FR2475313A1 - Automatic regulation of turbo-alternator sets - by sensing and processing alternator parameters, using reference circuit prior to supporting adjustable regulator - Google Patents

Abstract

Automatic control of a turbo-alternator set connected into a switching network. The set includes an independent speed regulation system operating on the input to the turbine via control of its alternator exciting current. The alternator output and input levels are represented by signal levels which are fed into a reference model. Signals taken from a tachometer (22), Watt metre (25) and Voltmeter (26) connected to the alternator output are separately processed (30,31,32) and summed from which an instantaneous mean value is obtained. Amplitude filters (50-54) supply an adjustable regulator (5) whose output is a function of a linear combination of the motor outputs modified by coefficients set up in the summing circuit. A signal image of the mechanical power (13) is processed (34) and also fed to the regulator.

Description

 The invention relates to a self-adaptative control method of the operation of a turbo-alternator group coupled to an interconnection network and equipped with an independent speed regulator acting on the intake to the turbine. The invention also relates to a regulator implementing this method.

 In stable steady state, the alternator rotates in synchronism with the network, receives from the turbine a mechanical power according to the flows and enthalpies of steam admitted to the different stages of the turbine, flows determined by the positions of the valves of the turbine. admission, delivers to the network an active electrical power equal to the losses close to the mechanical power received, and exchange with the network a reactive power equal to the product of the active power by the tangent of the phase angle between current and voltage, function of the phase between the angular position of the rotor with respect to the pulsation of the network.

 The contribution of the group does not significantly change the state of the network, given the number of interconnected groups. Seen from the coupling point, the network can be defined as a reversible generator having an electromotive force, a substantially reactive internal impedance (practically the impedance of the connection lines), and a frequency, these quantities being independent and immaterial and therefore inaccessible. It should be noted that the frequency also defines an origin of times (to a whole number of periods) which will be decisive for the exchange of energy with the groups, but is virtual and is not linked to an absolute time scale. .

For its part, the alternator is defined as a system with four input magnitudes, engine torque, rotational speed, excitation and origin of rotation of the rotating magnetic field vector created by the cited rotor. The input quantities are here set somewhat arbitrarily, and may be replaced by other quantities resulting from combinations of the input quantities stated above. In particular the mechanical power is the product of the engine torque by the speed of rotation0
However, in the steady state, there are four independent quantities of input, correlated with the four quantities which define the network. The coupling of the group to the network implies that there exists at the point of coupling four output quantities, independent of each other in steady state. but of course correlated with the input quantities on the one hand, and with the quantities which define the network on the other hand.

 State vector means a vector represented by the group of four quantities relating to the same point of the coupled system. Thus, there will be a group input state vector, a network state vector, and an output state vector of the group.

From a general point of view, the destination of the turbo-generator group is in the first place to deliver a determined active power to the network, and secondly under economic conditions which imply a determined reactive power exchange. The regulation will consist in its general definition of adjusting the input vector of the group so that the output state vector coincides with a determined target state vector so that, taking into account the vector of state of the network, the active and reactive power exchanges correspond to what is required. Or this regulation is strictly possible only for permanent states, since the state vector of the network is likely to vary, which implies - do variations of the reference vector, the components of the input state vector are independent only in a steady state, and finally the output state vector is linked to the input state vector by a non-linear matrix transfer function0
Furthermore, in practice, the input state vector can not be adjusted on all its correlating components. Due to the mechanical inertia of the group and the response time constants of the associated steam generators, the engine torque is adjusted in rotational speed function by an independent regulator acting on the positions of the steam inlet valves at the various stages of the turbine. On the other hand, the origin of rotation of the magnetic field vector created by the rotor with respect to the stator is not adjustable in itself, and can only be displaced with respect to the origin of the times defined by a reference frequency and by temporarily sliding the speed of rotation of the group relative to the reference frequency. Also the regulation, as envisaged in the context of the present invention, consists in the control of the excitation (magnitude of the inductive magnetic field) from the target state vector. Since only one input quantity is controlled by the control, the control will be carried out from the single-parameter comparison of the output and target status vectors. Due to the interdependence of the variations of the components of the rector d the input state, and the non-linearity of the group transfer function, the choice of the control parameter, and the transfer function of the control will determine the variations in the time of the vector of state of aortite, that is to say on the convergence in time of this output state vector to the target vector, which corresponds to a steady state.

 According to the oldest state of the art, the output state vector parameter chosen was the stator voltage (rms voltage at the coupling point) and a target voltage was set, based on the needs of the network. power, which corresponded to a desired perilous regime. The convergence of the output state vector to the target state vector was to be ensured by the interdependence of the components of the input state vector; the internal parameters of the turbo-alternator group and the coupling parameters of the network had to be adapted, so that this convergence is effective (natural stability). The parameters, such as inertial of rotating masses, armature reactance, internal damping, coupling impedances, were deduced from the coupled circuit stability equations. However, the adaptation of these parameters was insufficient during major disturbances of regime to avoid instabilities (pumping in particular) and possibly the stall of the group. Moreover, the increase in individual power of the groups and, in parallel, the increase in the length of the coupling lines to the network make it more difficult to adjust the natural stability of the groups.

 The instabilities have been palliative by correcting the difference between the stator voltage and the target voltage by introducing a parameter derived from the active power transiting at the coupling point, which amounts to correcting the reference state vector. This has resulted in practically acceptable stabilities for the powers of the turbo-alternator units currently in service.

