EP1116885B1 - Method and apparatus to control a turbo compressor to prevent surge - Google Patents

Description

The invention relates to a method for controlling a turbocompressor for preventing pumping, of the type specified in the preamble of claim 1, and a device for carrying out the method.

Pumps are referred to in compressors, the intermittent or periodic backflow of fluid from the pressure to the suction side. This condition occurs, for example, at too high final pressure and / or low throughput. In the map, therefore, a surge line can be defined that separates a stable map area from an unstable area to the left of the surge boundary. Operation in the unstable area to the left of the surge limit line is not permitted because serious machine damage can occur within a very short time. To avoid the pumping, ie the operation in the unstable region, a surge limit control is used, which controls a valve on the compressor outlet, which is connected as Abblaseventil with the atmosphere or as Umblaseventil with the suction side of the compressor. By opening the valve, the flow through the compressor is increased to such an extent that the operating point always remains within the stable map range. For such a regulation, a control line (blow-off or blow-by line) is defined in the characteristic field at a safety distance from the surge limit line. When approaching the current operating point to the rule line the blow-off or Umblaseventil (hereinafter called only blow-off valve) is opened more or less.

More specifically, such control operates such that the target value for the flow control is determined from the compressor pressure or from the pressure ratio between discharge pressure (discharge pressure) and inlet pressure, or from a quantity derived from this pressure ratio. This setpoint corresponds to the control line. The measured Kompressorsansaugdurchfluss is compared with the setpoint, and in case of a deviation occurs an adjustment of the blow-off valve. If the operating point of the compressor is on the control line, the control difference of the surge limit controller is zero and the blow-off valve remains in its position. If the operating point exceeds the control line in the direction of the pumping limit, the controller continues to open the valve, if the operating point is to the right of the control line, the controller closes the valve.

In normal operation of the compressor, the operating point of the compressor is clearly to the right of the control line (the design point is typically 20-30% to the right) and the blow-off valve is fully closed. When the operating point is shifted out of this operating state in the direction of the pumping limit, a conventional controller does not open until the actual value falls below the setpoint value, that is to say when the operating point has exceeded the control line in the direction of the pumping limit.

A method according to the preamble of claim 1 is known from DE-2828124 C2. In these known methods, the difference signal from the setpoint and the actual value is supplied once without delay and delayed in parallel to a subtraction point, from which the input signal for the regulator is tapped. This has the advantage that the control loop can also process fast, transient changes in the operating point with sufficient certainty. The effect of the system is that the setpoint of the controller an additional Signal is added, which causes a shift of the control line at transient operating point shifts such that when approaching the operating point to the control line, the safety margin between surge line and control line is increased and the controller responds earlier. The rule line is moved almost dynamically and effective is a new "dynamic rule line". As a result, the safety distance between the control line and the stability limit under transient conditions is significantly greater than under stationary conditions and the compressor is significantly better protected in such critical conditions.

The known method, however, has the disadvantage that, although the safety margin for transient working point shifts, which take place from a stationary state towards the surge line, is increased, but that the controller only delays the changes, with the time constant set on the delay element, can follow. The known method is only fully effective if the operating point shifts from a stationary operating point in the direction of the surge line. On the other hand, the known method only works unsatisfactorily in the case of disturbances which initially result in a displacement of the operating point away from the surge limit and then again in the direction of the surge limit. When moving the operating point away from the surge line, the rule line is initially transiently shifted to the left, with the tendency to return to the steady state value according to the set time constant, that is, usually over several minutes. Only when this condition has subsided can it be assumed that a new steady-state condition exists, and the known method can show its full effectiveness. Until the transient state dies, the dynamic (this is the effective) rule line is to the left of the stationary rule line. The surge limit control therefore intervenes only with a delay, since the operating point is transient must travel another way in the direction of the surge line until the controller intervenes.

US 4,938,658 A and DE 38 09 070.8 A disclose a method for controlling a turbocompressor to prevent the pumping, in which in addition to the continuous control of the Ab- or Umblaseventils depending on the control difference, a quick opening of the blow-off valve is provided which depends on the speed of the pumping limit directed working point shifts is triggered. In addition, a filtering can be provided, which causes the differentiation for speed determination of the operating point without delay responds to working point shifts towards the surge line or blow-off line. According to the present invention, it is about the optimization of the continuous control of the discharge or Umblaseventils and not a superimposed this rapid opening.

