the fuzzy control quantity of integral action coefficient increment △ ki (k) in kth moment, be multiplied by quantizing factor, wherein the quantizing factor of integral action coefficient increment adopts the method for gathering through test to obtain, its scope is 0.01 ~ 0.03, the conventional proportional integral controller integral action coefficient initial value that conventional proportional integral attitude conirol method is adjusted out is added after precision, obtain the integral action coefficient of kth moment fuzzy immunization pi controller
ki(k)=ki0+△ki(k)×α
Wherein, the integral action coefficient of the fuzzy immunization pi controller that ki (k) is the kth moment, ki0 is the conventional proportional integral controller integral action coefficient initial value that conventional proportional integral attitude conirol method is adjusted out, and α is the quantizing factor of integral action coefficient increment;
4) at the proportional action coefficient of line computation fuzzy immunization pi controller: design immune fuzzy controller, if the antigen levels in the i-th generation is ε (i), the output of antigen stimulated cells is T
h(i), T
h(i)=k
1ε (i), k
1for exitation factor, the output of antigen T suppression cell is T
s(i), T
s(i)=k
2f () ε (i), k
2for inhibiting factor, f () is nonlinear function, and total stimulation of cell antigen is S (i)=T
h(i)-T
s(i)=(k
1-k
2f ()) ε (i)=K (1-η f ()) ε (i), with kth moment outlet air temperature set value T
if(k) and delivery temperature mean value T
on averagek the deviation e (k) of () is as antigen levels ε (i), then immune fuzzy controller exports as K (1-η f ()) e (k), the proportional action coefficient of fuzzy immunization pi controller is K (1-η f ()), wherein, K=k
1control reaction velocity,
control stablizing effect, f () is selected nonlinear function;
Adopt method Nonlinear Function Approximation f () of fuzzy reasoning, the fuzzy immunization pi controller in kth moment exports and is designated as G (k), fuzzy immunization pi controller exports and obtains exporting change rate through differential, the value in kth moment is designated as △ G (k), G (k), △ G (k) is all as the input value of immune fuzzy controller, determine the fuzzy domain {-1 that fuzzy immunization pi controller exports, 1}, fuzzy language value { N, P}, also can be expressed as { negative, just }, the fuzzy domain {-1 of fuzzy immunization pi controller exporting change rate, 1}, fuzzy language value is { N, P}, also can be expressed as { negative, just }, the membership function of output and exporting change rate all adopts triangle, respectively as shown in Figure 5 and Figure 6, determine the fuzzy domain {-1 of nonlinear function f () value, 1}, fuzzy language value { N, Z, P}, also can be expressed as { negative, zero, just }, membership function adopts triangle, as shown in Figure 7, design nonlinear function f ():
The fuzzy immunization pi controller in kth moment exports G (k), exporting change rate △ G (k) is multiplied by quantizing factor obfuscation respectively, the quantizing factor that fuzzy immunization pi controller exports adopts the method for gathering through test to obtain, its scope is 0.01 ~ 1, the quantizing factor of fuzzy immunization pi controller exporting change rate adopts the method for gathering through test to obtain, its scope is 0.01 ~ 1, rule list is looked into after obfuscation, obtain kth moment nonlinear function f (, k) fuzzy control quantity, be multiplied by the quantizing factor of f (), the quantizing factor of nonlinear function f () adopts the method for gathering through test to obtain, its scope is 0.01 ~ 1, obtain after precision kth moment nonlinear function f (, k) exact value, thus obtain the proportional action coefficient of the fuzzy immunization pi controller in kth moment
kp(k)=K(1-ηf(·,k))
Wherein, the proportional action coefficient of the fuzzy immunization pi controller that kp (k) is the kth moment, K, η adopt the method for gathering through test to obtain, and the scope of K is the scope of 0.0005 ~ 0.0007, η is 0.1 ~ 0.3;
5) the fuzzy immunization pi controller calculating the kth moment exports G (k)
Wherein, T represents the sampling time, and value is 0.001 second.
6) the fuzzy immunization pi controller in kth moment is exported G (k) and send into fuel stroke selection of reference frame module, participate in the control of fuel flow rate, obtain the fuel quantity w in kth moment
fk (), makes valve opening reach corresponding fuel quantity by fuel valve control module, fuel is sent into firing chamber.
Embodiment:
With reference to certain single-rotor gas turbine model, actual exhaust air temperature is
Wherein, T
realfor delivery temperature actual value, unit DEG C, T
volumefor the specified setting value of delivery temperature, unit DEG C, w
ft fuel quantity that () is t, ε
tdfor turbine and exhaust system propagation delay time constant, n is rotating speed, L
iGVfor the stroke of compressor inlet adjustable guide vane topworks, T
afor environment temperature, unit DEG C.
Suppose that basic load stage gas turbine temperature controls to come into operation, the specified setting value T of delivery temperature
volumevalue 538.9 DEG C, at full speed, rotating speed n=1, compressor inlet adjustable guide vane reaches range L to unit
iGV=1, time delay ε
td=0.04 second, T
abe set to 15 DEG C, the fuel quantity w of t
f(t) be t fuel stroke benchmark G (t) after 0 ~ 100 amplitude limit, be multiplied by gain 0.77, add the value after compensation 0.23, outlet air temperature set value T
ifprovide step signal 548.9 DEG C to 538.9 DEG C at the initial time of emulation, add undesired signal at the intermediate time of emulation.
Use the parameter tuning function of MATLAB/SIMULINK proportional plus integral plus derivative controller, the initial parameter values kp0=0.000387 of fuzzy self-adaption pi controller of adjusting out, ki0=0.01749, arrange 0.001 second sampling time.
In design immune fuzzy pi controller process correlation parameter choose as follows, outlet air temperature set value T
ifwith delivery temperature actual value T
realthe quantizing factor value 0.01 of deviation, outlet air temperature set value T
ifwith delivery temperature actual value T
realthe quantizing factor value 0.01 of deviation variation rate, the quantizing factor value 0.03 of integral action coefficient increment, the quantizing factor value 1 that fuzzy immunization pi controller exports, the quantizing factor value 1 of fuzzy immunization pi controller exporting change rate, the quantizing factor value 1 of nonlinear function f (), K value 0.0006, η value 0.2 in proportional action coefficient formula.
Design fuzzy immunization pi controller, comprise design immune fuzzy to control, obtain line computation proportional action coefficient as shown in Figure 8, also comprise design fuzzy control, line computation integral action coefficient as shown in Figure 9, the output response curve obtained is as shown in Figure 10, fuzzy immunization proportional plus integral control has less overshoot compared with controlling with conventional proportional integral, higher control accuracy, stronger antijamming capability, better robustness.