But it appeared that with more powerful groups that must equip the proposed plants, the stability provided by these regulations would be insufficient. Moreover improvements of these regulations did not appear likely to provide the desired stability
For the regulation of systems with non-linear transfer functions, so-called self-adaptive regulation devices have been proposed. Since the stable control assumes that the transfer function of the regulator disposed between the output and the input of the system to be regulated is the inverse of the transfer function of this system, the transfer function of the regulator must also be non-linear, and adapt according to the values of the parameters defining the state vectors of the system. To do this, a reference model is established performing a crab transfer function of the signals representative of the input and output quantities of the system to be adjusted corresponding to the transfer function. of this system in a chosen behavior. The output signals of the reference model are compared to corresponding signals of the system to be adjusted to develop a behavior state vector, and this behavior deviation vector is applied to a controller which receives the output quantities of the system to be set to develop a control vector applied to the input of the system to be adjusted, the transfer parameters of the regulator being adjusted so that the deviation vector is asymptotically canceled in response to the regulation. Theoretical studies establish that the asymptotic cancellation, and the stability of the looped system, are ensured when the adjustment of the transfer parameters of the regulator is carried out according to a proportional and integral law of the behavior state vector. It is clear, moreover, that the cancellation of the behavioral difference vector corresponds to the coincidence of the actual behavior of the system and of the behavior chosen. It will also be noted that it is irrelevant in the abstract that the development of the vector of Behavior difference is made from the output state or the input state of the system to be set, ie the model is arranged in parallel with the system to be set with a direct transfer function, or series with the system with a reverse transfer function. Consequently, the choice of a parallel or series reference model will be guided by practical considerations, such as, in particular, the minimization of the number of components of the deviation vector. In the case of a turbo-alternator group, where the regulation acts directly on a single control variable, the excitation, the behavior difference vector will be reduced to a difference coefficient, the comparison being effected between signals component counterparts.

 The behavior of the system to be regulated is defined as the evolution of state of the system between two stable states. Because of the interactions of the different state quantities on the transfer function of the system to be regulated, the behavior defined by the reference model will correspond to a possible behavior of the system to be regulated only if the stable states between which the system evolves. state of the system to be adjusted are both close to the state for which the behavior is defined by the reference model0 In particular in the case of a turbo-alternator group which is expected to provide steady state powers variable in the same ratio at least, it is conceivable that the choice of a suitable behavior in all the extent of these variations, and the design of the structure of a reference model defining this suitable behavior, encountering virtually insurmountable difficulties.

 The subject of the invention is a method of self-adaptive regulation of the operation of a turbo-alternator group which ensures stability in practically the entire possible operating range.

 The invention also relates to a self-adaptive control method using a model consisting of imples operators, conventional layout.

 To these effects the invention proposes a method of self-adaptive regulation of the operating state of a turbo-alternator group coupled to an interconnection network and equipped with an independent speed regulator acting on the admission to the turbine, by action on the excitation of the alternator, the method according to which signals representative of an input quantity and output quantities of the alternator, the stator voltage, of which these representative signals are applied are taken. to a reference model defining a behavior chosen to develop a coefficient of behavior deviation, the representative signals are applied to a self-adaptive regulator developing a control signal function of the signals representative of output quantities according to a proportional action law and integral of the coefficient of behavior deviation so that this coefficient vanishes asymptotically in response to the regulation, and one compares the control signal to a reference signal for developing an excitation control signal, characterized in that, since the model and the regulator deliver output signals sum of the contributions of each of the representative signals, those ei are constituted, for each magnitude of output, by the difference between an image signal of the instantaneous value of this magnitude and the current average of this image signal established with a time constant T of the order of magnitude of the group's own time constant, and for the magnitude the difference between a nominal stator voltage and the current average stator voltage image signal, the excitation control signal being the sum of the adjustment signal and the integral thereof with respect to time, with the same time constant T.

 When the instantaneous operating state of the group deviates little from a steady state, the deviation of each of the quantities which define the instantaneous operating state to the corresponding magnitude in steady state is, as a first approximation a sum of contributions is independent of the differences of other quantities. It is well known that, for a function of multiple variables es by a set of relations, apart from singular points where indeterminacy intervenes in relations (zeros or poles), the first-order deviations of these variables are related between they are functions of the point of origin of these deviations. Here the vector of origin state chosen is the vector defined by the current averages of the quantities defining the successive instantaneous states, and the state deviation vector is defined by the deviations between each current instantaneous value of a quantity and the current average of that magnitude. By taking the current averages established with a time constant close to the group's own time constant. , it is ensured that the original state vector chosen corresponds to a steady state state around which the instantaneous state is moving, and which corresponds to the steady state state. It is closest to the current instantaneous being, so that the insan-determined deviations so determined are small and define a state difference within a domain where the approximation of the independence of the gap contributions is valid.

Moreover, the reference quantity which determined the stable state of state chosen as the origin (current state vector) is the imposed nominal stator voltage, while the reference quantity which determined the current state is represented in FIG. better by the current average of the stator voltage taken with the group's own time constant. Since, as previously stated, only one control variable (excitation) accessible to the control system exists in the system under consideration, the control vector will be a single component, or control signal, formed by a sum of contributions independent of the different size differences. For the same reason, the enslavement of the behavior ns can play only on one component, so that the model will define a coefficient of behavior deviation, composed of a sum of contributions independent of magnitude differences, each contribution being the product of the corresponding deviation signal by a factor determined for a selected standard behavior. On the other hand the adjustment signal obtained at the output of the self-adaptive regulator obtained by combining the deviations between magnitudes defining the state of stable regime and ocelles which define the current state reflects the difference between the current and current average magnitudes of control signal to be applied, so that the restitution of this control signal implies the inverse transformation of that which is applied to the quantities at the input of the positive control device, that is to say a signal composed of the sum of the adjustment signal and its integral with respect to time with the same time constant T.

 The representative signals are preferably filtered through first-order filters with a time constant of the order of 0.04 T to apply them to the model.

 A first-order filter delivers, in response to an input signal that is a function of time, an output signal which is the average over time with the filter's time constant of the input signal. The variations of the slow input signal with respect to the time constant are reflected in the output signal, while the rapid variations with respect to this time constant are planed. The choice of the time constant of the first-order filters at 0.04 of the current averaging time constant, which is of the order of magnitude of the group-specific time constant, has the consequence that that the filtered deviations remain significant from the operating state difference to a sufficient extent while being free from the influence of spurious transient fluctuations.