The invention is based on the object to improve a method and a device of the known type so that the advantage of increasing the safety distance is always fully usable, regardless of whether the operating point before the beginning of the disturbance in a stationary Operating state is or whether previously transient operating point shifts have occurred.

The achievement of the object according to the invention is specified in claim 1. The other claims relate to further advantageous features of the method according to the invention or the device for its implementation.

According to the invention, the time constant of the delay element is designed asymmetrically. At operating point shifts in the direction of surge limit, the delay element acts as described in the prior art. On the other hand, with a shift of the operating point in the direction away from the boundary line, the delay element operates with a much smaller time constant. This ensures that the control line follows almost instantaneously at working point shifts away from the boundary line, but with the known, much slower time constant when moving in the direction of the surge line.

In other words, according to the invention, the known surge limit control for turbocompressors with stationary set control line, which is effective at a fixed distance to the right of the surge line, extended by a dynamic rule line. This dynamic control line is implemented in such a way that it changes the effective position of the control line in the event of transient shifts of the compressor operating point in the direction of the surge limit, in such a way that it depends From the speed of approach of the operating point to the surge line, the effective control line is shifted to the right in the map in the direction of the operating point with the result that the safety margin between surge line and control line is increased and consequently engages the surge regulator earlier. For very fast shifts of the operating point in the direction of the pumping limit, the control line is shifted to the right, for example by half the distance between the operating point and the stationary control line, with slower operating point shifts, the shift of the control line is lower. If the operating point is moved very slowly, that is to say over, for example, 1 hour in the direction of the stationary control line, the dynamic control line almost completely coincides with the stationary control line.

The control method according to the invention is particularly suitable for applications in which a controlled variable, in particular the flow signal, is very noisy due to flow vortices at the measuring point. Classic PID controllers with differentiating algorithms fail in these applications because the derivative component responds to the rate of change of the controlled variable. In the case of high-frequency (a few Hz) signal disturbances with signal hops of a few percent, differentiating control algorithms are not permitted because they also lead to significant signal changes of the manipulated variable during stationary machine operation. They would also react in steady state operation near the design point and often prevent complete closure of the valve during steady state operation to the right of the control line. For reasons of economy, the valve is to be kept completely closed in stationary operation. The method according to the invention offers significant advantages here, since it does not show this disadvantageous effect even with extremely strongly noisy controlled variables.

Embodiments of the invention will be explained in more detail with reference to the drawing. FIGS. 1-7 each show a simplified scheme of a device according to the invention for controlling a turbo-compressor to prevent pumping.

According to FIG. 1, the pressure in front of and behind an orifice plate (not shown) in the suction line 1 of a turbo-compressor K is measured by measuring sensors 3, 5, from which a measuring transducer 7 forms the actual value for the suction-side compressor throughput V. At the compressor output, a measuring sensor 9 with a downstream measuring transducer 11 comprises the actual value of the final pressure P. Both actual values V and P are input to a computer 13. The computer 13 is provided with a memory in which the course of a stationary control line (blow-off line) in the compressor map, for example represented by P and V, e.g. as a polygon, is stored.

wobei

• κ = isentropischer Exponent
• R = Gaskonstante
• T₁ = Temperatur am Kompressoreinlaß
• P₁ = Druck am Kompressoreinlaß
• P₂ = Druck am Kompressorauslaß

The setpoint is preferably calculated based on the enthalpy difference (head) Δh of the compressor according to the formula $$\mathrm{.delta.h}\mathrm{=}\frac{\mathrm{\kappa R}{\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">T</figure-callout>}}_{\mathrm{1}}}{\mathrm{\kappa }\mathrm{-}\mathrm{1}}\left[{\left(\frac{{\mathrm{P}}_{\mathrm{2}}}{{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{1}}}\right)}^{\frac{\mathrm{\kappa }\mathrm{-}\mathrm{1}}{\mathrm{\kappa }}}\mathrm{-}\mathrm{1}\right]$$

in which

• κ = isentropic exponent
• R = gas constant
• T ₁ = temperature at the compressor inlet
• P ₁ = pressure at the compressor inlet
• P ₂ = pressure at the compressor outlet

The Kompressorauslaßdruck P ₂ can be used to calculate the setpoint or even the setpoint, if all other parameters are constant.