According to a preferred arrangement of the invention,
Adds to the control signal a signal developed from the speed controller and representative of the mechanical power variations assigned by the controller. When the speed regulator intervenes on the steam inlet to the turbine to adjust the flow rates due to variations in power demand translated by group speed variations, the power responses of the turbine are slow. Also the action of the signal developed from the speed regulator anticipates the response of the turbine and correct the adjustment signal accordingly.

 In another aspect the invention relates to a self-adaptive control device implementing the aforementioned method.

The invention thus proposes an analogue device for self-adaptive regulation of the operating state of a turbo-alternator group coupled to an interconnection network and equipped with an independent speed regulator acting on the intake to the turbine, for the implementation of the aforementioned method, wherein the output of the alternator is equipped with sensors delivering image voltages V, Pe and Q respectively proportional to the stator voltage amplitude, the electrical active power and the differential rotation speed of the group, an adjustable voltage source delivering a stator voltage setpoint Vc, characterized in that it comprises three average operators respectively receiving the image voltages V, Pe and Q and each constituted by an amplifier with a parallel capacitance resistance feedback loop forming said time constant T, three adders associated respectively with the operators of medium and receiving on the one hand the image voltage and on the other hand the average of this voltage so as to develop a deviation signal respectively AV, Spe e and An, a differential gain amplifier unit receiving as input the setpoint voltage
Vc and the average voltage Y and delivering a reference deviation signal AVc a reference model comprising three first-order filters with a time constant Tf of the order of 0.04 T respectively receiving as input the deviation signals AV, Aflet AVc and a summator with input attenuators recov- ering the first-order filter output signals and the deviation signal AV, and outputting a behavior difference voltage, a four-branch regulator combined into output by an adder through attenuators, each branch assigned respectively to said deviation signals and comprising in series a first multiplier receiving the respective deviation signal and the behavior deviation voltage, two parallel paths pa, a proportional and a other integral, these two channels meeting on an adder further receiving a clamping voltage, and a second multiplier receiving on its inputs the respective deviation signal and the signal is su- the branch summator, the output of the attacking regulator an integration means comprising in parallel a proportional channel and a time constant integral channel T, the output signal of the integration means constituting an excitation control signal.

 Those skilled in the art will recognize in the medium operators and the first-order filters analog devices of conventional constitution, amplifier with a feedback loop with a suitable time constant, so that the structure of the regulating device corresponding to the elaboration of the functions stated in the process. It will be noted that the input of the model assigned to the difference of Pg does not include a first-order filter g indeed the gain of the contribution of the difference of electric power to the elaboration of a voltage difference of compound relative to the power is -1, so that the signal eeart of electric power, equal to the source of the variation of deviation and the filtered difference value is equal to the difference of the deviation variation and the contribution of the electrical power deviation signal to the development of the behavior deviation voltage.

Preferably, at the output of the device, there is an excitation limiter consisting of two threshold comparators jointly driven by the signal Y and regulated respectively on a rated maximum and a minimum of the stator voltage, the output voltages of the two comparators being adding to the excitation control signal. As long as the voltage V is situated between the thresholds, the output voltages of the comparators compensate each other and the excitation control voltage is determined by the control device. When crossing a threshold, the corresponding comparator controls the excitation control voltage in order to stabilize the stator voltage independently of the regulating device. This arrangement was designed to avoid incidents during sudden power variations caused, for example, by a tripping. the group of the cost of the turbine, a loss of synchronism, a disjunction of the coupling to the network,.,
Also preferably there are noise reducing filters on the deviation signal inputs of the regulator. Indeed the integration of noise in the controller can cause a drift of the excitation control signal, these sounds appear as small value deviations. The noise-reducing filters are such that they have a zero amplitude response within an interval between two symmetrical thresholds, and proportional beyond this interval. By adjusting the interval to a width equal to four gaps types of noise on the signal considered, it virtually eliminates the drifts due to noise, without tainting a serious error regulation.

Such reducing filters can be constituted by arranging for the voltage values of each polarity a pair of limiting circuits, each controlled by a threshold comparator.
The features and advantages of the invention will become apparent from the following description, by way of example, with reference to the appended drawings in which
FIG. 1 schematically represents a turbo-alternator unit equipped with a regulating device according to the invention;
Figure 2 is a Fresnel diagram explaining the operating state of an alternator;
FIG. 3A is a functional diagram of a mechanical power interaction correction network;
FIG. 3B is an analog layout diagram of the network of FIG.
FIGS. 4A and 4B are respectively a block diagram and an analog layout of a current average and a gap operator 1
FIGS. 5A and 5B are respectively a block diagram and an analog layout of a model for developing a behavioral difference voltage
FIG. 6A is a functional diagram of regulator with adjustable coefficients g
FIG. 6B is an analog arrangement of a regulator branch according to FIG. 6A
Fig. 7A is an excitation voltage limiter diagram x
Fig. 7B is a response diagram of the device of Fig. 7A
Fig. 8A is a noise reducing filter diagram;
Figure 8B is a diagram of the response curve of the filter of Figure 8L;
Figure 9 is an analog layout diagram of an excitation control voltage feedback circuit.

 According to the embodiment chosen and represented in FIG. 1, a turbo-alternator group consists of a turbine 1, coupled to an alternator 2 as a whole. The turbine 1 is supplied with steam, supplied by a generator, not shown, through a power regulating device comprising a high pressure valve 5 and a medium pressure valve 12, both controlled by a non-speed regulator. represent. A computer device 13 sensitive to the position of the valves 11 and 12, delivers an image signal of the mechanical power assigned to the turbine. It is understandable that the solar power is linked to the high and medium pressure steam flows, as well as to the enthalpy losses of the steam as it passes through the turbine.

 The alternator 2 comprises a rotor 20 coupled to the turbine rotator and provided with an inductor winding traversed by an excitation current delivered by an adjustable source 57.

The inductor 20 rotates in the region of a stator 21, with three-phase winding shown here for simplicity in a single-line diagram.