The computer 13 determines from the position of the operating point in the characteristic field defined by the actual values of V and P relative to the control line a nominal value for the throughput V.

Setpoint value and actual value are fed to a subtraction point 17, which forms a difference signal x _d (control deviation). The difference signal x _d is fed to a further subtraction point 19 once via a signal path 21 without delay and once via a delay element 23. The difference between the instantaneous and the delayed signal formed at the subtracting point 19 is supplied as input to a regulator R, which generates a control signal for controlling a blow-off valve or blow-off valve 27 provided on the output line of the compressor K in accordance with a control algorithm implemented in it, by the surge limit control perform in a conventional manner. In that regard, the arrangement described corresponds to the method known from DE-2828124 C2.

The delay element 23 according to the invention is a first-order delay element which is asymmetrical in terms of its time constant. In shifts of the operating point to the left in the direction of the surge boundary line G, ie when the change in time of the differential signal x _d formed at the subtracting 17 has a positive slope, the delay element 23 operates with normal delay. If, on the other hand, the operating point shifts to the right, ie away from the surge limit line G, and thus the change in the time of the difference signal x _d formed at the subtraction point 17 has a negative sign, the delay element 23 operates with a significantly smaller time constant, typically approximately 1 second. This ensures that the controller R can also follow rapid changes in the operating point, which are directed away from the surge boundary line G and therefore "harmless", almost instantaneously.

In the method according to the invention, no differentiating controller is used. In particular, the control algorithm implemented in the controller R is not differentiating. As already mentioned, a differentiating controller can only be used if the input signals are largely free of signal noise. The non-discriminating controller method according to the invention can be used even if the input signals have a high proportion of signal noise, e.g. is caused by vortex formation in the area of the flow sensing sensor 35.

More generally, the circuit branch between the summators 17 and 19 containing the asymmetric delay element 23 may be considered as a "gradient sensor" detecting the direction and velocity of working point shifts towards or away from the surge line. The mode of operation can then be described as follows: In the memory 13, the stationary control line is entered as a polygon. In the summer 17, the difference between the stationary control line (setpoint for the surge limit control) and the current flow of the compressor is formed. This (current) control difference is applied to the controller R, which changes its output according to the implemented control algorithm. Furthermore, from the current control difference by means of the gradient sensor 23 and the adder 19, a virtual control difference is formed as the sum of the difference between the stationary control line and the current flow and the output signal of the gradient sensor 23. This virtual control difference is applied to the controller R as an additional input variable. This results in a virtual rule line, which follows a transient shift point with a set time constant (typically 20 to 60 seconds). After the decay of this time constant, the virtual rule line coincides with the stationary rule line.

For the gradient sensor represented by the member 23 in Fig. 1, various configurations are possible, which will be explained below.

According to FIG. 2, the circuit part 40, which is framed in dashed lines and functions as a gradient sensor, contains an integrator 32 whose output signal is fed back to the input and is added to the input signal of the integrator 32 with a negative sign by means of a summing element 30. The time constant of the integrator 32 changes depending on the movement of the operating point.

3 shows an embodiment in which an additional dynamic rule line is given by an additional function generator 41 (e.g., polygon generator) in addition to the stationary rule line given by the function generator or computer 13. The dynamic control line 41 is formed from the same input variables as the stationary control line 13, however, a "gradient sensor" 40 with a downstream summer 19 shifts the dynamic control line by an amount equal to the difference between the stationary control line and the current compressor flow formed in the summer 17 , changes dynamically. The gradient sensor remembers the stationary distance between the stationary control line and the current compressor intake flow and adds this size of the dynamic control line 41. In the adder 18, the difference between the dynamic control line and the current intake flow is formed and applied as a control difference to the controller R, which in turn adjusts the blow-off / blow-around valve accordingly. The gradient sensor is designed such that it shifts the dynamic control line only in the direction of larger flows, that is to the safe side for the compressor.

According to the invention, with a transient shift of the operating point away from the surge limit, the gradient sensor follows the new distance between operating point and control line without delay. In the direction of the control line, the output of the gradient sensor follows in a time-delayed manner, that is to say slowly, and thus causes a continuous settling into the stationary state.