Stalled on the same axis as the rotor, a two-phase tachymetric alternator 22 delivers a signal proportional to the difference between the speed of rotation of the group and the speed of exact synchronism on the network. The stator 21 is coupled to the network by a power transformer 27, and comprises conventional sets of stator voltage measurement transformers 24, and current 23. A power meter 25 is powered by the secondary transformers 23 and 24, and develops a signal proportional to the active power. A voltmetric device 26 connected to the secondary of the transformers 24 produces a signal proportional to the average of the stator voltages of the three phases. It will be noted that the wattmeter 25 and the voltmeter 26 have their own time constants which are of the same order of magnitude as the network period. Indeed an active power or an effective voltage have no meaning: considered on time intervals shorter than the period. Moreover, these quantities must be developed over a very short time interval with respect to the control time constants, in order to be considered as instantaneous values in the regulation process. As will be explained below, the resolution on time of the regulation is of the order of one-tenth of a second while a regime considered over ten seconds will be considered as a steady state. In other words the bandwidth to consider for regulation is between approximately 10 and 0.1 Hz.

 The frequency, power and voltage signals respectively from the sensors, tachometer 22, power meter 25 and voltmeter 26 are applied to difference forming circuits 30, 31 and 32 consisting of a time constant average operator T of the order of 5 seconds, the order of magnitude of the clean time constant, or the casting time of the group, followed by an adder which subtracts the average of the instantaneous value. These circuits will be described in more detail with reference to FIGS. 4A and 43.

 Furthermore, the adder 35 which receives a set reference voltage 35a and, in opposition, the average stator voltage developed by the average operator of the circuit 32, outputs a reference deviation signal.

 The deviations produced by the circuits 30, 31, 32 and 35 are applied to a linear model 4, the function and structure of which will be specified with reference to FIGS. 5A and 53. This model calculates the contributions of the ap plies with respective coefficients. according to a specific law for a selected operating state. These contributions are added, in the adder 40, to the effective variation of the difference in electrical power obtained by differentiating the difference coming from the circuit 32, in a differentiating circuit 33.

The signal coming from the adder 40, or the behavior deviation voltage, and the difference signals coming from the circuits 35, 30, 31 and 32 are applied, through amplitude filters respectively 50 to 54, whose structure will be described later, to an adjustable regulator 5, the function and structure of which will be specified with reference to FIGS. 6A and 6B. The output signal, the adjustment signal, results from a linear combination of the products of the difference signals applied. by adjustable coefficients according to a proportional and integral function law of the behavior difference voltage resulting from adder 40,
On the other hand, the assigned mechanical power image signal produced by the circuit 13, is transformed by the circuit 34 whose structure and operation will be specified with reference to FIGS. 3A and 3B, into an assigned mechanical power deviation signal. which corrects, in the adder 34e, the adjustment signal from the regulator 5.This deviation signal is representative of the mechanical power variation trend, in particular during regime changes due to abrupt power variations called , and provides an early correction to the setting signal.

 The control signal developed by the regulator 5 is applied to a circuit 55 comprising a proportional channel and an integral channel in parallel, the structure of which will be specified with reference to FIG. 9. The integral channel of the circuit 55 comprises a constant-state integrator. time equal to the time constant of the average operators of the deviation forming circuits 30, 31, 32, so that the circuit 55 has a transfer function inverse to that of these deviation forming circuits. The excitation control signal restored by the circuit 55 is coherent with the signals delivered by the sensors 22, 25 and 26.

 The excitation control signal passes through a limiter circuit 56, controlled by the voltage signal from the voltetric sensor 26. This limiter circuit 56, an embodiment of which will be described with reference to FIGS. 7A and 7B, transmits the signal of excitation commando restored by the circuit 55 to the conventional excitation generator device 57, as long as the voltage supplied by the voltmeter 26 remains between two thresholds, a maxiiai and a minimum. If the stator voltage reaches one of these thresholds the excitation command takes a corresponding assigned value. The excitation generator 57 includes conventional excitation with its corrected driver circuitry to reduce the time constant.

 By considering the Fresnel diagram of FIG. 2, an alternator developing an electromotive force E and having a transverse reactance Xt is coupled to an infinite network of voltage Yr and pulsation n0 through a coupling reactance Xq. Since the mains voltage Vr defines the origin of the vector arguments, the electromotive force R has an argument d, called the total angle, which is very substantially the corresponding argument of the rotating magnetic flux generated by the rotor, while the module of E is determined by the excitation intensity. As neglecting the dissipative losses, Xt and 1q are pure reactances, the voltage drops in Xq and Xt have the same argument, so that the stator voltage V, at the point of of coupling between the internal reactance Xp and the coupling reactance Xq at its end aligned with the ends of the vectors Yr and E. The vector V argument, or transport angle is +, and the angle Q between the vectors V and E is said internal angle.

The power exchanges between the alternator in the operating state corresponding to an electromotive force vector of module E and argument 8 and the infinite network Vr are expressed by
3 E Vr sin #
Active power P = (1)
Xt + Xq
3 Vr (E cos # - Vr)
Reactive power Q = (2)
Xt + Xq
Taking into account that, in steady state, the electric active power is imposed by the applied mechanical power, there exists a family of operating states corresponding to an active power value, defined by a constant value of E sin # #; the operating state is completely defined only if the reactive power Q is of chosen value.

 If the modulus of the electromotive force depends in a substantially linear manner on the excitation, it is a quantity which can not be measured directly. Moreover, the elements that define the infinite network, voltage Vr, reactance 1q 'are virtual elements, which can be apprehended only from the point where the alternator is coupled to the network.

At the point of coupling the operating state of the alternator can be expressed by the following formulas p = 3 EV sin Q (3) '
#t
Q 3 V (E cos # - V) (4)
xt
In addition to the infinite network
3 V Vr sin #
Pe = (5)
xo
3 Vr (V cos # - Vr)
Q = (6)
xo
Moreover the origin of the arguments is inaccessible so that # is not measurable directly.However, since the frequency of the network is a magnitude stabilized in time, so that the period of the network can be considered as a base of absolute time, and that the electromotive force vector is substantially fixed in the direction relative to the rotor of the alternator (considered to have a single pair of poles) the angular position of the rotating mass of the turbo-alternator group is defined by #ot + s. It is therefore the total angle that intervenes in the equations defining the dynamic equilibrium of the group. The internal angle, which would be accessible to a measurement, depends on a relation involving the variable parameters of the network (Iq and Yr).

The angular velocity of the group (considered to have a single pair of poles) a @ is given by
d
# = #o + # (7) dt
We chose the accessible output quantities Pe and
V to express the operating state, these quantities being directly representative of what is required of the group in priority.

 To obtain linear relations expressing the operation, one considers the variations of magnitudes in a narrow sector around a state of reference; in order to obtain valid linear relations in all the possible operating range, a reference state defined by the average values of the quantities is chosen over a time interval preceding the current moment, sufficiently long to correctly represent a steady state, and sufficiently short for that the differences between the current values and the average values remain in the sector where the approximation to the small variations is suitable. It will be understood that the time constant own of the greupe, also called time of launching of the group, represents the order of magnitude the minimum interval of time to consider, during which one can define a mechanical steady state.

il
l'équation 11-1 dérive de la relation 7 ; l'équation 11-2 exprime que l'écart des puissances mécanique et acti ve électrique est compensé par une variation d'énergie cinétique de rotation de la masse tournante (coefficient H) et une absorption d'énergie par les amortisseurs (amortissement de type visqueux) (coefficient D) ; les équations 11-3 et 11-4 traduisent le système des équations 1 à 6, en combinaison avec l'équation 7, pour le secteur aux petites variations. The current average Y of a variable variable Y in time, taken with a time constant T, is given in symbolic notation, with s Laplace variable, by s
#
lilfar (8)
AY
The difference between the instantaneous value of Y and its current average is therefore
Y (s)
Y (s) = Y (s) - (9)
or # Y (s) = Y (s) (10) l + Ts
In the sector where the small variation approximation is valid, with the definition of the above deviations, we can establish the following relations d ## = ## (11-1) dt d ## = H (#Pm - # Pe) - D ## (11-2) dt d # Pe = ### + ss ## + & gamma # Pe + ## E (11-3) dt
# V = h1 ## + h3 #Pe (11-4)

he
equation 11-1 derives from relation 7; Equation 11-2 expresses that the deviation of the mechanical powers and electrical activity is compensated by a kinetic energy variation of rotation of the rotating mass (coefficient H) and an absorption of energy by the dampers (damping of the type viscous) (coefficient D); equations 11-3 and 11-4 express the system of equations 1 to 6, in combination with equation 7, for the sector with small variations.

On the other hand, the adjustment of the excitation, in a substantially linear relation with the modulus E of the electromotive force, will depend on an input quantity U which will be defined later and output quantities n, Pe and
V. In the sector of small variations
#E = k0 #U + k1 #V + k2 # # + k3 # Pe (12)
The combination of 11-3, 11-4 and 12 gives
d #Pe = (# + # k1h1) # # + (ss + # k2) # #
dt
+ γ + # (k3 + k, h3) # Pe + # k0 # U (13-3)
The coefficients k0, k1, k2, k3, of the equation (12) are variable, and their adjustment reflects the self-adaptation of the regulator 5. In the system of equations 11, only the equation 11-3 is at coefficients (α, ss, & gamma, #) variables coefficients D and H of equation 11-2, and the coefficients h1 and h3 of equation 11-4, are determinable from parameters of the turbo group. alternator. Solving equation 13-3 by adjusting the coefficients applied to the deviations gives the proper fit of the coefficients of equation 12.

The system of equation 11 with substitution of 13-3 to 11-3 is in the form:
dX / dt = As X + Bs U (14) where As and Bs represent two matrices of coefficients. It can be shown that if two matrices Am and Bm define a steady-state solution to equation 14, there is a behavior gap vector # such that
# = Am X + Bm U - dX / dt (15) whose cancellation translates the equality of matrices Am and As and Bm and B8. The laws of adjustment of the matrices As and following g

donnent au système des propriétés d'hyperstabilité de Map ounov.

give the system properties of Map ounov hyperstability.

 The matrices As and Bs are the model matrices that define the imposed behavior of the operating state.

 It will be remembered that the input state vector has only one component that can be used as control quantity, namely the excitation control. Consequently, we will be able to control the behavior only if this control can result from the action on a single magnitude, and consequently if it can reveal a parameter of the behavior deviation vector whose cancellation coincides by one-to-one relation with the behavioral error vector.Or, referring to equations 11-1, 11-2 and 13-3 which define the components of this deviation vector in a reference system Pe, # and, we find that the next component 8 is identically zero, and that the next component n can be canceled by a prior adjustment, so that the next component Pe satisfies the single parameter requirement and can be likened to a coefficient of behavior difference E .

This coefficient E thus explained can be expressed as a function of AU, V? A n, A Pe and APe for
dt transformations involving 11-4 to eliminate the term in A8. The coefficients assigned to the deviations are computable either directly from the construction parameters of the group, or from quantities defining a chosen operating state, insofar as there remains an indeterminacy (choice of behavior).

It can also be shown that the system retains its properties of hyperstability when the differences taken into account are filtered through filters with a function of a strictly positive transfer such as
1 + Tfs establishes a weighting of the variations of the deviations on the time constant Tf. This reduces the noise by reducing the bandwidth, and furthermore the determination of the derivative with respect to the time of # P @ is facilitated. Indeed if 1 ds
# Pef = # Pe, (#Pef) = # Pe.

1 + Tfs dt 1 + Tfs
The operation of the regulating device of FIG. 1 can be summarized by taking again the notations of the preceding development; the circuits 30, 31, 32 elaborate the current averages of n, Pe and V, and the distances an # Pe and AV. The differences are applied to the model 4 which reproduces the matrices Am and Bm, so that the circuit 33 developing the derivative d / dt # Pe, the adder 40 elaborates the deviation coefficient or behavior deviation voltage. applying the deviations and the deviation coefficient to the regulator 5 which carries out the adjustment laws (16) and (17), the adjustment signal # E is obtained as output, which causes the system's hyperstability.