FIG. 4 shows a possible embodiment of the "gradient sensor" as it is inserted in block 40 of FIG. 3. A feedback integrator 32 is preceded by an asymmetric limiter 31. The output signal of the integrator 32 is applied to the summer 30 at the input with a negative sign. The asymmetric limiter 31 is set to reduce input signals generated by movement of the compressor operating point toward the surge line to very small values, e.g. 0.02, limited. Input signals corresponding to a shift of the compressor operating point away from the surge limit are hardly limited, e.g. by a limit of 1.

The effect will be explained by an example. Suppose a shift of the compressor working point in the direction of the surge line, for example, such that the operating jumps almost jump jump from an operating point at a distance of 20% from the rule line to a point at a distance of 10% from the rule line. Due to the agreement "control difference xd = setpoint minus actual value", this means a change in the control deviation xd from -0.2 to -0.1. The output of the integrator 32 is before the start of the disturbance -0.2, the input of the summer 30 jumps from -0.2 to - 0.1. The output of the summer 30 is +0.1. Since the limiter 31 limits positive values to a maximum of 0.02, the integrator receives an input signal of 0.02 and thereby integrates it with a time constant of 50 seconds. Only after the decay of this settling time, the output of the integrator 32 coincides with its input. In the summer 19, the difference between the input of the adder 30 and the output of the integrator is formed and the dynamic Rule line switched on. In the steady state, the output of the summer 19 is zero, with transient shifts of the operating point in the direction of surge line, the output of the summer 19 transiently assumes a positive value whose amplitude is proportional to the size of the operating point shift and proportional to the speed of the operating point.

When the operating point shifts away from the rule line, the lower limit of the limiter 31 engages and the integrator follows with a small time constant, e.g. 1 second. The gradient sensor is thus almost ineffective in this direction. However, this has the advantage that the integrator very quickly assumes the new steady state value. If the operating point changes transiently, e.g. first from -0.2 to -0.3, and then to go to -0.1, according to the invention, the output of the integrator follows very quickly to - 0.3. The full dynamics of the system is thus available. In a prior art technique, the integrator would follow the change from -0.2 to -0.3 with the same time constant as in the other direction. With a short succession of both disturbances, the movement away from the rule line would be ignored.

FIG. 5 shows a further embodiment of the gradient sensor 40. Instead of the limiter 32, two amplifiers 33 and 34 are used whose outputs are connected to the integrator 31 via a changeover switch 35. The amplifiers are set to different amplification factors, the amplifier 33, for example, to 0.02 and the amplifier 34 to 1. The switch 35 is controlled by a differentiator or a sign generator 36 and switches, depending on the sign of the input to one or the other amplifier , This ensures that at a shift in the operating point in the direction of the pumping limit, a small gain and thus a large time constant is effective, and with a shift in the direction away from the surge line, a large gain, ie small target constant.

A further embodiment is shown in FIG. 6. Here, instead of the limiter 31, an integrator 32 with a parameter-adaptable time constant is used. Depending on the direction of change of the input signal, the time constant of the integrator is switched over the adaptation block 37 between a large and a small value.

FIG. 7 shows a further embodiment whose gradient sensor 40 uses a special structure-switchable integrator NFI 32. Via a control input, which is connected to the output of the differentiator 36, the integrator switches between the two modes of integration and tracking. If the operating point of the compressor shifts in the direction of the surge line, the differentiator DIF 36 switches the integrator 32 into the operating mode Integrate. The integrator, with its set time constant (typically, for example, 50 seconds), follows the change in the input signal to the summer 30. On the other hand, if the operating point shifts away from the surge line, the differentiator 36 switches the integrator 32 to the tracking mode. The output of the integrator follows without any time delay the second input, that is, the output of the summer 17. As soon as the operating point moves back towards the surge line, the differentiator 36 switches the integrator 32 back into the operating mode Integrate. The output of the integrator follows from this state with its set time constant to the new value.

This effect is also explained in the example below. If the operating point changes transiently, for example initially from -0.2 to -0.3, in order then to go to -0.1, according to the invention the output of the integrator follows very quickly to -0.3. Since the differentiator 36, the shift away from the surge line, that is change in the control difference xd from -0.2 to -0.3, the integrator will always assume the same value as the control difference, ie the output signal of summer 17 during this process in tracking mode. The output of summer 19 is always zero and the gradient sensor is therefore zero ineffective at a work shift away from the surge line.