The circuit 34 is designed to intervene essentially during significant imbalances between xecanic and electrical power. In effect, the action of the intake valves 11-12, controlled by the regulator 13, is insufficiently fast to avoid on its own accelerations or decelerations of the group such as sin d stannule, which eliminates any possibility of regulation. We will therefore intervene in anticipation of the risk, on the excitation by the transiently de-regulating, so as to brake or discharge the alternator to counter in advance accelerations or decelerations. As the regulator 13 develops a picture Pm of mechanical power assigned, that is to say, the mechanical power that will provide the turbine when the action of the intake valves will become effective, the aignal Pm carries a trend information of the mechanical power. Furthermore, in the case of a corrective and transient maladjustment in addition to the play of the main regulation, it is not essential that it be strictly dosed. Also the circuit 34 can be a predetermined model. Theoretical considerations show that between signal Pm and excitation a transfer function of the form
1 8 - a
b0 1 + T1s (18) is well suited, the coefficients bo and a'3 being calculated from the known parameters of the group and the time constant T 'being of the order of 0.2 Tf (Tf time constant of the filters of the main model).

 But, as the tuning signal results. in taking account of the differences in the output image quantities of the group, the error signal U applied to the adder 34a must also be homogeneous. The signal Pm will therefore, in the circuit 34, be applied to an operator deviating from the current average of Pm.

The structure and function of the circuit 34 will be better understood with reference to FIGS. 3A and 3B. The assigned mechanical power image Pi is applied to an operator 101 with a transfer function Ts which outputs
1 + Ts a difference signal difference between the current value of
Pm and its current average 1 Pm taken with the constant 1 + e8 of time T used for image signals of output quantities. The operator 102 has the transfer function defined by the relation (18). The general transfer function of Pm to UA is the product of the transfer functions of the operators 101 and 102. In analog layout according to FIG. 3B, the difference operator consists of an active filter 104 composed of an amplifier with a Feedback resistance and attacked through a capacitance, resistor and capacitance defining the time constant T.

The transfer function operator defined by the relation (18) comprises a parallel capacitance resistance circuit 105 defining a time constant Tf and an input link for a filter 106 composed of an amplifier with a parallel capacitance feedback feedback circuit. at time constant T '. In an exemplary embodiment, the circuit 105 has 0.1 megohm and 1 microfarad, and the circuit 106 0.1 megohm and 0.2 microfarad.

The development of the current average values and the output magnitude deviations are elaborated by filtering circuits functionally represented in FIG. 4A, FIG. 4B being an analog arrangement. In FIG. 4A, the representative signal, iti V is applied to a transfer function operator 110 which
1 + Ts elaborates the current average V over the time interval T.

The signals V and V are respectively applied to the direct and inverting inputs of a differential amplifier 111 which thus produces the gap AY, according to the process already described. The arrangement of FIG. 4B shows a circuit with an amplifier stage 112 provided with a parallel capacitance resistance feedback network T. The input resistance and the feedback resistance of the stage 112 being equal, the function of transfer of the stage 112 1 is and at the exit of the stage we obtain - V. La som
1 + Ts me of the output and the input of the amplifier 112, obtained at the node of the adder 113 thus has a transfer function 1/2 # 1 - # = .In the example + Ts 1 + Ts ple of embodiment already mentioned with reference to FIG. 3B the time constant T is obtained by combining 1 megohm and 4.7 microfarads.

 A model for developing the coefficient of behavior difference is shown in functional representation in FIG. 5A and in analog layout in FIG. 5B.

In FIG. 5A, each of the four difference inputs # U, A n, AV and APe drives a first-order filter (121 to 124) with a time constant Tf, operating the transfer function 1/1 + 1TfS. In addition, the input #Pe drives a transfer function filter s / l + Tfs followed by an inverter 126. The adder 120 performs the sum of the filtered gaps, each assigned a coefficient b0, a'2, respectively. a'l, a'3. The coefficients b0 and a'3 are equal to the corresponding coefficients of the correction term construction circuit presented in FIGS. 3A and 33.

In FIG. 5B, the first-order filters 131, 132, 133 are constituted by amplifiers with a parallel capacitance resistance feedback circuit; the power deviation input Pe 134 does not have any
1 + T @ s filter, and thus performs a transfer function 1 + TfS comparable to the difference 1 + 1TfS - (~ Tf) 1 + 5TfS e The potentiometers of the adder 130 make it possible to adjust the various coefficients taking into account inversions provided by the amplifiers. The adjustable amplifier 135 makes it possible to definitively scale the coefficient of behavior deviation E. In the embodiment already presented, Tf is obtained by a resistance of 0.1 megohm and a capacitance of 1.1 microfarad.

The adjustable regulator shown in FIG. 6A in functional arrangement comprises four branches, respectively strongly 140, 150, 160, 170, each assigned to a difference signal respectively AU, An, V, APe which join on an adder 179. Each branch comprises similar elements that are referenced with the same unit number, and the tens and hundreds numbers of the branch, so that the description of a branch is appropriate for all four. Thus branch 140 includes, from the entry to the output, a multiplier circuit 141 which receives on a multiplicand input the difference signal (AU) and on a multiplier input the coefficient of behavior deviation E. The output of the multiplier 141 attacks two channels, a proportional channel 142 , with an amplification factor 0, and an integral channel 143, transfer function ss 0. The channels 142 and 143 meet, with
The output of the adder 145 drives the multiplier input of a multiplier circuit 146, whose multiplicand input receives the signal from a multiplier circuit 146. gap (AU). It is clear that the response of this branch is of the form AU + A UE (
8
For branches 150, 160 and 170 the coefficients
K and and respectively comprise the indices 1, 2 and 3.

It will be understood that the regulating output signal of the regulator # Ve follows the law of adjustment of the relations (16) and (17) which ensure the hyperstability in the Liapunov sense of the system.

 FIG. 6B shows an analog embodiment diagram of a branch 180, with an input multiplier 181, a proportional channel with a feedback gain amplifier 182, a potentiometer adjusting the input voltage, an integral channel 183 with a capacitive feedback amplifier, and a potentiometer 184 for introducing the constant coefficient. The adder 185 is constituted by the three-resistance convergence node connected to one of the inputs of the multiplier 186. The resistor 189 is an adder resistor and adjusts the branch's contribution to the development of the multiplier. In an exemplary application, the multipliers deliver an output signal whose voltage ol volts is one-tenth of the product of the input voltages expressed in volts; the integral branch has an input resistance of 0.01 megohm and a feedback capacitance of 10 microfarads.

It has already been explained that, the differences in magnitudes applied to the control device being taken with respect to a current average over a time interval T, with a transfer function TS 1 Ts, the restoration of the excitation control voltage V e from the setting signal 1 + Ts #Ve implies a reverse transfer function, the
Ts time constant T being the same in both transformations. The transfer function can be written 1 + 1 in
So that the voltage V e is the sum of the difference #Ve and its integral with respect to time with the time constant T.Figure 9 represents an analog circuit arrangement performing this operation, and comprising, after one stage buffer 220, two parallel channels, an integral channel 221 composed of an amplifier with a feedback capability and an input resistance, resistance and capacitance jointly determining a time constant T, and a proportional channel 222 with a resistive feedback amplifier the two paths joining through two resistors of an adder 223. In the embodiment already mentioned, the time constant of the channel 221 is determined by a resistance of 1 megohm and a capacity of 4.7 microfarads.

 In order to avoid, during disturbances of very great amplitude, that the regulating device continues an aberrant operation (such as, for example, that the internal angle reaches the value), an excitation limiter signal is available on the driver threshold of which Figure 7A has a provision. The amplitude limiter comprises two amplifiers 203 and 204 whose amplification factors are adjusted by feedback to an equal and relatively high value (for example). The outputs of these amplifiers are connected by a resistor adder 2o5, while the inputs are driven by a signal sum of the stator voltage V applied to the conductor 120 and a threshold voltage determined by the potentiometers 201 and 202 on the amplifiers 203 and 204 respectively, these threshold voltages being equal and of opposite sign at the limits, maximum and minimum, that one imposes on the stator voltage V. The operation will be explained with reference to FIG. 7B.

 The amplifiers 203 and 204 provide an output voltage which is limited in positive and negative values by saturation so that, in a narrow range around the zero of the input voltage, the output voltage is proportional to the voltage. input, and for input voltages beyond this range, the output voltage is blocked on the saturation value. Since the input voltage is an arithmetic average of the voltage V and the threshold voltage, the own response of the amplifier 204 is the same as the dashed curve 206, with a positive bearing up to the point 206a, a fall linear to the point 206b, and finally a symmetrical negative bearing of the original positive bearing. The proper response of the amplifier 203 presented by the dashed curve 207, also comprises a positive bearing up to the point 207a, a linear drop until point 207b, and beyond a negative level. The combined response at point 205 has the appearance of the solid line curve 208, which has a positive plateau to point 206a, a linear fall to point 208a of ordinate zero, the abscissa corresponding to that of the point 206b, a zero-voltage bearing to point 208b, of Swiss nome abs as point 207a, a linear drop to point 207b, and beyond a negative bearing.The voltage at node 205 is added to the voltage of excitation control, so that, as long as the voltage V on the conductor 200 is between the threshold values, maximum and minimum, imposed, the contribution of the limiter to the excitation is zero. Beyond the thresholds the limiter's contribution intervenes effectively to neutralize the action of the regulator.

 The sensors and the control device are the seat of inevitable noise. These noises have a statistically zero action until the exit of the model 4 of FIG. 1. However, because of the presence of multipliers and integral channels in the regulator 5, and of the integral channel of the feedback circuit 55 the noises may have a non-zero output at the output of the control device (correlated noise products), resulting in a slow drift of the operating state. To reduce the effect of the noises, it is possible to have on the inputs of the regulator 5 amplitude filters 50 to 54, which, as shown in FIG. 8B, have a linear or amplitude response 219 beyond two abscissa thresholds 219a and 219b, symmetrical with respect to zero, and a zero response between these thresholds. By adjusting the gap between the thresholds to four times the noise standard deviation at the point in question (filter input), 99% of the noise (Gaussian noise) is eliminated.

 Such a filter may have the structure shown in FIG. 8A. The input voltage is applied to the conductor 210, connected through the resistor 210a to the output 210b, and coupled to the direct inputs of two differential amplifiers 213 and 214 mounted in threshold comparators. The inverting inputs of these amplifiers 213 and 214 are respectively connected to the potentiometer sliders 211 and 212 which respectively define the positive (219b) and negative (219a) thresholds of the filter. The outputs of the binders 213 and 214 are connected by resistors respectively to the control points 213a and 214a of diode limiters 215, 217 and 216, 218. The diodes 215 and 217 have their cathodes connected in common to the point control, and their anodes respectively at ground and at the exit point 210b. The diodes 216 and 218 have their anodes jointly connected to the control point and their respective cathodes to ground and to the point of exit 210b. When the voltage of the conductor 210 is between the thresholds, the amplifier 213 is blocked in negative saturation and the amplifier 214 in positive saturation; the diodes 215 and 216 are passing and control points 213a and 214a are massed, so that the exit point 210b is rusted to ground. Beyond the threshold position, the amplifier 213 is in positive saturation and the diode 215 blocking; the control point 213a is at the positive saturation voltage, and the output point 210b is unlocked to the positive values and follows the input voltage on the conductor 210. When the input voltage is more negative than the negative threshold , the amplifier 204 goes into negative saturation, unlocking the exit point to the negative values.

 A regulating device corresponding to an alternator of nominal characteristics has been realized.

Voltage 11,55 KV
Active power 1485 MW
Reactive power 719 MVAr
Transverse impedance 0.66 ohm.