If, on the other hand, the operating point moves in the direction of the surge limit, that is to say if the control difference xd changes from -0.3 to -0.1, the differentiator 36 recognizes the change in direction and switches the integrator into the operating mode Integrate. The output of the integrator 32 follows the input with the set time constant (e.g., 50 seconds). As a result, the adder 18 transiently adds up a positive quantity which has the same effect as the described dynamic rule line.

The full dynamics of the system is thus available, since the gradient sensor is already effective with the beginning of the operating point shift out of the xd = -0.3 position. In a system according to the prior art, the gradient sensor would move the control line in the direction away from the surge line only at a control difference of -0.2 in fast successive operations. With a short succession of both disturbances, the movement away from the rule line would be ignored.

The embodiment variants of the gradient sensor 40 illustrated in FIGS. 4-7 can also be used in the embodiment according to FIG. 2. It only needs to be replaced in the arrangement shown in FIG. 2 of the integrator 32 and summer 30 existing gradient sensor 40 by one of the gradient sensors 40 as shown in FIG. 4-7. A separate drawing and description of such arrangements will be omitted here. The advantage of the arrangement according to FIG. 2 with respect to that according to FIG. 3 and its variants according to FIGS. 4-7 is that only one Function generator 13 (polygon generator) is required for the imaging of the stationary control line, while a dynamic rule line function generator 41 is omitted and the dynamic rule line exists only virtually since it is calculated as the distance between the stationary control line and the current operating point.

Claims (5)

1. A method for controlling a turbo compressor for so as to prevent surge,
   wherein a valve (24) branching off from the compressor outlet is controlled by a controller (R) on the basis of a supplied input signal and a predetermined control line,
   wherein the input signal of said controller is obtained from a difference signal (x_d) corresponding to the difference between a continuously determined actual value of an operating variable of the compressor (K) and a target value depending on the position of the set point in the characteristic diagram,
   and wherein the difference signal both undelayed and delayed by means of a delay element (23) is used for forming the input signal of the controller (R) to thereby obtain a dynamic shift of the control line in case of working point shifts,
   characterized in that said difference signal (x_d) is delayed with different time constants depending on the direction of its change (increase or decrease) in such a way that the controller (R) responds to working point shifts is the direction towards the surge limited line with a slower shift of the control line and responds to working point shifts in the opposite direction with a faster shift of the control line.
2. Method according to claim 1 wherein the difference signal is sent once via the delay element and once without delay to a subtraction point (19) from which the input signal for the controller (R) is taken.
3. Method according to claim 1 or 2, wherein the delay of the delay element is substantially or nearly zero in the case of working point shifts directed away from the surge limit line.
4. Device for carrying out the method according to one of claims 1 to 3 for controlling a turbo compressor to prevent surge, comprising measuring transducers (3, 5) for determining the actual value of one or more operating variables characteristic for the working point of the compressor (K), a target value generator (13) having a predetermined shape of a control line (A) in the characteristic diagram of the compressor, a difference element (17) for generating a difference signal (x_d) from the target value and the actual value, the controller (R) for generating an actuating signal for actuating a valve (27) at the compressor outlet, and a circuit supplied with said difference signal (x_d) and comprising a delay member (23) for generating the input signal for the controller (R)
   characterized in that said delay element (23) is an asymmetric delay element providing a delay which is greater in case of a change of said difference signal (x_d) corresponding to a shift of the working point in the direction towards the surge limit line (D) than in case of a change of the difference signal (x_d) in the opposite direction.
5. Device for carrying out the method according to one of claims 1 to 3 for controlling a turbo compressor to prevent surge, comprising measuring transducers (3, 5) for determining the actual value of one or more operating variables characteristic for the working point of the compressor (K), a target value generator (13) having a predetermined shape of the control line (A) in the characteristic diagram of the compressor, a difference element (17) for generating a difference signal (x_d) from target the value and the actual value, and a controller (R) for generating an actuating signal for actuating a valve (27) at the compressor outlet
   characterized by comprising a gradient sensor (40) determining the amount and direction of the change of the difference signal (X_d), the output signal of said gradient sensor (40) controlling a function generator (41) in which the dynamic control line is stored, wherein the sum of the output signal of the function generator (41) and the actual value of the compressor throughput (V) is supplied to said controller (R) as an input signal.

Images