 The determination of the model was made for a point where the powers are nominal powers, the voltage 0.95 of the rated voltage is 10.4 KV and the connection reactance to the network 1q = 0.12 ohm. The stator voltage limiter was set at Vnt 5%, ie between 10.4 and 12.13 KV.

The time constants used were those indicated previously. Simulation tests have shown that
At active and reactive powers and nominal voltage at a mains impedance of 0.072 ohm, a nominal voltage step # 5% causes a damped response with an overshoot of the order of 3%. On a 0.112 ohm network reactance the responses are similar, whereas the classical regulations are very unstable. On mains reactance = 0.096 ohms, the conventional regulations cause a divergent oscillation (pumping), whereas the regulation according to the invention dampens the oscillation on a few seconds, by substitution with the conventional regulation.

 The network load variations, corresponding to line openings and closures, representing reactance variation from 0.036 ohm to 0.056 ohm, then back to 0.036 ohm, give rise to substantially aperiodic reactions.

 Finally, the simulation of a three-phase lateral fault at the terminals of the group transformer, starting from the nominal speed with a 0.064 ohm network reactance, eliminated in 120 ms, shows that the adaptive regulation makes it possible to keep the synchronism, while with conventional regulation the group goes into overspeed.

 Of course, the invention is not limited to the examples described, but embraces all the variants of execution. In particular, the control device can be made in a digital arrangement, for which the person skilled in the art will find in the state of the art arrangements of calculating circuits capable of performing all the transfer functions exposed. It is obvious in this case that the circuits will have to be fast enough to operate in less than one network period, the ultimate resolution of the regulating device.

 Those skilled in the art will also find in the state of the art the theoretical support necessary for calculating the regulating device elements corresponding to particular cases of turbo-alternator groups.

Claims (8)

 1. A method of self-adaptive regulation of the operating state of a turbo-alternator unit coupled to an interconnection network and equipped with an independent speed regulator acting on the intake to the turbine, by acting on the excitation of the alternator, the method according to which signals representative of an input quantity and output quantities of the alternator, the stator voltage, of which these representative signals are applied to a reference model defining behavior chosen to develop a coefficient of behavior deviation, the representative signals are applied to a self-adapting regulator producing a control signal function of the signals representative of output quantities according to a proportional and integral law of the coefficient of behavior deviation in so that this coefficient vanishes asymptotically in response to the regulation, and the adjustment signal is compared to a signal 1 for setting up an excitation control signal, characterized in that, since the model and the regulator delivering the output signals satisfy the contributions of each of the representative signals, they are constituted, for each output quantity, by the difference between an image signal of the instantaneous value of this quantity and the current average of this image signal established with a time constant T of the order of magnitude of the group's own time constant, and for the magnitude of the difference between a nominal stator voltage and the current average stator voltage image signal, the excitation control signal being the sum of the adjustment signal and the integral thereof with respect to time, with the same time constant T.  2. Method according to claim 1, characterized in that the representative signals are filtered through first-order filters with a time constant of the order of 0.04 T at the input of the model.  3. A method according to claim 1 or claim 2, characterized in that a signal developed from the speed controller and representative of the variations of mechanical power assigned by the regulator is added to the control signal.  4. A self-adaptive analog regulating device for the operating state of a turbo-alternator unit coupled to an interconnection network and equipped with an independent speed regulator acting on the inlet to the turbine, for the purpose of the process according to claim 1, wherein the output of the alternator is equipped with sensors delivering image voltages V, Pe and n respectively proportional to the stator voltage amplitude, the electrical active power and the speed of the generator. differential rotation of the group, an adjustable voltage source delivering a stator setpoint voltage Vc> carac three in that it comprises three average operators respectively receiving the image voltages V, Pe etn and each constituted by an amplifier with a loop parallel capacitance feedback forming said time constant T, three adders associated respectively with the operators of average and re first, on the one hand, the image voltage and, on the other hand, the average of this voltage so as to produce a difference signal respectively A V, aPe etAn, a unit gain differential amplifier receiving as input the reference voltage Vc and the voltage average V and delivering a reference deviation signal A, a reference model comprising three first-order filters with a time constant Tf of the order of 0.04 T respectively receiving as input the deviation signals.  Au, an and AV c and a somxator with input attenuators receiving the first-order filter output signals and the difference signal AVe, and outputting a behavior difference voltage, a four-branch regulator combined into output by an adder through attenuatueurs, each branch respectively assigned to said deviation signals and comprising in series a first multiplier receiving the respective deviation signal and the behavior deviation voltage, two parallel paths, a proportional and the Moreover, these two paths are assembled on a soilator receiving a calarge voltage, and a second multiplier receiving on its inputs the respective deviation signal and the signal from the branch summator, the output of the regulator attacking a means. integration device comprising in parallel a proportional channel and an integral channel with a time constant T, the output signal of the signal constituting integration means of excitation cos-  5. Device according to claim 4, wherein the speed controller is equipped with a device producing a signal P11 representative of the power assigned to the turbine by the controller, characterized in that it further comprises a correction network receiving the signal. P and coupled to the output of the auto-adaptive regulator, and comprising in series a time constant average operator T, an adder receiving the signal P and the average signal P1 and delivering a difference signal aPs, and a formed filter by an amplifier with a parallel capacitance input lattice forming time constant Tf, and a parallel capacitance resistance feedback loop forming a time constant of about 0.2 Ti.  6. Regulator according to claim 4 or 5, characterized in that it further comprises an excitation limiter consisting of two comparatcurs threshold jointly attacked by the signal V and respectively adjusted to a maximum and a minimum rated stator voltage, the output voltages of the two comparators are added to the excitation control signal.  7. A controller according to any one of claims 4 to 6 having noise-reducing filters on the deviation signal inputs of the controller, characterized in that the reducing filters have an amplitude response, zero within a interval between two symmetrical thresholds, and proportional beyond this interval.  8. Regulator according to claim 7, characterized in that said reducing filters comprise for the voltage values of each polarity a pair of limiting circuits each controlled by a threshold comparator.